PC as a standard convertor

[Paul Stenning]

As others have said, if anyone can do it, Kat's the person!

[DoctorWho]

I totally agree. I really hope Kat can achieve this. I know that many of us would be very grateful.

[ppppenguin]

There's a reasonable chance of getting a graphics card to output RGB analogue 405. Some may go down to such low scan rates, others will not.

There are 2 further problems that I can see. The first is the sheer inconvenience of having your PC's output at 405! This might be solved by having 2 graphics cards in the PC or a dual head graphics card. The more interesting problem is producing decent 405 line mixed sync. Graphics cards generate separate H and V sync. While these could be simply added to the video the result would not have proper vertical sync serrations and the sync pulse widths might well not be correct.

I'm not trying to put anyone off the project. Honestly!

[Kat Manton]

Originally Posted by ppppenguin

There's a reasonable chance of getting a graphics card to output RGB analogue 405. Some may go down to such low scan rates, others will not.

Indeed, that's my findings with the preliminary research.

Worst-case I might have to swap a crystal or oscillator module on a graphics card and maybe hack the driver code; but I'm not sure it'll come to that.

There are 2 further problems that I can see. The first is the sheer inconvenience of having your PC's output at 405! This might be solved by having 2 graphics cards in the PC or a dual head graphics card. The more interesting problem is producing decent 405 line mixed sync. Graphics cards generate separate H and V sync. While these could be simply added to the video the result would not have proper vertical sync serrations and the sync pulse widths might well not be correct.

I'm not trying to put anyone off the project. Honestly!

It's possible to dual-boot Windows and Linux, though this takes some setting up. MythTV works best if you can dedicate a spare PC to it; it then basically becomes a (large and unusually-shaped) "set-top box", controlled from an IR remote; rather than a general-purpose computer.

My modern TV is working happily enough with a 2-transistor circuit combining H and V syncs, which is actually powered from the TTL sync signals from the VGA card, having told the driver to generate negative-going H and V sync pulses for both. That's just to produce composite sync; I'm feeding RGB + sync into the SCART on said TV. Addition of R G and B signals in the correct proportions then adding syncs isn't exactly rocket science, at least in this case where we're after a monochrome signal anyway.

The neat thing about doing all this with Linux is the greater flexibility available in configuration; it's possible to specify video timings and sync polarity etc in a detailed and fine-grained way; also the software's all open-source so if anything needs a tweak I can tweak it and redistribute the code. And as mentioned, pretty much all the software exists already and doesn't need modifying - once the graphics system is producing a 405-line signal the rest of it will just scale itself to fit the resolution the display's running in.

MythTV is quite a scalable system, it's possible to run anything from one PC with a single tuner card in it, with the display attached to that, right up to to a distributed networked system with multiple backend machines containing several tuner cards each, with multiple frontend machines connected to different display devices. If I get this working I'd probably set up a combined backend/frontend machine on my main large colour television and add another frontend-only machine set up specifically for 405-line.

The eventual aim would be to produce a CD allowing easy installation of the software on a PC containing appropriate hardware - it's not necessarily going to work with every graphics card and certainly won't work with every TV tuner card as there aren't Linux drivers for a lot of them.

But it's looking promising, and potentially cheap, even if it's necessary to purchase second-hand another PC along with specific tuner and graphics cards. It doesn't take the latest high-spec machines to run it; my current experimental system is an old 600MHz Pentium III I was given, and that's more than powerful enough.

As they say, watch this space

Regards, Kat

[Nickthedentist]

Originally Posted by Kat Manton

It's possible to dual-boot Windows and Linux, though this takes some setting up.

The most foolproof way of doing this (as far as I'm concerned) that I've seen advocated is to use 2 separate HDDs in caddies, one loaded with Windows and the other with Linux. Thus, no clever computer knowledge is required, you just insert the relevant drive before powering up, depending on which system you want.

Nick.

