anyone else seen the Panny AG-AF100, due in late 2010?

We were talking about this a few months ago...

more news about it here:

http://mirror.dpreview.com.s3.amazonaws.com/AG-AF100.pdf

CVP have a little info on it in their AVHC camera section. Somewhere it says about £6000
Sony are also planning a similar camera based on the new Nexus digital camera range with interchangable lenses, apparently due in December.

Both the Panasonic AF100 and the Sony one are pretty much vapourware at the moment, but the AF100 seems a bit more solid.

Certainly very interesting to watch the response from these two to the HDSLR craze.

IBC should be interesting!!

Dave

Yep, the Panasonic 4"/3 camera's been a hot topic here for several months now.

I'll be at IBC again, doing training sessions as last year. Plus, I'll be available via the stand that BPR is putting together, on historic TV.

The European version will be the AG-AF101. It will also have HDSDI output. Here is some more info. http://panasonic.biz/sav/broch_bdf/AG-AF100.pdf
I can't wait to get my mits on this, it's due to be shown at IBC and be available to buy in December.

Testing cameras at IBC is a real challenge, but, if there's a unit there, I'll be pushing Panasonic to let me have a play with it.

It's an ugly looking blighter! :)

But it's a big camera, it has to be, that's the whole point :)

What, it has to be ugly? It's not the normal, aesthetically pleasing design I'd associate with Panasonic.

Hmmm, I can think of a lot of big cameras which aren't ugly, but I note Alan's smiley face....

I could allow it being ugly if operationally necessary, but it seems to be exactly the opposite - an ergonomic disaster. Hand holding? Forget it. Please, please let Panasonic change the styling before final release? The pictures were only of a mock-up, weren't they?

I also think that price is a lot for a camera with an AVC-HD codec, and that must have been underlined by the Canon XF-305 with it's (fully broadcast approved) 50Mbs MPEG2 recording. OK, Panasonic may not want to go with that, but if it's going to be anywhere near £6,000 (dearer than the Canon) I'd expect AVC-Intra from them.

I don't care what a camera looks like as long as:

1) It is reliable (that includes been knocked about a bit).
2) It starts up fast.
3) It produces a great picture.
4) It is intuitive to use.
5) It is ergonomic (related to above)
6) It gives me recording options over and above the native system (ie NanoFlash etc).

Not nescecerily in that order. All equally important IMHO.

This has got me thinking a bit. Is there a limitation to digital sensors? I'm thinking about the resurgence of research into analogue computing to help speed. Perhaps an analogue sensor of some sort that allows control over the sensor area without much detriment on resolution? So smaller area for deeper DOF and vice versa. I'm not talking about an analogue system through and through, but just at the sensor level somehow. An organic sensor of sorts.

I'm with Simon in this, looks are irrelevant, performance matters. Had looks been an issue, we'd never have got RED ONE, ARRI D20/21, and lots more.

There's not chance of a camera having controllable image size, it's not on. All sensors are already analogue, they accumulate charge (measured in Coulombs) in charge wells, that charge is read out as a current into a capacitor which produces a voltage, which is then sampled before it can be quantised into digits.

The only way the sensor can be made independent of format/size/resolution, is to return to scanning. Vacuum pick-up tubes (Vidicon, Plumbicon etc) are all independent of resolution and format, those are imposed by the scanning process. Resolution is limited by charge leakage within the sensor layer and by the size of the scanning spot which discharges it to form the video signal current. No other way.

Interestingly film is part digital and part analogue. Each grain is either exposed or not exposed, a binary state. But each grain is far smaller than a pixel, so it's the aggregation of exposed grains that makes up the density for each pixel.

Couple of things worth bearing in mind:

DSLRs are truly ergonomic nightmares, it hasn't appeared to hinder this market.

Second, a retail of approx $6000 has been set, I can't seriously see it hitting the streets at £6000.

The landscape of cameras is changing dramatically (our personal itinerary is testament to that) with only one shoulder mount among them. Things are moving towards being more modular and inheriting some elements of the Red/Cinema concept (I.e you build the camera to your spec starting with a body and customise)

The final model obviously has not been completely spec'ed, there are huge lists over on DVXuser asking for intra/P2 options etc, so there is still time.

Its main competitor will not be the Canon or the Ex's, it will be scarlett and the subsequent promise from Sony.

My main concern is this could be a big nail in P2's coffin if they don't go there. Which would be a shame, but I reckon the market they are looking at (Indie film-makers) aren't bothered by this.

Ugly - well, I didn't marry my camera.

True, but isn't that just because they are so (relatively) cheap? You can get a video quality at a price which is a fraction of what it might be from a dedicated video camera. So it becomes worth adopting a make-do-and-mend attitude. Once you put down £1,000s for a dedicated camera, don't you expect better?

I was just going by earlier posts, we'll have to wait and see. I do suspect the price will depend quite a lot according to lens option. Is that price for body only? With basic lens package?

Yes, but it's only in so far as lens options that this camera can be considered modular. In other respects it looks like current AVCCAM cameras, and doesn't look to have much allowance for external modules.

I agree, but my reason behind mentioning the new Canon is a feeling that the XF305 will establish a sense that paying this sort of money entitles one to a fully broadcast approved codec. Which realistically means 50Mbs 422 XDCAM-HD or AVC-Intra 100 in this case. Panasonic have tried to dodge the issue by saying that "but that would mean P2, which would greatly increase the cost".

What Canon have also shown is that you can get a full broadcast codec in a camera without it meaning P2 or SxS - just use Compact Flash. (And Compact Flash is easily capable of handling AVC-Intra 100.)

Agreed, apart from the compact flash element.

I've been spitting and cursing at our Red's CF (Bent pins way to easily on the internal MCR of the computer, awkward to load into camera and reader.)

Also a R1 is hardly known for its ergonomic prowess and that costs a fair bit more thant most of the stuff we're discussing here. I do reckon you build your own support gear around a body these days!

Hi
This appears to be a new wait-for-it page:
http://pro-av.panasonic.net/en/af100/index.html

Thanks Paul looks interesting and I see some Nikon prime lenses there too.

I have asked panny if I can be a beta tester but I am sure I will probably have to buy one to do that!

At least we will then have a Black and a RED brick to chose from:D

I'm still not sure whether they'll get the resolution/aliasing balance right. It needs a proper optical low-pass filter, what the Canon 5D/7D hasn't got yet.

Could they not get a good result (from a multi-megapixel chip) IF they are able to read out the whole sensor at frame rate, then do a decent digital low pass filtration job before down-conversion?

Unlike an OLPF, that would still enable such a camera to have uncompromised stills ability.

Yes, absolutely right, that's what they SHOULD do,but the power consumption will be prohibitive, and the sensor chip will get very hot. You only have to look at proper large-format cameras (RED, ARRI) to see what that means. In a smaller way, the Thomson Viper and LDK8000 are vertically oversampled in order to do various scanning formats, but the simple process of adding together the sub-pixels is a poor way to do it, my tests on the cameras show just how poor that process is.

Its finally here and I await the test reports:http://blog.panasonic-broadcast.com/ibc-product-highlights/ag-af101/

Hi
"we were told that the anti-aliasing technology they are developing to give it a significant edge over HD DSLRs isn't finished yet, so any opinions you might see about the pictures won't be the final word. Street price is likely to be under £4,000"
http://urbanfoxtv.blogspot.com/2010/09/new-panasonic-af100af101.html

Barry Green's initial feed back

http://www.dvxuser.com/V6/showthread.php?t=222673

"I'm posting this from the show floor at IBC, where the first working prototypes of the AF100 are being shown to the public. I've had the luxury of working with the pre-production AF101, which is the European model of the AF100. Same basic camera, because all of them are NTSC/PAL switchable.

This prototype isn't finished yet, so I can't say some things definitively, but I can say this – I'm done with shooting on DSLRs. This thing is the bomb.

It is basically everything that I was expecting/hoping it to be, and more. I mean, I was expecting an HPX170 with DSLR-style shallow DOF, and yeah, that's it. What I wasn't expecting was variable frame rates up to 1080/60p. I was hoping for things like 2.35:1 frame markings in the viewfinder, and yeah, it's got it. I was expecting things like zebras, waveform monitor, vectorscope, uncompressed audio, XLRs, line/mic/phantom power controls, manual audio, pre-record, interval recording, and everything else that the HPX170 does, and it has pretty much all of that (I didn't see time-stamp for legal video, but … seriously, this isn't a legal videography camcorder! Although, it could be, I guess).

I wasn't expecting simultaneous HDMI and HD-SDI output, that's really pretty awesome. You can use a cheap HDMI monitor, and use the HD-SDI to record to a NanoFlash or KiPro or whatever, simultaneously. Cool.

I was expecting timecode. I wasn't expecting the ability to sync timecode through LTC (like the HMC150 has) so you can easily sync to timecode slates, to external sound recorders, or to other camcorders in a multi-camera shoot.

I was hoping for (well, demanding really, but it amounts to “hoping for”) a high-def LCD, and yes, that's what it has. High-def LCD and high-def viewfinder. But what I wasn't expecting is that they said that in the final model, the LCD will actually be better than the one on the HPX370. That's really good news!

I was hoping to be able to remove the top handle, and you can. I wasn't expecting to be able to remove the side handgrip, but you can – which lets you strip the unit down to a fairly small box, and also reveals three 1/4-20 mounting holes on the side, which I can just imagine people will be using to attach all sorts of accessories, Red-style. Now that I've seen those two items come off, I really wanted them to make the viewfinder/XLR pod removable, so you could strip it down to a totally svelte box, but that's not possible. I have asked them to consider this for a future version, as that'd be really cool.

I was pretty happy to be setting the sensitivity by ISO instead of by gain levels. But you can do both, at least in this prototype.

As for images – well, imagine a fully-hacked, fully-improved GH13 with better sensitivity and without the aliasing or low-light banding, and that's a good start. It crushes the 7D/5D for clarity and detail, and I shot brick buildings at every possible zoom setting, and there just isn't any rainbow moire. Finally! You can do a wide-angle, deep-focus shot without any fear of the camcorder ruining the shot! And your talent can wear corduroy, or fine-striped shirts, without turning into a huge purple/orange smear! But I can't really comment too much on the imagery, because it's not done yet – by their estimates, it's only at about 70% of completion, and they expect to be able to improve it and make it even better by release time. Which is really promising, because while it's not perfect, already it's better than what people are going to be comparing it to.

I put on my Zeiss ZF 85mm and 50mm, also used a hyper-sharp Olympus 14-35, and at the booth they have a Zeiss Compact Prime 35mm. And, I used the GH1's 14-140 and I even mounted on the GH1's compact pancake 20mm lens, which looked preposterously tiny on such a big body, but really it was pretty darn cool. The 14-140 makes the AF100 into basically a complete video camera, the autofocus is smooth and nice, the zoom is manual but not too shabby, and the range is great (even if the minimum f-stop isn't). With my 85mm and 50mm (both f/1.4) super-extremely-shallow DOF was extremely easy to achieve.

Can't speak to price yet, that's not been released and won't be until the press conference, which is scheduled for 7:00 p.m. tonight (which is about eight and a half hours from now).

Now, keep in mind that everything is subject to change. They might add more features, or they might take some away. It's a pre-production prototype so it's very early, and (to borrow a line from Red) “everything is subject to change.”

Skew? Hugely improved. Didn't get to do a comparison test, but the 24p skew of the AF101 felt like it was at least as good as the HPX370/EX1.

I am sooo not disappointed. They have built pretty much exactly what I was hoping for. There are things I'd like to see added before it's released, obviously, and no we didn't get some of the wild things we were hoping for (like a new 50-megabit 4:2:2 AVC codec) but what we did get is pretty much exactly what we needed – better-than-DSLR imagery in a professional, proper video body, with all the conveniences and features that pro video shooters are used to. "

The only thing I suspected here that appears to be true is a push towards being able to strip down like the R1, not as flexible but either way it looks like you've got a bit more modularity. What we've got is a new type of camcorder where red meets dSLR meets 171 but with no p2. :)

Video of Barry Green here:http://blog.panasonic-broadcast.com/ibc-video-gallery/?ref=nf

Astounding that they just slap a mic on the camera and expect it to pick up everything, the questions are great though!;)

Could somebody please tell me what the multiplication factor will be with a standard 35mm still lens fitted. i.e with 1/3 sensors the factor is 7.2

Kind regards

Mick

Hi
2
digified.net/focallength

Hi Paul,

Thanks for that.

Mick

A standard 35mm (still/ or PL) lens will cover the 4/3s sensor.

What it can be compared to is S35mm frame (movie) and FF35mm sensor (stills). It's a crop factor of 2 for FF35 sensor and I believe 1.18 for S35 size frame or sensor.

