VideoHelp Forum
+ Reply to Thread
Results 1 to 21 of 21
Thread
  1. Recently there was a discussion claiming that just because 720x480 (or 720x486) and 720x576 frame sizes are codified in Rec. 601, they somehow are "native" to analog video being digitized via ADC, and that one should never digitize with other number of samples per line to avoid errors generated by scaling/resampling (the number of lines is not being questioned here).

    It seems to me that the information provided in this excerpt from 2002 Conexant datasheet as well as the excerpt below throws the idea of "native" 720 samples per line out of the window.

    Video Sample Rate Converter - Flexible Video Timing Formats
    In the CX2388x, the Ultralock circuit has been enhanced by a Video Sample Rate Converter. This capability is required for applications such as video editing/authoring and time shifting that require decoded video to be sent out of the CX2388x for compression and simultaneous direct connection to a digital video encoder such as the Bt865A or Bt860/861.

    In combination with Ultralock implemented in the Video Decoder, the Sample Rate Converter allows a choice of video sampling rates and video decoding rates. This flexibility means that video can be captured at the highest possible sampling rate, using Ultralock, to acquire the maximum number of samples at the 10-bit Analog to Digital Converter(s).

    If decoded video is required on the CX23880’s output pixel port for connection of hardware MPEG compressors, TV encoders or hardware video deinterlacers, the Sample Rate Converter can be enabled to provide digital video with the following attributes:
    • A valid video data sample present with each output clock
    • Produces the same number of output clocks per line
    • Generates consistent sample point locations from line-to-line.
    These attributes are mandatory for interfacing to the aforementioned devices. Figure 2-42 provides a top level overview of the Sample Rate Converter and Video Decoder.

    From Figure 2-42 it can be seen that there is a clock domain for the analog to digital converters and another domain for the Multistandard Video Decoder and Scaler. Typically an 8x Fsc (video subcarrier frequency) crystal is used to provide a sampling clock for the video Analog to Digital Converters in order to obtain more samples per line than is required (see Ultralock Overview) for square pixel or 4:3 aspect ratio decoding.

    If no external video processing is required, the Sample Rate Converter can be bypassed and the video decoder clocked at the same frequency as the Analog to Digital Converters. The decoded video samples are DMAd to the PCI bus and the graphics overlay surface in the same fashion as the previous generation Fusion 878A decoder.

    If external video processing is required, then the Sample Rate Conversion register should be programmed for the desired frequency of video decode. This is typically 27MHz for ITU-R. BT656 decoding, although other formats such as square pixel NTSC (24.5454/12.2727 MHz) or PAL (29.5/14.75 MHz) may be supported. See Table 2-31.

    In order to use the Sample Rate Converter the ratio between the ADC sampling frequency (typically 8x Fsc) and the desired Video decoder core frequency (typically 24.54, 27.0 or 29.5MHz) must be derived, multiplied by 2¹⁷ and then converted to binary or hexadecimal.
    Image Attached Thumbnails Click image for larger version

Name:	fig-2-42.png
Views:	49
Size:	96.4 KB
ID:	64624  

    Click image for larger version

Name:	table-2-31.png
Views:	76
Size:	87.5 KB
ID:	64625  

    Quote Quote  
  2. Member Skiller's Avatar
    Join Date
    Oct 2013
    Location
    Germany
    Search PM
    720 is the logical consequence derived off the maximum amount of oscillations a standard def analog video signal is allowed to have around the world. (That alone should be enough not to question it. )

    With the right amount of pre-filtering, sampling at any rate is perfectly possible, but those are special applications. I don't see why you would want to introduce a level of uncertainty by your capture hardware and driver software when you can capture at a guaranteed standard and then do virtually anything you want in software after capturing.
    Quote Quote  
  3. Of course there's no such thing as a native pixel resolution horizontally in an analog signal. There are no pixels. It's a continuous waveform of limited bandwidth. It could be digitized with as many or as few samples as you like. If you use too few you will lose picture resolution. If you use too many you will waste storage space. 13.5 MHz sampling was codified because it was deemed enough to capture all the resolution of the highest quality analog sources, without being excessive. And it was a good compromise for PAL and NTSC, making hardware design simpler.
    Last edited by jagabo; 3rd May 2022 at 12:44.
    Quote Quote  
  4. using anything higher than 720 is equivalent of digitising crappy 45 single at the rate of 96khz
    Quote Quote  
  5. FWIW the attached document gives an impression of the demanding process for developping and eventually agreeing on a standard (Rec.601), which at the end ensures compatibility and interoperability between systems. The agreement on the 13.5 MHz luma sampling rate represented a breakthrough which - together with the scanline duration - is the base for the horizontal "resolution" of digital SD video. And yes, it is not "native" to analog video, it is a standard.
    One could of course take as a base the "native" spectral bandwidth of each source like analog TV signal, VHS, S-VHS, Video8, Hi8, laserdisc etc. and adopt a tailored lowpass filter and sampling rate for each of these (even manufacturer dependent), including color (sub)sampling variants, and end up with a billion of digital formats and frame sizes aka "resolutions". Technically possible yes (within practical and economical constraints), it would however not contribute to reducing the confusion, I think.
    Image Attached Thumbnails History of Rec601 - trev_304-rec601_wood.pdf  

