maximum tv output resolution for ATI All In Wonders ?

[lordsmurf]

I currently have ATI AIW 7200 cards. The maximum output resolution for these cards is 800x600 for tv output.

I want to upgrade to another ATI AIW card that does 1024x768. As usual, ATI.com's information sucks. So I have to rely on user information.

What ATI AIW card have you got? What is it's maximum TV OUTPUT resolution?

[Bodyslide]

On the ATI AIW 9800 Pro and the 9700 from the Manual it states You can set your display resolution as high as 1024x768.

[edDV]

I've been wondering about the output performance of the AIW S-Video output as well. Seems underwhelming to me.

The AIW-8500DV (Theater 200) outputs "max supported resolution" of 1024x768 on video out but that is frame buffer resolution, not what is on the S-Video.

The card samples from the output frame buffer to D/A and is analog filtered for my estimated analog resolution of ~450 lines of resolution (equiv ~550x480/576).

So how to measure it?

Here is my first idea.

Use the THX test pattern frequency sweep. Reference is a DVD Player S-Video output. Then compare the same DVD playback at 800x600 and 1024x768 from the ATI card.

How to measure? If I loop the S-Video back to a capture device, other errors may be introduced by the capture device. How about shooting the monitor with a still camera?

[swiego]

Interesting, I've been trying the same thing. I wanted to use my computer's s-video output as a means for testing a few different s-video cables that I have...

[lordsmurf]

I was only interested in getting 1024x768 of computer data on the tv screen, but what you're talking about is interesting too. I've thought about it before, but it was never something I needed to test. I do look forward to seeing somebody else's test, however.

[edDV]

OK first pass. Goal is to make this a test that can be repeated by others with various cards.

Consider this first pass a test of the test.

Test equipment:

Test reference: Star Wars IV "A New Hope" THX video test slide.
Monitor: Philips 27PT830H HDTV (S-Video in)
Digital Still Camera: Canon S60 (mode 2592x1944, flash off, auto exposure, tripod, 5ft back)
DVD Player: Pioneer DV363 (S-Video out)
Test Display Card: ATI AIW-8500DV Theater 200 (S-Video out - full screen)

Still Photos were too large to post, so I reduced them to 720x540 in Photoshop.

Reference Pioneer DV363 (S-Video out)


ATI AIW-8500DV 800x600 Full Screen (S-Video out)


ATI AIW-8500DV 1024x768 Full Screen (S-Video out)



Preliminary conclusions:

1. The DVD player easily outperforms the ATI AIW for frequency response (detail)
2. ATI AIW 800x600 mode looks better that 1024x768 for frequency response (detail)

Test flaw: Camera was framed and zoomed a bit tighter for reference picture. Next pass, the settings should be held constant. However, the pictures represent what was seen.

Any comments before I repeat the test later this evening?

PS: The same S-video cable was used to eliminate that as a factor. I plan to zoom into the slide for better web display next pass.

[jagabo]

Try putting the AIW in theater mode and enable Video mode. You should find the AIW's video output is much closer to a DVD player's. Of course, you only see the contents of the video on the TV.

Can that source image be downloaded from anywhere?

Try putting the AIW in theater mode and enable Video mode. You should find the AIW's video output is much closer to a DVD player's. Of course, you only see the contents of the video on the TV.

Can that source image be downloaded from anywhere?

I found a copy of the image online but I think it's actually a screen capture so it's probably not as clear as the one you started with. Here's the s-video output from the old Matrox Millennium G450 (not MCE) photographed (hand held) on my 32" Philips SD CRT (scaled and cropped with VirtualDub's Lanczos3 filter to match the size of your images):



I found a cleaner version of the THX test pattern. Here's the same test using the cleaner image:



The moire patterns in the color bars weren't visible on the TV screen. They're an artifact of photographing the screen.

I'd duplicate this test on another computer with a Sapphire ATI X300 card but it's only hooked up to a 13" TV.