[YC-156]

Originally Posted by Nickthedentist

The most foolproof way of doing this (as far as I'm concerned) that I've seen advocated is to use 2 separate HDDs in caddies, one loaded with Windows and the other with Linux. Thus, no clever computer knowledge is required, you just insert the relevant drive before powering up, depending on which system you want.

Even better would making your own single CD distribution, not unlike Knoppix. When you want your line converter, stuff CD in CD drive and reboot. The distribution could include MythTV and associated tools and drivers already set up, perhaps needing a bit of tweaking to parameters. Add a bit of spare space on an existing harddisk, and you wouldn't even need that.

There are tools out there making a custom bootable CD quite easy to configure and maintain. This basically means that if end users has supported hardware, their required Linux knowledge to install and use the system will be very close to zero.

People always drops their jaws, once you boot their PC using the latest Knoppix CD when they have their backs turned.

"Hey Joe, didn't know you had started playing with Linux."
"Uhm, I didn't..."


Best regards

Frank N.

[DoctorWho]

This is truly fascinating. Presumably if it works as planned you would then feed the signal to a modulator and away you go?

Peter.

[Kat Manton]

Hi,

I can report that after spending the best part of today on this little project I seem to be making good progress

Several hours spent with the System A specification, a calculator and several sheets of paper resulted in an X modeline that seemed promising. So, I added it to the X configuration file then fired up the system.

After starting the system up, I checked the X log file and found the following information reported:

(**) NVIDIA(0): Validated modes for display device CRT-0:
(**) NVIDIA(0):      Mode "SystemA": 12.1 MHz, 10.1 kHz, 50.0 Hz (I)
(II) NVIDIA(0): Virtual screen size determined to be 968 x 378
(**) NVIDIA(0): Display dimensions: (200, 150) mm
(**) NVIDIA(0): DPI set to (122, 64)

Note the slightly 'odd' resolution of 968 x 378 - the graphics card won't run a dot-clock lower than 12MHz so in order to get the rest of the timings right I've ended up with 978 pixels per line. This shouldn't be an issue; specifying the display size as 200 x 150mm 'tells' X that the display is 4:3 and the TV software will scale the picture to fit anyway. The unavoidably high dot-clock does mean the video bandwidth is higher than I'd like. I expect this will manifest as horizontal blurring of the programme guide and MythTV menus etc., but it probably isn't going to be an issue with off-air programme material. I'm intending to investigate alternate video cards; one which supports a lower dot-clock than 12MHz would be more appropriate.

Anyway, I checked the output from the graphics card with my 'scope. The seperate horizontal and vertical TTL-level sync outputs have the timings and polarity I specified in the configuration. Launching the TV software 'blind' (no display connected at this stage) and checking R, G and B outputs with the 'scope I can see that, as expected, the TV software appears to be interpolating and scaling an off-air picture to full-screen.

Success, then; I have what looks promisingly like a System A signal coming out the back of my computer.

Next, it's time to knock up a circuit to combine syncs, along with R G and B video in the correct proportions to get composite monochrome video. Once that's done I need to build a modulator.

Once I've got something resembling composite video I'll try to take some photographs of various waveforms as displayed on my 'scope and post them here.

Cheers,

Kat

[jjl]

Nice work Kat.
The numbers in your modeline sound a bit odd, could you let us know what the number of active i.e. visible lines is supposed to be in a 405 frame?
Have you managed to get your NVIDIA card to output interlaced video?
You may have noticed that NVIDIA cards seem to need horizontal timing specified in multiples of 8 pixel clocks, not sure whether this applies to ATi and others.
I have an idea that I mentioned in an earlier post that might allow a single colour channel output (R, G or B) from a graphics card to be used to output 405 line composite video thus requiring minimal external hardware for dealing with syncs etc. I'll try to investigate this when I have enough spare moments.


John

[ppppenguin]

Kat, sounds like you're making really good progress.

