Maker Pro
Maker Pro

Video grabber

G

Grumps

Jan 1, 1970
0
Hi
I'm looking at making a video grabber that will grab VGA video with a pixel
clock of up to 250MHz (approx).
I've used monolithic decoders in the past which work fine. But the frequency
does not go to 250MHz.
Before I start evaluating what I need to do, is anyone aware of a suitable
decoder that runs at this speed? Whether it's a monolithic part or even a
complete board, I don't mind.

And would I be correct in thinking that the solution to this is to have a
PLL/VCO to lock to the hsync in order to generate the sample/pixel clock.
Some sort of phase shift for the sample clock to get the centre of the
pixel. Then a few ADCs?
Thanks.
 
G

Grumps

Jan 1, 1970
0
Jan Panteltje said:
Why make a VGA video grabber?
In case of a PC grab the display buffer, in case of an FPGA the same.
If all else fails but a HD Canon camera and point it at the screen.

Thanks. All very good suggestions, but none are acceptable in this
situation.
It HAS to be from the VGA analog signal.
 
G

Grumps

Jan 1, 1970
0
Jan Panteltje said:
Anna Log has been dead for a few years :)

RIP poor Anna!
However, there are many users of Anna's VGA. And some of these want to frame
grab high res.
I can buy a grabber for $2k. But that's not a proposal I'd want to put to my
customer.
 
G

Grumps

Jan 1, 1970
0
Jan Panteltje said:
One thing you may consider, is if it is not a moving, but a static screen.
If it is static, you could just sample over a line, slowly moving from
line to line,
and rebuild the frame in memory.
That only needs simple H and V sync, and a variabel delay counter, fast
sampling
gate, and cheap ADC.

I wish life were that simple.
It's a moving image that we have to compress. The compression is ok.
Grabbing the image is currently the part that's giving me a headache!
 
G

Grumps

Jan 1, 1970
0
Just buy the 2k$ box. Assuming you're a consultant (you talk about
your customer), and you charge what we charge, 150$/hr, the 2k$ box
represents 13 billable hours.

We have to deliver several hundred of these, and a price of $2k + the rest
of the system is WAY too much.
Can you design a complete solution with gerbers and files ready to go
to manufacturing in less than two working days? With firmware and PC
front end?

Unlikely, even you're as good as me. :)
So, either you don't charge enough, or your customer or you are in
over their head.

You also talk about "monolithic decoders", which to me sound like
composite video-to-digital chips. All the hard work is already done in
the IC. It's not a great achievement to get those working.

No problem with the monolithic decoders. We've used them before and they
work well. But are limited to 170MHz (as far as I've found).
 
G

Grumps

Jan 1, 1970
0
Robert said:
There are video capture cards...

I've looked at many but have not found any (except expensive $2k units) that
will digitise a 250MHz VGA signal.
 
G

Grumps

Jan 1, 1970
0
Mike said:
Look at the chips they use as front-ends on LCD monitors.

That's an idea I've already had too. All of the monitors we have here are
only 1920x1200@60Hz which is about 150MHz pixel rate.
I'm going to have to get another monitor in and take it apart.
 
G

Grumps

Jan 1, 1970
0
Jan said:
Right, that is the correct way to think about it.
My old boss would say: 'I want to see the whole project,'
and then have me design a whole new system that does not need the VGA
to mpeg2 or whatever.

But these days I would buy a 150$ LCD monitor, a 600$ Canon HV20 HD
camera,
have wooden box made for 100$, charge the customer the 2k$, and have
1150$ profit for coffee, hehe
:)

I'm glad you put the smiley at the end. It would be my customer who had the
"hehe" right in my face if I proposed that solution.
 
M

Martin Brown

Jan 1, 1970
0
Grumps said:
I've looked at many but have not found any (except expensive $2k units) that
will digitise a 250MHz VGA signal.

If you can live with a low frame rate then one of the cheaper VGA2USB
cards might do the job eg.

http://www.epiphan.com/products/frame-grabbers/

Can't you specify another VGA card that has a digital output?

It seems to me like utter madness to redigitise a fast analogue video
stream that is originating from a computer.

