Test Pattern Generator software techniques and Display Adapter S-video Out

[cybertheque]

As part of my ongoing quest to do low-cost test and measurement, I was looking at the advertising
for 'Calibug' and the similar software-only 'NTSC Test Pattern CD-ROM' which both claim to
output on supported display adapters' S-video output port. Unfortunately, the documentation on
the Web is minimal and I didn't find a list of supported cards. I installed a Geforce2-based
display adapter with S-video out in a Wintel box and configured the display driver to do independent
dual displays, and I confirmed that the S-video output works by changing the 'background' image
to an arbitrary jpeg file for the NTSC monitor, however, when trying the 'output to hardware'
option in 'vscope', I got nothing. Perhaps 'vscope' doesn't grok the API and is married to some
other hardware.

Has anyone used the S-video output on Nvidia cards (or other makers') as the display device for
rendered output in applications like NLE? My Targa 2k card gives me that facility as do other
editing cards (like the DC10+ and higher from Pinnacle), but I have not found mention of
doing this with ordinary display adapters, even though they are capabable of independent
display management.

I would like to find software that uses the S-video out to deliver test signals and patterns, independent
of the desktop display settings, but it seems that the 'Calibug' and similar products probably just run a
standard full screen output to the default (and cloned) display device(s).

Michael

[edDV]

Has anyone used the S-video output on Nvidia cards (or other makers') as the display device for
rendered output in applications like NLE? My Targa 2k card gives me that facility as do other
editing cards (like the DC10+ and higher from Pinnacle), but I have not found mention of
doing this with ordinary display adapters, even though they are capabable of independent
display management.

I would like to find software that uses the S-video out to deliver test signals and patterns, independent
of the desktop display settings, but it seems that the 'Calibug' and similar products probably just run a
standard full screen output to the default (and cloned) display device(s).

Michael

Yes and I was appalled at the S-Video quality from typical NVidia cards. I haven't tested recent "PureVideo" cards. That is why I suggested you rent a test signal generator if you want a credible test signal source.

For system calibration I place test signals on my Vegas or Premiere timeline and then export as DV over IEEE-1394 to my Canopus ADVC-100 and output composite NTSC/PAL or S-Video from the ADVC. The ADVC also outputs an internal analog color bar with either zero or 7.5 setup that is useful for capture device calibration. The ADVC quality is very good compared to the NVidia or ATI cards.

For levels setting, the SMPTE or Belle-Nuit color bars are sufficient. Multiburst and zone plates are more problematic due to 8bit, 4:1:1 DV coding. That is where you need a quality test signal generator.

[cybertheque]

Has anyone used the S-video output on Nvidia cards (or other makers') as the display device for
rendered output in applications like NLE?
Michael

Yes and I was appalled at the S-Video quality from typical NVidia cards. I haven't tested recent "PureVideo" cards. That is why I suggested you rent a test signal generator if you want a credible test signal source.

In my case, the characteristics of the S-video output would depend on the specs of the bt869 chip,
the board layout, bypassing, digital noise pickup, etc., and I hope to qualify these. I've rented instruments
in the past, but could possibly borrow a TSG, vectorscope, etc. from our local public access cable operation.
I need however to have in-house capability on a tiny budget, ergo my exploration of the cheap solutions.

As for using a second Windows display, NTSC, for NLE rendering and preview, what is likely and how is it
done with these cards?

Michael

[edDV]

If you read the VideoQ documents, they explain the test and measurement problems caused by output through computer display cards. That is why they developed an internal PC measurement system to isolate the capture device. IEEE-1394 export is another strategy.

This largely depends on the types of measurements you intend to do and the amount of precision you seek.

[jagabo]

Matrox cards have pretty good s-video output, better than the ATI an NVIDIA cards I've seen. Even the old Millennium G400 era cards. Their "DVD Max" setting (like Theater mode for ATI cards) allow video playing in a window on the desktop to appear full screen on the TV output. They also have separate proc amp controls for the TV output.

[cybertheque]

This largely depends on the types of measurements you intend to do and the amount of precision you seek.

