HDTV Shopping List

[Richard Choroszewski]

on page 34 of What Plasma & LCD TV it says, and I quote... "A minimum screen resolution of 1080 x 1920"

Yet in this magazine and What Home Cinema as hard as I've tried I've only found reference to 1 large format model that actually achieves this 'minimum' resolution.

Is this a misprint?

[infocus]

It's more normal to talk of 1920x1060 (widthxheight) and this is the common image format, in practice the maximum resolution come across for most HDTV. In the future it may well become the norm, at present, I believe it is indeed the exception, and the EBU cites a lower resolution as a minimum to earn a display the HDready mark.

To quote from the "HDTV in Europe" thread (Chatter):

In essence, this HD-ready label can only be attached to displays that meet the following conditions:

have a minimum of 720 horizontal lines;
accepts HD inputs via:
analogue YPbPr, and;
DVI or HDMI;
can accept 720p/50, 720p/60, 1080i/25 and 1080i/30 inputs;
the DVI or HDMI input supports content protection (HDCP).
This breakthrough is very important because it demonstrates that the emergence of two competing formats (720p/50 and 1080i/25) for HDTV in Europe will NOT cause problems for consumers. It also shows that there is no need for a format war: the two systems can coexist peacefully with each other.

[Alan Roberts]

The common image formats for HDTV are 1920x1080 and 1280x720. 1060 doesn't come into it, I guess that's a slip of the fingers.

However, most of the cameras and lenses in use don't "fill" the channel bandwidths to the extreme. For example, the Sony HDCAM recording system limits at 1440 pixels and the Panasonic Varicam limits at 960. Vertically, HDCAM can support the full 1080 lines only when the camera shoots in PSF mode (film look), in iterlaced mode it rarely carries more than 720 lines of resolution, like the Varicam.

So, for most of HDTV production, 1440 pixels is fine, and 768 lines is enough. It's only when you get people using prime lenses and full-resolution recording that you get more detail than that. Expensive drama and film production are the only areas you'll come across that.

What's crucial for viewing HDTV is that the display can accept and dela with the signals. Infocus' commenst are 100% correct there. But the display itself need not exactly match those numbers, yet.

[Richard Choroszewski]

Many thanks for the great replies

"But the display itself need not exactly match those numbers, yet."

So with the all important ",yet" caveat in there what do the pundits think is the time frame before HDTV with 1920x1080 becomes generally available in the UK.

Are we talking weeks, months or years?

[Alan Roberts]

Months.

Sky's launch will happen so that they can broadcast the football next year. Sky's own testing has shown that 1920x1080 i/25 is preferable to 1280x720 p/50, largely because we're used to interlace, so I don't see that there will be much production or broadcasting at 720p, maybe 10% or so. The figure's less than that in the US at present. Germany and France are both launching HD this year. Other broadcasters will follow fairly quickly. The BBC's response is not yet published, but there's a lot of discussion and planning going on, they won't let Sky keep too much of a lead.

I'll explain why I added "yet". There are several things to consider.

For many years, SDTV has been made digitally, with pixel count of 720 (only 702 of which carry real picture) and 576 lines. Vertically, this system can carry detail up to about 420 lines because interlace gets in the way. The filtering used in cameras and the transmission system caused a sharp limit at about 5.5MHz. Since the sampling frequency is 13.5MHz, the system can carry (in theory) detail up to 6.75MHz (the Nyquist limit for sampling). The filtering is there to prevent aliasing, where higher frequencies are mistaken as lower frequencies (wheels going backwards is a good example). So the 702 pixels can carry detail up to 5.5/6.75*702=572 lines per picture width. In order to display that correctly, you need a display with 702 pixels, and a filter between the pixels and you that eliminates the aliasing caused by the resampling that the display gives. In practice, no such post-filtering is ever done, because it has to happen in the light world, not the electrical world. The trusty cathode ray tube effectively does some filtering because it's modulation transfer function is Gaussian, so it attenuates 5.5MHz by about 4.5dB anyway, just enough.

