Line Interpolation.

Forum for discussion of narrow-bandwidth mechanical television

Moderators: Dave Moll, Andrew Davie, Steve Anderson

Postby Steve Anderson » Sun Apr 29, 2012 9:45 pm

Gary, I did wonder whether to leave the "no flames" remark in or not just prior to pushing the submit button, but push I did, so once there what's the point of later deleting it?

An observation I have made, and I'm sure you too, is that sync pulse widths are not consistent, not only from different sources but within the same piece of video. Often being modulated by the adjacent video level before and following the pulse, and/or a longer-term APL modulation.

Add to this what often appears to be excessively slow edges to the sync pulses that they end up being sync 'bumps' rather than pulses. Any movement up or down (voltage-wise) in these pulses relative to the sync-slice voltage and you recover variable width sync pulses modulated by the rest of the signal.

So what to do? Take the centre of this sync bump as the time reference point? Graham Lewis has observed that a short front-porch of black prior to the sync pulse does improve matters, I assume he's using leading-edge sync detection here. (As in the negative-going edge, no reference to the technology used. My technology is somewhat trailing-edge).

As for LF pre/post emphasis I did a few experiments with this a few years ago with surprisingly good results, but I only tried post-emphasis as I was concerned about record overload. Graham again has experimented with this and there are a few items in past newsletters.

But I knew what post-emphasis to apply as I had built the complimentary filters and knew their characteristics to simulate a bogus piece of audio gear. Often this is nigh on impossible to determine on a commercial piece of equipment, especially when it's in someone else's house on the other side of the planet! So that idea was quickly binned, but it does hold promise. But heavy-handed application does bring with it noise and LF 'thumps'.

Steve A.
User avatar
Steve Anderson
"Fester! Don't do that to 'Thing'"
 
Posts: 5360
Joined: Fri Mar 30, 2007 10:54 pm
Location: Bangkok, Thailand

Postby gary » Sun Apr 29, 2012 11:36 pm

Steve Anderson wrote:An observation I have made, and I'm sure you too, is that sync pulse widths are not consistent, not only from different sources but within the same piece of video.


Indeed, I made a reference to that, albeit in a slightly different context, in Albert's "wavy line" thread.

I would imagine that implementation of a fly wheel sync mechanism should more-or-less remedy that as the long term average would be very close to spot on assuming a consistent, if not accurate, time base (an accurate time base would solve everything but I suppose that's asking a bit much).

Measuring the centre of a sync pulse has it's own problems due to a rather course sample resolution i.e. +/- a sample which is the best part of a pixel. I tend to use the sync tip which, regardless of the modulation should be temporally consistent. Alas, this is not true either because of the "smoothing" effect of the overall low pass filter (whether it be 10kHz as per spec, or nearly half the sample rate max) as this guarantees that some of the programme material is mixed in with the sync pulse thus changing it's shape.

As far as I can see this could only be overcome by transmitting the sync pulse separately to the programme.

Edit: sync tip in this context being define as the last sample of the sync pulse's nadir - the start of your trailing edge. This, of course, requires memory of at least the width of the sync pulse - I actually use a much longer buffer.

Edit: I am, of course, treating this as a purely digital exercise.
gary
 

Postby Klaas Robers » Wed May 09, 2012 12:33 am

Steve, this is a tricky stuff. I used always the down going edge of the sync pulse. However in my generators (e.g. EPROM-) I inserted one black (blanked) pixel at the end of each line, just before the sync starts. That is no more visible in the 10 kHz low passed video signal, but it gave much less sync jitter due to the video content. Anyway a flywheel (Nipkow disc or in software) in the line generator is eliminating the remaining line jitter.

I prefer active clamping over DC restoration. This presumes that you have sync pulses = clamping pulses. Then there is no R3 at all, but an active switch connects the video signal to a reference voltage during the clamping pulse. This includes no extra "sag" on the video line.

For the detection of the sync pulses some extra "sag" in the video line is of no importance. For that purpose the "perfect diode" can be used and a resistor that pulls the signal to a negative voltage, such that there is a almost constant "sag" in each line, undependant from white or black as video content.