I have two goals:

1. To characterize the performance of various video capture cards;

2. To have basic in-house setup and calibration capability all along the analog chain.

Types of measurements are those typical of a studio environment, for example as described in the Tek
document "NTSC Systems Television Measurements", standards and practices from RS-170A and
EIA-250-C amongst others. Precision should be adequate to ascertain deviation or compliance with
standards.

In evaluating capture cards, I had hoped to consult vendor specifications but sadly have found
essentially nothing; the detailed BT848 datasheet and app notes are silent on specifications
for S/N, distortions, crosstalk, etc. Even detailed reference documents for high-end capture cards
lack this data. It strikes me as odd that there isn't an Internet resource revealing measurement
data for at least the studio-grade products.

There ought to be low-cost but capable test and measurement solutions that leverage the PC platform
for video work -- the DPS VM-2000 and VM-3000 were a start (live scope on a card) and I like
my TBC-II as a TBC/proc amp on the cheap. There is evidence on the Web for some homebrew
PC-based TSG and scope project designs intended to meet performance standards, but URL rot and
Asian-site incompatibilities have made them unobtainium.

For now, I will try to borrow instruments to characterize my current capture cards, but I really want to
have a basic TSG and waveform display capability without additional chassis.

[cybertheque]

Matrox cards have pretty good s-video output, better than the ATI an NVIDIA cards I've seen. Even the old Millennium G400 era cards. Their "DVD Max" setting (like Theater mode for ATI cards) allow video playing in a window on the desktop to appear full screen on the TV output. They also have separate proc amp controls for the TV output.

I will try some Millenium cards that I have on-hand; in what ways did you find the V-out better than on other cards
that you tried? The Nvidia card driver/control panel for my current experiments does have proc amp controls for the S-video out.

Does the Matrox full-screen V-out with desktop window overlay depend on a proprietary player? Have you used this
feature in other apps like an NLE?

[jagabo]

I will try some Millenium cards that I have on-hand; in what ways did you find the V-out better than on other cards that you tried?

It's been quite a while since I've used s-video output but as I recall, it was much easier to get pixel-for-pixel output (ie, avoid digital scaling of the image before being sent to the DAC), handling of interlaced video worked properly (no deinterlacing, each field of the source comes out as one field on the s-video port), and the aspect ratio was correct. I could sometimes get an ATI or NVIDIA card to come close but after rebooting, or after a while, they would revert to scaling the source.

Does the Matrox full-screen V-out with desktop window overlay depend on a proprietary player? Have you used this feature in other apps like an NLE?

It worked with any player that used the graphics card's video overlay feature. Just about all media players. I don't know about NLEs. If they have preview functions that use video overlay they should work.

There were some limitations though. I think only top field first interlaced video came out in the right order on the older Millennium cards. With the newer Parhelia based cards you can change the field order but swithing is a pain because it has to be done via the graphics card's setup applet. There were also issues about what frame size you need to get perfect output. As I recall the older Millennium cards worked best with 640x480 sources. The newer cards work with 720x480 or 704x480.

[cybertheque]

In my case, the characteristics of the S-video output would depend on the specs of the bt869 chip,
the board layout, bypassing, digital noise pickup, etc., and I hope to qualify these.

There may be some hope for bt868/869 broadcast-quality video out; the datasheet, although
describing the chip's targets as consumer devices, describes capabilities well within the domain
of the full NTSC spec including programmability of the scaling filter chain (and optional bypassing
of it). Regarding subcarrier phase stability, excerpts below hint at high performance goals:

"The color subcarrier is derived directly from the CLKO (derived from
XTALIN/XTALOUT) CLKI when EN_XCLK=1 input, hence any jitter or
frequency deviation of CLKO (XTALIN/XTALOUT) or CLKI when
EN_XCLK=1 will be transferred directly to the color subcarrier. Jitter within the
valid CLKO cycle interval will result in hue noise on the color subcarrier on the
order of 0.9–1.6 degrees per nanosecond....

...Crystal-based clock sources with a
maximum total deviation of 50 ppm (NTSC) or 25 ppm (PAL) across the
temperature range of 0°C to 70°C produce the best results for consumer and
industrial applications. In rare cases, temperature-compensated clock sources
with tighter tolerances may be warranted for broadcast or more stringent PAL
(e.g., type I) applications."

