Slashdot Mirror


When is 720p Not 720p?

Henning Hoffmann writes "HDBlog has an interesting entry about many home theater displays.
Home theater displays around the resolution of 720p (most DLP, LCD, and LCOS displays) must convert 1080i material to their native resolution for display. No surprise there. But many displays do this by discarding half of the 1080i HD signal, effectively giving 720p viewers an SD signal - not watching HD at all! "

26 of 399 comments (clear)

  1. Reminds me of Sound Blaster by Anonymous Coward · · Score: 4, Insightful

    This sounds like the visual version of what Creative Labs has been doing for YEARS with their Sound Blaster audio cards. With most other cards if you want to record with a sample rate of 44.1 khz, you record at 44.1 khz, but even with the newer Sound Blaster cards it must be resampled to 48 khz first.

    It doesn't matter if you are sampling up or down, resampling is bad, your best b
    et is to find a device without it, or if it is necessary like in this case, the one that does the best conversions.

    If I bought one of these displays I would be pretty pissed, but I doubt there is much that can be done about it, if you COULD do something than companies like Creative Labs would be out of business.

    1. Re:Reminds me of Sound Blaster by Shkuey · · Score: 5, Informative

      Actually this is an issue of giving people what they want. In this case an HDTV that isn't a thousand bucks more expensive and doesn't have a video processing delay.

      The first incorrect thing in the /. post is that this is somehow standard definition. It's not, 540 lines is more than 480. Not only that but they process 1920 lines of horizontal resolution (scaled down to 1280 for a 720p display), which is quite a bit more than 640.

      Anyone who is serious about getting the absolute most out of their display will have an external scaler and a device to delay the audio. Frankly as digital display technologies take more of a foothold in the market I'm hoping these interlaced resolutions will become far less common.

      When I first read the headlines I thought they would perhaps talk about 1024x768 plasmas with rectangular pixels being marketed as 720p. That kind of thing is far more blasphemous in my opinion.

      So in summary of TFA: 720p is not 720p when it's 1080i.

    2. Re:Reminds me of Sound Blaster by steve6534 · · Score: 4, Informative

      :The NTSC video standard has 525 lines. :The PAL and SECAM video standards have 625 lines. :So where does 480 linrd come from? The specification is for 525 lines but there are only 480 lines of picture informtation - The other 45 are blank lines that were designed to let the electron gun get back to the top of the screen to begin drawing the next frame.

    3. Re:Reminds me of Sound Blaster by J+Isaksson · · Score: 5, Informative

      Due to interlacing, a single 525 line picture is split into two ~262 line frames for display on the TV screen.

      Lines 243-262 of each frame (off the bottom of the TV) start with 0.3V for 4.7us, and the rest is 0V. This tells the TV to prepare for a new frame.

      This leaves just 242*2=484 lines of effective display.

      http://eyetap.org/ece385/lab5.htm

  2. It's there by Anonymous Coward · · Score: 5, Funny

    No surprise there. But many displays do this by discarding half of the 1080i HD signal, effectively giving 720p viewers an SD signal - not watching HD at all!

    The HD signal's still there... you just have to learn how to read between the lines.

  3. For the inevitable /.ing by Anonymous Coward · · Score: 5, Informative

    Here is tfa for you...

    When is 720p not 720p?

    Tom Norton, in his coverage of the Home Entertainment expo, brought something up that I was unaware of.

    720p displays show native 720p signals directly, of course. They also upconvert SD signals (like DVD) up to 720p for display. And 720p displays must convert incoming 1080i signals to 720p before they can be displayed. No surprise there, this makes sense. But, Silicon Optix claims that most manufacturers do the 1080i conversion just by taking one 540 line field from each 1080i frame (which is composed of two 540 line fields) and scaling that one field up to 720p, ignoring the other field. Reason being, it takes a lot less processing power to do this than to convert the image to 1080p and scale that, which would use all the information in the original signal to derive the 720p signal. If you have a display like this, it means that you're watching 540 lines of resolution upconverted to 720p. This is not HD, just like watching a DVD upconverted to 720p is not HD. Sure, you'll get the full width of the 1080i resolution, but you're only getting half the height. While this is better than DVD, it's not HD in my mind. (Aside: Tom Norton mentions this in his review of the Screenplay 777 projector.)

