PC as a standard convertor

[Kat Manton]

Hi,

I have a cunning plan...

I've been messing about with computers far too much recently; which is why I've been a bit quiet on here for a while. One of the things I've been playing around with is 'MythTV'; a free open-source PVR system which runs on top of Linux, turning a PC into a TiVo-like PVR and more.

More info on MythTV can be found here: http://www.mythtv.org/

I'm pretty impressed with it so far; it's not taken much effort to set up on top of the Gentoo Linux distribution (well, easy for a well-seasoned Linux geek like me at any rate) and, among other things, I've got off-air television (which can be paused), a programme guide, can play DVDs and can record programmes on the hard-drive for later viewing. That's on an old 600MHz PIII box I had lying around and a few odd bits of spare hardware I had anyway. I still need to pick up a digital terrestrial card that'll work with the system, but I spot these cards going for around £35 on eBay.

But, what's all this modern stuff got to do with vintage television..?

Well, while I was messing around getting it set up a thought struck me - it should be possible to get a 405-line signal out of it. For DVD, off-air television (analogue, digital terrestrial etc, depending on the card you shove in) and previously recorded programmes it interpolates and outputs the signal full-screen to whatever resolution you have X running in.

So, in theory, with a graphics card that'll manage the right pixel clock and a bit of effort working out an X modeline we could have a fairly straightforward way of watching off-air telly, DVDs etc on 405-line televisions.

I'm going to look into this some more; get my calculator out and spend a happy hour or two working out the magic numbers for X; then hopefully I should get a video signal out of the thing with the right timings.

One minor problem; rather remiss of me I know; I've not got anything which will display a 405-line signal. Would anyone be able to lend me a 405-line monitor at some point, once I've got something on my 'scope which looks right?

If I can get this going, I think I could produce a modified version of KnoppMyth - http://www.mysettopbox.tv/knoppmyth.html - the aim being an easy installation on a computer with the right bits in it and you get a source of 405-line content.

More later, I have a few things I need to get out of the way this next week but I should be back in Manton Laboratories in the fullness of time and post some 'scope waveforms if I get anything sane happening

Kat

[ppppenguin]

Kat

Unlike so many folks who have suggested using a PC for standards conversion you have the technical knowledge to do it if it's possible.

The main problem as I see it is getting a sensible 405 waveform. Ignoring DVI which isn't relevant here, the outputs of graphics cards are either RGB or, in some cases, PAL/NTSC. Unless the PAL/NTSC coder in the graphics chip is very programmable* I think you will have trouble getting a sensible 405 line waveform out of it. For monochrome you would use the Y part of an S-video output but it's quite likely that the line and field lengths will be "hardwired". The RGB outputs are more promising since they are usually highly programmable. A small amount of analogue circuitry would be needed to add syncs and generating composite 405 sync might be a problem.

A colleague has a lot of experience of getting graphics cards to run at funny resolutions. He once made a BBC-B work at 405 (I tried this years ago when I had one) and has made DVI outputs do almost impossible things. He reckons it was hard getting the DVI output to run at 720*576, 50Hz (CCIR 601) though this varied between different graphics cards.

I'm not saying it's impossible, just pointing at some of the likely problems and wishing you all good luck.

*Some cards may have a separate PAL/NTSC coder. David Robinson has succeeded in making the ADV7171 do 405 out (not on a graphics card) but it took a lot of ingenuity. I tried and failed even though I knew the chip pretty well.

[YC-156]

Please excuse my utter ignorance on most things regarding video, but I cannot help wondering: If my math isn't completely off, then the pixel rate of a 405 line video signal is 'only' around 5.5MHz.

With the speed of todays PCs it ought to be possible to simply synthesize the video signal in real time using a program combining brute force, data reuse, double buffering and table lookups. Then the video generation exercise is reduced to finding a way of converting this stream of 5.5 million bytes per second into an analog signal. That might mean convincing a video card into lending us its D/A converter via DMA in some fashion. Maybe another forum member has a better idea of accomplishing this.

Just a thought...

Frank N.

[oldeurope]

Originally Posted by YC-156

Please excuse my utter ignorance on most things regarding video, but I cannot help wondering: If my math isn't completely off, then the pixel rate of a 405 line video signal is 'only' around 5.5MHz.