    Quote Quote  
  6. Member DB83's Avatar
    Join Date
    Jul 2007
    Location
    United Kingdom
    Search Comp PM
    Am I mistaken but that data-sheet refers to hardware MPEG compression and very few capture devices did that. And those that did, I guess, restricted their output to standard dimensions.

    Even then there is a vast difference between 'Myth' and 'Reality'. And while that sheet mentions theoretical possibilities it then proceeds to table the standard output.


    Someone, and I know who, will state that his capture device supports 640*480 'natively'. Yet that same device preports to support 352*288 for a NTSC source.
    Quote Quote  
  7. Capturing Memories dellsam34's Avatar
    Join Date
    Jan 2016
    Location
    Member Since 2005, Re-joined in 2016
    Search PM
    There is no myth in a standard, it's a piece of legislation, that's all what it is. What manufacturers do with a standard is another story. The good capture cards including the professional cards and devices, capture the entire 53.33µs scan line in 720 samples which protects the active video area (about 704 samples that floats from left to right, not centered, based on the media content). Some consumer cards and devices such as some DVD recorders crop off 8 pixels on each side, some blank 8 pixels on each side, this is okay if the 704 area is in the center but most of the time is not, some even resize to 640, others capture a small window (about 52µs) at 720 samples which gives a wrong aspect ratio.

    Here is how a NTSC frame looks like on tape with the scan lines already being sampled at 720, This is not the entire width of the frame recorded on tape, horizontal chroma and luma timing signals are excluded, If you want to see the entire frame look at VHS-decode project samples:

    Image Attached Thumbnails Click image for larger version

Name:	720x525.jpg
Views:	928
Size:	49.3 KB
ID:	64635  

    Last edited by dellsam34; 3rd May 2022 at 06:42.
    Quote Quote  
  8. One can approximately simulate, by horizontal down-upscaling, what would happen to this picture with prefiltering and sampling at a lower sampling rate (=fewer samples) than the standard 13.5 MHz: A loss of sharpness and loss of details. The effect would be more eye-catching the better the resolution of the original analog source (tape) is.
    Without prefiltering one would get ugly aliasing.

    Also to mention that cropping borders (or even cropping into the picture) reduces the frame size but does not allow to reduce the sampling frequency accordingly, because the spectral bandwidth of the cropped picture remains basically the same. Reducing the sampling rate would deteriorate the picture quality (sharpness, details, aliasing) of the remaining cropped picture similar as said before.
    Image Attached Thumbnails Click image for larger version

Name:	720x525 resampled.jpg
Views:	44
Size:	42.8 KB
ID:	64637  

    Last edited by Sharc; 3rd May 2022 at 14:16.
    Quote Quote  
  9. Originally Posted by ConsumerDV View Post
    Recently there was a discussion claiming that just because 720x480 (or 720x486) and 720x576 frame sizes are codified in Rec. 601, they somehow are "native" to analog video being digitized via ADC, and that one should never digitize with other number of samples per line to avoid errors generated by scaling/resampling (the number of lines is not being questioned here).

    It seems to me that the information provided in this excerpt from 2002 Conexant datasheet as well as the excerpt below throws the idea of "native" 720 samples per line out of the window.

    They are native as they are worldwide standardized and widely accepted in broadcast - uniformization of the sampling rate i.e. 13.5MHz (as this is most important outcome of ITU-R.BT 601) offering many benefits. You can sample video with any frequency you wish, you may even try to use sub-Nyquist sample rates but this lead you to non-compatible sampling rates and all related signal processing will be mandatory.
    Before ITU-R.BT 601 commonly 4*Fsc (8, 16 * Fsc) was used as this marginally simplify some signal processing - important before LSI/VLSI time but almost impossible to do digital format conversion between different digital video formats.