Originally Posted by jjl

You may have noticed that NVIDIA cards seem to need horizontal timing specified in multiples of 8 pixel clocks, not sure whether this applies to ATi and others.

I'm pretty sure this is true for Matrox cards such as the G550.

[oldeurope]

Originally Posted by Kat Manton

Hi,



Success, then; I have what looks promisingly like a System A signal coming out the back of my computer.

Next, it's time to knock up a circuit to combine syncs, along with R G and B video in the correct proportions to get composite monochrome video. Once that's done I need to build a modulator.


Kat

Hi Kat, I'd be happy to help you with the analog hardware if you want.

Kind regards
Darius

[Steve]

I started playing around with the idea of generating 405 under linux about 3 years ago but as I knew even less about video than I do now (which still isn't alot) I quickly got distracted playing around with other things and never revisited the project.

If you do get it working, it might be worth looking at using the Microsoft XBOX games console as a cheap source of standard hardware. The XBOX already has composite and RGB outputs available, built in network card and DVD drive. I use one as the front end for my myth TV setup and can't fault the video quality. As these boxes are available for around 70 pounds they make a cheap, convenient platform for running linux on.

Happy to give any configs a whiz on my xbox to see if it's viable.....

[Kat Manton]

Hi,

I thought I'd post the results of some measurements and calculations after further refinement of the System A X Modeline.

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. The pixel-clock range of this card can be specified in increments of 10kHz from 12MHz to 350MHz (determined by experiment), and as was suggested earlier in the thread, this is one of the cards which needs horizontal timings specifying in multiples of 8 pixel clocks. However, in spite of having to make a few approximations, I think I've got pretty close!

My source of System A timing information was http://www.pembers.freeserve.co.uk/W...Standards.html

The results of my calculations give the following; I'm compiling a .pdf document showing how I arrived at this which I'll post here later.

• Pixel-clock: 12.07MHz
• Vertical resolution: 377 pixels
• Horizontal resolution: 968 pixels

The modeline I've ended up with in my /etc/X11/xorg.conf file is as follows:

Modeline "SystemA" 12.07  968 992 1104 1192  377 377 385 405  -hsync -vsync interlace

This gives the following results:

• Line rate: 10.1279kHz (measured - 2.9Hz high - spec is 10.125kHz)
• Frame period: 20.00ms (measured - within spec - spec is 20.0ms)
• Active line period: 80.2us (calculated - 0.1us low - spec is 80.3us)
• Horizontal front porch: 1.99us (calculated - 0.29us high - spec is 1.7us)
• Horizontal sync pulse length: 9.3us (calculated - within spec - spec is 8-10us)
• Active field period: 188.5 lines (as spec - 377 active lines/frame)
• Field sync duration: 4 lines (as spec and confirmed by observation)

By twiddling with the trigger level control on my Gould OS250 I can get it to lock to alternate field sync pulses with the line sync on the other trace - and indeed, the field sync duration is correct and we definitely have an interlaced signal.

Apart from the slight timing errors noted above, about the only thing "wrong" is that the VBI won't contain the double-line-rate serrations as per the specification. However, if I XOR the vertical and horizontal syncs to get composite there will be line-rate serrations, basically horizontal sync pulses inverted during the VBI.

I suspect all this is close enough to work.

Now I'm at the stage of having near-correct System A timings; I'm happy enough with it to feel there's some point in building the electronics to get RGB+syncs combined into monochrome composite. I'll try to photograph some waveforms once I've got that far

Regards, Kat

[Courtney Louise]

Kat,

That is amazing work and results. I wonder if it would work on diffrent chipsets like ATI for example. I would pretty much imagine that anything higher than the GForce etc would handle the regular NVIDIA commands as they all use the same driver pack, and the higher cards just seem to add features rather than take them away. It would also be economical if this were to work to be able to build/scrounge a whole pc of a uitable standard to do this at little to no cost, in fact several machines that I think would have done this have been available on freecycle recently.

