PC as a standard convertor

[Courtney Louise]

If windows users with NVidia cards with GForce4 or higher look deep enough into their settings, they already have most of what powerstrip offers as standard...

Andi

[Steve]

Originally Posted by Kat Manton

Hi,
The card I'm using is an MSI card which uses the nVidia GeForce 4 MX4000 chip. This card supports interlaced modes, I bought it (for £17.71 including postage off eBay) to replace a GeForce2 card which doesn't support interlaced modes.

Hi Kat,

My mythbox currently has a "rivafb: nVidia device/chipset 10DE0150
rivafb: nVidia Corporation NV15 [GeForce2 GTS/Pro]" fitted. This has got a composite out on it, so is this likely to support interlaced modes?

Cheers,

Steve.

[jjl]

Hello Steve

The composite output would be provided by a TV encoder chip on your card, this is intended to provide PAL, NTSC etc. output from either progressive or interlaced video output by the graphics processor. It is unlikely that the TV encoder would support 405 line output, I haven't come across any that do. Having said that, it is worth a try, but please follow Kat's warning and don't connect up your highly prized 405 line set before looking at the output waveform with a 'scope.

John

[Kat Manton]

Hi Steve

Originally Posted by Steve

My mythbox currently has a "rivafb: nVidia device/chipset 10DE0150
rivafb: nVidia Corporation NV15 [GeForce2 GTS/Pro]" fitted. This has got a composite out on it, so is this likely to support interlaced modes?

It's the graphics processor and driver which determine whether interlaced modes are supported, my GeForce2 card with the nVidia 1.0-6629 driver release did not support interlaced modes; the same driver with a GeForce4 card does.

So it's likely that your card doesn't support interlaced modes - if you're using it in a MythTV box on 625 line you probably have 288 line progressive at present.

Have a look through the X log file (probably /var/log/Xorg.0.log or /var/log/XFree86.0.log.) This should contain information on whether the GPU supports interlacing, the range of the dot-clock generator, etc., along with some information on the calculated timings for whatever display modes you're using.

The card won't produce non-standard modes out of the composite or S-Video outputs; as John says this comes via a TV encoder chip which handles 625/50 PAL and 525/60 NTSC. The nVidia driver refuses to run anything other than 'standard' resolutions/timings on either S-Video out or composite out anyway; X will exit with an error if you try. This is why I'm doing this via the RGB+sync output; if it worked via S-Video or composite I'd be using it as it'd be less complicated and saves some effort

Regards, Kat

[Kat Manton]

Hi,

A little more progress; I've lashed up a little circuit to sum R, G and B from the output of the graphics card, here's the result of displaying colourbars on the computer. Y is 0.2V/cm, the output's a little high at 0.8V p-p. X is on 'expanded' locked to frame-sync so I could get the photo, the horizontal timebase could be anything with this old 'scope, but I already know the computer's generating correct timings anyway. The little 'glitches' before and after the greyscale are a 1-pixel wide grey border around the full-scren window displaying the colourbars, by the way.

Now, just add syncs...

Regards, Kat

[oldeurope]

Originally Posted by Kat Manton

Hi,

A little more progress; I've lashed up a little circuit to sum R, G and B from the output of the graphics card, here's the result of displaying colourbars on the computer. Y is 0.2V/cm, the output's a little high at 0.8V p-p. X is on 'expanded' locked to frame-sync so I could get the photo, the horizontal timebase could be anything with this old 'scope, but I already know the computer's generating correct timings anyway. The little 'glitches' before and after the greyscale are a 1-pixel wide grey border around the full-scren window displaying the colourbars, by the way.

Now, just add syncs...

Regards, Kat

Hi Kat, some things are looking strange to me.
First the line must start with white and end with black.
Making a black and white picture from a colour bar normally results in
different step size.
Can not say something about the level, need more details.

Kind regards
Darius

[ppppenguin]

Originally Posted by oldeurope

First the line must start with white and end with black.
Making a black and white picture from a colour bar normally results in
different step size.

By convention, colour bars have white on the left and black on the right. This is not essential! Looking at the levels of the individual bars I suspect Kat has used 3 equal value resistors to sum the RGB. Luminance is the weighted sum of RGB:

Y = 0.299R + 0.587G + 0114B

Hence the resistor values for R, G, B should be proportional to the reciprocals of the coefficients. Approximately 3.3 : 1.7 : 8.8. Don't worry too much about the exact values. Just scale to whatever is convenient.