Or you could use the clamped output to detect your sync pulses. However there seems to be a "chicken and egg" problem. I haven't experienced this contraption. The others are described in NewsLetter 26/1 and 26/3 and are in the PCB's of the club.
User avatar
Klaas Robers
"Gomez!", "Oh Morticia."
 
Posts: 1656
Joined: Wed Jan 24, 2007 8:42 pm
Location: Valkenswaard, the Netherlands

Postby Steve Anderson » Wed May 09, 2012 12:55 pm

Klaas Robers wrote:I used always the down going edge of the sync pulse. However in my generators (e.g. EPROM-) I inserted one black (blanked) pixel at the end of each line, just before the sync starts. That is no more visible in the 10 kHz low passed video signal, but it gave much less sync jitter due to the video content. Anyway a flywheel (Nipkow disc or in software) in the line generator is eliminating the remaining line jitter.


This agrees with Graham Lewis's findings as I mentioned before and is worthy of further investigation. It is possible that the same applied to the start of active video (providing a 'back porch') may improve matters further. But that is unknown at this stage. Traditionally within television broadcasting the leading-edge (the -ve going edge) of the sync pulse has been the time reference. Ditto the equalizing and broad pulses...but we don't have them.

Klaas Robers wrote:I prefer active clamping over DC restoration.


So do I. But what I'm driving at here is that detecting the sync pulses and thereby generating the clamp pulse as per 26/1 and 26/3 suffer from the same shortcomings as above. If you look at the second op-amp in both references you'll note that the time-constant is about 220ms, (330n + 680k), very similar to that illustrated above (190ms). If you can't detect the sync pulse you can't generate the clamp pulse. At least with this arrangement.

As I have mentioned previously, this is an extreme signal. But I noticed that with the "Eureka MkI" upon large changes in APL at scene changes it 'lost it'.This uses exactly the same methodology and also uses sync-tip clamping. (Vol. 35 No.4).

So my prime focus has be the extraction of sync pulses under this extreme condition, surely if it can handle this it should handle anything? Sadly the simple capacitor/resistor/op-amp/diode arrangement just can't.

Steve A.

P.S. There's been a lull in this due to trying to get an item off to Jeremy within the next few days. Things should resume soon.
User avatar
Steve Anderson
"Fester! Don't do that to 'Thing'"
 
Posts: 5360
Joined: Fri Mar 30, 2007 10:54 pm
Location: Bangkok, Thailand

Postby Klaas Robers » Wed May 09, 2012 8:12 pm

Steve Anderson wrote:
Klaas Robers wrote:I prefer active clamping over DC restoration.


So do I. But what I'm driving at here is that detecting the sync pulses and thereby generating the clamp pulse as per 26/1 and 26/3 suffer from the same shortcomings as above. If you look at the second op-amp in both references you'll note that the time-constant is about 220ms, (330n + 680k), very similar to that illustrated above (190ms). If you can't detect the sync pulse you can't generate the clamp pulse. At least with this arrangement.


You are right Steve, the time constant is almost the same, but the 680k is pulling to ground (0V) while the sync-tips are kept at about +4V for a video signal of 1.4V peak to peak. That implies that each line sags down and is restored to +4V at each sync bottom. Now the mean value of the video voltage is of little influence on the sag, the signal always sags downwards, unless there is a clipped and pasted DC jump in one line. But then still the video signal is sagged to +4V sync tips within a few lines.

This of course is an unwanted situation for video reproduction. That is the reason that there are two DC-restorers in these circuits, one optimised for video reproduction with a long RC and one optimised for sync separation.
User avatar
Klaas Robers
"Gomez!", "Oh Morticia."
 
Posts: 1656
Joined: Wed Jan 24, 2007 8:42 pm
Location: Valkenswaard, the Netherlands

Postby Steve Anderson » Wed May 09, 2012 9:29 pm

Klaas, I agree with your previous posting, but I don't think you've quite grasped what I'm trying to point out...perhaps it's my poor explanation.

First thing, the nature of the input video waveform, extreme though it might be, it is a legitimate signal. Four full frames of white followed by four frames of black.

I'll quote myself here...

