Maker Pro
Maker Pro

Testing a Video Signal

M

Mike Warren

Jan 1, 1970
0
We have just purchased a quantity of PCIe video cards with
TV out (S-Video) and can only get a monochrome picture
from the S-Video output.

I have so far tried 6 different cards from 2 batches and
all have the same fault. Here are my observations?:

-Another brand of video card that used the same video chip
(nVidia 8400GS) works fine using the same driver.

-The driver /is/ set up correctly.

- The fault is there in both PAL and NTSC modes. Only one
of my monitors supports NTSC so I haven't done any extensive
testing in that mode.

-These cards produce a B&W picture on 3 separate monitors
(2 LCD and one CRT) All monitors work fine with the other
brand. The 15 year old CRT monitor will produce a colour
picture from this card for a few seconds when cold.

- Feeding the signal into a Canopus ADVC-100 video capture
device produces a colour picture on my computer.

Here are some pictures of the output. In the oscilloscope
pictures the top trace is the composite signal after my
passive S-Video to composite converter (just a 1000pF cap)
and the bottom trace is the chroma signal from the card.
www.mike-warren.net/tmp/DSO1-ECS-web.jpg
www.mike-warren.net/tmp/DSO1-Leadtek-web.jpg
www.mike-warren.net/tmp/TV-ECS-web.jpg
www.mike-warren.net/tmp/TV-Leadtek-web.jpg

This is a vectorscope signal produced using Adobe Premier
from the same colour bar signal. I haven't recorded the
output of the good card since I don't believe the
vectorscope will help with this problem.
www.mike-warren.net/tmp/Vectorscope-Leadtek.jpg

What tests can I do to try and work out what is wrong with
the signal? I tried to read the frequency of the colour
burst but I haven't been able to lock the DSO (Tek TDS1002)
on it so far.

The supplier claims there is nothing wrong with their cards
as no one else has ever complained.
 
M

Mike Warren

Jan 1, 1970
0
Michael said:
S-video has seperate Luminance & Chroma channels. You need both to
get a color image. If you are feeding a composite monitor, you need a
S-video to composite adapter.

Thanks for the reply. As I explained, I'm converting through
a 1000pF capacitor. This have been sufficient for me for over 6
years, and anyway I have the same problem feeding into a
S-Video input to the monitor.
 
P

Phil Allison

Jan 1, 1970
0
"Mike Warren"
We have just purchased a quantity of PCIe video cards with
TV out (S-Video) and can only get a monochrome picture
from the S-Video output.

I have so far tried 6 different cards from 2 batches and
all have the same fault. Here are my observations?:

-Another brand of video card that used the same video chip
(nVidia 8400GS) works fine using the same driver.


-These cards produce a B&W picture on 3 separate monitors
(2 LCD and one CRT) All monitors work fine with the other
brand. The 15 year old CRT monitor will produce a colour
picture from this card for a few seconds when cold.


** The symptoms fit with a colour sub carrier frequency that is too far off
for your monitors to lock onto.

The allowed error ( for a broadcast signal ) is only 5 Hz in 4.4MHz or
about 1ppm - which is way better than an un-trimmed crystal can provide.

Being only a 10 cycle burst means measuring it directly is not possible.

If the PCI card uses the computer's clock frequency for all timing, then you
are stuck.

If there is an on board 4.43MHz crystal - maybe you can tweak it a bit.



..... Phil
 
M

Mike Warren

Jan 1, 1970
0
Phil said:
"Mike Warren"


** The symptoms fit with a colour sub carrier frequency that is too far
off for your monitors to lock onto.

That's what I was guessing, but hoped there would be an easy way to confirm.

The allowed error ( for a broadcast signal ) is only 5 Hz in 4.4MHz or
about 1ppm - which is way better than an un-trimmed crystal can provide.

Being only a 10 cycle burst means measuring it directly is not possible.

If the PCI card uses the computer's clock frequency for all timing, then
you are stuck.

If there is an on board 4.43MHz crystal - maybe you can tweak it a bit.

There is only one crystal. I'm guessing it's the master clock and all the
other clocks are PLLs.

I don't fancy getting involved in hardware mods to 100 video cards so I'm
hoping there is a software trim available. I've asked the manufacturer and
are awaiting their reply.

Thanks for your reply, Phil.
 
M

Mike Warren

Jan 1, 1970
0
Jan said:
start:
The burst is way too small, possible teh color killer activates.

I don't believe this is the case here since the good card has exactly the
same signal level. Also, I've had others with even lower colour burst still
work fine on my test monitors.

Thanks for your reply.
 
M

Mike Warren

Jan 1, 1970
0
Jan said:
Your S-Video interface sucks.

It certainly does, but the same type of circuit is very commonly used,
and I've used it myself in hundreds of these machines since 2003
without a single complaint of no colour.
 
