Building a Standards Converter

[ppppenguin]

Adaptive interpolation can work, it's complicated to get it to work well. Just look at the de-interlacing on many flat panel TVs; it can be really horrible on things that are moving. Just watch something that's slowly moving and then stops. And snaps into focus. Better sets do a respectable job but they are stuck with the facts:

1: Most motion is already aliased (inherently by the field rate, akin to wagon wheels going backwards in cowboy movies) which makes 2: even harder
2: On an interlaced image motion and vertical detail can occupy exactly the same 2 dimensional frequencies and hence impossible to separate reliably

It can be done very well indeed, some modern TV chipsets are excellent but they can be caught out on occasion.

Add all the data rate compression used for recording and transmission, which is taking liberties with motion anyway, and you can end up with revolting pictures on occasion. On the whole it works OK provided the broadcasters and mastering engineers don't push the compression too hard.

If you yearn for analogue PAL, just remember the cross-colour effects on fine luminance detail. They weren't very nice either though comb filter decoders could minimise the problem.

[Cardigan]

Add all the data rate compression used for recording and transmission, which is taking liberties with motion anyway, and you can end up with revolting pictures on occasion.

e.g. 25PsF. Looks horrendous.

I'd also add that motion vector compensation is used in the uncompressed video domain as well - see the SAM "Alchemist" converters and the better-specced "Kudos" converters.

https://s-a-m.com/store/alchemist-ph.c-hd/c-24/p-122

[`FRANK.C]

I finally got the test card done, it is similar to the PM5544 style one I did wit a PIC but was much easier to do with a FPGA. Gratings are the usual 1, 1.5, 2, 2.5 and 3 MHz.
For what it is it uses quite a bit of resources so if I run short it will get the chop and I will use something simpler instead, although I expect that wont need to happen.  
I have also done a 400Hz tone using PWM  to go with it.
Photos below.

Frank

[ppppenguin]

Is that testcard done with logic or is it an image being read out of memory?

[`FRANK.C]

Hi Jeffrey
It is done with logic, it is clocked from the pixel clock. When in 5:4 mode most of the side castellations gets cropped which keeps the dimensions of the circle correct.

Frank

[peter scott]

Very impressive!

Peter

[Murphyv310]

I'm with Peter.
Impressive

[`FRANK.C]

Hi All
I haven't posted for quite a while just other things had to take priority.
I have been dipping in and out of this project any chance I got, hopefully this post will bring it somewhat up to date.

The video decoder defaults to NTSC at start up if no signal present at its input. As I am using a Frame sync pulse that is derived from the field indicator signal this causes a problem as the frame sync pulse resets the 405 sync generator and when it arrives at 30 Hz instead of the expected 25 Hz it really messes up the syncs and the test card don't display as it should.
There are probably signals from the decoder that indicate when a PAL signal is present but instead I decided to detect PAL by count the lines in each frame and when  PAL is not present the Frame pulse is disabled allowing the 405 sync generator to free run.

I have populated the RF modulators and they are working OK. All 13 channels appear to be working good except for the sound on Channel 13. There is a continuous steady tone on it of maybe 200 or 300 HZ.
It is there even when the Video modulator is disabled or when the audio input is grounded.
Viewing the Sound RF on the scope there is no visible AM modulation and when tuned into the center the tone is at lowest, tuning either side of center increases the volume of the tone a lot. This makes me think it must me some form of FM modulation.

I have started to finalize the PCB but before I could do this I needed to chose a case for it. The maximum size of the PCB I limited to 100mm X 100mm as once over this size the cost off manufacture increases dramatically. I would like to mount all the components including the switches and connectors on the PCB. They wont all fit along one edge of the PCB so will have to used two edges, because of this the distance between the front and back panels of the case will need to be 100mm or less and to facilitate fitting the PCB into the case the case will have to have removable front and rear panels. Another criteria is the FPGA development board will need to fit between the PCB mounting posts of the case.
I found one that suits the bill it is made by multicomp part No. MCRM2015M, it will allow a PCB size of 100mm X 89mm which is large enough to to have everything PCB mounted. On the back panel will be 3 X phono sockets for audio and video,  a power socked and a F connector for RF out. The front panel will have 3 X 3 position and 3 X 2 position toggle switches for options, a hex switch for Chanel selection, a LED and a phono  socket for video out.

Frank

[ppppenguin]

Not surprised the decoder defaults to NTSC but can it be programmed to do otherwise? Possibly forced to PAL? If it gives an indication of absent input this can be used to switch to stored testcard. Simply ensure that its pulse outputs are not connected to the counters when input is absent.

Again this reinforces the great joy of programmable logic. If you encounter a problem like this it's easy to do a few trials to find a solution.

I have no idea about the modulator sound problem.

[`FRANK.C]

Hi Jeffrey
I don't believe the TVP5150 can be forced to PAL. The TVP5146 that I was using I am fairly sure could.
I have decided to use a switch to select the test pattern as opposed to have it atomatic on no signal, I thought it is easier to flick a switch than to have to unplug when the test card was required.

I still find it hard to believe how fast, powerful, versatile and flexible FPGA's can be, what a truly dream device if there was ever one. They may be a little harder than PIC's to program but the pay back is HUGE and that is putting it very mildly.


The photo below shows the scope trace (50mV/2mS Div.) on pin 14 (RF PLL loop filter) of the MC44BS373 modulator for sound when channel 13 is selected, it should be a flat line and is a flat line on all other channels. On the vision modulator it remains flat for all channels including 13. I swapped all the configuration  settings between the vision and sound modulators and the problem switched to the vision modulator, so not a hardware problem. I don't think it is a configuration problem either as apart from frequency the configuration is the same for all channels.  I will have another look at it tomorrow. I built a MC44BS373 based modulator a couple of years ago and it had the same problem with channel 13, I never looked into it as channel 13 is not one I have much use for.  It might just be a spooky 13 thing

I haven't used a 3.5 MHz trap before the modulator as I don't think there is need for it. I tried the converter on a modulator with a trap and detuning the trap made no noticeable difference on sound

Frank