I demonstrated this to myself a little while ago, by presenting a resolution chart on my 15" laptop screen, and on the 15" crt on my desktop machine at the same time, side by side. The laptop looked a lot sharper, simply because the crt display has a Gaussian-shaped spot which nicely atenuates higher frequencies.

It's when we come to pixel-based displays that life gets harder. Again, there should be a post-filter in linear light to remove the pixel structure (a diffuser) but nobody ever does that, so we rely on the modulation transfer function of the eye to do it for us. And that means placing the display sufficiently far away for us not to see the individual pixels easily. Then, the picture looks right.

Human vision has an acuity limit of about 1 minute of arc, which means that that subtends and angle of less than 1/60 of a degree at the eye appears to be a point. So you need only just more than 60 pixels per degree. If the display subtends an angle of 20 degrees at the eye, then it needs just more than 20*60=1200 pixels or you can see the pixel structure. My eyesight is a little better than that, I can see 0.7 minutes of arc, so I need more pixels.

So, to work out how many pixels your display needs, you have to measure the distance between it and your normal viewing position, and calculate the anmgle it subtends at your eye. This is simple geometry. Mutiply the angle in degrees by 60 and you get the number of pixels that the average viewer needs. Doing the sums for my practical viewing at 3 metres (the distance in my lounge) my 28" Panasonic wide tele (26" diagonal picture, 22" wide) subtends an angle of 2*ARCSIN(22"/2/3m) (sorry about the mixeed units) = 10.69 degrees. So the tele needs 10.69*60/0.7=916 pixels across it for me not to see the pixel structure. I can clearly tell the difference between analogue and digitaltv on that set (I've got a very good analogue signal) so I'm seeing the extra sharpness that digits provide, but could happily have some more than the 702 that digits manage.

If you do this sort of sums for your own situation, you can work out what you need for HDTV. I have a calculator chart that takes out all the hard sums (not that they're hard) if you're interested.

[hedleyw]

Try quoting that lot to a Dixons salesman :D

[Alan McKeown]

Alan,
In the absence of a satisfactory optical anti-aliasing filter between the TV picture and the viewer, would not a practical equivalent be to substantially increase the number of pixels in the display and up-convert the input video (by interpolation) to match this display? A form of spatial oversampling.

That way you could sit far closer to the screen than your 10H for SDTV and 5H for HDTV (which do seem inordinately large distances).


The viewing distance target for HDTV has traditionally been 3H.

(H = picture height)

For a screen height of H and a viewing distance of 3H, the vertical viewing angle subtended by H is:

2 * arctan [(H / 2) / (3H)]
= 2 * 9.46 degrees
= 1135 minutes of arc


For HDTV (1920 X 1080)
The pixel spacing is H / 1080 , so (for 3H viewing distance) each pixel spacing subtends:

1135 minutes of arc / 1080

= 1.05 minutes of arc

which is very close to the accepted 1 minute of arc for 20:20 vision. So a viewing distance of 3H seems about right for HDTV, even without display oversampling.


Alan

[Alan Roberts]

You're absolutely right, and that's the correct way to look at it (no pun intended). An oversampled display is inherently the best solution, always.

The 3H viewing distance came from research done in NHK over 20 years ago and has been largely superseded by events. What has happened in practice is that people have simply replaced their displays with bigger ones and put them in the same place, or wherever makes physical sense. So the actual viewing distance is the more important figure now, rather than the relative distance. Recent surveys in the US produced a figure of 3.5m, in the UK it's nearer 2.75m. In reality, people don't move the furniture around as they get bigger tellys, they always end up sitting at about the same distance because that's where the furniture is, so you have to do the sums using the real distance measurements to get it to make sense.

BTW, I have tried explaing it in Dixons, the results are always hugely entertaining :D