Steve Anderson wrote:The test signal is a 'Club Standard' NBTV signal comprised of four frames of white followed by four frames of black complete with missing syncs continuously repeated. This DC-coupled signal was then passed through a single CR hi-pass filter with a -3db cut-off of 10Hz to simulate the output of an audio source. Now -3db at 10Hz is not a shabby performance of any audio gear, and here only one hi-pass stage is employed. Two waveforms of the test signal are below, (Blk2Whi 1.gif and Whi2Blk 1.gif) the yellow trace being the TSG output, the cyan being after the hi-pass filter. One shows the black-to-white transition, the other white-to black.

Now I'll be the first to admit this is an extreme waveform, but it is legit, and although unlikely to happen on real progam material, it could and the system should cope with it.

In most cases there's another CR hi-pass filter interposed, some form of buffer amplifier, here is no exception, but I did set it at a very long time-constant of one second, -3db at 0.16Hz.

The results shown in (Wht2Blk 2.gif) are woefully inadequate...which is not surprising. You can twiddle the sync-slice pot nut you end up slicing genuine video and missing the otherwise valid sync pulses.


The point I'm trying to make is that at the junction of the capacitor and resistor in all these similar designs is going to look like the waveforms below, maybe not exactly the same, but very similar, syncs will be missed. This has nothing to do with DC restoration, clamping etc., just simply sync extraction in a difficult waveform such as this.

I haven't repeated all the waveforms from previous, just the pertinent one to this below. The Cyan trace is what you'll see at the C/R junction aforementioned with this source waveform, a cut from a scene with a high APL to one with a low APL. You end up with teeth missing. (Green trace).

Steve A.
Attachments
whi2blk_2_574.gif
whi2blk_2_574.gif (6.58 KiB) Viewed 12755 times
User avatar
Steve Anderson
"Fester! Don't do that to 'Thing'"
 
Posts: 5360
Joined: Fri Mar 30, 2007 10:54 pm
Location: Bangkok, Thailand

Postby Klaas Robers » Thu May 10, 2012 4:03 am

Steve, the first thing I see is that your input signal has a high pass RC-time of 15 msec. That is a -3 dB point of 11 Hz. Such signals are difficult to handle anyway. The -3 dB frequency should be much lower.

Then I see that the sag-speed of the DC-restorer has a time constant which is much longer. That is asking for problems. That will never work. You see that happening. It should be the other way around.
User avatar
Klaas Robers
"Gomez!", "Oh Morticia."
 
Posts: 1656
Joined: Wed Jan 24, 2007 8:42 pm
Location: Valkenswaard, the Netherlands

Postby Steve Anderson » Thu May 10, 2012 12:32 pm

Klaas Robers wrote:Steve, the first thing I see is that your input signal has a high pass RC-time of 15 msec. That is a -3 dB point of 11 Hz. Such signals are difficult to handle anyway. The -3 dB frequency should be much lower.


Yes, true, as I pointed out previously, this is to simulate a bogus piece of audio equipment, few go any lower than that. In their intended application there's no reason to do so, indeed it's better they don't. Sure, you can modify a piece of audio gear to extend its LF performance, but how many want to knife-and-fork their expensive audio equipment or a PC motherboard? (Ever tried it? Without a manual? I have, don't ask.).

There are those amongst us that will have a go at this but I feel that's going to be a minority. So for 'maximum appeal' to those considering getting involved in NBTV this does need to be a plug-and-play exercise, at least as far as the recording media is concerned.

This is all an aside to the main theme of this meandering thread (nothing unusual there), but I found this shortcoming in the "Eureka MkI" and want to address it. The fact is that it appears to apply to other examples of this method of sync extraction also.

Klaas Robers wrote:Then I see that the sag-speed of the DC-restorer has a time constant which is much longer. That is asking for problems. That will never work. You see that happening. It should be the other way around.


I'm not sure what you're referring to here, the "Eureka MkI"? If so what point in the circuit? Or which components?

Steve A.
User avatar
Steve Anderson
"Fester! Don't do that to 'Thing'"
 
Posts: 5360
Joined: Fri Mar 30, 2007 10:54 pm
Location: Bangkok, Thailand

Postby Klaas Robers » Fri May 11, 2012 6:15 am

Steve, preferably R3 should be a constant current source that sources a current such that the video level drops 1/3 sync length during one NBTV line. So each sync tip tries to get 1/3 sync length lower, however will be restored to the reference level by the "perfect diode".