[edDV]

The THX patterns are on many commercial DVDs.
http://www.thx.com/mod/products/dvd/dvdFind.html
Click "entire list"

I'll be back to this in a couple of hours.

[edDV]

Second pass.

The AIW8500DV is in Theater Mode now.
I cropped the interesting part of the photos rather than resizing.
But I still had to downsize to 720 for fit the web page.

Moire patterns create some of the blur but this is constant over the 3 images. Moire is caused by the digital camera and is not visible to the eye on the screen.

Test equipment:

Test reference: Star Wars IV "A New Hope" DVD - THX video test slide.
Monitor: Philips 27PT830H HDTV (S-Video in)
Digital Still Camera: Canon S60 (mode 2592x1944, flash off, auto exposure, tripod, 5ft back)
DVD Player: Pioneer DV363 (S-Video out)
Test Display Card: ATI AIW-8500DV Theater 200 (S-Video out - full screen)

Reference Pioneer DV363 (S-Video out)


ATI AIW-8500DV 800x600 Theater Mode Full Screen (S-Video out)


ATI AIW-8500DV 1024x768 Theater Mode (S-Video out)



Preliminary conclusions:

1. The DVD player still outperforms the ATI AIW-8500DV for frequency response (detail)
2. ATI AIW 800x600 mode looks better than 1024x768 for frequency response (detail)
3. Theater Mode improved the AIW performance.

Later I'll try a Radeon 9550.

PS: To complete the picture here is what the AIW-8500DV looks like on the 19"Hitachi computer monitor at 1024x768.

[jagabo]

edDV, did you enable "video mode"? Enabling it is a little tricky. You have to start playing a video before you bring up the video settings dialog or the button doesn't show up.

1) Start playing a video with WMP -- or any player that uses overlay.

2) Right click on the Windows desktop and select Properties.

3) Select the Settings tab.

4) Press the Advanced button.

5) Select the Displays tab.

6) Press the TV button.

7) Press the Adjustments tab.

8) You should now see the Video Mode button at the bottom of the dialog.



On my X300 card with Windows running at 800x600 or 1024x768 (and not in theater mode) the tv output is the full desktop scaled down to ~640x480. At resolutions over 1024x768, the TV output is scaled down from a 1024x768 window that follows the mouse cursor. So you can't see the entire desktop. I would actually prefer it used a 640x480 window so that there would be no scaling taking place but I haven't found a way to force that.

[swiego]

I've gotten some amazing results by setting my resolution to 720x480 and enabling video mode. Windows desktop looks almost as sharp as the real thing over s-video to my Dell 2405 LCD. The 2405 has a side-by-side mode for two inputs so I ran a pair of 720x480 signals (one DVI-D, the other s-video) to it and couldn't see much of a difference, except for minor flicker with s-video output.

Keep in mind that even if you don't turn on video mode, there still is a svideo sharpness control (as well as flicker control) available, at least with the latest catalyst control center plus a Radeon 9800 Pro.

[jagabo]

I've gotten some amazing results by setting my resolution to 720x480 and enabling video mode.

How do you set 720x480? I can't go below 800x600 from the main settings dialog.

[edDV]

OK, I'll do it again and make sure video mode is enabled. I've kept sharpness and flicker control off so that we can see the true path performance.

[dLee]

I've tried several times to set up different TV's (SD) as computer monitors using the S-Video connection. The results have all been unacceptable.

I have had success with my Sony KV-36XBR800 CRT TV using the DVI-HDTV In (Video 7). My ATI AIW 9800 Pro has a DVI Out.

I eventually found a resolution (1152x648 @ 60Hertz) that works well with the 16:9 HD letterbox desktop displayed on a 4:3 CRT.

MMC TV and File Player apps look great. In full screen mode (i.e. 16:9 on 4:3 display), I get an excellent approximation of my captures.