This is not magnification either, it's crop or field of view. The focal length stays the same.

I have an adapter for my GH1 which allows Nikon 35mm lenses to be fitted, it will be the same field of view as the GH1 with Nikon.

Thanks for the info

Mick

1080p at 60fps. That's upped the ante a bit. Also sounds good that you can remove the hand grip. I never hold these types of cameras with that, so it would be good to be able to get rid of it.

Will be interesting to see how closely Sony is watching. Panasonic have put some really good features in there. The simultaneous HD-SDI and HDMI is a seriously good addition.

Not quite. The recording (in 1080 mode) is 1080psf for 25 and 30Hz frame rates, and 1080p for 24Hz - the codec doesn't support 1080p/60.

What I believe it WILL do is full overcrank in 1080p (or rather psf) mode, something which has previously been limited to 720p. So it seems you may be able to shoot 1080p at 60fps, but only play it out (in slow motion) at 24fps.

Sorry Infocus, I should have been clearer, I realised it was undercranking for slowmotion :)

I still think it has upped the ante as a result. No more having to switch modes to 720p to get slow motion. Sony really are going to have to push the boat out to match this camera with their offering. My only problem is at what point would be best to sell my EX3!

Will be interesting what this camera does to the market in general.

I've seen some initial prices of £3300 plus vat. Given I already on a Gh1 lens, this will be a no brainer for certain types of work. All this said I'm still not 100% comfortable with the codec or the cards. So may wait (for a change) to see if Panansonic bring out an Intra/P2 version - it would be worth the extra for me.

I've got to say I love the look of it now, almost looks like a super 8mm camera!

Funny - I was chatting to Alan Roberts at IBC and he said he was told that you couldn't do simultaneous HDSDI and HDMI, as it uses the same output port in the chipset...or something!

Since the Panasonic manager from Japan was in attendance, It's likely that Alan got the right info.

Regards

Dave

Will be interesting to hear what Barry Greens comments relate to (ie did he actually use that function, or was he given misinformation?)

On the HMR-10 AVCCAM Recorder, when you output SDI HDMI is cut off and vice versa so I think it likely this will be the case.

Overall, it's very interesting, and the feature highlighted by Simon is obviously a big plus. I'd like to see a better codec on a camera at this price (sorry, but Canon have raised the bar. End of story.) but I'm really reserving judgement until they demonstrate the finished product. Will they really be able to overcome the aliasing problems from high res chips without massive power consumption......?

Does it really matter not having both HDMI and HDSDI available at the same time? What about such as a loop through via the recording box?

It also sounds like it won't be available for purchase roughly the time that the equivalent Sony model is available. I don't think I'd buy either of them until there's a chance to compare the two side by side.

I don't think these cameras are going to be in the same price range. Apples and oranges in terms of products.

It's a bit too early to say but isn't the Canon quite a bit more expensive even without the lens of the 100? And personally I've no love for MP2 any longer, it's ancient, and despite the 4:2:2 of the Canon (which I admit I've not tested as the camera just not relevent to me) - my own tests of mp2 v AVC(mp4) in terms of output don't come close. MP2 is much dirtier even with double the datarate.

If they do put an AVC-I/P2 version out which will add approx £1000-2000 very vague estimates (according to JC of Panasonic) then it will fit the bill. Also Sony are quite happy enough to put 4:2:0 on a £12,000 350, so maybe Panasonic could easily justify a more expensive version.

Of course this doesn't really have anything to do with the camera, just the fact that Canon are five years late with SS and quite old technology, even if it's well implemented.

I think we can safely say there is now a camera for pretty much everyone and every pocket.

(As a complete aside the great news is even RED are now leaving CF flash behind in lue of SSD, it can't come soon enough - it's the weak link of the R1 in mine and many eyes. Awful in terms of reading technology and docking when you've got to be swapping cards a plenty. I've got through three different card readers with different levels of success, including bent pins, miss-reads, slow and occasional transfers. I think the stills guys may have managed as they don't have loads of cards constantly on the go! )

No. It depends which Canon you refer to (and how much you pay for lenses to put on the AF100). Canon have recently announced four models, two (the 300/350) are fully BBC approved and are between ?5,000-?6,000, the other two (the 100/105) are substantially cheaper than an AF101 even without lens. Their front end may not be broadcast approved, but they have shown you can have a fully approved codec in a camera costing much less than the AF101.

That should be a lesson not just for Panasonic, but Sony as well. It's time to put the 50Mbs codec into the EX series. It's not that the 35Mbs codec is bad - rather that the 50Mbs codec is better.

All I shall say to that is refer to the EBU codec trials, done highly scientifically, and accepted by the BBC and other European broadcasters. For their purposes, XDCAM 422 50Mbs and AVC-Intra 100 are approved for general acquisition - all flavours of AVC-HD aren't. That doesn't sound as if the EBU think "MP2 is much dirtier even with double the datarate", does it?

Frankly, that sounds to me very like marketing excuses. What users really care about is primarily the codec - the AVC-Intra part - not what it gets recorded on to. P2 may add a lot to the overall cost, but there is no technical reason why AVC-Intra 100 couldn't be recorded to something like Compact Flash. There may be marketing reasons why they don't want to use that, but that's a different matter.

I'm confused here, are we talking about comparing the price of cameras that are fully approved or the just the codec? Are the 100 and 105 approved or not?

Maybe but Panasonic already do make a camera that has a "fully approved" codec in the shape of AVC-I. Panasonic's variants of 4:2:2 have been in their cameras for ages. The front-ends may have a given compromise at a given price point such as the 300s. But the point is it takes Sony to nearly ?12,000 before they get there. (And don't tell me I can stick a box on an ex1...)

MP2 is old as the hills, old codec technology. I'm not a fan that's all I will say. This doesn't mean that AVC-HD is great too but it's here to stay - I don't think I can say the same for MP2. This choice is pretty personal and yet again I re-iterate the old debate about the BBC standards being largely irrelevent for a lot of production companies and shooters, unless you're working for them day in day out.

Of course it's marketing, they invested heavily in a broadcast technology that has become a pretty hefty standard!!

Compact flash - please be realistic. Again it's old as the hills and designed for stills shooters. The fact that you can make it work for video doesn't make it a solid choice.

Yet again I cite the cost for P2 as solid proven scalable aquisition versus an open ended pin-bending nightmare that just happens to be cheap as illogical.

Besides SD based technology is where they've gone on the cheaper side of things, but at least you can purchase Japanese SD cards produce for/by Panasonic that work.

Back on track I don't think the AF101 has BBC shooters in mind either. That's not it's market.

The likelihood is that the 100/105 won't be fully approved, but that's because of front end - not codec. (It seems the coders are the same as in the 300/305 - fully approved for BBC acquisition.) The point I originally made was "I'd like to see a better codec on a camera at this price (sorry, but Canon have raised the bar. End of story.) "

It's too early yet to say how the AF101 front end will perform, especially as regards aliasing. If it does get an approval, why stunt the whole camera with a non-approved codec? Not when a better one need not cost more?

The Canon range are between £3-6,000, all with fully approved codec - that's what is meant by "Canon having raised the bar". You may not like CF - but do you like SDHC? Or the cost of P2? Or AVC-HD? Nothings perfect. I doubt we'll ever see a P2/AVC-Intra version of this camera - though a SDXC/AVC-Intra version may be highly likely before too long.

Yes, but not in the ?3-6,000 price range, and you have to pay quite a lot to get it mated with a good front end. If Panasonic were to put AVC-Intra 100 in the AF101, that'd be fine.

Sometimes "old" is good - a bit like a solid antique may be considered better than a new bit of flat pack furniture. In codec terms, there is a question of suitability, MPEG4 may be highly more suitable as a transmission codec than MPEG2. For acquisition it's far less clear cut - all it will do is give the same quality at a lower bitrate, but at a cost. For some purposes the oldest "codec" of all (uncompressed) may be easily the most suitable.

The MPEG2 old tech thing is a bit generalised. Once you go above a certain bitrate (can't remember off hand roughly where the threshold is) codecs like AVC don't hold any advantage compression wise over MPEG2. It's at lower bitrates that the advantages are seen. So as an acquisition codec MPEG2 is fine. Ish.

Now we've got devices like the Aja Ki Pro Mini coming out that can record to Prores formats, as well as the NanoFlash which can record to very high bitrate Intra as well as Interframe formats, I'm not so sure that the in built camera recording codec is that relevant anymore.

As long as the front end chip size is big enough and the camera has a pre compression HD-SDI or HDMI output you can pretty much make any camera within reason broadcast acceptable (please note the "within reason" bit).

In fact I'm not sure why camera manufacturers bother including a recording codec any more, unless they are simply looking at it as a base level, a bit like giving a free recording card away with the camera to get you up and running. They are providing us with HD-SDI outputs and HDMI outputs on most base level pro camcorders now. So they know full well devices like the Nano etc will be used extensively.

Yes, the Nano is 2 grand extra on top of a camera. But I look at it a bit like buying a long term item like a tripod. A camera will be replaced by new tech quite quickly. But a Flash Nano and Ki Pro Mini will be useable over many different cameras. It is unlikely that we will surpass the need for a 220Mbit Intra frame codec soon, something that both the Nano ad Ki Pro are capable of. So they should last for a good few years over a few camera purchases.

Hi
I've been reading the interesting and long thread on dvxuser where Jan Crittenden (Panasonic technical marketing USA) has discussed with the Russian guy who's hacked the GH1 firmware/codec the continuing development of long-GOP algorithms - and the importance of IBP-frame motion prediction, which the AF101 does, as opposed to the IP-frame older codecs as used in the GH1. And presumably other older long-GOP implementations...
dvxuser.com/V6/showpost.php?p=2103331&postcount=99

Interesting, Paul - nothing to do with codecs, but tucked away amongst a lot of words in that thread is this comment from Jan Crittenden:

Now, at first sight that may be seen as a negative ("what! a high res chip, but can only manage 1920x1080 for stills!) but I think it may be a big plus for the camera. (As long as you are primarily getting it for video.)

Pure speculation, but you have to ask why? Why not full sensor resolution for stills, as for DSLRs? What if it's because they have put an optical low pass filter on the chip to greatly reduce the aliasing compared to DSLR video? Highly unsuitable for a DSLR, of course - the prime function of that is to take quality stills - but may be a very good idea if you want to use a large, megapixel sensor for HD video.

Not as good as full pixel readout and high quality downconversion, but for the reasons Alan gives earlier in this thread, that may just be too taxing for current technology at this price point at the moment.

Hi
Interestingly the new iPod Touch seems to produce 'better' video than the iPhone 4, because it has a chip matching its 720p video resolution rather than the better stills resolution chip of the iPhone 4 - less jello etc:
http://www.youtube.com/watch?v=qkOq2qgaSHo
youtube.com/watch?v=_D54OYZqJwE

As far as I can see, there are four ways to get video via a large format chip.

1/ Have a large format chip, but match the number of pixels roughly to HD output resolution. (De-Bayering makes it a bit more complicated.)

2/ Use a mega-pixel chip, only read out a fraction of the available pixels for video.

3/ As 2, but put an optical low pass filter in front.

4/ Use megapixel chip, read out all pixels at frame rate, and perform high quality filtering and downconversion.

No 4 is arguably the best, certainly if you want to use an off-the-shelf megapixel chip, but is likely to be impractical with current technology. (See posts 20 and 21.) No 2 is what happens with DSLRs, and gives uncompromised stills, but aliasing on video. No 3 will improve the video - but limit still capability to HD video resolution.

It would be nice to think method no 4 was to be used, that Panasonic had overcome the challenges, but I think it far more likely (especially after Jans remark) that it's no 3. It would certainly be an improvement on method 2!

But it is old tech. I was editing on 4:2:2 Mp2 10 years a go. Capturing BETA. That's old in my book.

Okay, it's not acquisition but I did lots of tests producing MP2 v AVC(MP4), every combination (different encoding engines) and even with them both on 25-30mb/s - MP2 had artifacts and was 'dirtier' than AVC. I didn't want it to be like that because of the encoding speed (which that said isn't slow on i7 anyway).

I know this doesn't prove anything about the cameras but it showed me that MP2 was never as clean as AVC no matter what I did.

Whether MP2 is more than good enough for your own choice of aquisition then that's personal, and it's hard to argue against a 4:2:2 codec that is fast and easy to edit but it's certainly antiquated in my book. But that's Canon - they're not innovators in camcorder tech.

I'm personally not interested in bolting things on to small-form cameras to make them bigger more expensive cameras. Which is why this doesn't appeal. I can see why folk do this but I try and have a type of camera for each job and each market rather than have a jack of all. I suppose having the option is always good though.