You deserve a few for all this work and even tho many have said it, keep it up

Regards

Andi

[Kat Manton]

Originally Posted by Jarl Ayari

That is amazing work and results.

Thanks - it's also good fun, though several friends now think I've 'lost it' completely... "You're trying to get *what* out of a PC?!?"

I wonder if it would work on diffrent chipsets like ATI for example.

That depends entirely on the range of possible dot-clocks available; I think there is an ATI card which will run down to 8MHz, which might be more appropriate, but I'm not sure how small the 'steps' in dot-clock are. I'd have to get one and play with it, I think. I suspect it'd need a completely new modeline calculating after determining the specification of the card.

I would pretty much imagine that anything higher than the GForce etc would handle the regular NVIDIA commands as they all use the same driver pack, and the higher cards just seem to add features rather than take them away.

I'm pretty sure about that myself; I have a lot of generic documentation for nVidia cards here and it looks like GeForce4 and higher-spec cards should work with the modeline I've calculated above.

It would also be economical if this were to work to be able to build/scrounge a whole pc of a uitable standard to do this at little to no cost, in fact several machines that I think would have done this have been available on freecycle recently.

The MythTV system is set up assuming it's the sole system running 24/7 on a dedicated PC; you don't need a keyboard or mouse on it as it's controlled through an IR remote, and it retrieves programme guide information during early morning - so it's best to dedicate a PC to this and leave it going.
My experimental system is a 600MHz Pentium III system I was given, it currently has a 40G hard disk and 256M RAM. With the addition of the GeForce4 card off eBay I've spent less than £20 on it. I still need to add a digital terrestrial card (around £35 off eBay - must be a Hauppauge Nova-T as this is well-supported) I'm currently using a very basic 'frame-grabber' analogue terrestrial card which is less than satisfactory; the machine hasn't enough processing power to encode captured video, buffer it to disk, then decode it in real-time at a decent resolution (it does this as it's possible to pause and rewind live tv etc) But, the DVB-T card produces an mpeg stream so no encoding is necessary, meaning the system is more than powerful enough as it only then has to buffer to disk and decode/display - and the GPU on the GeForce4 card does a lot of the hard work.

You deserve a few for all this work and even tho many have said it, keep it up

heh, thanks I'm not stopping now; this is looking too promising... I can't wait to see it running - and I've not even got a 405-line TV to try it with yet I think I understand what my friends were thinking...

Regards, Kat

[ppppenguin]

Again, congratulations to Kat!

All we need now is a member within reasonable distance of Leeds who has a 405 monitor. Alternatively a modulator and 405 receiver would do though I would always prefer a monitor for experimental work like this. Some monitors also have separate sync inputs (composite, not separate H and V) which would make the tests even simpler as you not need to combine video and sync.

Just a couple of resistors to mix the H sync and video together would be enough to prove the picture and horizontal lock and if you XOR the H and V syncs as Kat suggested you should get a locked picture with no more than an XOR gate and a few resistors. An ideal arrangement would have a properly buffered output but who cares for initial trials.

[oldeurope]

Hi,
sorry I have very low kwnollege about computers and software.
What is neccessary?
Is it an additional graphic card and a CDROM with your software
and I get (RGBHV) 405 at the output?

Kind regards
Darius

[Kat Manton]

Hi Darius,

Originally Posted by Darius

sorry I have very low kwnollege about computers and software.
What is neccessary?
Is it an additional graphic card and a CDROM with your software
and I get (RGBHV) 405 at the output?

At present, I'm using a PC with just one graphics card in it - I have several machines here, networked; it is easy enough for me to administer and configure the MythTV machine remotely from another machine as it's all a bit experimental and I keep changing things.