I'm able to sit at a normal distance (couch to TV) and use MPEG Video Wizard and VideoReDo Plus to edit my MPEG2 captures. I can also browse the web or read email.

The ATI AIW 9800 Pro has two resolutions (@ 60Hz) that are exactly 16:9 - 1152x648 and 1920x1080. It has two other resolutions that are close to 16:9 - 1360x768 and 1776x1000.

1152x648 is the resolution that I prefer. I can sit at a normal distance and edit video, read text etc. At 1920x1080 text is very small and hard to read. The other two resolutions result in small parts of the desktop edge being out of the 16:9 display area. Text is also harder to read.

[swiego]

I've gotten some amazing results by setting my resolution to 720x480 and enabling video mode.

How do you set 720x480? I can't go below 800x600 from the main settings dialog.

When I enable a TV as a s-video advice, it switches to 720x480 and I am able to both see and set that resolution in the "TV properties" page of the catalyst control center as well.

I can't get lower than 800x600 in the main display properties, either.

We really should be having the video card's frame buffer running at 480 lines so as to avoid any scaling artifacts.

EV - I've seen a definite improvement by turning the sharpening level to More.... at least, I haven't seen any downside.

[edDV]

DVI and YPbPr (with HDTV adapter) are the preferred methods for desktop display. When I finish wringing out the S-Video connection, I intend to compare the result with the YPbPr path.

The YPbPr HDTV adapter will scale the output frame buffer to 720x480i, 720x480p, 1280x720p or 1920x1080i. This contrasts to the 800x600 and 1024x768 frame buffer size for Theater Mode over S-Video. The sweeps above show S-Video analog horizontal detail is compromised between 3-6 MHz (right half) compared to the DVD player direct (720x480 sourced). S-Video will always have 480 or 576 interlaced vetical lines.

S-Video is important for displaying or recording video. S-Video should be used for desktop display only as a last resort but it's nice to know how the methods compare.

[swiego]

I must say that in video mode with s-video feeding my LCD (which has a s-video input) I was stunned by the monitor's ability to nicely resolve a single pixel at 720x480 framebuffer resolution. I didn't think s-video had quite that resolution. I was using a bluejeanscable s-video cable and I also noticed the monitor image to be somewhat cleaner (less shake/flicker) with it than with the super cheapo one that I have. However both were clearly resolving more than I'd ever expect to be able to pull off either a SVHS or a Hi8 tape.

[edDV]

I must say that in video mode with s-video feeding my LCD (which has a s-video input) I was stunned by the monitor's ability to nicely resolve a single pixel at 720x480 framebuffer resolution. I didn't think s-video had quite that resolution. I was using a bluejeanscable s-video cable and I also noticed the monitor image to be somewhat cleaner (less shake/flicker) with it than with the super cheapo one that I have. However both were clearly resolving more than I'd ever expect to be able to pull off either a SVHS or a Hi8 tape.

720x480 equates to about 6.75MHz maximum analog bandwidth. S-Video (Y) is good out to 8-9MHz if your source and display are good enough. S-video horizontal resolution in theory can exceed the DVD spec. See this discussion:
https://forum.videohelp.com/viewtopic.php?t=294517&highlight=ikegami

[swiego]

If you don't mind my asking, when you say "S-Video (Y) is good out to 8-9MHz if your source and display are good enough" are you referring to physical properties of the cable itself? I think you are saying that a quality s-video interface should more than exceed ITU Rec-601 requirements, is that fair to say? (If so, then what is the advantage of component for any NTSC interlaced material? Is there any?)

[edDV]

If you don't mind my asking, when you say "S-Video (Y) is good out to 8-9MHz if your source and display are good enough" are you referring to physical properties of the cable itself?

I was referring to the analog Y connection and the connection cable only has to be good enough, for say under 25 feet, before it becomes an issue.

