What HDTV should i get?

For many people, the extra isn't worth it, but if you're after the best image quality, and can recognise it when you see it, it's well worth the extra money. In broad terms, Sony and Panasonic use good quality scalers, the others are much less predictable. The scaler is the circuit that converts the incoming signal from whatever standard it is (720x576, 1280x720, 1920x1080, etc) to that of the panel (typically 1366x768 for 32"). Sony's and Panasonic's scalers use filters with enough taps to do a good job, the others use smaller filters with variable results. I've seen one panel that used no scaler at all, just copied or dropped lines to make it fit, with the expected appalling results.

If you're going to buy a flat panel for HD or other viewing, insist on seeing signals of all the types you're going to show on it before you buy. Then use your own judgement.

My understanding is that the long term goal is for displays to be 1920x1080 native. Any guesses as to when we might start seeing, say 50%, of screens to be such resolution?

Two issues here:

1920x1080 display resolution isn't really needed until you start viewing pictures at around 3xpic height, and for the average UK viewing distance of 3 metres that means a display that's 1 metre high, 2 metre diagonal (80 inches). That's just because the pixel size matches the human acuity limit at that size and distance. Abviously, if you watch at say 2 metres instead, the 1920x1080 limit comes at 2/3 of that, 53 inches. There are panels of these sizes, and bigger, but most of the consumer market tops at around 42 inches, for which 1920x1080 makes sense only when you view at 1.5 metres.

Exact pixel-matching of the display to the signal is relevant only if you eliminate overscan. For some reason that I still don't understand, all the panel-makers still insist on incorporating overscan (which was a requirement only on crt displays when stable power supplies were impossible). So each panel has a scaler that resizes the incoming picture. The better the scaler, the less damage it does. Sony and Panasonic have good scalers, others are less good, some are disatrous. And the closer the pixel-count ratio of signal-to-display is to unity, the more important this gets.

So, native 1920x1080 would be lovely, but it's not really important in consumer displays for a while yet. The current size of 1366x768 is a decent compromise, it's ok for 720p and for HDV (just), and is fine up to around 37" from 3 metres.

Does that help?

Yes, in terms of the background, though I'd understood that a move to 1920x1080 was seen as eventual policy, and just wondered what the timescale was likely to be.

I've been doing a little measuring up in preparation for (hopefully) an eventual flat screen. At present our viewing distance seems to be about 2.6m - very close to the 2.7m that a link Alan previously gave (BBC R&D document) assumed as a UK norm. Currently our screen size is limited more by the size of the TV and stand than screen itself, and is a 32" widescreen.

Looking ahead, I picked up one of the sheets that a big retailer was given away - paper foldable to any of the most common big screen sizes. With the new room layout, the viewing distance stays at roughly 2.6m, but I was quite surprised to find that with the screen flat against the wall, even 50" didn't seem oppressive. (Any bigger would, and probably about 48" would be optimal.) Currently 42" seems to have the best size/price compromise - but maybe in a years time........? I just hope our current set lasts until HD broadcasting is more commonplace, and 50" prices are more affordable.

Again referring to the R&D paper, for those specs HD becomes highly desirable, and according to the given table probably 1080. Interestingly, the paper seems to have been written a couple of years ago and the rapid increase in available screen sizes underestimated. Hence some conclusions about the adequacy of 720p may now be seen as slightly outdated. (Though the reasoning was sound at the time.)

I've always held that the conclusions in that paper are flawed in that they grossly understimate the public's want for bigger displays. I'd been asked to review the paper and said just that, and it got me in an awful lot of trouble in certain parts, which bothered me not a jot :)

The experiment on which that paper was originally based had been designed (by me) to measure human acuity to colour, not luminance. The luminance part of the experiment was to have been a calibration process, to be completed before starting on the colour work. But the colour work was never done, and the whole project was used to provide evidence to convince the EBU that 720p was what's needed in Europe. Scandinavians had already decided on 720p, so confirmatory evidence was needed. Hence the furore at IBC 2004. Although I took part in the measurements (I'm the one that always wanted far more resolution than anyone else, ignored by the conclusions), I dissociated myself from the document's conclusions. The measurements are correct and correctly reported, but the conculsions are wrong.

Well, that's what I think, anyway.

That's the conclusion I've come to from my somewhat less scientific tests. (I certainly won't quibble about the bulk of the underlying science.) For anybody else with interest in this, the paper referred to is at http://www.bbc.co.uk/rd/pubs/whp/whp-pdf-files/WHP092.pdf with conclusions on page 7.