Regards,
Martin Brown
 
G

Grumps

Jan 1, 1970
0
Martin said:
If you can live with a low frame rate then one of the cheaper VGA2USB
cards might do the job eg.

http://www.epiphan.com/products/frame-grabbers/

It'd have to be the VGA2USB-pro. Actually, even that has an issue. Epihpan
do have a real 2k x 2k unit for $8k.
Can't you specify another VGA card that has a digital output?

That was my first hope, but they say a categoric "no".
We can capture DVI at this resolution ok. Tried and tested.
It seems to me like utter madness to redigitise a fast analogue video
stream that is originating from a computer.

Well, if our end price is more than the cost of replacing VGA cards, then
maybe they'll think again.
But as it stands, it is a very arduous route for them to take to change
hardware/drivers etc.
 
G

Grumps

Jan 1, 1970
0
Jan said:
Your customer is an idiot anyways if he does want to digitise VGA,
so who cares.
And he would not know the difference!

Thanks. I'll forward your reply to him :)
 
N

Nico Coesel

Jan 1, 1970
0
Grumps said:
Hi
I'm looking at making a video grabber that will grab VGA video with a pixel
clock of up to 250MHz (approx).
I've used monolithic decoders in the past which work fine. But the frequency
does not go to 250MHz.
Before I start evaluating what I need to do, is anyone aware of a suitable
decoder that runs at this speed? Whether it's a monolithic part or even a
complete board, I don't mind.

And would I be correct in thinking that the solution to this is to have a
PLL/VCO to lock to the hsync in order to generate the sample/pixel clock.
Some sort of phase shift for the sample clock to get the centre of the
pixel. Then a few ADCs?

Analog Devices has several chips that do the A/D conversion and
synchronisation. These are intended for flat panels so they are pretty
cheap as well.
 
G

Grumps

Jan 1, 1970
0
Nico Coesel said:
Analog Devices has several chips that do the A/D conversion and
synchronisation. These are intended for flat panels so they are pretty
cheap as well.

That sound useful. Have you any links to the faster devices? >250MHz
Thanks.
 
G

Grumps

Jan 1, 1970
0
250 MHz "pixel clock"?
How does that compare to current refresh rates on televisions?

It's much faster than HD TV (1080p etc).
Or fancy computer monitors?

There are several that run a 4Mpixel display at 60Hz.
140 MHz ADC (12 at a time?) speed is fastest on the Analog Devices
page.
How does that translate to refresh rate?

http://www.analog.com/en/analog-to-digital-converters/video-decoders/products/index.html

ADV7403 12 140 10 NTSC; PAL; SECAM Composite; RGB; RGB SCART; Y/C;
YPbPr 10 & 20-bit YCbCr 4:2:2; 12-bit RGB DDR; 24-bit YCbCr/RGB 4:4:4;
8 & 16-bit YCbCr 4:2:2 12 550 100-Lead LQFP $21.14

The refresh rate on video is nowhere near that high is it?

It can be, yes.
120 Hertz refresh is fast for video isn't it?
Yes.

Isn't 250 Megahertz overkill?

That depends on the number of pixels as well as the refresh.
Are there video cards generating that fast of a signal?

Yes. Plenty can do over 400MHz.
Are there displays capable of changing that fast?

It's not a refresh rate issue in this respect.
I don't pretend to be proficient at this.

Are you?

I'm getting there!
1080p/120 = 1920x1080 at 120 Hz through a VGA connector?

Don't restrict yourself to 1080p. There are monitors that are twice that
resolution (and more).
Wouldn't such a signal be HDMI and not VGA?

Both. Customer has an installed base with mainly VGA.
2073600 pixels per frame

Half of what I need.
The 140 MHz A/D does 12 bits at a time. (12 input channels?)

It has an input mux. Only one (set of) ADC.
/12 = 172800 12 bit segments
The A/D can convert a frame 810 times per second..
But you want it to go almost twice that?

Check my math..

Don't / by 12. There's effectively only 1 ADC, not 12.
What kind of source and display does 1620 frames per second?

I must have miscalculated somewhere because I can't
believe that A/D would make these parts 800 times
faster than needed for the fastest TV's...

Where did I go wrong?

See above.
 
G

Grumps

Jan 1, 1970
0
This is WAY beyond somebody's bad ass CAD/CAM system isn't it?

I guess it could be used in this application. It's not our target though.
At that resolution is there a commercial frame grabber available?

