PC as a standard convertor

[oldeurope]

Hi Jeffrey,
thanks for your schematic.
I soved the problem with the fast rising edges of the sync by using
a HCF40..BE driever.
The problem is you add the signals and the noise of the black level signal
is added to the sync tips. Maybe you noticed the spike in Kat's oscilloscope
screen shot. This will be added to the sync edge and it will destroy it.
In Kat's and my design the sync is independent from noise at the black level.
The DC level of the input RGB is unknown, thus the DC level of the sync tips
at the output is unknown. In the modulator you need an additional DC restorer
and biasing circuit.

Kind regards
Darius

[ppppenguin]

Originally Posted by oldeurope

I soved the problem with the fast rising edges of the sync by using
a HCF40..BE driever.
The problem is you add the signals and the noise of the black level signal
is added to the sync tips. Maybe you noticed the spike in Kat's oscilloscope
screen shot. This will be added to the sync edge and it will destroy it.
In Kat's and my design the sync is independent from noise at the black level.
The DC level of the input RGB is unknown, thus the DC level of the sync tips
at the output is unknown. In the modulator you need an additional DC restorer
and biasing circuit.

The RGB outputs from graphics cards normally have black level at or near 0V. If not, it's easy to bias the output opamp. A modulator MUST have a DC restorer or clamp since you cannot guarantee the DC offset of any video signal that you might feed to it.

The RGB signals from a graphics card should not have any significant noise. If there are large spikes in the blanking interval then some kind of reblanking will be needed.

[oldeurope]

Originally Posted by ppppenguin

A modulator MUST have a DC restorer or clamp since you cannot guarantee the DC offset of any video signal that you might feed to it..

Hi Jeffrey,
in the MK II and MKIII converter/modulator every input signal gets a new clean sync
and the discribed stages are always in use.

Kind regards
Darius

[Kat Manton]

Hi,

Originally Posted by oldeurope

Maybe you noticed the spike in Kat's oscilloscope
screen shot.

The two 'spikes' are at start and end of the active period of the line - the windowmanager I'm using on X (the very minimalist "evilwm") puts a 1-pixel mid-grey border around application windows. So, using 'display' to get my full-screen 'colbars.png' image on screen, the windowmanager puts this border around it. In a way it's useful, I get grey first and last active lines, and the left and right borders are handy 'markers' showing the length of the active line. MythTV actually suppresses all window decoration when running, so on off-air TV pictures this quirk should be absent.

Jeffrey, I think I'll get hold of a video opamp or two and have a play. If all goes well, there's probably good cause for getting boards made up for this as something that could be sold complete or as a kit. With my 'production engineer' hat on I regard reducing the number of components as a Good Thing (tm) as construction is simplified and there's less to go wrong etc... It looks like a circuit with a video opamp would offer better performance and simplified construction. I'd guess with the high bandwidth of these wee beasties I'd need to pay careful attention to grounds, supply decoupling, etc - similar to the sort of issues involved in RF and high-speed digital logic design with which I'm more familiar.

Incidentally, I modified the output stage of my rough and ready mess with the wandering black-level; set the X modeline back to 625-line and got a reasonable stable monochrome picture on my 25" Mitsubishi "modern" (slightly dim and a little lacking in HF response, but I can probably tweak that). So it'll do for now until I make a 'Mark II' with a video opamp

Regards, Kat

[jim_beacon]

Kat,

your link to your original combiner diagram is broken - do you have a copy of the diagram anywhere else?

Secondly, when playing with the X modeline, did you just put the 405 version in, and remove the others, or at least comment them out?

Finally, does the sync come out of the card as a combined signal, or as seperate H and V signals? (the VGA data suggested either was possible).

I do at least understand how you arrived at the numbers for the Modeline - an interesting little set of calculations!

Many thanks

Jim.

[Kat Manton]

Hi Jim,

Originally Posted by jim_beacon

your link to your original combiner diagram is broken - do you have a copy of the diagram anywhere else?

Umm... I removed it as the video waveform it produces is terrible; I'm currently working on a new one (DC coupled, other improvements) which I hope to finish before NVCF and which I'll publish once it's finished. Cutting it a bit fine, I guess

Secondly, when playing with the X modeline, did you just put the 405 version in, and remove the others, or at least comment them out?

I actually fairly radically modified the file, such that it's easy to switch between 800x600 on the monitor, 405 and 625 on the VGA output and 625 on the S-Video output as I'm using the system on our main 25" colour set in the living room as well.

I'll tidy the file up and comment it a bit better then let you have a copy.

Finally, does the sync come out of the card as a combined signal, or as seperate H and V signals? (the VGA data suggested either was possible).

This depends on the card - the GeForce4 doesn't produce composite sync - hence having to combine (in my present circuit, with a XOR gate) the H and V syncs.

I do at least understand how you arrived at the numbers for the Modeline - an interesting little set of calculations!

Looks scary at first; easy when you know how...