With the AF100, I can just about see the logic but why not just add the cost of the lens and Nano and just purchase and up-spec camera in the first place? The market it's aimed at is unlikely to be phased at its limitations of codec anyway. Despite what the current debate says it's hardly a Red or Alexa replacement - different markets even if panasonic are tying to extend it.

In fact rather than strap a Nano to it Panasonic could just fit P2 to it up the cost, no extra box or cable handing out and you've a tidy solution. That makes more sense to me.

I played with a couple of 101s at IBC. List price is going to be around 4.9k in Euros, although I saw only prototypes. I like it, the weight is right, the handles come off so other goodies can be bolted on instead. All the controls seemed to be in the right place, and it outputs both HDSDI and HDMI, uncompressed, although not both at the same time, which is fine.

The sensor is 19.2x10.8 mm as far as I can tell, and the lens flange is so close to the sensor that adaptors can be fitted to take any lens. And Panasonic say that they will be making adaptors for any lens, plus a range of lenses spefically for this camera.

I also saw the prototype Canon 105, which has a couple of really good features, and leads to the possibility of a truly amazing 3d camera later in the year.

At the same bitrate, I would fully expect MPEG2 to be of less quality than MPEG4, though the latter to need more processing. (Though how much will likely vary with differing coders.) That shouldn't be in debate. But it's not what we're referring to in this context.

We're comparing AVC-HD at an average bitrate of about 21Mbs, with MPEG2 at about 2.5 times that. We are also talking about hardware encoders, not software, and except at the top end they are unlikely to be able to implement all the refinements that a software encoder might. And the more complex the technology ("new", if you like), the more true that is likely to be.

"New" technology has most relevance in transmission, streaming etc where datarates are of most importance. For acquisition it's of far less importance - certainly if you can use lower cost memory - and the ease of use of what you call "old" technology may be more significant. If it allows a fully approved codec to be incorporated, I'd say that was significant enough that I'd be happy enough if it was from the Stone Age! Can we leave this subject and move on now?

Do you have any early opinions about how well/badly it performs in respect to aliasing etc?

And the comments so far have revolved around features most suited to digital cinema type usage, is there any indication of availability of a lens which may more suit more traditional video usage? (Longer zoom range, motorisd zoom etc.)

Given that the camera has been deliberately aimed at the DSLR market, I reckon they've got it right. I couldn't do a sufficiently reliable test on it to establish the absence of aliasing, but then again, I didn't see any either, we'll have to wait for a production model and a proper test. They wouldn't even say (perhaps didn't actually know) what the precise image dimensions are (I've only guessed 19.2x10.8 for 16:9) or pixel dimensions (but I suspect it's quite a bit more than 1920x1080. it ought to be at least 2715x1528 in order to do a reasonable down-conversion to 1920x1080, or 2880x1620 to get it right, like in the Alexa).

As for lenses, they had a case-full of new Micro 4/3 lenses on show. Nothing in video format, but that's not the market they're aiming at. The lenses looked rather tasty (Zeiss, I think), but I'm pretty sure that Panasonic are going to be making lenses directly (not sure whether they'll buy an existing business or start a new one).

I'm pretty sure I've heard it confirmed it's the same sensor as used in the GH1. (Which would make sense.) If we assume that, we're looking at a four thirds sensor (17.3 x 13.0 mm) with 14 megapixels total, 12.1 megapixels effective, according to the GH1 website.

But all that tells you is that the MPEG2 encoder you tested doesn't produce as nice MPEG2 files as the H264 encoder you used produced H264 files. It tells you nothing about MPEG2 or H264 themselves. Also, unless you were using uncompressed sources, it only tells you about how those encoders cope with footage already containing artefacts -- not how it'd cope with uncompressed footage live from the camera.

The thing you have to remember about H264 and MPEG2 is that they only define how the decoder will interpret the incoming bitstream. Crucially, they don't specify how that bitstream is to be formed -- that is purely down to how the encoder is implemented which varies from vendor to vendor (and from camera to camera). If you read the link PaulD posted, you can see that the H264 encoder in the GH1 is not using all the features of an H264 decoder (specifically, it isn't making use of B frames -- so massively reducing its ability to compress the footage) -- this means quite simply that at a given bitrate it is quite likely to look worse than a decent MPEG2 encoder would produce at the same bitrate.

MPEG2 and H264 aren't that dissimilar as codecs -- H264 is an evolution of MPEG2 with bells and whistles to make it able to better encode footage at low bitrates. And these extras come at the cost of significantly more processing power to both compress and decompress the footage -- which means more heat and less battery life.

Suffice to say, the choice of codec isn't important -- its the quality of what comes out the other end that counts and, even more importantly, how it interacts with the rest of your post-production chain. There's no point using an encoder that produces great footage straight out of the camera if it sends the rest of your post-production chain mad and ends up producing worse quality at the output...

Steve

I couldn't have said any of that any better, you've read my book :)

What matters is the pictures at the viewers' display, not how they got there. What we should all be doing is ensuring that what we shoot survives best throughout the programme chain; individual components of it are important in that they shouldn't wreck the final result.

Fair comment.

Point still stands, it's a real-world test for me, If I've got access to several of the mainstream encoding engines and test them all and I get similar results that tell me to choose the codec that produces the better results... On balance there is a degree of consistency in the output that points me to a reasonable conclusion. I.e I'm taking very high quality 4K R3D down to 1920x1080 and I need the best output I can achieve and under my own circumstances H264 output has produced the best results which is what I would expect given its advancement. I'm not sure how else to test which is the best codec.

I understand implementation, trust me I own a pre and post hacked GH1 and this is a great demonstration of what you're talking about but somewhere along the way we have to decide whether implemetation or theory takes the debate or whether it just becomes semantics.

Plus its difficult to argue that if you took the best available implementation of MPEG2 and the best implementation of an AVC codec such as Ultra or Intra that MP2 would stand a chance subjectively or technically?

But it isn't the codec that's important, it's the coder. The decoder is fully specified in the standard, it's only the coder that can be developed.

Coder performance is highly affected by the input content. If the input signal contains any spectral content which is the result of spatio-temporal aliasing, then the coder will perform dramatically worse. It's difficult to identify this in practical pictures, it's only repeatably possible to assess coders using test sequences, and the major broadcasters, through the EBU, have defined such test sequences and methods. That's why the major broadcasters and the EBU have a view on which coders are acceptable and which aren't. Subjective tests on one camera on general subjects can't hope to get to the root of the matter.

But then, is there a better way that I can get my hands on - going from .R3D 4K via CS5s new encoder (Main-concepts H264) to play-back on blu-ray?

I can't help feeling sometimes that these Broadcasting standards tangibly offer very little for the average production guy on here who wants the best he can get/afford and at the end of the day practical pictures is what we make.

But you are talking about software encoders working in non real time, a camera needs a hardware encoder working in real time.

An H264 software encoder may draw on a large number of tricks and a powerful power hungry processor to achieve it's magic - it's highly unlikely that a hardware encoder in a camera costing a few thousand pounds total will do anything like as well. That's the point Alan makes - it's not the codec, it's the coder. I'm not querying your results - but it is not possible to draw the conclusions from them that you are doing.

The more I learn about the whole subject, the more complicated I realise it is. That's why the EBU recommendations are so valuable - you don't need to try and work it out for yourself. My attitude now is that as far as the back end goes, if it's XDCAM 422 50Mbs or AVC-Intra 100 it's OK. If it's something like AVC-HD, HDV, or whatever, it may not be. That is why they get given so much attention, and why they should be taken notice of, even if you're not working for an EBU broadcaster.

As far as the AF101 goes, that's why however good the front end may turn out to be, it would be far better overall with one of those approved codecs as opposed to AVC-HD. And Canon have proved you can get a fully approved codec for the sort of money Sony and Panasonic are charging for AVC-HD and 35Mbs MPEG2.

For that particular case - software encoding for Blu-Ray - then it's highly likely that H264 is the best way to go, especially since Blu-Ray encoding assumes a fixed target bitrate. I don't think anyone here is saying in principle that it isn't.

What is (strongly) disagreed with are the further conclusions you are drawing - that therefore H264 based coding must ALWAYS be better than MPEG2. That's especially true of 50Mbs XDCAM versus 21Mbs AVC-HD.

H264 has most point when bitrate must be limited - transmission, Blu-Ray encoding, etc. It makes far less sense when bitrate is less important - and that includes acquisition.

I like to think of the standards as a sort of quality assurance symbol - bit like the "HDready" logo when HD TVs first came out. Buy a set with that on and you knew it would have a certain minimum performance. A set without it MAY be good enough - but a bit like buying a pig in a poke. For the average production guy, a camera with an approved codec means he knows what he's getting in that respect - the front end is a different matter, of course!

How an encoder performs on a computer with several gigabytes of memory and that doesn't have to work in realtime is no indication of how well you can get an encoder to work when it has to compress in realtime (with no ability to go back and steal bits from certain parts of the stream to increase the quality of others) and in hardware (where you've got all sorts of interesting considertations -- from do I have enough gates on my FPGA to can I cool this).

But you are not comparing like with like then -- AVCIntra and AVCUltra are more like DVCProHD than it is H264 or MPEG2. It (and HDCamSR for that matter) only uses the part of the H264 spec for encoding single frames -- there's no inter-frame prediction going on. This means that, on average, each frame of video is allocated 4Mbits (for AVCIntra100) -- that's a 10:1 compression ratio for each frame. On the other hand, long-GOP 50Mbit/s 4:2:2 MPEG2 is allocating 25Mbits for each group of 12 frames which is where things get interesting -- the long-GOP encoder does not need to allocate those bits equally between each frame (AVC-Intra or Ultra does), if something is identical in every frame then a long-GOP encoder can throw six times as many bits at encoding it than AVC-Intra can.

I have absolutely no doubt that long-GOP MPEG2 @ 100mbit/s would look identical to AVC-Intra in the worst case, and quite possibly a whole lot better. For that matter, so would Panasonic's own AVCHD compressors (which are using long-GOP H264) -- assuming they could get the silicon to run that fast -- the interesting question is whether the long-GOP H264 would look better than long-GOP MPEG2 -- which basically boils down to the question of whether the additional ways H264 can break down the image sequence for encoding actually buy you anything in terms of picture quality....

Steve

Good pics of it and more details on this video:http://exposureroom.com/members/dvwarrior/5b1064261d774782b535cc9d28300419/

Comparing AVC-I and Ultra whith DVCPro is not comparing like with like, it's comparing one element, ie the lack of inter-frame prediction. I also thought DVCPro was an MJPEG variant?

Really? Despite being more efficient and 10bit? Any way you could demonstrate this under any circumstances?

No it really is... AVC-Intra and DVCProHD are both DCT-based intraframe codecs, AVC-I has a couple of extra bells and whistles -- mainly a better arithmetic encoder to allocate the data from the model into bits. That's not to say AVC-I isn't a much better codec than DVCproHD just that it is in the same class as it.

But AVC-Intra isn't more efficient than MPEG2. AVC is, at low bitrates and when it uses all the features that the decoder supports but not AVC-Intra, which only makes use of one part of the AVC compression toolkit. The way to think about any compressor is as a 'model' to describe the data, and an 'encoder' which compactly encodes that model into bits. To get better compression, you can either improve the model or improve the encoder, or both. AVC-I has a better encoder than MPEG2, but a much simpler model -- only working on frames not GOPs. Any efficiencies in the intraframe part of AVC (as used by AVC-I too) over the MPEG2 intraframe part will be made up for by using long-GOP part of the MPEG2 model.

The 10-bitterness would give it a slight advantage if you want to grade the footage, although that depends as much on the cameras SNR. There's no point recording in 10bits if the last couple of bits are just noise... (Hence, the BBC archiving its D3 recordings to 8bit MXF -- the transform PAL decoder only gives 9 usable bits of output, the last bit of which is easily folded back into 8bits).

No one ever said video technology was simple... And don't get caught out by marketing arms spinning the buzzwords (in this case, the AVC part of AVC-I) -- they are notoriously bad at doing it -- witness Panasonic making a huge deal of the fact that DVCProHD was 4:2:2 on the HVX200 compared to 4:2:0 with HDV cameras while quietly ignoring the fact that none of the cameras could fill a chroma bandwidth higher than 4:2:0 anyway... Learn the science behind it and it all becomes much simpler (if slightly less magical)...

Steve

So if it AVC-I is less efficient and better looking than looking than Long GOP MP2; why are we bothering? We may as well have the old license free compression system. (Though I really doubt you can have better looking and more efficient in this example.)

Or are we saying what is good for aquisition is not good for editing, let's be correct there is no free lunch.