Yes. Expensive. Even if we bought their technology, it's still too much.
If there is, aren't you in a rarified specialty market that
makes it reasonable to just PAY THE MAN for the product?

Our customer has already asked existing manufacturers for a part-custom
design. Then they came to us.
Installed base, at that resolution?

Motion picture animation or what?

It's not, and I can't say what the primary app is.
When we get it working, and our customer is satisfied, then we'll go for
other markets (I can think of two more that would find this useful).
 
T

Tim Williams

Jan 1, 1970
0
Greegor said:
250 MHz "pixel clock"?
How does that compare to current refresh rates on televisions?
Or fancy computer monitors?

My monitor is burning 1600 * 1200 * 85Hz right now. That's right around
163.2MHz pixel clock, not counting refresh times (maybe 10% more). Monitors
come even higher and faster, so a 250MHz clock is certainly concievable.

Incidentially, if it's an analog CRT, the video output needs a bandwidth at
least half the pixel clock. The scary part is they're doing it into the CRT
cathodes or grids, which require on the order of 100V full contrast. I
don't think even Tektronix ever even attempted that much bandwidth at that
voltage -- they manufactured special CRTs with higher deflection factors
instead!

Tim
 
T

Tim Williams

Jan 1, 1970
0
Jan Panteltje said:
The issue of driving a wehnelt at that frequency to control beam current
is very different from deflecting an electron beam with a given speed.

The capacitance might be slightly lower, but voltage is voltage. The
voltage required for modern CRTs is comparable to the deflection voltages on
old electrostatic CRTs (e.g. 5AQP1, et al.).

Looking at old datasheets, this TV tube:
http://www.mif.pg.gda.pl/homepages/frank/sheets/079/1/14RP4.pdf
shows a requirement of 30 to 100Vp-p drive over all conditions, and an input
capacitance of 5pF cathode or 6pF grid. It might be lower in modern CRTs.

This tube:
http://www.mif.pg.gda.pl/homepages/frank/sheets/041/5/5AQP1.pdf
shows typical conditions of about 10V/cm deflection sensitivity, requiring
about 100Vp-p across the ~10cm screen, and maximum 6.3pF per 'vertical'
deflection plate, so they're actually pretty close.

Since HV HF transistors were hard to come by, Tek built better tubes, with
deflection sensitivity on the order of 1V/cm vertical. You can drive a load
like that with little more than regular RF transistors. (Sure, the
impedance is lower, but by constructing a lumped transmission line, the
impedance is flatter, so it's not a big deal.)

I guess CRT transconductances never got beyond the microsiemens, so they
still had to deal with big voltages, hence the HV HF transistors which are
so convieniently cheap today.

Tim
 
N

Nobody

Jan 1, 1970
0
250 MHz "pixel clock"?
How does that compare to current refresh rates on televisions?
Or fancy computer monitors?

To give a single data point, my monitor runs at:

(II) CHROME(0): Modeline "1920x1440"x0.0 297.00 1920 2064 2288 2640 1440 1441 1444 1500 -hsync +vsync (112.5 kHz)

Translation:

1920x1440 screen resolution (displayed pixels)
297.0MHz pixel clock
Horizontal:
1920 pixels displayed
2064 pixels from hblank-end to hsync-start
2228 pixels from hblank-end to hsync-end
2640 pixels in total (including hsync and hblank)
Vertical:
1440 lines displayed
1441 lines from vblank-end to vsync-start
1444 lines from vblank-end to vsync-end
1500 lines in total (including vsync and vblank)
Negative hsync, Positive vsync
112.5 kHz horizontal frequency (line rate)

Calculating 112.5e3 / 1500 reveals that the vertical frequency (frame
rate) is 75Hz, hence 2640 * 1500 * 75 = 297MHz pixel clock.
 
N

Nico Coesel

Jan 1, 1970
0
Grumps said:
That sound useful. Have you any links to the faster devices? >250MHz
Thanks.

Use 2 devices with the second device using an inverted clock.
 
G

Grumps

Jan 1, 1970
0
Nico Coesel said:
Use 2 devices with the second device using an inverted clock.

Thanks. I'm evaluating that idea at the moment with a couple of NXP devices.
Unfortunately these devices are now obsolete (they were triple ADCs at
270MHz). Next best is a set of 500MHz single 8-bit ADCs.
 
Top