With the speed of todays PCs it ought to be possible to simply synthesize the video signal in real time using a program combining brute force, data reuse, double buffering and table lookups. Then the video generation exercise is reduced to finding a way of converting this stream of 5.5 million bytes per second into an analog signal. That might mean convincing a video card into lending us its D/A converter via DMA in some fashion. Maybe another forum member has a better idea of accomplishing this.

Just a thought...

Frank N.

Yes I agree with Frank.
The question is who is first on the marked. The low cost aurora or the
PC solution?

[Station X]

Originally Posted by YC-156

Then the video generation exercise is reduced to finding a way of converting this stream of 5.5 million bytes per second into an analog signal.

I'm straying into an area I know nothing about here. Isn't 5.5 Mbit/s just the number of pixels per second? ie 405 X 405 X 4/3 X 25 = 5467500?

A pixel needs to specify brightness information. Not a problem with analogue, you just vary the amplitude, but with digital you'll need say 8 bits (just a guess) to specify the brightness level, so the PC or what have you will need to work at 40 Mbit/s.

It seems to me that when designing a convertor generating and synching the line and frame sync pulses from 625 to 405 is probably comparitively easy. A to D conversion is more difficult and generating say 2 lines of video from 3 is very difficult indeed.

Please feel free to shoot me down in flames.

[oldeurope]

The question is what do you want.
If you want to watch test patterns you need 3 MHz.
(TV set via RF input, not video monitor!)
And this is a standing picture.
If you want to watch moving images like films or so,
everything over 1,5MHz is enough.
The films we got as an xmas present from the BVWS are definately
below 1 MHz at 405. But like all old filmes they are noisy and this
is difficult for digital equipment and makes it a bit more noisy than before.

Kind regards
Darius

[jjl]

I have developed a commercial product that uses PC hardware and software to perform standards conversion, although in the case of the product I worked on, input was interlaced PAL, NTSC etc. video and output was progressive RGB.
The suggested approach ought to be feasible but the main issue to overcome is that, assuming the PC graphics card would be used for output, these cards really are not designed for outputting this type of video signal, the TV encoder devices fitted to some graphics cards are not programmable enough to allow 405 line output.
I think that the most promising approach would be to use one of the analogue RGB colour channels to output a composite 405 line video signal with this signal synthesized in software, hopefully this makes some sort of sense. Of course it would not be possible to display such a signal on a normal PC monitor so a 405 line set would need to be available when debugging.
As to performing interpolation etc. in software, I can state that this is most certainly possible and that more or less any PC built in the past 4 or 5 years has enough CPU horsepower and memory bandwidth to perform the computation required to capture modern standard definition video and convert it to 405 line output.

John

[YC-156]

Originally Posted by Station X

I'm straying into an area I know nothing about here. Isn't 5.5 Mbit/s just the number of pixels per second? ie 405 X 405 X 4/3 X 25 = 5467500?

I did say 5.5 million bytes per second, Ie. 44 Mbit/s. That is still not a problem for a modern PC, as each byte is processed in parallel within the PC. So the internal calculations will still 'only' need to happen at 5.5 million operations per second. In fact for many types of operations the PCs processor can handle 4 bytes at a time, 32 bits.

Once you have a frame calculated and ready in memory, blanking plus sync pulses and all, you push one byte (8 bits) at a time to a fast Digital to Analog (D/A) converter at a rate of around 5.5 million per second. This needs to be done all the while you calculate the next frame.

The only way I can think of this becoming cheap enough to compete with a device like the Aurora (£145) mentioned by Darius, is if one can reprogram an inexpensive, old graphics card, so that we can use its video capable DACs for our sinister purposes. I'm not even sure if the hardware on graphics cards allows us to use it this way.

Best regards

Frank N.

[Station X]

Originally Posted by YC-156

I did say 5.5 million bytes per second

Sorry Frank. So you did. I misread bytes as bits. Easily done when the value 5.5 million stays the same and only the units changes. With my telephony background I tend to think of bit streams being specified in Mbit/s rather than Mbyte/s

[Paul Stenning]

Originally Posted by YC-156

I did say 5.5 million bytes per second, Ie. 44 Mbit/s. That is still not a problem for a modern PC....