    So i need to ask - what is your point? IMHO it is at least 10..15 years too late for such discoveries...
    Quote Quote  
  10. Member DB83's Avatar
    Join Date
    Jul 2007
    Location
    United Kingdom
    Search Comp PM
    ^^ Don't be daft.

    There is no point. If there was the OP would have replied many moons ago.


    Dare I add. This is a classic trollish post. Make a statement and watch every respondent add their 2 cents without a single input from the initial poster.


    But do prove me wrong and contribute to the discussion and prove that everyone's opinion is buried in the sand.


    BTW That data-sheet was dated 2002. My calculator tells me that was 20 years ago. Anyone new who wants to digitise analog sources to digital does not give a flying f**k of what might have been achievable back then but rely on accepted standards. And if it was then why did not the subsequent hardware manufacturers provide for it ?
    Quote Quote  
  11. Member Cornucopia's Avatar
    Join Date
    Oct 2001
    Location
    Deep in the Heart of Texas
    Search PM
    Threads with "myth" in the title are almost always shit-stirring troll posts.


    Scott
    Quote Quote  
  12. Going to feed the troll only because my 525 line clip showed up again!

    Fun fact, it was captured at 8x oversampling (108 Mhz)... which I am a fan of cropping to 704 active width then scaling to 640 for 1:1 PAR and bonus noise averaging in the final product.
    Quote Quote  
  13. @pandy, what is the point? Well, as I said, 720 samples per line are not "native" for the case of digitizing low-res analog video like VHS or Hi8. I don't know why I have latched on this word, I think I've seen it used, but maybe it is just a figment of my imagination. Rec. 601 was created to allow digitization of broadcast analog formats 625/50 and 525/60 and facilitiate the exchange between the two. VHS and Hi8 are not broadcast formats, so Rec. 601 has as much to do with these domestic video formats as ATSC or DVB-T with Netflix streaming formats.

    Another point was to make a counterpoint to the claim that most or all modern ADCs sample at one rate, and any resampling is done later programmatically either in the firmware of the ADC or by the capture program. This is a minor point simply to make a point.

    @Sharc, I see what you are saying. But I do not suggest sampling into 640 pixels right away, neither I suggest converting 720 pixels into 640 pixels. Please, see attached picture that I cobbled up from two sources and allowed myself to add to it.

    A certain ancient AD chip is shown using 4x Fsc to sample 3 successive lines of video and decode into square-pixel "NTSC" output. Because of time base errors, the lines have different duration and hence different length and different number of pixels (should be 909). So, the chip either removes extra pixels or adds interpolated pixels to obtain the same number of pixels per line, then resamples each line into 780 pixels.

    Then, if I understand it correctly, 780 pixels are resampled again into 720 pixels and sent to the capturing program, where the user is happy to capture them as, um, "native", non-resampled, crispity crisp, just what is recorded on tape... not.

    So, there are two resamplings before the "native" 720 pixels arrive to the capturing software. In this case, why being so hung up on this magic number? I am just saying that I don't see a huge difference between converting 762 pixels (out of 780) into 704 or into 640 pixels, this is what a good combination of ADC/software does - it strips the overscan and converts 4:3 frame into square pixels. It would be a third conversion either way. Will the effect on sharpness and detail be so bad?

    The above was just an abstract talk, and next is a practical rant, stemmed from watching too many youtube videos with incorrect proportions.

    Even if there will be a noticeable drop in quality, I consider it the price for correct FAR. So many people capture into 720 pixels only to forget about non-square pixels, or not knowing about them beforehand. Or they make an error when setting up PAR. To me, the choice between slight loss of resolution and, um, butchered AR is an obvious one.

    Maybe do not insist on capturing into 720. If people don't want to bother themselves with FAR and PAR, let them capture with square pixels and run through a simple "bob" deinterlacer - this is what every cheap TV set does -, at least it will be watchable. And when they upload their video on YouTube, it will not look stretched, like so many YouTube uploads do (and not deinterlaced as well).

    I guess someone will retort that the forum offers the best practices, so after capturing into 720 the next best practice is to deal with overscan and PAR... Ok then, I suppose this whole rant should be placed on YouTube, not on a video-geek forum. I guess I digged myself into a hole from where I cannot argue.

    P.S. @DB83, I am sorry, I did not notice your message. Did you ask something or suggest something? It is hard to figure it out, getting through personal remarks and social commentary. May I suggest being more concise with your wording next time. As a non-native English speaker, I may miss intricacies of the Royal language. 'Thank you'.