    If this is indeed the case, most people with 720p (or similar) projectors (and most DLP, LCD, and LCOS home theater projectors are exactly that) are not seeing what their displays are capable of. They're not, technically, even watching HD. This is crazy! How can this be? Why haven't we heard of this before? How are manufacturers getting away with it?

    Over-reacting? Well, if you're an owner of a 720p (or any similar resolution) projector you're either gonna be really upset by this or you're just gonna be laisez-faire about it because there's nothing you can do and you're enjoying your projector just fine thank-you. But me, I don't even own any such projector and I'm a little ticked. But I guess I should really wait for evidence of how properly-done conversion looks in comparison before making any snap judgements. I'm sure that the people selling HQV (a processor chip that does it the RIGHT way) will set something up.

    To me, this is a serious issue. Comments are welcome.

    from: http://www.hdblog.net/

    1. Re:For the inevitable /.ing by Overzeetop · · Score: 4, Informative

      That's all well and good, but I'm afraid I tend to agree with them. If content providers want to "do it right" they should ditch the 1950's interlacing and get with the 1980s.

      He's leaving one step out. 1080i is 540 lines scanned 60 times per second, offset by half a vertical pitch. 720p is 720 lines scanned at 30 times persecond.

      To try and take two frames which are not occuring at the same instant, stitch them together, remove the motion artifacts, resample, and then display is just plain silly. And frought with errors, as you are expecting a computer to determine which parts of the motion (over 1/60 of a second) to keep and which to throw away.

      If you wanted high fidelity, you'd spend the money for a 1080p60 system. Then it wouldn't matter. Except that you would complain about the quality, because each frame you see was upsampled from only 540 lines of resolution.

      It all comes back to the fact the the FCC let the industry choose this "18 formats is good" spec.

      Personally, I'm in favor of an olympic standard mayonaise, but...no...wait...awww hell, I give up.

      --
      Is it just my observation, or are there way too many stupid people in the world?
    2. Re:For the inevitable /.ing by Phreakiture · · Score: 5, Informative

      720p is 720 lines scanned at 30 times persecond.

      Mostly incorrect.

      There are 18 recognized MPEG stream formats for HDTV.

      • 640x480 24 fps progressive narrow
      • 640x480 30 fps progressive narrow
      • 640x480 30 fps interlaced narrow
      • 640x480 60 fps progressive narrow
      • 704x480 24 fps progressive narrow
      • 704x480 30 fps progressive narrow
      • 704x480 30 fps interlaced narrow
      • 704x480 60 fps progressive narrow
      • 704x480 24 fps progressive wide
      • 704x480 30 fps progressive wide
      • 704x480 30 fps interlaced wide
      • 704x480 60 fps progressive wide
      • 1280x720 24 fps progressive wide
      • 1280x720 30 fps progressive wide
      • 1280x720 60 fps progressive wide
      • 1920x1080 24 fps progressive wide
      • 1920x1080 30 fps progressive wide
      • 1920x1080 30 fps interlaced wide

      In presenting these on a monitor, your receiver/settop box/whatever is supposed to turn them to a format that your monitor can handle. This will typically be one of these four:

      • 480 row 60Hz interlaced
      • 480 row 60Hz progressive
      • 720 row 60Hz progressive
      • 1080 row 60Hz interlaced
      • It is noteworthy, though, that some videophile monitors can handle, and set-top boxes deliver, 1080 row 60Hz progressive.