As an aside, it's interesting what you cite as science also appears in a sony marketing 'white paper' http://pro.sony.com/bbsccms/assets/files/micro/xdcam/solutions/MPEG-2_Long_GoP_vs_AVC_Comp-Strategies.pdf It's also interesting that they use NLE's to demonstrate the codec credentials.

DVCPROHD, as well as being intraframe, codes only 1440 luma samples per line. So the 4:2:2 nature of it records 1440:720:720. That's for 1080/50. The 60Hz equivalent records 1280:640:640. The first DVCPROHD coders were made out of an array of four standard DV coders. Later cameras had/have custom coder chips. Thus each frame stands alone. the coding algorithm is quite crude, being old, hence it needs to screw every bit of saving it can (like not doing it at full resolution).

MPEG2 and MPEG4 (which includes the H.262 coder of MPEG and the H.264 coder of MPEG4) are interframe coders, they code groups of frames together. They cannot be directly compared with any intraframe coding scheme, because they're different.

AVC-I is an intraframe version of AVCHD, which itself uses the H.264 coder of MPEG4, but with some custom tweaks that are allowed within H.264. Each frame stands alone as in DVCPROHD, but at full resolution. The efficiency comes through H.264's ability to vary the size and shape of the macroblocks according to image content, and the resulting reduction in the coding coefficients.

The busi9ness of assessing and comparing coding schemes is highly complex, because it critically depends on the image content. The EBU has specialist groups from research institutes across the world working on this, you can't hope to compete with them at home.

Different coding schemes are good at different things. Intraframe is good for capture, because frames are not polluted with content from adjacent frames by interpolation or prediction. Interframe schemes are good for distribuition and final delivery, where the footage will never be edited or post-produced, so any compression or motion artefacts cause visual distrurbances which cannot further affect compressors (because there are no further comoressors).

That's why we bother.

I not pertaining to compete at home (work) with the EBU (though I can't deny it's an interesting and complex area) . I'm interested in what's the best quality for my OWN output. That's my whole point. If I was a broadcaster it would be different.

Getting back on track. US list on this $4995. They could easily afford to put AVC-I on this.

As an aside your book, can it only be purchased from EBU technical or is there another outlet?

The EBU is the only official retailer, although I have a private stock for sale in my office (actually, now down to 2 copies, and two people have said they want one recently, but not done anything about it yet) which I can sell for the price the EBU charges me, £72 per copy, + about £4 for p&p.

You have to distinguish between final delivery (which is what you've tested), and acquisition, and what we're all saying is that it's wrong to draw too many conclusions from what you've observed and apply them to acquisition coding.

The whole point about the EBU approvals is that they take all the work away. As a user, you just don't need to worry about the camera codec. They effectively say use A or B, and for general material the coding can be regarded as "quasi-transparent" - it'll accurately record whatever the camera front end can throw at it. I'd say that's as relevant to you in respect of wanting a guarantee of best quality for own output as it is for a broadcaster. And AVC-HD most certainly cannot be said to be "quasi-transparent".

However good this camera turns out to be, there is no getting away from the fact it would be far better with AVC-Intra 100 or XDCAM 422 50Mbs. And I suspect the sort of users to whom it will most appeal will be those who don't consider AVC-HD adequate, and use it with something like a nanoFlash. Far more satisfactory (and cheaper) to have just built one of the former in from the start.

And I'll also recommend Alan's book..... :)

MPEG2 and MPEG4 (which includes the H.262 coder of MPEG and the H.264 coder of MPEG4) are interframe coders, they code groups of frames together. They cannot be directly compared with any intraframe coding scheme, because they're different.

At 100mbs and beyond MPEG2 is definitely intraframe. DVCPro HD is now a very old codec that is not full resolution at either 720 or 1080. MPEG2 definitely uses DCT within the macroblocks and so it has some superficial similarities with older DV/DVCPro codecs. The big problem with DV based codecs is mosquito noise.

I have done extensive testing of both MPEG2 and MPEG4 (for post production) and I believe that MPEG2 is better within the macroblocks than MPEG4 (especially in dark broad areas and on dissolves and fades). Whilst it is generally given the MPEG4 is twice as efficient at compression than MPEG2 this is not necessarily the full story. The BBC have had some problems with their HD broadcasts in MPEG4 compared to the intial MPEG2, I believe, for these reasons.

The big advantage for AVC-I seems to be 10bit - I am not aware that any MPEG2 acquisition codec, no matter the bit rate, is anything other than 8 bit (Nanoflash for example)

I think it would be more accurate to word that "the big potential advantage for AVC-I seems to be 10 bit......"

It's of little use an acquisition format being 10 bit if the noise level of the camera is such that the least two significant bits are only coding noise. Hence the 10 bit nature of AVC-Intra may well be an advantage when coupled with a high quality (ie expensive) front end, but not give any practical benefit at all (except in marketing!) in a lower priced camera like the HPX301.

Post production is another matter, and even if acquisition is with a "noisy" camera and/or an 8 bit codec, there may be advantages in a transcode to a 10 bit system if there is heavy post work to be done.

"I think it would be more accurate to word that "the big potential advantage for AVC-I seems to be 10 bit......"

I'm only refering to the codec not any particular camera. The 3700, for instance is a different animal to the 301. The advantage of AVC-I in the 3700 is not a potential it is here and now.

That's exactly what I mean. In the case of the 3700, the 10 bit nature of AVC-I may indeed be a here and now advantage, in the case of the 301 I doubt it makes any difference at all because of the front end noise. All 10 bit coding does in the latter case is effectively waste 20% of the bitrate.

The 10 bit nature is a potential advantage, whether it is an actual advantage depends on the specific camera it is used with.

10-bit recording is valid when the PSNR is 50dB or better. Very few HD cameras get to that figure, at any setting. BUT 10-bit post-working of 8-bit material is perfectly valid because it preserves what's there and doesn't play games with it.

10-bit recording is valid when the PSNR is 50dB or better. Very few HD cameras get to that figure, at any setting. BUT 10-bit post-working of 8-bit material is perfectly valid because it preserves what's there and doesn't play games with it.

Yes this is often never quoted reliably by manufacturers. The Red camera raw footage quotes 12bit, has this camera over 60db of PSNR, Alan? The 3700 is quoted at 54db BTW.

In post production it is always better to use robust codecs like DNxHD or Apples Pro res which are engineered for the purpose, they come in 10 or 8 bit flavours.

As acquisition or final broadcast there is little perceptual difference between 8 or 10 bit for most purposes. Note I say most purposes. For intensive post work 10 bit has strong advantages as you say Alan.

My tests, crucial for some of the work flows we're developing, were concerned at preserving MPEG2 as an acquisition format trans-coded to DNxHD and testing final HQ output to both MPEG2 and MPEG4 delivery formats. I found that no matter the ratio to I, P and B frames; CBR, 2 pass VBR, CABAC coding, slices, etc. etc. MPEG2 outperformed MPEG4 in smooth dark areas and fades/dissolves, even at identical bit rates. I've a lot more testing to do with a greater variety of encoding software but MPEG2 if well implemented is still a very resilient acquisition and delivery codec. One thing I must add though is that MPEG4 can deliver greater detail than MPEG2, meaning to me at least it looks perceptually sharper for a given supposed comparable bitrate (i.e. 10mbs MPEG4 versus 20 mbs MPEG2).

Creative now have the UK price listed:http://www.creativevideo.co.uk/public/view_item_cat.php?catalogue_number=panasonic_ag-af101

Noise in the HPX3700 is about -46dB, but has the right profile, so they're not playing any games in the processing. That's using the Film Rec curve, which has lower gain in mid-tones, so, using a conventional gamma curve should worsen that figure by about 2dB.

Noise in a RED ONE is about -50dB at 500ASA and about -44dB at 2000ASA. Noise profile is as expected at low ASA figures (i.e. low gain), but noise levels and profiles change significantly with ASA figures, indicating that there's some form of noise reduction going on.

Beware that there are many different implementations of MPEG-4 , few use all the coding tricks, so performance can vary wildly between coders. MOPEG-2 coding is much more developed, and is much more stable, so performances tend to be more consistent. WAhat this means is that you can't measure one coder and predict the performance of others from it, unless you know exactly which tolls the individual coders are using, and they never tell you.

Hmm, so in that case it appears that a 10 bit codec doesn't even add a lot to the HPX3700? I was pretty sure it was a waste of bitrate in such as the HPX371, but I'd expected it to be far more worthwhile in such as the 3700.

If you wind the gain down to -6dB, 10-bit makes sense. Otherwise, it doesn't make a lot of difference.

I think with the 301 the advantage of having the AVC-I is the full raster recording rather than the 10bit which is part of the codec - Edius (5.5) doesn't manipulate in 10bit anyway for its AVC-I ingest.

As an aside I've just added a humax T2 HD box to my 1080p Ae2000 projector set up. This is the first time I've seen terrestrial broadcast in 1080p on my 60" screen.

Given all the technical debate here about the quality of certain cameras being applicable for HD broadcast it's a shame that the final quality isn't that great. It certainly helps me re-focus my business intention that I tend to get excited about my own output rather than for broadcast and waste time debating cameras when our own end to end quality is subjectively much much better than what I'm seeing on the HD channels.

I'm with you there. AVC-I is well worth it.

And, given that I seem to be the only one testing cameras and reporting as a scientist, I have to report on suitability for use by the people who pay for the tests, and so far, that's been exclusively broadcasters (with only 2 exceptions in over 60 camera tests). Nevertheless, you've never seen or heard me say that a camera is or isn't acceptable for broadcast (apart from the DSLRs), I only measure and recommend settings with which to get the best performance. The broadcasters decide on acceptability, albeit with my test reports as evidence.

Incidentally, my test methodology is being enshrined in an EBU code-of-practice recommendation (R118) supported by an EBU Technical paper (EBU Tech.3335), or at least it will when I get the writing finished.

Good review here:http://www.dvuser.co.uk/content.php?CID=234&ref=nf

Have I missed something, he says come back on the 19th for the review??

Mind you I'm not going to hold me breath the review on the Miller Tripod and previous reviews have never lined up with my experience.

I thought this was pretty interesting, with some video too. http://crews.tv/blog/2010/10/12/af100-storm-gathers/

I know that people in the BBC have tried it. Anecdotal responses include:

which is exactly what I'd expect.

Has anyone else noticed that in the latest pictures of the camera featured in the link in post #81 that the large carrying handle has sprouted an accessory shoe and some other intriguing lumps and bumps.

It looks as though it might now be removable too.

Just an observation. NL

I believe it is removable.

The top handle is removable, as is the side handle. Under that, you find a set of fixing screws for attaching whatever you want. The top v/f isn't removable. I talked to several Panasonic people at IBC and they said that the physical appearance wasn't then finished, so things like shoes may well be added before the launch, they have this annoying habit of listening to the potential users :)

I've had a message from Top Teks announcing that they'll have "a production model AF101... to handle, test and compare" at some kind of event at their show room in Harefield on Tues 23rd. They claim that this is the first time a production AF101 will be on show in this country. I have no connection with the firm except that a friend is on their mailing list.

They will also have examples of three other new cameras on display: Sony PMW-500, Panasonic AJ-HPX3100 and Canon XF305.

Show room: Bridge House, Royal Quay, Park Lane, Harefield, UB9 6JA
Telephone: 01895 825619
Fax: 01895 822232
email: sales@top-teks.co.uk
www: Top-Teks.co.uk

Got an update from panasonic this morning:http://www.panasonic-broadcast.com/en/products/high-definition/avccam/AG-AF101.php

Still no mention of the sensor's native resolution.

Steve

Prestons have 120 on order apparently :eek:

Eh? The AF100 isn't a 3D camcorder. The 3DA1 is the one designed for that. Besides, it is the operator and director that is mostly to blame for bad 3D. Then again I don't think I've ever seen really good 3D, except for one or two of the IMAX films. And that was probably because IMAX films inherently don't have any silly high motion camerawork, the screen fills the entire view, and the quality control standards are exceptionally high.

Sky TV on the other hand...

The link Gary posted quotes (under specifications):

Combine that with the fact that in a photo of the AF100 launch at NAB, Joe Facchini / VP of sales, Panasonic, is pointing to a Powerpoint slide proudly proclaiming “Single-chip micro 4/3-inch imager – 12.1-megapixel 16:9 MOS". and I think we can be pretty certain that it's 12.1 megapixel.

Which is why it's so odd that Panasonic are now being so tight-lipped...... Though conspiracy theorists would have you believe a crack team from Canon and Sony hacked his presentation.;)

Fundamentally, Panasonic are making use of an existing sensor designed for stills use, and made far more suitable for video by adding an optical low-pass filter. There's nothing wrong with that - it will give far better results than current video DSLRs. But then they realise their main competitor is bringing competing products to market at a similar time - with a sensor purpose designed.