My experimental system is as follows - but I do not recommend that anyone try this unless they have fairly extensive Linux experience. I have installed the Gentoo Linux distribution http://www.gentoo.org/ on it, then in turn installed MythTV http://www.mythtv.org/ on that using Gentoo's 'portage' package management system. This isn't for the faint-hearted though; it takes a couple of days just to install and configure Gentoo Linux, never mind getting MythTV working on it - but I've done this for my experimental system I need some degree of flexibility and I know my way around Gentoo as I use it on several other machines here. I would not recommend this approach to a novice, though.

But, eventually, my intention is to either produce another distribution based on 'KnoppMyth' (see http://www.mysettopbox.tv/knoppmyth.html and http://knoppmythwiki.org/) with all the necessary configuration changes for 405-line operation (and maybe other standards as well.) Alternatively, I may produce a website with enough detailed information for someone to modify a basic KnoppMyth installation for 405-line work, along with a list of tested hardware configurations that are known to work correctly.

Once this is done, all that is necessary is a PC of suitable specification containing a compatible TV capture card (analogue, digital terrestrial, cable or satellite); compatible sound card; a graphics card tested and known to be capable of producing correct 405-line video; and a small interface to convert RGB+sync to composite.

Once you've got a suitable PC together, the idea is that you then install KnoppMyth (or possibly a modified version should I decide to do this) from a single CD - the stated aim of the KnoppMyth folk is to make this easy. Next, you configure a few things (you need to tune the TV card in, etc) and it is done. The plan is to provide a means of installation that is easy and straightforward and doesn't demand advanced Unix knowledge

You then end up with essentially a dedicated "HTPC" (Home Theatre PC) tailored for 405-line television. The PC boots up straight into the MythTV system with the display and user interface on the television, and controlled entirely through an IR remote. No keyboard or mouse needed; no second graphics card or monitor needed. Think of it as being something like a set-top box, only bigger and a different shape

This approach may not be appropriate for some; it's best to dedicate a PC for the job and leave it running; and it's likely there will some 'mucking around with Linux' involved in getting it to work. Not a problem for me, I've been at it a few years now, and it seemed like a fun way to get off-air 405-line TV to me. But for others there are always standalone standards convertors

Regards, Kat.

[ppppenguin]

For most of us analogue hardware types I wonder if there's any way to do a quick reconfigure of the graphics card to get a taste before sorting out everything else. Don't really care what information is on the screeen provided it's 405!

[Kat Manton]

Originally Posted by ppppenguin

For most of us analogue hardware types I wonder if there's any way to do a quick reconfigure of the graphics card to get a taste before sorting out everything else. Don't really care what information is on the screeen provided it's 405!

If you've already got a Linux system with an nVidia GeForce4 card (or newer) in it and you are using the closed-source binary drivers from nVidia, then the X modeline I've calculated ought to work, you'd then end up with a rather odd-looking desktop due to the severely unbalanced X and Y resolution. As I don't want to get heavily involved in generic Linux tech support it is assumed that anyone wishing to try the modeline knows what to do with one.

If you're using Windows, you could give "PowerStrip" a try. http://www.entechtaiwan.com/util/ps.shtm. Note that I have never used this myself, I'm just aware of it as a means to get Windows boxes to do weird resolutions and timings. The same applies to this, if you're using a GeForce4 and my timings, expect a very odd-looking desktop. If you're not using a GeForce4 I have no idea if you'd be able to get correct timings at all.

Now, a general warning to anyone who feels like playing with 'funny' video timings on a PC - if it all goes wrong you could end up feeding weird timings into either your computer monitor (which might not like it) or a vintage television (which really might not like it) - you might end up stuck with an unreadable display or worse, escaping magic smoke.

I'm doing my testing with no monitor/TV connected to my experimental system at all, I'm logged into a bash shell remotely from another machine on my LAN, and I'm checking timings very carefully before I connect anything.

(I don't want to scare anyone off, but I also don't want to get the blame for blowing up a nice old television or a computer monitor...)

Be safe and don't let the smoke out

Kat