Regards, Kat

[Sean Williams]

Does anyone know if I could use a twin processor unit for this purpose - there are some fairly well specd models on Ebay at present, and at a reasonable price, located a few miles from me - any ideas?

Cheers
Sean

Itching to get some progress on this idea......

[Kat Manton]

Hi Sean,
MythTV itself doesn't take advantage of SMP systems; though it would run the backend and frontend on different processors, gaining some advantage as Linux will run on SMP systems and spread the load out accordingly.

Regards, Kat.

[YC-156]

Originally Posted by Sean Williams

Does anyone know if I could use a twin processor unit for this purpose - there are some fairly well specd models on Ebay at present, and at a reasonable price, located a few miles from me - any ideas?

Well, I haven't used MythTV for ages, but I did look into this idea a while back when Kat's results first emerged. This because I presently happen to have two dual processor machines here at home.

What I found was:

*) The MythTV documentation doesn't mention SMP (Symmetric MultiProcessor) with even a single word when it comes to hardware selection. In my experience this means that the program will not utilize two processors at once, if present. So if a single, slow CPU cannot encode/decode as fast as one would like, two of them will not be of any help.

*) Some people on the MythTV reported driver problems when running SMP. Ie. the drivers were not thread safe. Not good.

So unless you need to run two CPU intensive programs at once while running MythTV, SMP probably won't help much.

Additionally my experience with using SMP machines as workstations and number crunchers are that with older hardware you usually run out of other resources before the CPU power is exhausted. Ie. harddisk, memory or DMA speed is insufficient to fully saturate two CPUs. The exception is very CPU bound applications like raytracing, where you often can run two tracing instances at once for close to a 100% speed advantage.

SMP may help when you need to run two dissimilar programs at once. Like if something happens in the background while MythTV is playing, a second CPU might help prevent 'stuttering'. Notice emphasis on 'might'.

When the program of interest really is limited by the CPU, a single slightly faster CPU is often way ahead of two relatively similar processors at half speed or so. This is due to all the other enhancements, which accompagnies the faster CPU/motherboard. As an example of this, then a single 1GHz P-III is so much faster than 2x 400MHz P-IIs that it isn't even funny. The P-III has a larger cache, faster RAM interface and other architectural enhancements, so it is way faster than twice the combined CPU horsepower of the P-IIs.

Another case I have personally experienced was a single 1.3GHz P-III 'Tualatin', which obliterated a pair of 550MHz P-III Xeons! The latter pair was housed in a very expensive piece of server hardware (assembled by yours truly), yet give it a few years and a cheap beige box and a motherboard from Asus made toast of the Xeons...

What types of CPUs are you looking at for SMP?

Best regards

Frank N.

[Sean Williams]

Ah, right, well, the unit I was looking at runs dual 800Mhz processors, and has 512mb of ram - so on paper it would appear to be ok for Myth with just one of the processors.

However, if the consensus is that this might cause issues, then I will accept the hit on postage, and source a single CPU machine (just being tight fisted really!)

Daft really, I'm more than happy to lash out large chunks of cash for test gear, but wince when spending money on PCs!

Cheers
Sean

[YC-156]

Originally Posted by Sean Williams

Ah, right, well, the unit I was looking at runs dual 800Mhz processors, and has 512mb of ram - so on paper it would appear to be ok for Myth with just one of the processors.

As long as they aren't Durons or Celerons.

Originally Posted by Sean Williams

However, if the consensus is that this might cause issues, then I will accept the hit on postage, and source a single CPU machine (just being tight fisted really!)

My suspicion from reading about other people's experiences is that you might have to run the machine as a single CPU configuration to avoid a lot of hassle, and if money isn't really the issue, then I would dive in for a serious piece of second hand gear. Something like a 2533MHz P4 with dual channel memory wil smoke the 2x 800MHz setup, no matter what markings are on the latter. P4s that fast has been available for more than three years, so they ought to be available for relatively little money by now.

Originally Posted by Sean Williams

Daft really, I'm more than happy to lash out large chunks of cash for test gear, but wince when spending money on PCs!

The difference being that the test gear isn't disposable and outdated after a few years. I like ex-mil spec HP gear myself.

Best regards

Frank N.

[Kat Manton]

Hi Sean,

I found a single 800MHz Duron was a bit marginal decoding off-air digital telly; as the frontend would run entirely on one processor you'd probably have the same problems with dual 800MHz processors.

I'm running a single-processor 2GHz Athlon XP box now (that was the black machine at NVCF) and that's definitely ok; it should have enough thrutch to handle dual DVB-T cards; possibly an analogue tuner card as well. A rough guess is that a single 1.3GHz processor should be just enough to handle a single analogue 'framegrabber' TV card; 'framegrabber' cards need more thrutch as the images are encoded and dumped to disk by the backend then read from disk and decoded by the frontend. Digital telly takes a bit less crunch as it's already an mpeg stream so the encoding stage is skipped; but then you have no choice but to decode full-resolution mpeg2 streams as that's what you get.

Processor requirements are a bit tricky to nail down as it does depend on your source of video/off-air signal, how many tuner cards you intend to use, what type of cards you use, etc, etc...