    P.P.S. @Cornucopia, hello to you too!
    Image Attached Thumbnails Click image for larger version

Name:	ultralock.png
Views:	63
Size:	52.6 KB
ID:	64646  

    Last edited by ConsumerDV; 4th May 2022 at 02:17.
    Quote Quote  
  14. Capturing Memories dellsam34's Avatar
    Join Date
    Jan 2016
    Location
    Member Since 2005, Re-joined in 2016
    Search PM
    Originally Posted by ConsumerDV View Post
    Anyway, Rec. 601 was created to allow digitization of broadcast analog formats 625/50 and 525/60 and facilitiate the exchange between the two. VHS and Hi8 are not broadcast formats, so Rec. 601 has as much to do with these domestic video formats as ATSC or DVB-T with Netflix streaming formats
    Wrong, All capture cards designed around rec.601 don't care what video format is, rec.601 has little to do with what tape format is and everything to do with its standard (525 or 625).
    Last edited by dellsam34; 4th May 2022 at 02:11. Reason: typo
    Quote Quote  
  15. Member DB83's Avatar
    Join Date
    Jul 2007
    Location
    United Kingdom
    Search Comp PM
    "VHS and Hi8 are not broadcast formats"

    Agreed. But they were created as a means of recording, in the case of VHS, the analog broadcast and then displaying it on a standard television. Just imagine the reaction if JVC had insisted on a bespoke display the user already having made a not inconsiderate financial outlay in buying an early vcr. Equally, analog video cameras would also require a means of viewing on a standard television set.


    Capture devices generally followed the ITU standard for the digitizing of the analog source. Here follows a link and one should read post #10 which gives a potted history of both broadcast analog and consumer analog digital conversion.


    http://www.digitalfaq.com/forum/video-capture/12307-vhs-blackmagic-intensity.html


    But maybe someone would also like to dispute this.


    BTW This is not a "video-geek" forum. This is a "Video-help" forum. No help is offered when you try to be 'clever'. That approach really is best left to soap-boxes such as Youtube.
    Quote Quote  
  16. Rec. 601 was created to allow digitization of broadcast analog formats 625/50 and 525/60 and facilitiate the exchange between the two. VHS and Hi8 are not broadcast formats, so Rec. 601 has as much to do with these domestic video formats as ATSC or DVB-T with Netflix streaming formats
    Well, Rec.601 is the accepted basis for capturing analog SD video. And the referenced Conexant Datasheet has the title "PCI Audio/Video Broadcast Decoder".

    @Sharc, I see what you are saying. But I do not suggest sampling into 640 pixels right away, neither I suggest converting 720 pixels into 640 pixels. Please, see attached picture that I cobbled up from two sources and allowed myself to add to it.
    Yes, I understand that the "UltraLock" of the Conexant chip is a means to stabilize the picture horizontally. A more or less successful attempt of a "TBC" substitute. It has however nothing to do with questioning Rec.601 or questioning the 720 horizontal pixels.
    Maybe do not insist on capturing into 720. If people don't want to bother themselves with FAR and PAR, let them capture with square pixels and run through a simple "bob" deinterlacer
    If this is the conclusion - why not, if one needs square pixels (not DVD compliant BTW), and provided that the capture HW and SW support capturing in square pixels by adjusting the sampling frequency and pre-filtering properly (most don't), or eventually apply some manufacturer dependent post rescaling or whatever..... I remember the widely used Brooktree BT848/878 chips with its proprietary 17.735 MHz PAL sampling rate and all the scaling and driver issues and discussions.
    So either one agrees on a fixed sampling frequency (aka "720", Rec.601) and ends up with a catalogue of Pixel Aspect Ratios, or one captures to square pixels and ends up with a catalogue of sampling frequencies and prefilters. Choose your poison.
    Last edited by Sharc; 5th May 2022 at 02:35. Reason: Last sentence added
    Quote Quote  
  17. Capturing Memories dellsam34's Avatar
    Join Date
    Jan 2016
    Location
    Member Since 2005, Re-joined in 2016
    Search PM
    Originally Posted by Sharc View Post
    I remember the widely used Brooktree BT848/878 chips with its proprietary 17.735 MHz PAL sampling rate and all the scaling and driver issues and discussions.
    http://www.arachnotron.nl/videocap/doc/Karl_cap_v1_en.pdf Page 10.
    Quote Quote  
  18. Originally Posted by ConsumerDV View Post
    @pandy, what is the point? Well, as I said, 720 samples per line are not "native" for the case of digitizing low-res analog video like VHS or Hi8. I don't know why I have latched on this word, I think I've seen it used, but maybe it is just a figment of my imagination. Rec. 601 was created to allow digitization of broadcast analog formats 625/50 and 525/60 and facilitiate the exchange between the two. VHS and Hi8 are not broadcast formats, so Rec. 601 has as much to do with these domestic video formats as ATSC or DVB-T with Netflix streaming formats.