M

Mike Warren

Jan 1, 1970
0
Michael said:
That may be too much capacitance. You are working with siglas in the
3.5 MHz range and may be coupling too much of the Luminance signal into
the chroma channel. I've seen as little as 47 pF used.

I did experiment years ago and found 1000pF to be the best compromise in
practice from the values I normally keep in stock. I have never had the
need to revisit the situation. I already /know/ that the converter is
not the problem here because I get the same fault if I plug directly
in to the S-Video input of the monitor.
 
P

Phil Allison

Jan 1, 1970
0
"bg"
If the frequency of the burst was off, the color hue would shift, but you
would still have color.

** Completely WRONG.

The phase of the colour sub carrier is what produces hue shifts.

Explains why NTSC is so prone to hue shifts in transmission and why PAL was
invented to solve the problem.

If the TV receiver or monitor cannot phase lock onto the sub carrier burst,
then normally no colour will appear.



..... Phil
 
P

Phil Allison

Jan 1, 1970
0
"Mark"


Regarding the freq, the color bust freq and the V and H should all be
locked together. you can probably count the H and it should be
15.734. kHz for NTSC.

** The OP is in Australia an so using PAL monitors.


Another trick is to take a known good video
signal and loosly couple the UUT video to it and you will see the
video sync drift through. If the H sync drift less then 1 frame per
sec or so, it is probably close enought that a monitor should lock to
it.

** Horizontal synch is NOT the problem.

Colour sub carrier synch IS !!



...... Phil
 
P

Phil Allison

Jan 1, 1970
0
"bg"
Phil Allison
"bg"

If the frequency was off but close enough to allow the oscillator to
partially lock to it, wouldn't that cause the hue to alternate?


** Oh dear....

Go look up PLL on Google an learn something.



..... Phil
 
P

Phil Allison

Jan 1, 1970
0
"Mark"
** The OP is in Australia an so using PAL monitors.

Another trick is to take a known good video
signal and loosly couple the UUT video to it and you will see the
video sync drift through. If the H sync drift less then 1 frame per
sec or so, it is probably close enought that a monitor should lock to
it.

** Horizontal synch is NOT the problem.

Colour sub carrier synch IS !!


yes agreed, but since Hsynch and subcarrier bust are phase locked to
each other, the error expressed in % should be the same for both.

** Meaningless drivel.

A monitor's H-synch has a wide tolerance while the colour carrier synch
range is very small.

You have no clue how any of the circuits work.

Piss off

- you FUCKWIT GOOGLE MONKEY !!



..... Phil
 
M

Mike Warren

Jan 1, 1970
0
Jan said:
I would take the card back as I do no want to sell to idiots.

LOL! Thanks. I hadn't been called an idiot today, yet. :)

You are indeed fortunate to be able work in an area where you can use
the best theoretical techniques at all times and are not required to
do things by the cheapest method that gets the job done.

I'm not sure why you are fixated on the combiner when I have stated
several times that the problem still exists when I connect the S-Video
cable /directly/ to the S-Video input of the monitor.

Some more information for anyone who's interested:

I was able to get lock on the colour burst and measure it with the
cursors on the Tek. I get 4.44MHz on the good card and 4.42MHz on the
bad ones. Unfortunately this is only a one step difference of the
cursor position so is not an accurate indication.

I could discern /no/ difference in the timing of the H-sync between
the 2 cards.

The chroma level is within the measurement error of my Tek between
the 2 cards as well.

The manufacturer is looking into the problem.
 
M

Mike Warren

Jan 1, 1970
0
Paul said:
Didja try another s-video cable? Maybe the chroma pin is loose in a
connector.

Yes, early on, but it is not necessary since I can see the chroma
signal at the other end of the cable and the same cable was used
for both the good and bad cards.

My comparison is between one card made by ECS which works fine
and 6 cards made by Leadtek which do not work properly.

Both cards look nearly identical in layout and have probably just
been copied from the reference design.

I have previously used 35 of the ECS cards without any problems,
but can't get any more of them.
 
P

Phil Allison

Jan 1, 1970
0
"Mark Fucking Idiot Kolber "

true but irrelevent..


** It is in fact the ** MOST ** important issue here.

You fucking GOOGLE MONKEY !!

if you reacall I was suggesting the op measure the H sync error in
order to determine the color subcarrier errors.

** Wot a completely asinine suggestion -

from a FUCKWIT GOOGLE MONKEY !!

They will be the same when expressed in %


** TOTALLY WRONG.




...... Phil
 
P

Phil Allison

Jan 1, 1970
0
"Mike Warren"
I was able to get lock on the colour burst and measure it with the
cursors on the Tek. I get 4.44MHz on the good card and 4.42MHz on the
bad ones. Unfortunately this is only a one step difference of the
cursor position so is not an accurate indication.

** You bet is not accurate enough.

As I alluded to before, the lock-in range of typical PAL TVs and monitors is
only as far as the internal 4.43MHz crystal can be pulled by a small amount
of capacitance - ie maybe +/- 25 Hz.