However good a job Panasonic have managed to do of adapting a stills sensor, it is virtually certain to be outperformed by the purpose designed Sony one, which seems to be 4 megapixel Beyer - optimum for 1080p video. The real question is by how much, and we'll have to wait for real world measurements to answer that one.

I suspect that they try and limit technical info about the sensor to try and prevent web based rubbishing of the product before its even been launched!

I'll wait until we can look at some zone charts before drawing any conclusions about sensor design, i'm not a optical electronics engineer, so my speculation about sensor technology is fundamentally meaningless.

I also think that it is misunderstanding the market to claim that this level of camera will be chosen on image quality alone. There is so much more to things at this level, and i think many people who initially had put themselves in the "potential buyer" bracket will now be waiting to see what Sony do with their Super35 NXcam.

For me I'm quite seriously thinking that i will own one or other of these cameras within a year, but my choice will have as much to do with the overall ecosystem of the camera as it will be about output quality and sensor design. I'd rather own a camera which works well with a wide range of lenses (ideally including a servo zoom ENG style lens) and can interface into multicam / studio shoots. Which of these new cameras will be best for this? Nobody can know currently (i've seen rumours that the NXCam will not have any HDSDI output for example, which would be a total deal breaker for me)

Even simple things like what voltage the camera uses, compatibility with existing pro battery systems, ergonomics and of course cost seem more important than the highly technical differences between sensor designs.

But what's so odd here is that they were far more open early on, than they are publicly now! I agree with the above in principle - but the AF101 was effectively launched at IBC, working models have been demonstrated for some weeks, and they've been happy to take people's money for a while. Yet be very coy over some of the most fundamental of specifications.

Dealers are currently running sessions with a full production model - http://www.creativevideo.co.uk/public/panasonic.register.php - for example, and presumably happy to take orders. If anyone goes along, I just hope they ask about the sensor details - it does say "You will be able to quiz the two Phils about all aspects of the camera, both technical and practical....". And at this stage, if any answer about sensor details gets fudged, be very suspicious.

It's like going into a car showroom to buy a newly launched car. You ask the salesman "what's the engine size?", and he replies that he can't possibly say. That wouldn't be a surprise if it was a concept car, but on a finished model, about to be shipped, and that orders are being taken for!?

As far as the rest of your post, then I fully agree. There's far more to it than sensor specs. I certainly wouldn't rubbish the AF101, but don't think it's a good idea to rush into a purchase now unless you really have to. Very sensible to wait a few months.

This was shot at Holdan last week by Martin Kay.

http://vimeo.com/17692141

I haven't seen it in HD yet - this old P4 don't go that fast!

The spec says that the sensor is 12.4 million pixels, but not how it's organised or decoded. Until I test one, I suspect there will be no other disclosure about .

Manuals are on-line for download here:https://www.yousendit.com/transfer.php?action=batch_download&batch_id=RlRxU2V1Z2o4Q1N4dnc9PQ

But still no mention of how it gets from 12.4 million down to HD.

It says 12.4 million and that the sensor is 16:9, which makes it about 4688x2637 pixels (12,362,256), about 2.44 times over 1920x1080. That's easily enough to get it right, a factor of 1.5 is enough (viz ARRI Alexa). 4704x2646 is also a possibility (12,446784).

So, it should be perfectly possible to get it right, but I'm waiting to see and test one, a finished production model.

Fundamentally, it can take one of two basic approaches.

1] Read all 12.4 million pixels from the chip every frame, de-Bayer the result, then do a high quality downconversion.

2] Read out *ENOUGH* of the pixels from the chip each frame, then process similarly to above. The vastly decreased data load makes the process far easier and less power hungry. How many is enough? On a Bayer chip, something like 4 megapixels is a reasonable figure, so reading 1 of 3 should give good results for resolution on a 1080 output.

Does it matter which approach is used, and how can we tell?

The best way is the first - but is likely to be far more power hungry. As long as an OLPF is used - and all the indications are it is - the second way should still give comparable resolution figures, but the sensitivity/signal-noise to be well down. (As the majority of the 12.4 million pixels aren't being used.)

The specs give a pretty big clue as to which approach is being used. Way back in this thread (post #20), exactly this point was raised, and Alans reply then was:

So....... what's the power consumption of the AF101? The specs quote 12.4 watts when recording. Consequently, it's virtually certain that it takes the second approach, it reads the sensor by pixel-skipping.

DSLRs have given 'pixel-skipping' a bad name, but there's no real reason that should be the case - not with an OLPF. It doesn't mean the results won't be good - but does put it at a basic disadvantage compared to the two new Sonys.

I'm not going to join in any guessing game. The over-sampling from a 2.4Meg sensor is 2.44:1, while the ARRI Alexa is 1.4:1. In principle, the 101 could be doing it right, the power consumption is quite significant, the Canon XF105 I've just finished testing takes only 6.5 watts, amnd I'd expect Panasonic to be able to do it all properly inside 12.4 watts.the DSLR problem is that they can't afford the power consumption from the much-smaller batteries they use, that plus the heat dissipation. In a video camera, all that can be managed properly, particularly if it's physically small.

So, I'm still not going to join in any guessing game, I'll wait until I've measured a production model. After all, if they use a 12.4Meg sensor, and the camera doesn't do a stills mode (or, at least I can't find mention of it in the manual), then subsampling would simply throw away the light which lands on the never-used pixels, which would be a very silly thing to do.

Well, the EX1 is about 13 watts, the HMC151 10 watts, which may be better benchmarks, so I'd reckon the AF101 power consumption is pretty average for a camera in this class. If it was doing it "right" (reading all photosites each frame) I'd expect a figure considerably higher. I think there's also the question of cost. Doing it "right" is likely to be cost prohibitive for a camera at this price level.

The assumption is that a stills mode would be pointless due to the use of an OLPF which limits the resolution to HD video. In a way, doesn't this add weight to the likelihood that it *is* reading the chip by pixel skipping? If it was doing it the "right" way, then why not use an OLPF more appropriate for 12 megapixels, and lose the detail that would cause aliasing in the downconversion?

And I'm not sure I would necessarily call Panasonics approach a "silly" thing to do, anyway. Yes, it throws light away (so impacts a lot on the sensitivity), but it keeps cost, power consumption, complexity etc well down, which I'd expect to be of far greater importance given the aimed for market. They are using a sensor fundamentally designed for still operation. The approach has got over the worst of the DSLR problems, whilst keeping the cost reasonable. That said, it won't be as good as a large format sensor designed specifically for video with a pixel count to match.

And really, that's the problem for Panasonic. Less that there's anything technically "wrong" with the AF101 approach, but that Sony have brought out two large sensor video cameras which do have purpose designed sensors. In the space of a month they've gone from hoping for a competition free market niche, to seeing two competitors!

However good the AF101 is, all the indications are the F3 is better. But, yes, it's a lot more expensive. If you've got the money, the F3 is the one to have. If you haven't got the money, it's the NXCAM with the F3 sensor that must be the one to watch, and the expectation is for a full announcement at NAB, product a month or two later.

I haven't called Panasonic's approach a 'silly thing to do', because I don't know what Panasonic's approach is, and nor do you until we've tested it. We have no idea whether it throws light away or not, there's nothing in the spec to indicate either way.

And, while we're at it, I have actually pointed an F3 at a zone plate chart, and have seen things that need explaining. I haven't done the same with an AF101 yet. Until both cameras are properly tested, none of us can give any verdict on how good either is, particularly relative to the other.

There really isn't any point in this guessing game, it's all a silly waste of time.

Well, neither of us may have tested it, but that's not to say others haven't. No, I've not heard of anybody doing such as you do in such a scientific, exact, or precise manner, but it would be utterly wrong to therefore ignore such efforts. Very often, an intelligent and informed - albeit approximate - answer NOW is worth far more than a precise answer in the future, and that applies to many other things in life than camera specs.

Alan, we've been through this all before, and as far as I'm concerned "guessing" means just pulling a wild figure out of the air. (Which would be a silly waste of time, yes.) This is not what's happening. What you call "guesses" are far more than that. Rather a mixture of previously announced details, measurements, and intelligent deduction. Look back to the F3 thread ( http://forums.dvdoctor.net/showthread.php?t=52215 ), post 5 onwards, and I said:

None of that was simple "guessing", though you dismissed it as such at the time. Six weeks on, it's pretty well all been confirmed, with the exception of the point about the AF101 pixel skipping. I have every confidence that will turn out to be as accurate as the rest of what I was previously told.

In terms of evidence, the results of (approximate) tests seem to indicate the F3 has a far better signal/noise and/or sensitivity than the AF101. The information I have is "much better than a stop, maybe better than two". OK, we'll wait for you to eventually tell us exact figures, but in the meantime it's being widely interpreted as mainly due to the AF101 pixel skipping, slightly due to the larger size of the F3 sensor.

I'm certainly not ignoring what others do with cameras, far from it. But, neither am I trying to guess the performance without properly testing it. I'm quite happy that the AF1010 has a 12.4Meg sensor, no argument, but I'm certainly not going to be drawn into a guessing game as to how it gets from that to 1920x1080, simply because there's no point. As I've already said, many times, it's a silly waste of time and effort, I've got far more important things to do, like earning money.

And, on the subject of the F3, I do actually have some information which you don't have access to, but am not going to let it out before I've confirmed or rejected it, and that needs testing.

Why is the F3 being mentioned in the same context as the Panasonic.

Isn't it about 3 times more expensive?

It is but I think they are going to pop the same sensor in a lower priced AF101 type camera, my bet is that panasonic will then pop the 4/3 chip into an HPX601..........I'll get me coat!

It's a technical comparison, rather than a "buyer" comparison. It gives a reference to compare the AF101 to. They both have similar sized sensors, both use similar technologies (Bayer type single CMOS chip), and both give a similar range of outputs. The expectation is that certain performance aspects would therefore be similar, and underlying sensitivity should be one of them.

It seems to be the case that this is not so, at least in one respect. The F3 seems to have a big sensitivity advantage over the AF101, so the question then is "why"?

It's been believed since NAB that the AF101 was to use a chip of about 12 megapixel, derived from the still camera range, and get video from it by pixel skipping. The use of an OLPF means the pixel skipping *WON'T* give the aliasing problems of a DSLR, but theory predicts it *WILL* affect native sensitivity because (as Alan puts it) light is being "thrown away".

It's a good case of practical measurements backing up the expectation.

And yes, as Gary says, there will be an NXCAM little brother to the F3 in a few months time, and it seems it will have the same chip at least as the F3. The rumours are it will be about the same price as the AF101.

It's also worth saying that the quoted prices are body only, and to make the most of such cameras you're really likely to need to spend a lot on lenses, and probably an external recorder. Hence package prices will show a far smaller comparative difference than body only.

For the NXCAM I would presume that it would come with a basic zoom lens. The difficulty comes when you need fast glass, then it becomes expensive.

The shallow depth of field thing is overrated quite a bit because unless you have lots of ND (or one of those new fangled vari-ND filters) you rarely get the advantage of using it. Low light is also another area where these big chip cameras take a bit of a hit too. Unless you have very fast glass much of the low light advantage over normal 2/3" cameras etc is lost because of being limited to f4 or 5.6 with cheap lenses.

So, you need fast glass, which in turn causes more issues because focussing an f1.2 lens on the fly isn't fun!

And that's one of the oddnesses about this trend for large format cameras. You get short DoF only if you use wide apertures, so why get a large format camera (say, the 101, 4/3 inch) and then fit an F/5.6 lens to it? You'll get the same DoF as, say a 2/3 camera with an F/2.8 lens, which would be a lot cheaper anyway, or even a 12" camera with a F/2 would give the same DoF.

If Red ever release the 2/3" Scarlet it will be interesting to see the take up of it.

The main lesson that manufacturers can learn from DSLRs is that it is possible to have decent video quality (witness the GH2) in a very small camera body. Portability for me is a big issue these days. And to be honest I do not see a need for the big shoulder mounts any more.

The PMW-350 for example is pretty much an empty body! There's hardly anything in it. With some ingenuity I'm sure that someone can design a much, much smaller camera that is still ergonomically sound for run and gun.

Let's not forget that the reason that we have the shoulder mounted brick design in the first place is nothing to do with being able to comfortably balance the camera on the shoulder. It was because when they first arrived they had to be that big to fit all the tech inside them!

With likes of the AF101 Panasonic and Sony are coming late to the party to ride the big chip bandwagon. Perhaps too late, at a time when many are getting tired of every clip on the web showing interview subjects with one eye in focus and the other out. A 2/3" sensor is a good compromise, and very useful for general shooting.