In the final arrangment I would strongly recommend a buffer amplifier after the resistor summing network. There are many low cost video opamps, some of which are happy on a single +5V supply. For example Analog Devices AD8052. I'm sure that Darius would design something using a couple of transistors.

[oldeurope]

Originally Posted by ppppenguin

I'm sure that Darius would design something using a couple of transistors.

Yes!
And make sure that the sync tips are at 0V because this makes the
modulator very easy. (pnp emitterfollower makes biasing for the npn
modulator transistor.)

Kind regards
Darius

[Kat Manton]

Oops...

Somewhere I'd got it into my head that colourbars in monochrome == greyscale with even steps... It was 2am...

I'm more at home with programming and digital electronics; I've done very little with video. Still, it's good fun and I'm learning

Apart from horizontally flipping my colourbars image, I've tweaked the resistors a little; (they weren't equal - if they were the red, green and blue bars would be the same level.) They're in the correct ratios now so this ought to look right to those who know what looks right.

Regards, Kat

[ppppenguin]

Originally Posted by Kat Manton

they weren't equal - if they were the red, green and blue bars would be the same level

Apologies. I applied fingers to keyboard without properly engaging brain. Mea culpa.

Now looks about right. Actually in this application it really doesn't matter within reason. I have designed equipment for broadcast use where I was using carefully calculated 0.1% tolerance resistors to ensure that everything was very accurate indeed when converting RGB to YPbPr. In your application you could forget about the blue completely and it would hardly notice on real pictures. Even if you just used the green and called it luminance it would probably look OK.

[Kat Manton]

Originally Posted by ppppenguin

Now looks about right.

Ahh, excellent

Actually in this application it really doesn't matter within reason. I have designed equipment for broadcast use where I was using carefully calculated 0.1% tolerance resistors to ensure that everything was very accurate indeed when converting RGB to YPbPr. In your application you could forget about the blue completely and it would hardly notice on real pictures. Even if you just used the green and called it luminance it would probably look OK.

I'd thought of doing just that but I can be something of a perfectionist - if Y is defined as 0.299R + 0.587G + 0.114B then so it shall be with my circuit. Also missing off colours might be ok for off-air pictures, but bits of the software's TV programme guide and menus might disappear or become less readable if I missed off a colour or two. I'm possibly going to have to re-skin it anyway once I've seen it in monochrome.

Regards, Kat

[ppppenguin]

Originally Posted by Kat Manton

I'd thought of doing just that but I can be something of a perfectionist - if Y is defined as 0.299R + 0.587G + 0.114B then so it shall be with my circuit. Also missing off colours might be ok for off-air pictures, but bits of the software's TV programme guide and menus might disappear or become less readable if I missed off a colour or two. I'm possibly going to have to re-skin it anyway once I've seen it in monochrome.

Get it within 10% and that will be absolutely fine. Even with my reviewers hat on* I would be most unlikely to spot anything wrong.

*I have reviewed all the 405 line converters that have ever been commercially available for the BVWS Bulletin or 405 Alive. I hope to be reviewing Darryl's new design soon. I may have a go at reviewing KAT405 but strictly speaking this is not a commerically available product. It may need a postscript to the article I wrote about the history of 405 converters a few years ago. I'm happy to say that this passage may need updating:

Some have suggested using computers or digital signal processors (DSP). Conceptually there is no problem. I’m a pretty lousy programmer and I could write you the conversion algorithms in a few lines of a high level language such as BASIC. These would take a 625 line image, already in the computer, and convert it to a 405 line image, also within the computer. And there’s the snag. You still need to get the picture in and out and you have to ensure that the computer can keep up with the data. Video comes at you continuously so real time means exactly that – you cannot put your hand up, take time out and catch up later. So when you say you have built a converter with a cast off 486 PC I won’t believe you. Modern DSP chips are entirely capable of doing the digital parts of the processing though you will still need to get the signals in and out. Maybe I'm prejudiced but I would rather use programmable logic than DSP for this job.

[ppppenguin]

In case anyone's curious, here's the text (without pictures) of my 2002 article. I hope most of it is still valid though I'm feeling an urge to write a postscript. I look forward to adding Kat (and Darius and Darryl) to the hall of fame.

[oldeurope]

Hi Kat, answering your PM...
Hope it is helpfull for you.

Kind regards
Darius

[Kat Manton]

Hi Darius,

Originally Posted by oldeurope

Hi Kat, answering your PM...
Hope it is helpfull for you.

Thanks very much; I'll rebuild things (again...)