It's the table on page 6 which interests me, since at the given viewing distance of 2.7m (very close to what it really will be in our house) it predicts that even for a 42" screen nearly 30% of observers will require a 1440x1080 or better display, the figure rising to about 45% for a 50". The conclusion that 720p is good enough really doesn't seem valid anymore, and since that report was only published in September 2004 that must be a sign of how screen sizes are growing far faster than many expectations. Yet another reason for maybe setting a standard better than what seems to the minimum that can be got away with at the moment?

Couldn't agree more. WP92 assumed display sales based on advice from Eurpoean manufacturers, but virtually all the flat panels are coming from the Far East and sizes are going up at about the same rate as prices are coming down. Experience in the US is (indeed, was at the time) that sales go up far more quickly than the EU prediction, and that's exactly what I said at the time. As a piece of prediction, WP92 was out of date when it was being written, let alone when it was published. If you leave out the conclusions regrading 720p, WP92 is a good piece of work; shame it was never allowed to do the job I designed it for, to establish whether 4:2:2 is sensible and whether the Cb/Cr axes are the most efficient.

The world's broadcasters, with a few exceptions, are all agreed that 1080p is the final goal (Japan can't see any problem with 1080i and Scandinavia can't see any problem with 720). The problem is the processing data rate. 3Gb/s is a lot to handle. Compression will come to the rescue, H.264 can do it now, so can Dirac, and Richard Russell has a cunning compressor that drops 1080i data down to that of 1080i while maintaining viewability (i.e. a 1080p signal can be digitally compressed 2:1 and the resulting 1.5Gb/s data stream decodes as a recognisable picture in a conventional 1.5Gb/s dac). So, people are worki ng on ways of doing 1080p, but there's no hardware yet. Sony have a prototype camera but no recorder. If you can afford Panavision Genesis or Arri D20, you can do 1080p now, but only by recording MXF files on hard drives, it ain't tv yet.

I reckon we've got at least 5 years of 1080i/720p before there'll be any possibility of 1080p getting broadcast anywhere in the world. That gives enough time for the manufacturers to have go a return on current investment, so it looks sensible.

My 2 pen'orth.

Help my thread has been hijacked!

Is that the HDC-1500? Since I was reading some reports elsewhere about people using that with a HDCamSR (with a frame buffer board?) to get varicam-like offspeed recording in 1080p24. The camera operating in 1080p60, and the buffer board rearranging the frames to tape to record in 24p as I recall.

Found a web version of the post -- http://blog.cameraman.at/index.php/film?blog=5&s=HDC%201500&page=1&disp=posts&paged=2 (last post on that page)

Steven

Apologies to darrenson for the hijack, but I think Alan really said it all with his first reply. Generalisations can be made, but beyond a certain point it's up to the buyer what suits them. Don't forget about connectivity as well as quality, though.

Why not 4:2:0? I'd have thought that a system with chrominance resolution half that of luminance in both horizontal and vertical directions was sensible? (4:2:0) Or does your above remark apply to an interlaced system when 4:2:0 would put vertical samples four lines apart, Alan?

With all apologies for the hi-jacking, but all the answers came early on, so why not?

NTSC originally rotated the colour axes by 33 degrees then filtered them differently, exploiting what was considered to be the different spatial resolutions of the eye to those axes. A great idea, but dropped very quickly because it was too complicated. So NTSC has identicfal filtering for the two axes, and the whole reason for rotating the axes evaporates. However, some brilliant, seminal, work was done at BBC R&D by Ken Hacking in the 1950s (IIRC, I can look up the paper if need be) which investigated the minimum bandwidth needed for colour transitions on several axes, with fascinating results. Some transitions needed 70% of luminance resolution, others were ok with 30%. But the problem was that he did his tests in linear light and with real colour slides in projectors. And most of the colours he used were outside the colour gamut of television. Nevertheless, his results have been quoted and used ever since.

The decision to use 4:2:2, and then 4:2:0, was based on engineering simplicity rather than colour science; it's far easier to use alternate samples (with appropriate filtering) than to use any other reduction process. What I wanted to do was to present colour transitions on several colopur axes and at several luminance levels, all via television displays, and to assess the bandwidth requirement that way. Filtering in linear light (Ken used optical difusion) doesn't come near the right process of non-linear television, even if it were a constant-luminance system. So the experiment needs doing. That was what I wanted to establish, whether 4:2:2 or any other simple variant actually makes colorimetric sense, and we still don't know.

Ho hum.

I bought the Samsung LE26E51 about 8 weeks ago. When should I expect a recall.