Wide bandwidth baseband analog component video is expensive to generate but fairly cheap to distribute over medium distance. At an extreme I was describing in the other post, a SD broadcast camera like this $90,000 Ikegami HK-388 that can generate 8MHz +/- 0.1 dB analog video at -60dB S/N. A monitor capable of displaying 8Mhz at that quality might set you back another $25-45,000.

____
http://bssc.sel.sony.com/BroadcastandBusiness/DisplaySubCategory?m=0&p=8&sp=20073

A 25' cable that runs between them only needs to cost $25. No big deal unless the cable run goes long, then you need to equalize and it gets more expensive.

I think you are saying that a quality s-video interface should more than exceed ITU Rec-601 requirements, is that fair to say? (If so, then what is the advantage of component for any NTSC interlaced material? Is there any?)

Now we get into the apples and oranges of analog vs. digital video. ITU Rec-601 specifies a Y sample rate of 13.5MHz. Analog bandwidth cannot exceed half of that (6.75 MHz theoretical) so the camera and monitor above, exceed ITU Rec-601. That is a long way from saying and an AIW S-Video output will perform at those levels. So far, I'm seeing some problems over 3MHz but there is something there out to 5.5MHz or so. The THX sweep isn't calibrated but I assume the middle is ~3MHz and the right side is ~6MHz. A good real world indicator of quality is the legibility of the fine print at the bottom of the slide.

In theory analog YPbPr (or VGA) should perform near equal to DVI-D but high quality digital is cheaper to build than high quality analog. This becomes increasingly true as bandwidths and display resolutions go up.

Broadcast facilities use YPbPr, but few use S-Video (Y/C) for anything serious. The Y spec is the same but C is a modulated (3.58MHz subcarrier), bandwidth limited (about 500KHz) form of PbPr. It is minimally adequate for consumer camcorders, but is totally inadequate for computer desktop displays. VGA for example can exceed 30MHz per component.

Consumer game boxes like the XBox and PS2 are designed for S-Video as the target monitor connection. The desktop and games are optimized for that resolution.

[swiego]

Interesting and, I think, it makes sense.

Then, as an aside, what is the real challenge in getting the AIW's s-video output to generate a signal clean out to 8 MHz on par with that $90K camera? After all, it's right next to a RGB output on the same card capable of sending out a clean 2048x1536x60p signal. It seems to me that s-video output from a computer video card should be pushing the limits of NTSC 480i itself without breaking a sweat... what am I missing?

[edDV]

Interesting and, I think, it makes sense.

Then, as an aside, what is the real challenge in getting the AIW's s-video output to generate a signal clean out to 8 MHz on par with that $90K camera? After all, it's right next to a RGB output on the same card capable of sending out a clean 2048x1536x60p signal. It seems to me that s-video output from a computer video card should be pushing the limits of NTSC 480i itself without breaking a sweat... what am I missing?

Good question. There is no theoretical reason the Y output shouldn't be as clean as the Y on the HDTV YPbPr adapter at 480i/480p. The S-Video color components are inferior by definition because they are NTSC encoded (approx 500-700KHz).

In the real world, I suspect the S-Video Y is output by "TV grade" chips. I intend to measure both outputs and compare. I need to wait until dark before I can shoot the photos.

I have to point out that computer display cards are currently poor for HD analog video out. The HD Cable box is an order of magnitude better over YPbPr comparded to a MPeg2_TS playback through the AIW-8500 or 9550 with the HDTV adapter. I've tried a NVidea 5xxx card and there is little difference.

I haven't tested the latest x1000 series cards. That will have to wait until I build a new computer.

[swiego]

I await your results with bated breath!

This makes me wonder what kind of s-video output quality we get from various SVHS, Hi8, camcorder and other sources, if we are to believe that there is some variability here.

[edDV]

Tests on the 8550DV in Theater + Video Mode did not show different results.

Next I tested a clone NovaTech RADEON AI-9550. As far as video output features, the Radeon 9550-98xx and 3 digit x cards appear to have similar specs.

