1. Join Now

    AVForums.com uses cookies. By continuing to use this site, you are agreeing to our use of cookies. Learn More.

Question about oversampling

Discussion in 'Projectors, Screens & Video Processors' started by bonzobanana, Apr 17, 2004.

  1. bonzobanana

    bonzobanana
    Member

    Joined:
    Aug 21, 2001
    Messages:
    3,404
    Products Owned:
    0
    Products Wanted:
    0
    Trophy Points:
    86
    Location:
    Yeovil
    Ratings:
    +216
    Does a dvd player lets say with a 54mhz 10bit DAC do 4x oversampling with an interlace image and 2x oversampling with progressive scan?

    My understanding is;

    no ovesampling dvd interlace image = 13.5mhz
    2x oversampling dvd iintelace image=27mhz
    no oversampling dvd progressive image =27mhz

    etc

    Do players always maximise use of the video dac.

    So would a 216mhz 14bit dac be doing 8x oversampling of a progressive image to create even more fine gradiance of colour or would it only be used for resolution scaled up images?

    What I'm basically asking is if a 54mhz 10bit player can create 1024 shades of red due to 4x oversampling with interlace does that effectively mean its progressive image can only be 512 shades of red (of course the other field being shown may have some additional colours)?
     
  2. cybersoga

    cybersoga
    Active Member

    Joined:
    Apr 7, 2002
    Messages:
    1,628
    Products Owned:
    0
    Products Wanted:
    0
    Trophy Points:
    48
    Location:
    UK
    Ratings:
    +25
    Don't forget you don't really get 512 shades of red. DVD is an 8 bit format so that means 255 shades of grey, but the chroma is in lower resolution than the luma so not only is it 255 shades but each chroma sample needs to be spread over more than one luma pixel. Oversampling is of course creating in between shades, to try and esimate the lost dynamic range, it's most effective at smoothing transitions (like the blue in the sky, finding nemo etc). Your numbers for oversampling seem about right to me.
     
  3. bonzobanana

    bonzobanana
    Member

    Joined:
    Aug 21, 2001
    Messages:
    3,404
    Products Owned:
    0
    Products Wanted:
    0
    Trophy Points:
    86
    Location:
    Yeovil
    Ratings:
    +216
    If dvds are stored in component form, what actual form is that because green is also the sync and the luminance. I assume the colour resolution is the same for green though with 256 levels of brightness of which I think about 220 are actually used on screen. So effective colour resolution is 220x220x220 I suppose minus repeated colours. When I refered to 512 shades that was in the context of 2x oversampling but then I'm not 100% solid on this so. I saw 1x oversampling somewhere and don't understand what that is. There is no chroma with dvds is there? Thats a combined colour signal and dvds are stored in component form.

    There seems to be little information on this about.

    I was also curious if more oversampling caused a loss of colour resolution. I.e. a softening of colour.

    If you have a 27mhz 10bit DAC on a progressive scan dvd player is it a big loss that the player can't do any oversampling in progressive mode and you are effectively viewing the original dvds as 220x220x220 colour resolution instead of 880x880x880 of a 54mhz 10 bit dvd player? Thats if a 54mhz dvd player maximises use of its DAC circuitry.

    Alternatively do these SOC chips like the mediatek 1379 process in 10bit colour at interlace level and then de-interlace in a seperate embedded de-interlacer? This would effectively mean the dvd chipset is as good as a chip that creates a progressive image directly at 54mhz 10bit.

    I know some people like s-video because it gives you a full resolution luminance image (in fact its the only connection type where the luminance travels independantly of other signals) but the chroma/colour signal is more processed and you often end up with a more linear effect due to this processing. I.e. gradience can look better on s-video than RGB purely because the chroma signal causes a softening of colour definition. Some people think this looks more natural colour wise.

    Do plasmas and LCDs suffer most from obvious colour gradience problems. I mean the colour resolution of either is quite low compared to CRT which is a linear/analogue device. The reason I mention it is because I saw a plasma screen in Dixons. LG I think it was and it had colour stepping on a mainly red image.

