Request for measurements: 6MHz sweep response

[cybertheque]

Request for analysis of 6MHz sweep bitmap image (test card).
Edited to speed image loading -- please see followups

[edDV]

Again, you need to separate capture measurement from display card characteristics.

Can you show that the Sweep file to O'scope shows flat response?

Here are a couple of old threads that discuss 6MHz sweeps out display cards using the THX test slide as a free sweep reference. If you have questions about those, ask here in this thread rather than opening the old ones. I can duplicate those tests if you want.

https://forum.videohelp.com/topic296772.html

https://forum.videohelp.com/topic239205.html

[edDV]

This slide shows the what a reasonably good sweep display looks like from an ATI 9550 display card to a wide band "HD-Ready" CRT display that has a 3D comb filter for Y/C separation. Ignore the moire' caused by shooting the screen with a simple digital still camera.



This shows the same sweep feeding a cheap TV with a simple notch filter for Y/C separation. Your waveforms look similar. I'm guessing your capture card lacks a quality comb filter.



See http://members.aol.com/ajaynejr/vidcomb.htm

Save Ajay Nejr's excellent files. It looks like AOL is going to dump them.

[cybertheque]

I'm guessing your capture card lacks a quality comb filter.
See http://members.aol.com/ajaynejr/vidcomb.htm
Save Ajay Nejr's excellent files. It looks like AOL is going to dump them.

Thanks for the link and the images; indeed I found that the composite input on
the Targa card has a notch at about the color subcarrier frequency which is clearly
visible when looped-back to the Targa composite input, however the problem I
have is with a perceived notch on digitized Targa s-video output which I suspect
is a result of improper sampling in the various scope programs that I am testing
and/or problems with the test signal. To explore this further, here are some
waveform displays and vidcaps.

In order to view the s-video y/c signals in pass-through, I built this breakout box:

The BNCs connect to the vert. inputs of a dual-trace scope.

Here is the scope display for the 6MHz sweep waveform:

The scale isn't calibrated as I am only interested here in the frequency response.
Except for some garbage at the end of the sweep, the frequency response and
linearity of the sweep is quite good. The trailing garbage is of concern; my request
in this posting is to have others look at the test image with other signal analysis programs
and report the results. Inspecting the waveform with delayed and magnified sweep
shows that it is quite flat and linear out to about 5MHz and rolls-off steeply to 6MHz before
the garbage appears. Here again is the url for the test image:
http://www.cybertheque.org/tmp/vidtest/scope-targa-svid-loopback-6MHzSweep-1l.jpg

Here is a waveform display of that image from the program 'Photoscope' (a spatial analysis of
luminance):

To me this looks like bad sampling in the program (undersampled, aliasing and lack of a
bandstop filter); I have no other still image photo analysis programs to try at the moment.

Here is a waveform display of this sweep signal as digitized by a Bt848 capture card on a
different host, from the 'Vscope' program:

Remember that the input waveform here is the one shown in the scope photo above, however
the notch appears here as if the s-video input on this card has a notch-filter on the Y channel
(which is not the case); again I am suspecting some kind of sampling error in the 'Vscope'
program.

Here is a captured frame from the Bt848 card of the sweep signal; to me it looks pretty good:


FWIW, here is the 'Photoscope' waveform display of that captured frame:


I really need to know how good the source image is at the highest frequencies; I should think
that the Nyquist limit for 640x480x30fps should be around 10MHz (alternating light and dark
pixels), so 6MHz should be well within the permissible bandwidth.

If someone could analyze the test image in an imaging program with proper tools and/or
connect a scope to the NTSC video out of a good output card and photograph the waveform
for a single line of that image, it would be a great help.

Many thanks,

Michael

[guns1inger]

FYI your images take forever and a day to download. I have yet to see any of them because after 10 minutes I close the screen out of boredom. If you upload them to the forum instead they will open much, much faster.

[2Bdecided]

Images work fine here.

Nyquist limit of 720x480 is 6.75MHz. (13.5MHz sampling)
Nyquist limit of 640x480 is exactly 6MHz. (12MHz sampling)

No chance of getting 10MHz from either system.

Cheers,
David.