My wish; A 2/3" camera, perhaps a similar size (or smaller if possible) to the JVC HM700. My big question is why has nobody done this and instead stuck with the large shoulder mount design when technology is fast enough and small enough to do away with it? I realise that the whole 3 chip block in current 2/3" cameras takes up a lot of space, but single chip CMOS design appears to have come a long way in recent years.

Spot on Alan but sadly the availability of cheap camera's that can do shallow DOF has made it the latest fashion but as Simon says I think it's time is passing and people are getting sick of seeing every shot with 90% out of focus.

The body-size limit on cameras these days is the lens. The bigger the format, the bigger (and heavier) the lens gets. The old shoulder mount was a good way of carrying a large body (because of the tape mechanism) and a big lens, and the big battery because the electronics needed it. The only part of all that that hasn't, and can't change, is the lens. We're rapidly coming to the stage that stills photography wity big lenses hit decades ago, where it was common to fit the tripod to the lens and put the camera on the lens, rather than mount the camera first.

Miniaturisation of electronics is still going on, but physics can't change the size of the lens.

And I really do hope that this fad for small DoF is just that, and we can get back to seeing what is actually in the picture. I've seen a few documentaries recently where the DoF has been made so small that I've had no idea what the shot was about.

Jan the panasonic product manager just posted this on the DVI forum:

"Actually the 4/3's sensor is a little larger than 17.3mm but what is used in the still cameras is that size, 17.3 X 13mm. When the engineers designed the AF100, they decided to use a little wider on the chip to maximize the height measurement as well, so the image size of the AF100 is 17.8 X 10mm. This gives approximately 12.4 million pixels to work with"

True, the lens does cause issues, but only I would say when it comes to ENG style lenses with lots of electronics in them. After all, look at the size of most stills camera zooms (the general purpose ones, not the extreme telephotos). They are much smaller than any ENG lens, even the 1/3" ones, yet they are designed for a much bigger imaging area.

I doubt I ever used ay of the electronics functions on my J17. In fact unless I was threatened with a gun I never even use the zoom rocker! I much prefer to operate the zoom manually because I never zoom during a shot, except to snap to a new position. I could do that much better manually.

So while some lenses have to be big, ie if they have lots of electronics, are sub f2.0 capable, and extreme reach, they don't all have to be. It depends on the spec.

Even if you had to use a full size 2/3" ENG lens the camera body can be designed with this in mind so that balance is still achievable. For example being able to adjust the viewfinder much further forward. being able to adjust the lens palm holder position etc. It just needs someone with some ingenuity to come up with it and make a step change.

I didn't mean the electronics in the lens, but the longish zoom range that's customary in video cameras. Make a 17:1 lens at F/2 and it's going to be quite big, even for a small format, and especially if its interchangeable.

Exactly similar thoughts occurred to me as well.

But it could make sense in a sort of either/or way.

EITHER you use it as a shallow depth of field camera (effectively meaning fast primes), OR use it more conventionally, with a slower zoom lens, but at least the smaller aperture meaning the size/cost is kept reasonable.

If that meant (say) f5.6, first thought would be that the smaller stop would be too limiting for low light use. But the larger sensor should mean better sensitivity, and the hope would be that would counteract the smaller f stop. As a first approximation, the S35 sensor is about 8x the area of a 1/2" sensor, which all else equal should mean about a 3 stop gain in sensitivity. Then knock off about a stop due to single versus 3 chip. So the expectation would be for such as an F3 to be about 2 stops more sensitive than an EX1 - and that seems to be very roughly the case in practice.

Hence, an F3 with an f5.6 lens should be only about a stop down in low light performance compared to an EX1 at f2 - not three stops - with similar dof characteristics.

Unfortunately, the same doesn't hold true for the AF101. It's native sensitivity seems only to be about the same as an EX1 - this is the main evidence that it uses pixel skipping. This may not matter too much when used with primes, but is likely to make it not very good in low light with the vast majority of zoom lenses. That sort of sensitivity is OK with f1.8-f2 lenses - much less so when the max aperture is f4-f5.6. It'll be fine as the shallow dof camera - much less able to be used in a more "general purpose" way as well.

And frankly, I mainly echo what Simon and Gary say. For most filming a 2/3" chip is probably optimal. Even if the size of lens can't be shrunk, it should be possible to shift such as the shoulder rest and viewfinder forward with a smaller body to maintain handling balance.

(Incidentally Gary, the measurements you quote seem to indicate the AF101 probably uses the same sensor as the GH2. (Early reports said "based on the GH1".) I believe it's 4:3 in shape, but with the corners outside the lens coverage angle, and different sections being used according to required aspect ratio. Total pixel count is about 16 megapixel, and by selecting a 16:9 ratio, that's where the 12 megapixel comes from - 3/4 of 16.)

Hmmm

I wish that there was a future for zoom rockers, it seems that everyone has jumped on the cinema bandwagon of "no zoom in shot" yet for much of the work i do its essential to have the ability to zoom in shot (IMAG video screens at concerts) we rarely have enough cameras to cover the whole stage properly and so shots which drift from one band member to another zooming in to the area of intertest are essential.

I'd be very pleased by a smaller format 2/3rd inch camera, JVC sized or even a little bigger. 3 Chips or a single chip design which worked with B4 mount lenses. CCU control via triax with onboard card based recording would be perfect. Priced at a sensibly affordable point rather than the £50K+ solutions.

Chances of getting this? maybe about zero i guess as the entire world seems obsessed with silly digital cinema issues, creating cameras which work with toy photography lenses and have depths of field so short that almost nothing is in focus.

It has never been a bandwagon for me. It is how I have always shot things. A zoom in shot in most instances looks naff unless it is done in a purposeful, stylistic way. Although I accept for live performances on a screen you do need it. But that is an exceptional circumstance.

The good lenses are hardly toys. The optical quality of most of them absolutely trounces the pitiful efforts I have seen from Canon and Fuji in their ENG/EFP lines yet they cost far, far less! I cannot see any reason why I should be paying £10k for a lens that has inferior optical performance to a lens that costs £2k!

Now, if someone points out how much harder it is to make ENG glass with the zoom range that they have, then that is fine. But I would expect that £8k difference to give me a lens with equal or better performance, not less. And I know that I am far from alone in having these sentiments about current B4 ENG/EFP glass price/performance issues.

My company enters a big talent show production from January. We commissioned Panasonic HPX3100 to cover ENG for that and two AF101 for some special jobs and as a testing opportunity for eventual future drama applications. We took delivery of both models last week and just enter testing.
The AF101 has three interesting abilities that can possibly enhance the production values:
"cinematic" DOF , real overcranking, ultrawide/fisheye optics (from Panasonic micro 43). In this competitive world of TV production those three tricks alone can be of interest to enhance a show's values.
It is absolutely true that the choice of proper fast lenses (f2 for "academy" frame) is a precondition to employ selective DOF language successfully. On the other hand in the fast paced world of TV entertainment that f2 better be zoom based, no time to change primes. Tough call. The only products that approached our needs was theoretically Olympus Zuiko Digital line of lenses. We did not know how they would perform on the Af101 but risked the order anyway: 14-35mm, 35-100mm and 150mm f2 SHGs. Used on Olympus stills bodies the lenses are known to deliver exquisite resolution wide open and across the frame with topnotch contrast. In many tests the Zuiko zooms beat top primes from Leica and Zeiss.
Fortunately the first tests show we made the right decision. The Zuikos play an excellent front end to the AF101 body. The AF and IRIS work (although not in continuous mode and somewhat noisily). f2 on a cinematic lens (PL mount) would cost us much more than the 2000euro those zuikos typically go for. One shortfall is the visible breathing when (de)focussing on subjects. As for stills lenses it is really small,
, but by Hollywood measures probably unacceptable (for those "move focus from victim to killer" shots) .
Other than that, it's a match made in heaven, especially the 14-35mm/f2, which covers a popular range of angles in cinematography. Get them before they are gone.
Olympus just announced that the 4/3 line will not be developed any more....

Af101+Zuiko 35-100mm/f2
Looks serious, doesn't it...
Definitely some serious rails and rigging needed to stabilize it.

That looks a good lens Piotr and thanks for posting.

This may be useful for future reference and tests:http://101crew.co.uk/

Good link, thank you.

Hello everybody, I just realizeed my last post here was four years ago yesterday.
I was implementing then the "BBC setups" from Alan Roberts onto our SDX900.
Four years passed and....
How about those for AF101? and HPX3100?
Any chance to see them soon? TIA Alan.

Piotr

I'm wondering about how you get to control iris and focus? Does the lens have true manual rings for those - or are you reliant on auto - I note you previously say "The AF and IRIS work ....."?

My immediate thought is that it's a complete non-starter for hand holding. Whilst it indisputably does give shallow depth of field I wonder if a lot of interested people are aware of just how much they will have to sacrifice to get that......?

After a week with the cam I must say the whole 'ergonomics experience" is very oldskool. It's like one of the first "filmboxes" , rather awkward to handle and extremely tiring after several minutes. Its place is definitely on the sticks. I cannot think of any hand-held use without serious modifications from specialist companies.
The iris is controllable from the big dial/knob on the left/front/bottom corner. The Zuiko don't have manual aperture rings (which contemporary dslr lenses have them?) Yes, the panaleica summilux D 25mm/f1.4 does . Great lens on this cam too.
The focus is push only, ie. the lens focuses on the target after...pushing the proper button. There is a focussing ring on the Zuikos so manual is also possible. The sensor (and the whole micro 4/3 concept) use contrast detection for focussing, so older 4/3 lenses like Zuiko SHG cannot do continuous focus well or at all (they are tuned for phase detection mostly). All m43 panasonic glass (and such new m43 Olympus) are built for contrast detection and work 100% on the Af101 by default. Except they are rather slow ("dark"), which kills the idea of selective DOF.
To me it's strictly a drama production camera. Put it on the sticks, dolly, steadicam and it can be a poor man's Alexa. Attempts to use it in docus, ENG, corporate/wedding--anything requiring speed and mobility is a total misunderstanding. BTW Sony F3 looks very similar in concept and "ergonomics". Nothing wrong with that, as long as would-be users realize it's not your next Sony 150 or Panny 100.

Hi Piotr. All the Pana micro 4/3 lenses require firmware updates to get the best results and this update can be done on the AF-101 body. Once you have registered the camera you will get access to these.

Hi Piotr - you are right but you need to register for the 3 year warrenty and to get automatic technical updates about firmware revisions etc so it is worth it. Also some software like AVCCAM Viewer requires registration.

I'm testing cameras for the BBC tomorrow, the 101 may well be part of it, not sure yet. I also have the Sony F3 to look at when somebody gets round to it.

The Panasonic HPX3100 was only in 'engineering sample' condition when I tested it some months ago, as was the Sony PMW500. I expect to be testing both as production models sometime soon. There are test reports up to #40 posted on the web (Daniel Browning's site is the best at present) and there are 23 more awaiting publication. I can't release details of these until I'm told I can. Sometime soon, I expect them all to migrate to an EBU website and become EBU Tech documents, because I've nearly finished writing EBU Tech.3335, on how to measure cameras. There's also some work to do on R.118 which will be the parent document for it, plus some real research work still outstanding.

I am curious about AF101. In my own rather amateurish tests I found two interesting if a bit worrying characteristics. The camera/codec is fairly noisy in shadows at the supposed "native" 400 ISO. The resolution with the excellent zuiko 13-35/f2 lens is a disappointing 650 TVL (the lens in photography resolves 60lp/mm at MTF 90). Maybe the native ISO is not 400 after all but eg 200, maybe my resolution tests were not properly done or the OLPF is really strong. Will be interesting to find out.

A lot of people are finding the camera noisy and some tweaking of the scene files improves this, there is also a few reports that certain quick time settings can make this a lot worse during viewing in post processes.

There have a also been reports of the jello like imaging that happened on the HPX370 when it came out so no doubt there will be a firmware update to address some of these.

I am waiting to see with this camera but since I now have a RED ONE M-X available it may be that I don't need one at this stage but in the future I am sure the nikkor primes I have will be put to some good use!:D

When I played with the pre-production unit at home noise wasn't an issue. I think people using badly set up LCD monitors is the part of the problem. Images on my Plasma were very clean. Time will tell of course.

I'm getting our production demo camera for an extended period in 2 weeks time so I'll know more then.

Scene file set up should be done to get the "look" you want from the camera - it should NOT be thought of as a noise adjustment. Tweaking of the files to improve the noise situation will inevitably mean you are "off-tweaking" the camera in some other respect - robbing Peter to pay Paul. Increasing coring levels (for example) will certainly improve the noise level - but will also mean a loss of low level detail, and give a "puddingy" appearance to such as faces.