Fortunately, this is a well-documented part and not proprietary; there appears to be
an open-source effort of some magnitude for *nix drivers -- it may be worth
putting a TCXO and some signal conditioning on a card with this chip (e.g. an Nvidia
card) as the basis for a decent TSG.

[jagabo]

You're also at the mercy of the analog section of the output: low pass filtering, bad isolation (noise from the digital section leaking through), etc. This is where design is important and a lot of cards fall short. I believe both ATI and NVIDIA have done a lot in the last few years to help the manufacturer do it better. Five to ten years ago TV out was often total crap.

[cybertheque]

I installed a Geforce2-based display adapter with S-video out in a Wintel box
and configured the display driver to do independent dual displays, and I confirmed that the S-video
output works by changing the 'background' image to an arbitrary jpeg file for the NTSC monitor,
however, when trying the 'output to hardware' option in 'vscope', I got nothing. Perhaps 'vscope'
doesn't grok the API and is married to some other hardware

FWIW:

Digging a little more, using the author's name 'Gary Bonham', I discovered a rec.video.desktop post
announcing the product which stated it used the Azeenavision 500 for video output.

I also discovered a Mac OS8, 9 and X based vectorscope/waveform monitor, called 'VideoScope',
by a different author, which seems to be still supported.

[edDV]

This largely depends on the types of measurements you intend to do and the amount of precision you seek.

I have two goals:

1. To characterize the performance of various video capture cards;

2. To have basic in-house setup and calibration capability all along the analog chain.

Types of measurements are those typical of a studio environment, for example as described in the Tek
document "NTSC Systems Television Measurements", standards and practices from RS-170A and
EIA-250-C amongst others. Precision should be adequate to ascertain deviation or compliance with
standards.

In evaluating capture cards, I had hoped to consult vendor specifications but sadly have found
essentially nothing; the detailed BT848 datasheet and app notes are silent on specifications
for S/N, distortions, crosstalk, etc. Even detailed reference documents for high-end capture cards
lack this data. It strikes me as odd that there isn't an Internet resource revealing measurement
data for at least the studio-grade products..

All these analog measurements would apply more to the finished product on PCB rather than the chips alone. You might find specs for a reference implementation in the application notes.

"Studio Grade" isn't the goal for the BT848. Instead it is intended for application in consumer/PC grade products such as console games or set top boxes. Serious studio gear would be at least 10bit I/O with SDI (SMPTE 259M or SMPTE 292M) interface.

[cybertheque]

"Studio Grade" isn't the goal for the BT848. Instead it is intended for application in consumer/PC grade products such as console games or set top boxes. Serious studio gear would be at least 10bit I/O with SDI (SMPTE 259M or SMPTE 292M) interface.

Agreed, but I have a habit of trying to push the envelope on hardware when the economies are favorable.
I looked at a number of high end analog capture cards (admittedly not recent ones) and found that most had
eight-bit flash ADCs, so I wasn't too put off by the sample width on the BT848 but certainly the accuracy,
noise figure, freq. response, etc. are important, and these data are not provided for the chip. Your suggestion
of locating a reference design with its performance report is good and deserves some effort; I may ask on
comp.arch.embedded. I am not trying to put lipstick on a pig and certainly wouldn't include BT848 products with
serious studio gear, but if the part can be made to work to broadcast specs, I will use it.

Since it is very difficult to get low noise and high dynamic range in a mixed analog and digital IC, I am very
curious to see real data for this and other analog video decoder chips, however I suspect that much of this
data is purposely unpublished or provided only to card makers.

[edDV]

Check out the Aja Xena LS/LSe as a similar product intended for a true broadcast SDI/analog environment. These 8/10 bit PCI/PCIe cards are used to adapt Sony Vegas or Adobe Premiere Pro into a studio environment. The data sheet PDF lists full performance specs.

http://www.aja.com/html/products_windows_xena_LSe.html
http://www.planetdv.net/Pdfs/XENA_LSe.pdf

The Xena 2K model supports HD resolutions.
http://www.aja.com/pdf/XENA_Line.pdf