    Another point was to make a counterpoint to the claim that most or all modern ADCs sample at one rate, and any resampling is done later programmatically either in the firmware of the ADC or by the capture program. This is a minor point simply to make a point.
    Everything depend - "720" i.e. in context of the 601 where sampling rate 13.5MHz for luma is used is standard for all ITU-R.BT601 compliant devices i.e. for sources (your ADC) and players (i.e. DAC) - in this chain (almost 100% of the broadcast/consumer) "720" is native, standardized format.
    There is no standard resolution for "low res" analog formats albeit resolutions such as 352,480,576 are used quite frequently to address reduced bandwidth however all of them using "720" as primary standard.

    All modern ADC (i.e. all post 2000) are designed with 601 primary compatibility on mind, some of them allow to use non 601 sampling rates but rather limited programability in terms of delivered resolution (usually all SD capture IC'c are incapable to deliver more than 2047 pixels for whole 64us video line i.e. incapable to use sampling rates faster than 31.5MHz usually divided by 2 due using same bus for delivering luma and chroma samples in 4:2:2 format).

    Using 4Fsc (8Fsc) may provide marginally better quality (due possibility to do synchronous/coherent sampling) but then resizing is unavoidable if your goal is 601 standard ecosystem.

    So to conclude my point - "720" is native for whole chain - from capture to playout - it can be used to capture VHS, Hi8 or any other standard TV signal. With some limitations it may be used also to capture non standard video (such as generated by some digital sources like computers, game consoles etc) albeit usually with loss of quality.
    Quote Quote  
  19. Originally Posted by pandy View Post
    All modern ADC (i.e. all post 2000) are designed with 601 primary compatibility on mind, some of them allow to use non 601 sampling rates but rather limited programability in terms of delivered resolution (usually all SD capture IC'c are incapable to deliver more than 2047 pixels for whole 64us video line i.e. incapable to use sampling rates faster than 31.5MHz usually divided by 2 due using same bus for delivering luma and chroma samples in 4:2:2 format).
    Which AFEs can do the 2048 sampling? Renesas? TI? Pretty sure the SDP in the ADI chips are stuck at, basically, 858x525 and 864x625...
    Quote Quote  
  20. Originally Posted by Sharc View Post
    I understand that the "UltraLock" of the Conexant chip is a means to stabilize the picture horizontally. A more or less successful attempt of a "TBC" substitute. It has however nothing to do with questioning Rec.601 or questioning the 720 horizontal pixels.
    I think it has in that two re-samplings take place before arriving to 720 horizontal pixels, as such I see no big harm arriving to 640 (or 768) instead of 704 (cropped out of 720) for the sake of square pixels and correct DAR. But as @DB83 pointed above, such a soap-boxy topic better suits YouTube.

    Originally Posted by Sharc View Post
    So either one agrees on a fixed sampling frequency (aka "720", Rec.601) and ends up with a catalogue of Pixel Aspect Ratios, or one captures to square pixels and ends up with a catalogue of sampling frequencies and prefilters. Choose your poison.
    Often one agrees for 720 only to forget to correct for pixel aspect ratio, if they knew about this peculiarity at all. But I am getting on my soap box again... I probably should have titled this thread "I hate digitized VHS videos on YouTube with DAR 1.5" or something like this.
    Quote Quote  
  21. Originally Posted by energizerfellow View Post
    Which AFEs can do the 2048 sampling? Renesas? TI? Pretty sure the SDP in the ADI chips are stuck at, basically, 858x525 and 864x625...
    Some AFE's are 576p compliant, some offer even HD support (SD+ED+HD) but still - my comment was on non standard video sampling rates i.e. some HW offer desired level of flexibility to support such non-601 modes but still this is limited use...
    For those pursuing non-601 i think best approach is to use some universal ADC with more than 8 bit hooked to some USB3 bridge and decode video in software. With massive oversampling even goldeneye videophiles shall be happy.
    Quote Quote  



Similar Threads

Visit our sponsor! Try DVDFab and backup Blu-rays!