    Is it really plasmas and LCDs that beneift from 12bit and 14bit oversampling?
     
  4. cybersoga

    cybersoga
    Active Member

    Joined:
    Apr 7, 2002
    Messages:
    1,628
    Products Owned:
    0
    Products Wanted:
    0
    Trophy Points:
    48
    Location:
    UK
    Ratings:
    +25
    As CRT's have more dynamic range they should benifit more from oversampling, but Plasma's can benifit too, because they are working within tighter limits (and higher reolutions) it's important that the source has as many shades as possible so the plasma has more to work with. It's not just about oversampling but how accurate the DAC is, it doesn't matter how much you oversample if the DAC isn't very accurate. Your right, there isn't a lot of information about this on the 'net.
     
  5. cybersoga

    cybersoga
    Active Member

    Joined:
    Apr 7, 2002
    Messages:
    1,628
    Products Owned:
    0
    Products Wanted:
    0
    Trophy Points:
    48
    Location:
    UK
    Ratings:
    +25
    S-Video doesn't oversample the chroma, it puts the chroma down one wire leading to loss of chroma resolution and interferrence between colours, it's really not a desirable way of doing it. S-Video isn't the only connection type that the luminence travels seperately YPbPr component does this.
     
  6. bonzobanana

    bonzobanana
    Member

    Joined:
    Aug 21, 2001
    Messages:
    3,404
    Products Owned:
    0
    Products Wanted:
    0
    Trophy Points:
    86
    Location:
    Yeovil
    Ratings:
    +216
    I'm pretty sure with RGB the luminance is combined with the colour on each pin so Red+luminance, Blue+luminance and Green+Luminance. Where as component is red, blue, green+luminance+sync. So only s-video has a seperate luminance signal although it also has sync but I don't believe sync has a negative effect on image or interferes with the luminance signal.
     
  7. cybersoga

    cybersoga
    Active Member

    Joined:
    Apr 7, 2002
    Messages:
    1,628
    Products Owned:
    0
    Products Wanted:
    0
    Trophy Points:
    48
    Location:
    UK
    Ratings:
    +25
  8. cybersoga

    cybersoga
    Active Member

    Joined:
    Apr 7, 2002
    Messages:
    1,628
    Products Owned:
    0
    Products Wanted:
    0
    Trophy Points:
    48
    Location:
    UK
    Ratings:
    +25
    YPbPr "Component" is not "red, blue, green+luminance+sync" The 'Y' signal carries the luminence (black and white information) + sync, The 'Pb' and 'Pr' signals carry the color difference signals. With Y/C, C is PbPr combined with color burst, but Y is the same as Y in YPbPr.
     
  9. bonzobanana

    bonzobanana
    Member

    Joined:
    Aug 21, 2001
    Messages:
    3,404
    Products Owned:
    0
    Products Wanted:
    0
    Trophy Points:
    86
    Location:
    Yeovil
    Ratings:
    +216
    Why then is component described as RGB? Also why does component socketry have 3 colours (RGB) for the phono connections? None of the information in the link suggests otherwise? The colour difference is surely the actual colours. I looked at a diagram for a circuit that converted component to RGB scart and it basically split the luminance of the green connection and added it to the red and blue channels plus seperated the sync off. There was possibly inversion and amplification of certain signals going on but the basic premise is there. If you stick a green component connection into a composite socket you get a green fully synchronised image which shows it to be green+luminance+sync surely?
     
  10. cybersoga

    cybersoga
    Active Member

    Joined:
    Apr 7, 2002
    Messages:
    1,628
    Products Owned:
    0
    Products Wanted:
    0
    Trophy Points:
    48
    Location:
    UK
    Ratings:
    +25
    YPbPr component, the Y, or the green plug, does not carry any colour, only black and white information. Colour is derived from the Pb (blue plug) Pr (red plug) using a mathematical equasion.