To put it into perspective, that's less than half the speed of a standard 100Mb Ethernet adaptor.

Since a modern PC can capture composite video with audio, compress it to MPEG or DIV-X, and write it to the hard drive in real time, I don't think performance would be an issue. We don't even need to save this to disk (which is the part that often causes video capture to choke).

Could the converted data be sent out on USB2, FireWire, Ethernet, SCSI or something and then converted to the required video signal by a fast D-A converter? This may be easier than trying to pursuade a graphics card to do something it really doesn't want to do?

[channel405]

I wish you all the best Kat with your innovations and look forward to hearing of your progress. However, as already pointed out, the difficult bit is getting the 405 data off the PC into a form that a telly can use. Graphic cards present their own problems, this is I why I suggested outputting via a USB2 route to a simple device based on a PIC chip controller and D/A. Firewire video tends to be compressed and would involve another layer of realtime processing.

[yagosaga]

Originally Posted by channel405

However, as already pointed out, the difficult bit is getting the 405 data off the PC into a form that a telly can use. Graphic cards present their own problems, this is I why I suggested outputting via a USB2 route to a simple device based on a PIC chip controller and D/A.

Gary Milliard has programmed a software converter for AVI videos into the NBTVA 32 line format. See: http://www4.tpgi.com.au/users/gmillard/nbtv.htm
Perhaps there might be only a small step for converting videos into 405 line format. Output via USB seems to be the most promising way.

[ppppenguin]

Originally Posted by Paul Stenning

Could the converted data be sent out on USB2, FireWire, Ethernet, SCSI or something and then converted to the required video signal by a fast D-A converter? This may be easier than trying to pursuade a graphics card to do something it really doesn't want to do?

This raises the classic problem of the computer industry's interpretation of "real time". In the world of television, real time means exactly that. Every pulse and pixel has to arrive at its rigidly fixed time. All the time. If this isn't the case you get, zits, glitches and bounces on your pictures. This is a far cry from the meaning of real time in the world of computing.

All the computer interfaces you mention can never provide this totally even uninterrupted flow of data. Although they can sustain the average data rate. Hence you need buffering and by the time you've done this and stuck a DAC on the end you have almost built an Aurora converter. The only time all this buffering etc becomes very simple and cheap is when an application is sufficiently widespread to attract a chip designed for the purpose. For example USB speakers where the USB interface, buffering and DAC are presumably all integrated into 1 or 2 chips.

32 line video is easily represented as audio and so should be relatively easy to handle with standard computer hardware.

[oldeurope]

Originally Posted by ppppenguin

...on the end you have almost built an Aurora converter. ...

From the pic I've seen, the low cost aurora is one IC and the modulator. Do you want to brake it into two peaces? hi hi.
The problem is you need the kwnolledge and equipment to programm it.
And you must be able to assamble this smd divice.

An other solution is to put the low cost aurora as is to the 625 vision output
of the computer. Than you have 405 from the computer.

Kind regards
Darius

[ppppenguin]

Looking at the low cost Aurora picture I think Darryl has chosen to put chips on both sides of the PCB. From what I know about it, I can't see all of it in the photo. I'm not sure why he has decided to push down the size. Surely it doesn't reduce the cost by much.

An other solution is to put the low cost aurora as is to the 625 vision output of the computer. Than you have 405 from the computer.

I agree 100% with Darius.

[Kat Manton]

Hi,

I'll just reply briefly to a few points; the problem regards the pixel clock on graphics cards is more of getting it to run slow enough for 405; I need to do more research on this but it looks likely there are cards in existence for which Linux drivers exist, and which ought to be able to be made to work.

The card I have (nVidia Geforce 4) has a driver which refuses to output anything other than 625 PAL and 525 NTSC from the S-Video output; but it will output anything you like and which the card can do from the RGB output. So it looks like the best approach is to combine R, G and B from the VGA output and add syncs. I've already done this partially to feed RGB and composite sync into the SCART input on my television, bypassing the PAL encode/decode process that happens using the S-Video input.