        As for the presence/absence of interlacing, I agree that it is very bad to use interlacing at the strem level. This should be eliminated. I would make an exception for the 480 modes, because the material may have been originally captured on NTSC videotape, in which case some sort of conversion would have to take place to get a progressive image, and I feel very strongly that conversions should never be done for broadcast unless absolutely necessary (as when showing PAL/SECAM native material).

        On the other hand, at the monitor level, if you have an interlaced monitor, I don't think that is a major issue. In 1080 mode, the best picture that can be sent is the 30fps progressive stream. This can be interlaced for presentation on a cRT.

        Now, someone commented that CRT's are dead. Not if you have a budget, they're not! I've owned an HD set for over three years now, and it only ran me $700. It is a CRT. It has a beautiful picture.

        Further, I would put forth that CRT's, in addition to being significantly cheaper than the alternatives, also put out a better picture than LCD (view from any angle; accurate color rendition; no lag), are less susceptible to burn-in than plasma (which will be killed by network bugs) and do not exhibit the rainbow effect of DLP (which, in fariness is not really all that bad). Their major failings are their physical size and power consumption.

      --
      www.wavefront-av.com
  4. Resampling by FlyByPC · · Score: 5, Insightful

    There's got to be a fairly straightforward formula relating inherent resolution loss when performing any noninteger upsampling, or any downsampling. Any other change in resolution must necessarily degrade the signal, yes? (Except perhaps if a clever algorithm could losslessly encode the original data in a 1.5x-upsampled version, without distorting it.)

    --
    Paleotechnologist and connoisseur of pretty shiny things.
  5. Which Models? by goldspider · · Score: 5, Interesting

    Is there any way of telling which sets do this? This is certainly something I'd like to know before I dropped that kind of cash.

    --
    "Ask not what your country can do for you." --John F. Kennedy
    1. Re:Which Models? by David+Leppik · · Score: 5, Interesting
      Is there any way of telling which sets do this? This is certainly something I'd like to know before I dropped that kind of cash.
      Yes. Go to a showroom and look at the displays. If you see some that have greater vertical resolution than the non-HD models, there you go. If you can't see a difference, then it doesn't make a difference.

      If there is a difference you can't see but could learn to see, don't learn; it will not bring you joy, it will only make you miserable or annoying. Long ago I learned to see the FFT distortion in JPEG and MPEG images. Has it made me happy? No. I end up making the JPEGs on my website bigger than everyone else's so I won't see wrinkles on people's faces that are apparently invisible to everyone else. And I can't stand to watch satellite television on a big screen TV because of the annoying compression artifacts.

    2. Re:Which Models? by Total_Wimp · · Score: 4, Funny

      Yes. Go to a showroom and look at the displays. If you see some that have greater vertical resolution than the non-HD models, there you go. If you can't see a difference, then it doesn't make a difference.

      This is not a solution.

      At Best Buy and Circuit city I've seen lots of SD signals on HD displays. How on earth am I going to know if it's the set or the signal that's producing all those jaggies? Ask? At Best Buy? I might as well ask them to build a moon rocket while they're at it.

      Knowing the stats won't neccessarily guarantee you a better picture, but it is a better place to start.

      TW

  6. Workaround is to use an HTPC... by Rectal+Prolapse · · Score: 5, Informative

    A Home Theater PC with good quality parts, drivers, and decoders will preserve the 1080i signal - it will combine the 1080i field pair into a single 1080p signal, and then downconvert (ie. downscale) to 720p.

    As a reference, my Athlon XP running at 2.4 GHz (aproximately equivalent to an Athlon XP 3400+) with a Geforce 6800GT and TheaterTek 2.1 software will have (little) trouble achieving this, assuming the 1080i source isn't glitchy itself.

    Alternative is to use the NVIDIA DVD Decoder version 1.0.0.67 ($20 US after 30 day trial) and ZoomPlayer 4.5 beta ($20 beta or nagware) for similar results.