If the camera is too noisy, it's a sign the basic sensitivity is set at too high a nominal level, and (as Piotr says) the only real solution is to use it at an effective lower ISO setting.

None of this should be a surprise, I've said for a while that the AF101 does what it does by pixel-skipping, or only using a percentage of the total photosites. The best information I have to date is that it is most probably discarding 3 out of every 4 (more than the 2 out of 3 that was thought earlier), so the somewhat disappointing noise/sensitivity figures that people are starting to report are exactly as expected.

Now 1 in 4 used still means over 3 million (a quarter of 12.4 million), which even allowing for processing SHOULD be reasonably OK for close to full 1080 resolution. Hence I am a little surprised by Piotrs comments about what he has seen. (Piotr - I'm reading that as 650 line pairs across the horizontal, is that correct?)

The criticisms I heard about the HPX370 were less about jello (rolling shutter), but more about noise trails after moving objects. It appears that the 370 was introduced to meet criticisms about the high noise levels of the 300, and at first seemed indeed to be far less noisy. Unfortunately, it seems to have achieved it not by better sensors, but by aggressive electronic noise reduction - including inter-frame. That seems to work OK on static images, but give a trail of noise when the object moves, as might be expected.

It's just about OK much of the time on normal pictures - but severely limits the amount of post work that can be done before it becomes totally unacceptable.

The firmware update you mention doesn't seem to have really fixed the problem - just changed the balance between noise and trailing. There's less trailing - but a higher basic noise level.

Totally agree with what you say infocus but you put it far better than I can and from what I have seen the tweaking of the scene files are to reduce noise but do soften the detail etc.

The jello I meant was what you say the image streaking, as you say they claimed a new sensor design but it all all went very quiet once a firmware update basically gave you an HPX301 with two settings for wobble on or off!:p

Now the latest whinges re the AF100 are coming from the autofocus brigade as they want a fully auto focus shallow DOF camera that shoots in next to no light, not doubt there will now be pages and pages of why it doesn't work or track perfectly, sometimes I wonder who is buying this kit and what they expect it to do!

The 650 lines figure I quoted was read off a photographic chart, so I am not really sure I did this test properly. It was the central resolution trumpet consisting of horizontal lines, therefore I tought it represents "lines per picture height" resolution or the equivalent of TV lines (TVlines). It was somwhat consistent with the reading of HPX3100 which gave here ca 800 lines*. Also a bit low, should be really around 1000, I know. So maybe the chart, or my method were not really proper. Panasonic claims for the AF101 800TVL.
I hope Alan Roberts will give the real figures from his tests.
* read on reference monitor via SDi, cameras in progressive mode

I have now done formal tests on the AF101. I have yet to do the file processing, which will take about a day, and then I write it up as usual. When that report has been delivered to the people who are paying for it, and they say so, then I'll be free to comment. Until then ...

Well, I tried to dampen down some of the expectations earlier in this thread, but unfortunately too often people believe what they want to believe - not what's likely to be true. It also seems to be the case that Panasonic have over hyped this camera, so now, even if it's not actually "bad" it seems at the very least it's nowhere near as good as expectations.

OK, I think you're probably right in so far as that goes. These sorts of figures don't really mean a lot without defining units, and it's good to talk in terms of something like "lines per picture width" (or line pairs per picture width). Hence, for a 1080 system, the theoretical maximum is 1920 lines, or 960 line pairs. The advantage of making the definition thus is that the same figure is applicable regardless of whether we're talking about horizontal, vertical or diagonal resolution. In which case, the 650 lines you talk about would translate to 1155 lines related to picture width, or about 575 lines pairs. That would be equivalent to about the resolution of the 720 system. Frankly, I'd expect a lot more.

Even if Alan can't speak about todays tests specifically, then I hope he may like to talk a little about zone plates etc in general? I'm sure he'll do a far better job than me.....

In the interim, there are at least two problems with a chart such as you describe. First is simply that it is very coarse, if the measurements are just 800 and 650 - how do you tell the difference between 680 and 760, say? Second is that I don't think it is possible to tell the difference between real detail and aliases on such a chart, with just horizontal and vertical lines. In your example, you know you are giving an input equivalent to 650 lines, and the camera is giving an output - but is that output "real", or aliases. It also doesn't tell you what the diagonal performance is like.

The answer is zone plates, and with Alan around, I'm not even going to attempt any further explanation myself ......... :)

I don't know what Alans results will show, but I'm now hearing enough stories that make me feel I'd want to see his tests or an equivalent before parting with my money.

All I know is that from what I have seen, the grass certainly isn't greener on the other side. After seeing what I have I do not feel that the camera gives me anything at all over my EX3 apart from more depth of field control.

Can you get to PL mount without too much of a crop on the sensor to S35? I think the AF get's folk this.

I agree not a massive dealer breaker. But it's sensor size and acess to cinema type fast primes give the target purchaser a reason to buy.

Still you already own a camera that suits your needs.

Yeah I can't get my head around this.

Why buy a budget "cinema" camera and talk about auto-focus or noise or anything.

It does what it does for around 3K.

People expect too much for their money.

I've had to fork out (and struggled through a crap year) the best part of 30K to get anything like a no-compromise image quality.

Totally agree Tyrone but it tends to be our american cousins that do this, we used to get it all the time when I worked for AMS Neve, they would buy the product and then spend endless hours complaining that it didn't do what they wanted it to.

Trouble is, if I could afford to rent or buy PL lenses on a whim for general freelance camera operating, I'd be buying an Arri Alexa, not an AF101. It's nice to know it can be done, but just like buying a supercar that can do 240mph, but never really get to drive it at that speed, ever, it is more of an ego trip for the purchaser than anything else.

Let's be honest, the vast majority of people who get an AF101 will either be using their stills lenses (which will give a crop factor of 0.5 for APS-C lenses), or inexpensive kit lenses. At f4 - 5.6 that doesn't really give a massive DOF advantage.

I agree. But I also think that things have become too camera orientated. The skill of actually being able to put together a video that gets results and delivers something far more than a vacuous series of beauty shots is now quite underrated.

Perhaps Simon, but I think (especially with the US) they can afford to own the camera and have basic lenses for no-budget use and then rent glass. I think that's the consensus.

Yes, right, but as said before, it's been overhyped. Just look at the reviews Gary and Richard previously linked to earlier - http://forums.dvdoctor.net/showthread.php?t=51452&page=9 - (posts 81 and 83) for example. They're full of remarks like "the results were stunning" and "The detail was incredible in the dark hair.....". Combined with what Panasonic have been saying, it led a lot of people to expect a bargain - a Red at a knock down price. Look back at the account you quoted from earlier on in this thread (post 24), the official word when it was shown at IBC - http://forums.dvdoctor.net/showpost.php?p=372103&postcount=24 .

It's beginning to seem that the results are at best no more than you'd expect from a camera at the price point. (Hardly surprisingly.) That doesn't make it bad - but it doesn't seem to be the bargain that people had been led to believe.

And that's the problem with marketing. Undersell, and you don't get the interest you might. Oversell, and it can lead to disappointment and a backlash. In Garys case, I wonder what the American marketing people had promised the customers before they bought? "Yeah, yeah, it'll do that...." They make the sale (and get the commission), and poor people like Gary have to pick up the pieces afterwards.......

Maybe it's American salesmen you should be blaming, Gary, not the users?

Yup I seem to remember most of our customer support headaches at AMS Neve were down to the salesmen, well actually one in particular and he used to sell cars!:eek:

Seem to recall he was sacked for not filing his expenses claims on time and the company had to pay the inland revenue fine!

To be fair the jury is out for me whether it's a bad camera for the price. Hyping is nothing new at all and can be found on anything.

As for the noise there is lots of debate about this and doesn't appear to be conclusive. And let's face it most cameras have some noise, it's just about where it appears in the chain.

I have the Panasonic GH stills/video cam, with the similar? same? sensor. The camera gets some noise in shadows even at 200ISO--when you pixel-peep at 100% or 300% on a monitor screen!
That is nothing unusual for that size of sensor and pixel pitch (very small). So frankly nobody should be surprised that it's carried over to the videocam based on that sensor.
In normal viewing circumstances, on prints, this kind of noise is not disturbing at all.
Similarily if one watches video on a flat screen from an average roomviewing distance.
Pixelpeeping is not a normal viewing behaviour. Perhaps it was unreasonable to expect a noiseless sensor on the AF101 after all.

But forms of post production etc ARE normal behaviour in the industry, and that covers everything from editing, colour correction, chromakeying, effects work etc to compression for distribution. And such things can magnify flaws in the original, such that something only visible via "pixel peeping" in the original may be objectionable in the finished product.

I can't comment directly about the AF101 here, but I've certainly seen that principle with the HPX370 and the noise trail issue. On some material it's not noticeable, on some it's acceptable, on a chromakey test I saw it's bad enough to be considered unusable. Let's hope Panasonic aren't using the same type of noise reduction on the 101 - if nothing else, it destroys subtle differences that may be needed in a post process.

The AF101 is being marketed directly at the digital cinema market. As such, the expectation is that at least some of the material is likely to go through a post process, and "just about OK on normal viewing" is quite likely to mean "not OK after post production and compression".

It's the same principle as to why 10 bit working may be desirable. Even assuming the front end of the camera is up to it, you won't see any big difference between 8 and 10 bit on normal viewing - but 10 bit may allow better post working.

The other point is that a camera with higher intrinsic noise may be OK at a reasonably low ISO setting - but unacceptable with even slight amounts of gain. Hence a dead loss for low light working. How good or bad the AF101 is still remains to be seen, but noone should judge solely on normal pictures.

Yes, on our HPX301 we had some strange noise on a chromakey - didn't stop a good key though and the client would never begin to notice, and they still paid up.

Fixations with picture quality and the transmission chain are vastly over inflated espescially if you're into corporates/films for yourself. If you can buy a camera - get good images - and make money, then pixel peeping largely doesn't coming into it.

It's much more of a problem on the 371. The 301 got a bad reputation for general noise, and the 371 quite soon came out as a replacement to "fix" the noise issues that had been complained about. The "fix" seemed to involve heavy electronic noise reduction, which relied on frame by frame processing - fine on static shots, but led to noise trails on movement. That made it even worse than the 301 for processes like chromakey, the effect stood out like a sore thumb after processing, even if it didn't look too bad before.

At first Panasonic tried to downplay the issue, then (as Gary has previously said) brought out an "upgrade". That seems to turn greatly reduce the effect of the noise reduction (and hence the trailing) - but mean it's nearly back to square 1 in terms of general noise!

It's quite possible that the problems Panasonic seem to be having with 1920x1080 1/3" chips may be why it's taking them so long to upgrade the HVX200/171. Hence a change of tack to promoting a big sensor camera.

To add to that most of the noise problems with the 300/301 were in NTSC land, I've never had a problem with noise on my 301's but I always keep to the BBC settings in Alan's report.

I just go the graded Pro Res HQ master of a feature film I did last year and I am struggling to see any noise at all on my 37" LG TV and there was a lot of night shooting done on that film.

I appreciate that is not a scientific test but I will see what it looks like projected on a cinema screen later this week.

The camera I saw was a 371 in "PAL land", and the noise trailing was visible even at 0dB, certainly once you knew what to look for. OK, that might be considered pixel-peeping, but the point was that processing magnified the effect to an objectionable level, and the same if you'd wanted to use gain. It was more objectionable than a much higher level of straightforward noise, so I'm not surprised that the "upgrade" is simply to turn it off.

It doesn't mean you'll always get problems, but if I had this camera I'd be reluctant to use it with gain. I don't know what you did, Gary, but for a night scene, I'd also feel obliged to work at a higher lighting level than with other cameras - simply to avoid the use of gain.

Any news on when you may be able to release your reports about this camera, Alan?

I have to say I'm just hearing some quite disappointing reports about it in a number of respects, quite apart from noise issues. In particular, the resolution appears to be more in line with what you'd expect from a 720 camera, and whilst the aliasing is nowhere as bad as a DSLR in video mode, the reports I've heard are not as good as may be expected from most video cameras in this class.

I have to say this surprises me. I was pretty sure from the start it was using a 12 megapixel sensor (which has been confirmed), and was fairly sure it was sub-sampling the sensor - that explains the sensitivity not being what you'd expect. But even if a 1 in 4 sub-sample, that still gives 3 million used pixels out of the 12 million. Even that should give substantially better performance than what is being measured.

I hope to see more soon, but these results do tend to back up what Piotr mentioned earlier -

A few days more, be patient. While the spec says the sensor is 12.4 million, it also says that the image comes from a 17.8x10 mm part. Another document I have from Panasonic, unpublished, says that the sensor is 4:3, but doesn't make clear whether the part used for video is 12.4M or the whole sensor. My tests seem only to produce a bit more confusion. I'm trying to get that cleared up by Panasonic before releasing the document, because my measurements don't see to back up what would be the more logical situation..