For what it's worth, here's the present state of evolution of the mess I've been working on myself; it started off just suming with resistors, then the followers appeared, then the common load resistor turned into a current source, then I cobbled together a mess of DC restoration and sync circuitry. Then gave up after seeing what came out of it and started thinking about video opamps...

http://www.technogoths.demon.co.uk/temp/RGB-conv.png

It sort-of works; the sum of RGB is spot-on, the syncs are reasonable, but the black level wanders about a bit...

I'll freely admit I'm not too good with these new-fangled 3-legged sandy fuses

I'm a lot more at home with microprocessors and other digital stuff, programming and audio as long as I can use either valves or opamps...

I guess I ought to stick to what I'm good at, and avoid video circuitry...?

Regards, Kat

[ppppenguin]

Ideally this sort of circuitry is DC coupled but the wandering black level caused by AC coupling is harmless. A clamp or DC restorer in the monitor or modulator sorts out the black level. In most professional video circuits it's usual to use +ve and -ve power supplies. This makes it easy to use DC coupling and sit the black level at 0V. Industrial and domestic video circuits don't usually have a -ve supply and either end up AC coupled or with the black level above 0V. Neither of these points matters in these applications.

[oldeurope]

Good evening

Hi Jeffrey, I often read in your statements that you like black level at 0V.
Why?
Sync tips at 0V makes sense to me because of the easy modulator biasing.

Hi Kat,
thanks for your schematic.
Lets start at the input. R12/13/14 must be 75 Ohms.
In this case you get 0.7 or 1Vpp. If it is 0.7Volts you don't need the
R? in my schematic.
At C4 the signal has an impedance of ca.400 Ohms. Than the clamp Diode
followes. This causes that the black level changes depending in the bright
parts of the picture. C11g 1uF is the equivalent in my schematic to your C4.
Q311 does the same job as your diode D1. But note it is only switched on
during the sync tip. This makes sure that the black vision components
are untouched during the line and there is no change in the black level.
R11f 470Ohms is like your R16 4K7. The value of C4 is much to high and
the R16 is so high that you get losses in resolution.
Your output is 0 Ohms and TR5 will die if there is a short at the output.
Your R6 does not make a 75 Ohms output. Please note in my schematic
you have a collector output!!!
The simple serration pulse generator works, I use this in my modified camera.
In flywheel TVs it may generate the first few lines a little bit moved to the
right position.

Kind regards
Darius

[Kat Manton]

Hi Darius,

Thanks for your comments, I'll take another look at both circuits again and see what I can figure out.

I should have noted, the 3 video inputs are 0.7V p-p video only, no sync present on any of them.

The output of my 74LS86 XOR gate is giving TTL-level composite sync, with serrations of a kind - they're at line-rate and are actually inverted V-sync pulses as these run all the time from the computer. The video signal is supposed to have around a 0.9V DC offset by the time it arrives at TR5 base, TR6 was then bodged on as an attempt to get the syncs added.

Also, what's on the diagram is partly due to what I've got in my junkbox; I've not been out to buy more components for this yet so some values are a bit of a compromise - also, the output has only ever been connected to a 'scope; I don't count it as 'finished' yet; and it was missing a resistor on the diagram. I've just amended it a bit to reflect what I've actually got hanging off a blank bit of copperclad.

Jeffrey, I know what you mean about split supplies; i usually run my audio circuits off split 15V or 18V supplies; and of course AC coupling doesn't matter much. With this thing I was trying to get it to run off a handy 9V wall-wart; split supplies would make things a bit easier though

Regards, Kat

[ppppenguin]

Originally Posted by oldeurope

Hi Jeffrey, I often read in your statements that you like black level at 0V.
Why?
Sync tips at 0V makes sense to me because of the easy modulator biasing.

No equipment should ever rely on black level at a particular fixed voltage. In the broadcast and professional worlds, black level should be at 0V. When switching between video sources this prevents sudden changes in black level which can be difficult for the clamps in downstream equipment. Except for some video routers, all video inputs have clamps or DC restorers.

I've split the supply from a wall wart on occasion. In one design I regulated the wall wart output to 12V and split it to +5V and -7V with a 5V regulator. In this configuration you need to make sure that the regulator always sources current - they can't sink current - and I have used a ballast resistor in at least one design. Professional audio usually needs at least +/-12V, often up to +/-18V, to allow +18dBm or more.

[ppppenguin]

This is a very rough sketch of how I would implement the RGBS combiner in a professional design. I haven't bothered to calculate any component values nor have I checked to see if it is feasible with a single 5V supply. It will be perfectly happy with +5V and anything from -3V to -5V. The opamp type is not critical - any video opamp will do but remember that not all are suitable for +5V only.