    TheaterTek is roughly $70 US and includes updated NVIDIA DVD Decoders - too bad NVIDIA hasn't updated their official DVD decoders with the bugfixes that is present in the TheaterTek package.

  7. When is 720p Not 720p? by t_allardyce · · Score: 5, Insightful

    When the un-washed masses can't actually tell the difference (they can't even see DCT blocking) and you can get away with selling this crap to them..

    --
    This comment does not represent the views or opinions of the user.
  8. Good Ol' CRT by goldspider · · Score: 5, Insightful

    When I upgrade to an HD idiot box, I plan on sticking with tried-and-true CRT. IMHO, you can't beat the picture quality/price, and I have yet to hear a compelling reason to fork out thousands of dollars for the trendier offerings.

    --
    "Ask not what your country can do for you." --John F. Kennedy
  9. Should have bought a 1080i screen then! by node+3 · · Score: 5, Insightful

    If the broadcast is 1080i, and your display isn't 1080i, I don't think it's logical to assume the quality of the downsampled video will be equivalent to a true 720p broadcast.

    When I get around to buying a HD television (not any time soon, I do all my televisioning on my computer), it will be a true 1080i (are there 1080p televisions?) display so I'll know I'm getting the full potential of HD.

    Unless I'm strapped for cash, of course, in which case I'll just suck it up and know my 720p won't be the best thing for watching 1080i content on.

    On the plus side, it's important to get the facts out there for the consumer, who will likely (although not logically) assume he's/she's getting more than they really are.

  10. Hey, Bloggers... by Gruneun · · Score: 5, Insightful

    Why not submit a link to the original article, rather than a link to your blog, which consists only of a link to the original article?

    Otherwise, people might assume this is a shameless attempt to draw traffic to your site.

  11. spatial vs temporal resolution by Phearless+Phred · · Score: 5, Informative

    Ok, here's the skinny. 1080i is 1920x1080 @ 59.94 fields / second, meaning at any one instant in time, you're looking at a 1920x540 image made up of every other line of the picture (the odd fields, if you will.) Then, ~1/60th of a second later, you see the even fields. 720p is 1280x720 @ 60 FRAMES per second, meaning at any given instant you're looking at EVERY field of the image...not just the odd or even fields. If you were to try and take all 1080 lines from the original signal, they wouldn't really map properly to 720 at any given second because half of data would be from that same ~1/60th of a second later. Scaling the fields up is really the best way to go, at least for stuff that's been shot interlaced.

  12. The clever algorithm is a "Fourier transform" by roystgnr · · Score: 5, Informative

    And when you use it to upsample data, it is a lossless encoding that doesn't degrade the signal (unless you deliberately throw away data - discrete Fourier transforms are also used in lossy encoders).

    It's not a distortion-free transform, since high frequency signals (e.g. sharp edges) in the original image get interpreted as smooth changes and can get blurred between multiple pixels in an upsampled signal. But then again, that's exactly the sort of thing that happens when you digitize a picture in the first place - if you have a sharp black/white edge that passes through the middle of a pixel, the most accurate thing you can do is make that pixel gray.

  13. Re:Resampling: Imagine a 1-pixel-wide line by G4from128k · · Score: 4, Informative

    There's got to be a fairly straightforward formula relating inherent resolution loss when performing any noninteger upsampling, or any downsampling.

    Its a bit messy. Imagine 1080i image with a 1-pixel wide sloping black line that is nearly horizontal on a white background. If you throw out half the data, you create an image with a dashed-line. Gaps in the line occur where the slanting line cut across the rows that were discarded. If you upsample from 540 to 720, you will find that the remaining dashes become fattened non-uniformly. In places where the row in the 720-row image falls directly on top of the 540 row image, the line will be thin and dark. In places where the 720-row image falls midway between rows in the 540 row image, the line will be wide and less dark. The end result is the thin black uniform line is converted to a dashed line of varying thickness and darkness -- not pretty.