According to your testing, a CRT monitor locked briefly when it was cold -
so the sub carrier burst is likely to be on the low side of the correct
frequency.

The only way to measure it is to tweak a monitor's crystal so it DOES
lock - then measure the frequency of oscillation with a counter.


...... Phil
 
M

Mike Warren

Jan 1, 1970
0
Jan said:
Wonder what a 'theoretical technique' is, about the same as your S-video
interface technique?

You tell me. How would you do it?

Do you not think that before complaining about a thing you at least should
connect it in the correct way?

Again, I did. I connected the S-Video output from the card to the S-Video
input of the monitor using a commercial S-Video cable and got a monochrome
picture. From all six cards under test. The other brand, and others before
that, work fine.

Here analog video is long of the past, only some very small cameras I
have still use it.

Here too, and this is causing me problems because video cards with
S-Video or composite outputs are nearly impossible to find now. However
my customers have large numbers of old TVs installed in various venues.
It would bankrupt them to have to replace all of them and the wiring
etc. to get HDMI. And absolutely no one involved would care about the
improvement in quality.

I'm on the lookout for HDMI to S-Video or composite converters, but
it seems no one makes one.

To measure the frequency difference you can put one signal on X and
and one on Y.
A Lissajous figure will appear that rotates the faster the bigger the
difference.

I only have 10 cycles to view. If I had both cards running simultaneously
I still wouldn't be able to sync the colour bursts to do that.

Or you could external trigger on one, and measure the other, and see
how many periods pass by in a second.
If there is a lot of difference that would be difficult.

See above.

Here is a cheap frequency meter, but the resolution is only 32 Hz:
Should be enough for this purpose.

I have one that is accurate to within a few Hz too. The problem is getting
it to read the burst. The only way I can see to do that is the way Phil
suggests and measure it at the monitor's oscillator. That's more work than
I want to go to at this stage, unless the manufacturer comes back to me
and says there is no problem. Then, I will need to offer proof.

H sync frequency has nothing to do with it.

I know. I just did this to answer the suggestions of a couple of posters.

Tek?? Oh well.

Do you seriously think that a monitor will have trouble with a chroma
/level/
variation of a few percent?

I get that you don't like Tek, and I'm not all that impressed with them
either, but I don't have the budget others do, especially if this is to
help correct someone else's error.

All I needed to do here was connect the card to my monitor via the S-Video
connector and tell the manufacturer that it didn't work. What I wanted to
do was get a little more information as long as it didn't entail too much
work so I could reduce some of the inevitable to and fro while the
manufacturer tried to tell me things like "Make sure you have the card
set to PAL, not NTSC" and "have you tried another cable".
 
M

Mike Warren

Jan 1, 1970
0
Phil said:
The only way to measure it is to tweak a monitor's crystal so it DOES lock
-
then measure the frequency of oscillation with a counter.

Clearing enough space to disassemble the monitor is more work than I want to
do unless I have to, so I'll wait and see what Leadtek have to say first.
:)
 
P

Phil Allison

Jan 1, 1970
0
"Mike Warren"
Phil said:
Clearing enough space to disassemble the monitor is more work than I want
to
do unless I have to,

** You mis-comprehend.

I am NOT telling you what you to do - only pointing out what measurement
method would likely work.

so I'll wait and see what Leadtek have to say first.


** You already have your evidence.

The PCI video cards do not work with the monitors you need them to work
with.

Therefore, they are " unfit for the purpose intended " as the TPA says.

Even if you had a figure for the burst frequency error - it would not
prove you case any better than the evidence of people's eyes.



..... Phil
 
M

Mike Warren

Jan 1, 1970
0
Phil said:
** You already have your evidence.

The PCI video cards do not work with the monitors you need them to work
with.

Therefore, they are " unfit for the purpose intended " as the TPA says.

Even if you had a figure for the burst frequency error - it would not
prove
you case any better than the evidence of people's eyes.

At this stage I'm hoping that the error can be corrected by an update to the
firmware as I can't find any other cards with S-Video or composite outputs.
Everything has HDMI only. If I had another source, I'd not have even
bothered
going as far as I have.
 
M

Mike Warren

Jan 1, 1970
0
Mark said:
Do you see see more then one clock ref on that board, if not then the
H and V and color are phase locked.

This is a video graphics processor running (I guess) at somewhere
between 200 and 500MHz. There is only one external clock, a 27.0MHz
crystal. I expect that is only be being used as a reference for a
number of PLLs inside the chip, but I don't actually know anything
about what happens inside the chip.

It's all just guesses. The only thing I do /know/ for sure is that it
is not doing what it's supposed to, and another brand of card which
appears to have been built from the same reference design works fine.
If you can't measure the burst error, then measure the H error.

As I posted a few hours ago, I can not measure any difference in the
H-sync, but do see a very slight difference in the colour burst
frequency.
 
Top