I'm pretty sure I've seen it officially stated that the sensor is the same as used on the GH2 still camera, and of that dpreview says:

I believe that 16 megapixel resolution applies to the chip in 4:3 mode, and used in this camera, in video mode, it's roughly a 16:9 letterbox from this, so you'd expect about 12 megapixel effective.

In fact it seems more complicated, since the chip seems to be oversized, and about 18 megapixel total. So, in 16:9 mode the width of the chip used is a bit wider than for 4:3. Because of the reduced height, the corners of the frame are still within the circular coverage area, hence the 4:3 and 16:9 frame areas have close to the same length of diagonal. I believe for 4:3 the frame size is 4608 x 3456 (15,925,248), and for 16:9 it's 4976 x 2800 (13,932,800) - so it's actually nearer 14 megapixel than 12.

The logical situation would seem to be reading 1 in 4 pixels, so giving an effective value of about 3.5 megapixel. Which should give very nice 1080 resolution. Unfortunately, the measured values seem to show far less - "more in line with 720 resolution", is what I've been told. The best theory I've heard suggests that 1 in 4 photosites are being read, but in clumps of 4, (2 green, and 1 each red and blue) and a three colour pixel derived from those. Consequently an expected resolution of more like a quarter the effective sensor dimensions (4976 x 2800) or 1244x700 on the horizontal and vertical axes, and better on the diagonal.

The tests certainly confirm that. there's no evidence of pixel skipping, either from the resolution or sensitivity. The blurb talks of the pixel counts being in the ratio 1:2:1 for R:G:B, but don't mention Bayer pattern. Don't bother with guessing at this stage, all will be revealed very soon.

While i will be interested to see the tests, I already know that the camera does not perform a patch on alternatives. This camera most certainly is not a golden bullet. Lens choice and DOF flexibility are about the only advantages as I am sure Alan has seen. Resolution is severely limited compared to what one would expect from the sensor and the way that this camera is focussed on video acquisition. While a lot better than the Canon DSLRs there is still objectionable aliasing. As a camera operator and purchaser that is all the information I need to know. It's a deal breaker for me and I'm keeping my EX3 thanks.

I suppose - Simon, that is reflected in the price point entry - if you're going down the still prime lenses route for the AF101.

I agree - but we come back to the subject of hype. The advertising talks of "enables users to achieve cinema-like results"and "can record video in full HD". That led to expectation that technically we could expect a quality level at least on a par with something like an EX1. Not a direct competitor, because the larger chip size was seen capable of giving advantages and disadvantages, but resolution etc fitting what is marketed as a 1080 camera - on a par with the EX or Canon cameras.

I don't know if Simon is basing his comments on charts he's seen or real life pictures, but from what I've now seen I agree broadly with what he says. Having just actually seen the evidence, the earlier comments that were made to me ("more in line with 720 resolution") seem fully justified, as do other criticisms. It's not that it's bad - lot's of productions have been made with 720 cameras in the past! - rather it hasn't lived up to the hype.

I confess I'm mystified as to why the resolution does seem so relatively low. It's starting off from something like 12 (or 14?) megapixels on the used part of the chip - yet the resolution seems worse than (say) the XF100 with 2 megapixels.

I'm also surprised to hear Alan think it doesn't pixel skip. If it is reading out the entire sensor every frame, the downconversion must be truly abysmal!? Isn't pixel skipping, along the lines of what has been suggested to me earlier, a far more likely explanation? And one which is backed up by matters such as low power consumption. I also note the comments about the CMOS skew being reasonably good - this is also presented to me as a strong sign that all the pixels are not being read every frame. Only read out a percentage of the total pixels, and you can read a frame faster - hence less skew.

I think the main thing to bear in mind is that most camera's these days are designed, sorry sold to the USA market and 720p with shallow DOF tends to be their demands.

As we all know there are all sorts of people who will whinge and hack these types of camera's and then wonder why the pictures are A: Soft B: Noisy or C: why can't I do ten different frame rates with varicam to get my 24 frame (23.987564736352416245240) cinematic pictures at 4k like the RED Scarlet will do if it is ever invented and costs $3,000!
:D

We in the UK tend to get on with making content rather than thinking that 4k is more needed than 1920x1080i 25np unless that is if you are one of those 40,000 noo mediah stoodents who think that doing their course and buying a canon 5D makes them a DOP!
(dopey overrated prat)

I don't think it doesn't pixel skip, I know it doesn't. Stop guessing and wait for the report. It won't be long now.

We're talking about a camera I *assume* you haven't played with (apologies if you have) or I haven't played with so why the concern - it's maybe not the camera for you. Getting rid of aliasing and having SDOF is where the is camera is targeted. If anything it's good news (IMHOI) because it means panny may bring a higher spec P2 version out.

As for marketing, we've done this before - it's not unique to panasonic - all companies big up their products. And it does *record* in FULL HD, even if it doesn't resolve all the detail. Many 1080 (even expensive ones) cameras don't.

Thing is - companies can't control people's expectations (i'm not saying they can't influence) but if you stay away from the hype - wait till it gets tested - you wouldn't have the expectations in the first place.

There had to be some short comings and I think Panny have perhaps aimed their development costs at aliasing control - I don't know because I haven't had one in my hand. Sounds like we are going to find out though!

They should sack the people they paid to do it then!

Okay, we'll wait. But considering this camera can record 1080p at 60fps you can't exactly be surprised when the following questions are asked, expecially now you have said that it doesn't pixel skip.

1) If it isn't pixel skipping then is has to be reading the full sensor out. How on earth does it manage this at 60fps without cooking itself?
2) How on earth is it dealing with power consumption if it is reading the full chip out and dealing with the resultant processing?
3) How on earth is it managing such mediocre results if it is taking a reading off every pixel of the chip?

Alan, while I know you keep saying that we should wait for the report and stop guessing, you have only fuelled the fire. By stating that you know it doesn't pixel skip there is really only one other way it could be doing things. And that method doesn't correlate what you have said yourself in the past about high resolution, high framerate sensors not being possible in such cameras due to cooling and power issues.

I asked US product manager Jan Crittenden directly if it pixels skips and she answered directly "no". That was all I managed to get out of her.

That makes a lot of assumptions, some of which are true, but by no means all. Once you start trying to design such a camera you quickly see the many different ways of doing the same thing, and then you have to home in one the ones or one that is economically feasible.

For example, the ARRI Alexa has a spectacularly good way of scanning the sensor, which consumes a lot of power, but works beautifully. The Panasonic has gone in a different direction in an attempt to get there more cheaply. And, if you compare the prices of the two cameras, you maybe see why one is better than the other.

I'm trying really hard to avoid stoking the rumour mill, because it's a waste of all your time.

Goodness me, Sir Alan Roberts lives here! How many camera purchases have I made over the years after reading your verdicts on them to the beeb!

I've only recently heard about this camera and I'm quite excited about it, but realise there is a lot of hype out there, of course. I too am looking forward to Alan's test results on it before getting too excited.

Yes, I am using the 5D2 for a lot of personal work (I know, I know), but the portability of it and the ability to snap on and off nikkor/zeiss/canon primes makes it very useful. It also meant I could throw away my Letus Elite (always hated the set up time for that).

What interests me most about this camera, besides interchangable lenses, is the ability to over/undercrank it. I'm really looking for something decent at this price range that will give me nice slow motion, so I'm keen to see how this camera really fares in that department. I've been so underwhelmed with the efforts of most DV cameras attempting overcranking, so I'm keeping my fingers crossed.

Very nice community and forums, I'm surprised I haven't noticed them up until now.

Kris Koster

I had a little play with this camera at Prestons today and it looked extremely good to my eye, but just not for me(weddings!)

Great to meet Richard Payne from Holdan and had a good chat though!:)

Cheers.

I wondered at the time why Alan said that – I now assume it is based on the following sentence in his test results:

“It is very obvious that the scaling has not ignored (skipped) sensor pixels, since that would have invoked coloured aliasing, both horizontally and vertically.”

A friend asked to be kept in touch with this subject, and I asked them about this specifically. I’ve just had the following reply:

There is a method of reading the sensor which could involve pixel skipping, but would not give any colour aliasing. To my mind, it's a more likely explanation for measured results than reading at high resolution and downscaling.

It will also give equal R,G,B resolution, each at one quarter the native sensor resolution and would mean reading in the following manner:

<------------------4700 photosites----------------- >

o o R G o o R G o o R G o o….. ^
o o G B o o G B o o G B o o….. 1322 photosites
o o o o o o o o o o o o o o……… total vertically
o o o o o o o o o o o o o o…..
o o R G o o R G o o R G o o…..
o o G B o o G B o o G B o o…..
o o o o o o o o o o o o o o………
o o o o o o o o o o o o o o…..

Where R, G, B are read red, green, blue photosites, and o are skipped ones.

The basic unit becomes :
o o R G
o o G B
o o o o
o o o o

And it should be obvious that there will be a quarter the no of those (horizontally and vertically) compared to sensor photosites. Hence, basic camera resolution defined by the number of basic units, 4700/4 and 2644/4, or 1175x661 – exactly as measured.

Such an explanation would :

Give no colour aliasing
Give equal resolution for each colour (and each at 1175x661)
Explain the sensitivity/noise
Be consistent with low rolling shutter
Be simple to implement (it’s probably what the GH2 uses so no extra R&D)

All consistent with what is measured.

What makes me favour such an explanation? Have you ever heard of “Occams Razor”? It’s a guiding principle in science which roughly states “…..we should tend towards simpler theories until we can trade some simplicity for increased explanatory power”

In this case, the razor points towards an explanation similar to above, and away from any explanation that suggests anything far more complicated, and a poor downconversion.

It's by far the simplest explanation which fully explains measured results.

-------------------------------------------------------------------------------------------------------

I cannot fault any of the logic there.

If anyone (and especially Alan) can find anything wrong with the principle, then please tell us !

Would this also account for the resolution apparently being better on a 45 degree angle compared to the horizontal and vertical directions as demonstrated in Fig 2 of Alan's report?

In ANY sensor with horizontal/vertical alignment of the photosites you will get the effect of diagonal resolution being better than horizontal or vertical.

Mathematically, I've been told it's because you can break the resolution down into vector components, so for a diagonal line it's partly due to horiz, partly due to vert characteristics of the sensor.

The Panasonic marketing claim is: "• Large 4/3" "Best-in-Class" MOS imager with fast image scanning and optical and low pass filters that eliminate aliasing and moire while minimizing skew" : http://catalog2.panasonic.com/webapp/wcs/stores/servlet/ModelDetail?storeId=11201&catalogId=13051&itemId=473159&catGroupId=112502&surfModel=AG-AF100&displayTab=O

In the brochure, they show vDLSR aliasing by showing chroma moire, which from Alan's tests appear to be absent from the Pannasonic, however, the camera demonstrably does alias quite strongly....

The resolution, clean resolution, in the AF101 is exactly the same as you'd get in a 3-sensor camera with identical (but limited) resolution sensors. My tests clearly show that the delivered R G and B resolutions are identical (near enough), which must mean that the camera has two separate down-scaler designs, dealing with red and blue via one scaling filter design, and green via the other. The implication is that, rather than decoding the Bayer pattern at the full resolution of the sensor (which is the 'right' way to do it) and then downscaling the RGB results to 1920x1080, they've taken the red, green and blue patterns separately and downscaled each to a common denominator directly from the sensor, i.e,. the Bayer pattern isn't actually decoded at all.

If the sensor resolution (pixel dimensions) were precisely 3840x2160, then this approach would work well (because each of RG and B would have at least 1920x1080 pixel count, but the G would need some simple interpolation), but, again, my tests show that this isn't so, which is a shame. It isn't easy or cheap to get this sort of processing right, it's actually very hard. I've got filters (I designed them, and Christian Lett now knows how to do it as well) which would do it right, but the number of calculations needed for each output pixel would be a couple of orders of magnitude, or more, greater than what's actually in the AF101, and that would mean more power consumption, more heat, more weight, a cooling fan .....

A native sensor resolution of 3840x2160 would also downscale well to 720p, as it's exactly three times as much pixel density.

One thing that's interesting about the RGBG Bayer Filter grid is that its RGB geometry can be mapped with a one-to-one correspondence to the 4:2:0 YCC color sampling used in AVCHD encoders. With a 3840x2160 sensor, the encoder's individual YCC planes could be directly scanned from the sensor rather than converted from a downsampled RGB scan. This could make it practical to use the native 12-bit resolution of the sensor's photosites at the input to the encoder, providing enough color depth to support 10-bit and/or 4:2:2 color resolution in the encoded output.