Test equipment:

Test reference: Star Wars IV "A New Hope" DVD - THX video test slide.
Monitor: Philips 27PT830H HDTV (S-Video in)
Digital Still Camera: Canon S60 (mode 2592x1944, flash off, auto exposure, tripod, 5ft back)
DVD Player: Pioneer DV363 (S-Video out)

Test Display Card: NovaTech RADEON AI-9550 (S-Video out - Theater + Video Mode)

800x600 NovaTech RADEON AI-9550 (S-Video out - Theater + Video Mode)


1024x768 NovaTech RADEON AI-9550 (S-Video out - Theater + Video Mode)


720x480 480p scale (NovaTech RADEON AI-9550 +ATI HDTV YPbPr Adapter)


1920x1080 1080i upscale (NovaTech RADEON AI-9550 +ATI HDTV YPbPr Adapter)





Conclusion:

The Radeon 9550* S-Video results are clearly superior to my 8500DV for frequency sweep and comparable to the Pioneer DVD player. YPbPr output results were similar.

There may be differences in S-Video output hardware. The 8500DV does have the long input/output dongle. Mine could be faulty. Both have similar video related specs including the all important Video Immersion II that provides deinterlacing, filtering and HDTV upscaling. ATI describes it as:

"Video Immersion™ II
ATI’s Video Immersion™ II technology integrates industry-leading digital video features, including advanced adaptive de-interlacing algorithm, temporal filtering, and video gamma enhancement for unprecedented video quality and integrated digital TV decode capability. In addition, there is component output support for HDTVs (High Definition Television) at 480i, 480p, 720p and 1080i."

NVidea's High-Definition Video Processor (HDVP) is described here.The PDF goes into more detail.
http://www.nvidia.com/object/feature_hdvp.html


*although the NovaTech RADEON AI-9550 performed well for the frequency sweep and with the HDTV adapter, the S-Video output had a severe crosstalk problem that would reject this particulr card for S-Video uses. I'll post a picture here when I can.

PS: Crosstalk problem went away after cycling the S-Video port a few times.

[edDV]

Back to Mr. Smurf's origial question.

Here is what a desktop looks like on the S-Video port. TV picture looks slightly sharper than this. I captured S-video with the ADVC-100. DV max analog bandwidth is around 6MHz.

1024x768 over S-Video (NovaTech RADEON AI-9550)


1024x768 over S-Video


800x600 over S-Video


800x600 over S-Video


For comparison, here is a 800x600 original desktop sized the same. Notice the type resolution.

[edDV]

So in summary,

VGA uses short 6 ft coax cable, wide bandwidth (30-200MHz) RGB to provide a clear desktop.
YPbPr is optimized for SD to HD video over longer cable runs (Y 6-40MHz, Pb and Pr 1.5-3.8 MHz bandwidth)
S-Video luminance is 4-8 MHz, chrominance ~500-1500KHz.

Viewing a computer desktop over S-Video should be done as a last resort. Special desktops can be designed for S-Video as can be seen on XBOX, PS2 and cable menus. Computer desktops can be enhanced for display at low broadcast bandwidths as done on shows like TechTV or demo videos but this relies on restrictive techniques, special hardware and large fonts.

XBOX

[swiego]

Dunno.

It just sounds like analog RGB from the fairly early days of desktop computing was superior to the very best component video, at least as far as video data transport is concerned. It seems as though luma s-video should be all but indistinguishable from, say, DVI-D, at 720x480 resolutions. If the cables can handle it, the sockets can handle it and the NTSC protocol provided enough bandwidth to handle it, why couldn't s-video have been a perfectly legitimate transport for 480i at 720 lines of horizontol resolution?

[jagabo]

Dunno.

It just sounds like analog RGB from the fairly early days of desktop computing was superior to the very best component video, at least as far as video data transport is concerned.