Regards, Kat.

[Sean Williams]

Hi Kat,

Ok, Sorry if I seem a bit thick on this, but computers to me are a sort of spawn of the devil!

I take your point on proc speed - perhaps I ought to downgrade the system I use in the shack, and use that PC for this project.

I am curently unsure if I will ever use the system for real time off air decoding - SWMBO does not appreciate the "horrible wooden box" in the dining room!

I am currently bidding on a Nvidia ge force 4 video card, and I am looking into a DVB card - much to my amazement PCworld stock the one you mention!

So, perhaps the shack logging PC gets the chop - it is a 1.7G proc with about a gig of ram, so I guess this is really a no brainer!

Cheers
Sean

[Paul Stenning]

If the shack PC isn't doing much, maybe that dual 800MHz on eBay could go in there instead? It'd be happy with Windows 98 or 2000 but might be a bit sluggish with XP.

[jim_beacon]

Originally Posted by Sean Williams

I am curently unsure if I will ever use the system for real time off air decoding - SWMBO does not appreciate the "horrible wooden box" in the dining room!

Mine can't stand 10.125kHz I have to experiment while she's out

Originally Posted by Sean Williams

I am currently bidding on a Nvidia ge force 4 video card, and I am looking into a DVB card - much to my amazement PCworld stock the one you mention!

Don't forget that the FX5200 works as well. If you have onboard video, it may be worth looking what video chip set is fitted, a lot of newer boards have an nVidia chipset. MythTV also works with Matrox cards, but neither Kat or myself have tested one yet (I have a line on a couple of Matrox cards, and I'd like to try the system in "dual head" mode - VGA on one port, 405 on the other).

Originally Posted by Sean Williams

So, perhaps the shack logging PC gets the chop - it is a 1.7G proc with about a gig of ram, so I guess this is really a no brainer!

I have a system running with a 1.4GHz P4, with 630MB RAM, though knoppMyth documentation says that it can only use 512MB. My system is happy to run live analogue tv, but I haven't tried a digital card yet.

Jim.

[Kat Manton]

Hi,

Originally Posted by Sean Williams

I am curently unsure if I will ever use the system for real time off air decoding - SWMBO does not appreciate the "horrible wooden box" in the dining room!

Originally Posted by jim_beacon

Mine can't stand 10.125kHz I have to experiment while she's out

Guess I'm lucky, not only has Cobweb found 10.125kHz something you can ignore after a while but it was Cobweb who said "yes" to a console set we were offered. Once refurbished and reliable, it's going in the living room, and we're looking at collecting classic B&W movies on DVD to watch on it (Just got "Metropolis - now that is a classic movie, and off-topic for this thread... cue me starting another thread, methinks...)

Originally Posted by Sean Williams

So, perhaps the shack logging PC gets the chop - it is a 1.7G proc with about a gig of ram, so I guess this is really a no brainer!

That, as they say, would do nicely - though it's not so much MythTV can only use 512M, more it doesn't really need any more, not at these display resolutions, anyway. Watching the IRC channel #mythtv-users on Freenode I saw a discussion where it was noted that runniing at high resolutions into an HD set it does need more RAM. But 405-line is more LDTV in this context really.

Regards, Kat

[Kat Manton]

Hi,

"The Fools" have been at work again, and have some very interesting news...

Jim Beacon discovered that the later nVidia driver we're both now using for our cards permits pixel clocks below 12MHz.

I've worked out a new modeline for 405-line using a pixel clock of 6.48MHz, which gives me even closer 405-line timings (line and frame rate are exactly 10.125kHz and 50Hz respectively) and has a rather more respectable 536 x 378 resolution.

One upshot of this is the MythTV user interface is a lot more readable as the text isn't squashed; the user interface settings still need a bit of a tweak, but it's a vast improvement.

Details of this will be added to the website in due course.

Regards, Kat

[ppppenguin]

It;s amazing what "The Fools" can do. Has anyone done a dual head 405 and VGA yet?

That Pye monitor looks like it's doing a good job too.

[jim_beacon]

Originally Posted by ppppenguin

Has anyone done a dual head 405 and VGA yet?

I haven't tried it yet, but I do have a couple of Matrox cards at work which are capable of this mode. I may try it later in the week, but I'm not sure if the Matrox driver will work with the sub-12MHz dot clock we are currently testing.

Jim.

[ppppenguin]

I've got Matrox G550 and Parhelia cards (both dual head) around as part of the work I do for one of my clients. I know that my colleague who does the software runs them in non-standard modes though still with fairly high dot clock speeds. The G550 in particular has always seemed very tolerant of any abuse we've hurled at it. In particular we have meddled with the master 27MHz clock and the G550 never complains. The Parhelia on the other hand is a touchy little devil and will crash the PC at the slightest provocation. The Parhelia is an expensive card so it's not really suited to the FothTV project unless you have one lurking in the junk box. At least FothTV won't involve mucking around with the hardware on the graphics card.