The rest of it's already there - the MythTV software already handles video capture, tuning, channel guide, and importantly interpolation, I can watch TV and DVDs fullscreen on 320x200, 720x576 interlaced PAL, up to 1600x1200 on a huge high-resolution monitor - it's just a question of working out the timings and writing an X modeline - and having a card which will run those timings. Once X is running in a particular resolution, MythTV will just use it.

So, all that needs to be done is to get X driving a video card at the resolution and timings we want, a little bit of analogue circuitry hung off the VGA connector, and the rest of it will just work.

So - I think it looks promising... I'll see what I can do.

If you'll excuse me, I'm not likely to be working on this for a little while having recently lost a very dear friend; but I'll get back into tinkering with technology at some point and let you know what happens.

All the best,

Kat

[YC-156]

Originally Posted by Paul Stenning

Could the converted data be sent out on USB2, FireWire, Ethernet, SCSI or something and then converted to the required video signal by a fast D-A converter? This may be easier than trying to pursuade a graphics card to do something it really doesn't want to do?

I tried looking into this a bit today, and while for instance USB 2.0 is plenty fast, there appears to be no generally available and sufficiently fast D/A devices out there. Most of the modern busses would be quite good enough, but we still need to come up with a generally available yet inexpensive device to use in the end. There almost certainly are some ££££ devices for realtime data acquisition and generation for, say, SCSI or simply PCI. But the price kills that idea.

Originally Posted by yagosaga

Gary Milliard has programmed a software converter for AVI videos into the NBTVA 32 line format. See: http://www4.tpgi.com.au/users/gmillard/nbtv.htm
Perhaps there might be only a small step for converting videos into 405 line format. Output via USB seems to be the most promising way.

I'm afraid that will not work, as the limiting factor for Gary's program is that he uses the generic audio interface inside modern PCs. That is enough to simulate mechanical TV standards, but audio cards are about a factor of 100 too slow for 405 line TV.

Does anyone have a link to a page, which describes the 405 video standard in detail, please?

Best regards

Frank N.

[tubesrule]

Jeff,
There are indeed two IC's on the bottom of the converter; the second modulator chip, and the image FLASH ROM. It was a tough decision to go double sided mounting, but the circuitry dictated it. By putting the second modulator chip directly under the first one, both output pins are positioned as close as possible to the RF connector/combiner circuitry. The FLASH is in a VSOP32 package, and it would have wasted a great amount of top side area to connect it to the VQ100 fpga. By placing it directly under the fpga, the pins on the FLASH practically connect directly to the pins on the fpga directly above it. These short interconnects also help with the high frequencies used in this converter. The FLASH bus will run as high as 36MHz, with internal frequencies of the fgpa reaching over 106MHz for the 819 line output. These high frequencies were required to oversample the output enough to reduce filter requirements.
Engineering for the converter is complete, and it will be possible to program a unit for the following standards (each unit can only support one standard and one RF System at a time):

PAL->405/25
PAL->441/25
PAL->455/25
PAL->625/25
PAL->819/25
NTSC->343/30
NTSC->441/30
NTSC->525/30

I have also already calculated tables for all System A, E, F, and early/late M frequencies. I can add System L if someone requires this. So far, 4 switch positions are dedicated to setting 15 channel frequncies plus modulator off, one switch is for equalization pulse enable/disable for standards that did not originally have eq pulses, and one switch for disabling the converter so the unit can be used as a stand alone modulator. In this mode, a sound trap filter is provided on the video input, but no additional video filtering is done.
That leaves two switch positions left for future options. If anyone has suggestions for these, possibly more channels, I welcome the feedback.

Darryl

[Courtney Louise]

I must admit to have looking at the settigs on my new graphics card. It has a custom mode (nvidia) which allows you to change scan freqency by degree and pretty much everything else about it. Refresh Frequency, positive or negative trigger, also there is a special version of DOs for arcade emulation so you can run an arcade monitor (some use odd rates) from a video card.

Andi

EIT : The card was an NVideo FX5600XT

[DoctorWho]

This sounds highly promising. This is the first time I've read a viable way of using a PC as a standards converter, this sounds superb!