Yes, you can easily get 400 MHz VGA cables. Good video cards now have 400 MHz ramdacs which can output progressive 2048x1536 85 Hz., far beyond HDTV.

[edDV]

Dunno.
It just sounds like analog RGB from the fairly early days of desktop computing was superior to the very best component video, at least as far as video data transport is concerned.

VGA is analog RGB. VGA is great to 6ft or so in all its extended forms up to 1600x 1200. VGA is good for a local desktop monitor but sucks for moving video very far.
http://en.wikipedia.org/wiki/VGA

It seems as though luma s-video should be all but indistinguishable from, say, DVI-D, at 720x480 resolutions.

S-Video was designed for interconnecting high end consumer video equipment (e.g. S-VHS VCR, S-VHS camcorder, Hi8 camcorder, computer capture cards).
http://en.wikipedia.org/wiki/S-Video

S-Video as a connection technology can exceed 480p for Y but falls way short on chroma UV (PbPr) bandwidth. But ignoring that, 720x480 makes fine DVD video, but makes a very poor desktop for computers. For desktop, you want to read fine text. DVD sucks for documents and graphics. You need 24 bit 1024x768 RGB or up.

If the cables can handle it, the sockets can handle it and the NTSC protocol provided enough bandwidth to handle it, why couldn't s-video have been a perfectly legitimate transport for 480i at 720 lines of horizontol resolution?

S-Video connections can do 540 "lines of resolution" equivalent to 720 pixels easily and more so long as the equipment on both ends can handle it. Y for S-Video (Y/C) is identical to Y for YPbPr and can go to full up 1920x1080 HDTV no problem. The C part of S-Video is NTSC or PAL modulated subcarrier with resolution in the 80x480/576 range.

http://en.wikipedia.org/wiki/NTSC
http://en.wikipedia.org/wiki/PAL

PbPr has 1.75-3.5 MHz bandwidth for SD and ~2-3x that for HDTV

Best to understand analog before getting sucked into digital.

[edDV]

Dunno.

It just sounds like analog RGB from the fairly early days of desktop computing was superior to the very best component video, at least as far as video data transport is concerned.

Yes, you can easily get 400 MHz VGA cables. Good video cards now have 400 MHz ramdacs which can output progressive 2048x1536 85 Hz., far beyond HDTV.

QXGA 2048x1536 (2048x1152 16:9) isn't "far beyond" 1920x1080P. (+6.7% h)
Pretty much the same thing if you ask me.

RAMDAC (RAM Digital to Analog Converter) is an analog output technology to analog RGB. RAMDAC speed is not the same as cable bandwidth. RAMDAC speed determines refresh rate at various destop resolutions.

At the limit analog progressive RGB over VGA can use 400MHz D/A and cables over short distances. This is only used at extreme screen sizes and refresh rates.

Analog HDTV 1920x1080i or 1280x720p uses about 37MHz to the monitor and 59.94MHz refresh.

http://computer.howstuffworks.com/graphics-card3.htm
http://grafi.ii.pw.edu.pl/gbm/matrox/ramdac.html
http://home.aol.com/ajaynejr/bandwid.htm


When you look at digital DVI-D connection to the monitor, wideband RGB is used for both computer displays and HDTV.

Single-Link DVI-D maximizes at the following:
* WUXGA (1920 x 1200) @ 60 Hz (154 MHz)
* HDTV (1920 × 1080) @ 60 Hz with 5% LCD blanking (131 MHz)

Dual-Link DVI-D supports larger sizes and refresh rates.
* QXGA (2048 × 1536) @ 75 Hz with GTF blanking (2×170 MHz)
* HDTV (1920 × 1080) @ 85 Hz with GTF blanking (2×126 MHz)
* WQXGA (2560 × 1600)

http://en.wikipedia.org/wiki/DVI
http://www.mycableshop.com/3rd_Level/Video-Flat.htm