    Even if you resample directly from 1080 to 720, you still run into problems where the 720-row image pixels fall between the 1080-row pixels. At best, you can use higher-order interpolation (e.g. cubic) to try and fit a curve through the original data and try to estimate what was in the middle of the pixels so they can be shifted half way over. But the result wil never look like an image that was taken with a 720-row camera in the first place.

    --
    Two wrongs don't make a right, but three lefts do.
  14. Re:This is what you get..... by Zed2K · · Score: 4, Insightful

    Kind of hard to "do your research" when you can't find out how the conversion takes place or can't understand how it all works. This is not a consumers fault kind of thing. This kind of information is not made known unless people ask the question. Who would have thought to ask a question like this?

  15. Not news - Buy a scaler. by GoRK · · Score: 4, Informative

    This is not particularly news. Some "blogger" discovers something because he never bothered to ask and screams something about the sky is falling.. I'm kind of sick of this "news" reporting. Incidentally, this same issue affects owners of most plasma and LCD tv's with native resolutions below 1920x1080 too.. depending on how you look at it as a problem or not.

    Anyway, it's fairly well known that the internal scalers in many devices suck. That is why there is a market for good external scalers. If you are paranoid about watching a lot of 1080i on your 720p projector or LCD TV or Plasma, go buy a scaler. They cost about $1000 but will improve scaled display a lot.

    At least if you have an external scaler you will have some options about how you convert 1080i to 720p. The article makes it sound like splitting the fields is a huge sin -- and it is if you discard one field per frame (Half field deinterlacing), but it's perfectly acceptible to scale EACH 540-line field to a seperate 720-line frame and double the framerate. This is called bob deinterlacing and is often the best for converting 1080i video to lower resolutions. If you are watching a 1080i upconvert of a film or something, though, you can have the scaler do full field deinterlacing and inverse telecine for you and see a nice 720p/24fps picture. Scalers also generally have internal audio delays for various types of audio feeds so you won't have to worry about AV sync issues either.

    If you have any questions about how your device does this, you should try to find out before you buy it. Most devices don't publish how they do it, though, so your only option may be to derive it -- and that will not be an easy job.

  16. Comment removed by account_deleted · · Score: 4, Insightful

    Comment removed based on user account deletion

  17. How to prove/disprove it easily. by Stavr0 · · Score: 4, Interesting
    Simply provide a 1080i test pattern made of alternating horizontal lines where the left half is black lines on even scanlines, white lines on odd scanlines and the right half is reversed (white-even,black-odd)

    The results may be one of the following:

    You will get a screen full of tiny, shimmering horizontal lines that shift in the center of your screen Congratulations! Your HT gear is showing a true 1080i picture You will get a full screen of gray, possibly with a line in the center Not bad, your gear is properly downscaling the signal Half your screen is black, the other is white Uh oh. Your gear is taking the easy way out and dropping half the scanlines to downconvert (Bele and Lokai) I call that the Cheron Test.
  18. Shameless Atari / Amiga Plug by tjstork · · Score: 5, Interesting

    Back in the day, on the Atari and Amiga, you could actually do stuff in that interrupt time. The most common thing was to swap display buffers for double buffering. This made for rock steady hardware scrolling, an effect that still lacks somewhat in today's PC's, believe it or not, as there was absolutely no tearing of the display whatsover. Just a beautiful effect.

    --
    This is my sig.
    1. Re:Shameless Atari / Amiga Plug by tjstork · · Score: 4, Informative

      The difference is that EGA/VGA you had to poll the card, but on the Amiga and even the Atari, you actually got notified by an interrupt. When you got the interrupt, you could be reasonably assured that there was no latency between the interrupt trigger and the event as interrupts ran at such a mega priority. And on those systems too you could also specify the memory window versus the display window and also scroll by adjusting the memory pointers as well as some shift registers. It just looked really, really good.

      --
      This is my sig.