    Component is technically any form of video that splits the video into it's component parts, be that (RGB, RGsB, RGBS) or (YUV, YCbCr, YPbPr ). The "component" connection on the back of most kit carries YPbPr, but technically RGB+(Composite video (sync)) via scart is component video too. The difference between RGB and YPbPr is colour space. YPbPr was designed to save bandwidth with no reduction of picture quality compared with RGB, and you can convert one to the other by using mathematical equasion. The reduction of picture quality comes when YCbCr 4:4:4 (full colour resolution) is converted to YCbCr 4:2:0 as stored on a DVD, or rather the other way round in a DVD player, and different dvd players upsample the chroma in different ways, some look better than others. Add to that mpeg artifacts and there's plenty to go wrong!
     
  11. Zacabeb

    Zacabeb
    Active Member

    Joined:
    Jan 17, 2002
    Messages:
    474
    Products Owned:
    0
    Products Wanted:
    0
    Trophy Points:
    21
    Ratings:
    +9
    To get back to the issue of oversampling and more bits:

    More bits than 8 are needed because the 8 bits per component stored on DVD are reserved for the picture content alone (Y can have values in the range 16...235, Cb and Cr in the range ±112). Above and below there needs to be room for sync and other stuff such as Macrovision. So, you need at least 9 bits to retain the full range of values in the outgoing signal while keeping room above and below for the other signals.

    If I understand correctly, oversampling is needed because the borders between pixels are theoretically of infinitely high frequency, and the signal must be filtered so that the transitions are under control. Oversampling solves this problem mathematically before the signal is converted to analog, rather than having a more aggressive analog filter after the DAC. I guess it also permits smoothing pixels where they do not represent contours. More bits also allow for the added samples to take on levels intermediate to the original 8 bits.

    So if the above is true, the higher the oversampling, the more of the sharpness existing in the image can be retained, and the more bits, the more precisely the oversampling can shape the new slope between the original pixels and keep it steeper or smoother depending on what's needed.

    And of course, by definition in progressive scan the DAC needs to output twice as many samples given the same relative oversampling rate.

    But that is about as much I know this far, and still I feel that it's not enough...

    As for S-video, what happens in there are three things:

    1.) During encoding the color components are lowpass filtered, because they need to remain below half the frequency of the color subcarrier in order not to interfer with each other when carried there. This lowpass filtering virtually eliminates both mosquito and block noise in the color components.

    2.) During decoding the remaining traces of the subcarrier and any crosstalk between the components needs to be filtered out, so additional bandwidth is lost during decoding.

    3.) In the case of PAL signals, the color components on each line are blended with those from the preceding line in the field, to cancel out phase errors that would lead to tints of opposing color in each line (Hanover bars). This causes a loss of vertical sharpness, in addition to that caused by the 4:2:0 color subsampling on DVD Video. However, the blur also helps hide the subsampling artifacts (the inevitable color combing in interlaced images, and its ugly cousin The Chroma Bug).

    Add to that the fact that many displays use Color Transient Improvement to shrink color transitions and sharpen them, and to many people S-video may look better. Maybe two wrongs can make a right... at least with poorly encoded source material.
     
  12. Chris Muriel

    Chris Muriel
    Well-known Member

    Joined:
    Jun 14, 2002
    Messages:
    7,094
    Products Owned:
    0
    Products Wanted:
    0
    Trophy Points:
    136
    Location:
    Manchester
    Ratings:
    +711
    We used to have a useful applications note on my company's website that included explanations of the various formats as well as most of the formulae (AN-548).
    (I work for Analog Devices who wrote the original apps note).

    I just had to Google for it to find someone with it on his personal website :

    http://nick.bradley.org.uk/an548.pdf

    Chris Muriel, Manchester.
     
  13. cybersoga

    cybersoga
    Active Member

    Joined:
    Apr 7, 2002
    Messages:
    1,628
    Products Owned:
    0
    Products Wanted:
    0
    Trophy Points:
    48
    Location:
    UK
    Ratings:
    +25
  14. Timmy B

    Timmy B
    Guest

    Products Owned:
    0
    Products Wanted:
    0
    Ratings:
    +0
    I always remember the (now dated) secrets test:

     

Share This Page

Loading...