Suppose we have a video signal of 0.7 volt black to white with 0.3 volt sync pulses. In this case the voltage should drop during one line by 0.1 volt, that is 1/7 of the video black-to-white, and is shifted up again by 0.1 volt during the sync.

When the missing sync pulse occurs the video drops 0.2 volt, but the black level is still not dropped at sync level at the end of line 1. So if you detect the leading edge at 0.05 volt everything is Ok.

Now, if you have an input signal that rises 0.1 volt each line then still nothing goes wrong. Only the DC-restoring diode has nothing to do. If it rises more than 0.1 volt per line you will start to miss syncpulses.

Your resistor R3 is not a current source. However if you raise the Vref it gets better. A higher resistor to a negative voltage will work even better. The voltage over R3 should be really higher than the video peak-peak voltage. Then the current through R3 is almost independant from the mean DC level of the video signals content.

Of course this is far from optimal for the video signal itself. An active clamp is better for this and as soon as you have the sync pulses separated you can incorporate that.
User avatar
Klaas Robers
"Gomez!", "Oh Morticia."
 
Posts: 1656
Joined: Wed Jan 24, 2007 8:42 pm
Location: Valkenswaard, the Netherlands

Postby Steve Anderson » Fri May 11, 2012 4:11 pm

Klaas, all points taken on board. It may be worth knocking this version up on a breadboard sometime, I have some CA3240s here which are the dual version of the CA3140 less the offset null and strobe pins. (Unused). I can get CA3140s locally, but it's a drag right across the other side of Bangkok, something I try to avoid. I would use RS or Farnell as both are represented here, but their delivery costs are a joke! (There is no 'Trade Counter', you cannot go and collect).

As for a current source not requiring a negative rail, a simple current-mirror using an LM394, SSM2212, BCV62 or even a CA3046/3086 should do the trick. Worth investigating. In the references mentioned where the sync tips are clamped to +4V there's enough voltage to use a more conventional one-transistor version.

Once what is cluttering up my workbench at the moment is removed, I'll return to this.

Steve A.
User avatar
Steve Anderson
"Fester! Don't do that to 'Thing'"
 
Posts: 5360
Joined: Fri Mar 30, 2007 10:54 pm
Location: Bangkok, Thailand

Postby AncientBrit » Mon May 14, 2012 5:50 pm

The only problem with active clamping is that if the video clamp point has noise on it then this will be converted into line by line dc variations visible as line striations.

A better method would be to use a feedback clamp with LF averaging as used in some 625 kit. I don't think I've seen an implementation of this for NBTV.

Returning to the subject of sync separation I did carry out some experiments in software using a form of 'profile' detection of the most negative part of the video waveform by using a peak detector.

The idea was to subtract this then from the video waveform, or use it as a reference for the sync sep.

If you are attempting this in hardware you will need to delay the video (maybe a PIC ADC-DAC) for a few pixels so the detected sync profile matches the video.

The problem remains though that if the 'dv/dt' of negative video content approaches that of a sync pulse then it's not possible to accurately detemine a true sync pulse.

Cheers,

Graham
AncientBrit
Green padded cells are quite homely.
 
Posts: 858
Joined: Mon Mar 26, 2007 10:15 pm
Location: Billericay, UK

Postby Klaas Robers » Thu May 17, 2012 11:53 pm

Yes indeed Graham, some resistance or resistor should be in the switch circuit, to ensure that the time constant with the coupling capacitor is not too short. An RC in the same order as the duration of the clamping pulse will do. Then the HF noise is eliminated.

However in NBTV the sync pulse and thus as well the clamping pulse is rather short. May be a variable resistor to find the compromise between clamping noise and sag would be a suitable solution.
User avatar
Klaas Robers
"Gomez!", "Oh Morticia."
 
Posts: 1656
Joined: Wed Jan 24, 2007 8:42 pm
Location: Valkenswaard, the Netherlands

Postby AncientBrit » Fri May 18, 2012 5:19 pm

Thanks Klaas, I'll try the CR combination.

I've seen something similar in 625 use where a parallel tuned circuit resonant at subcarrier frequency is used to allow clamping in the presence of the sc burst.

Regards,

Graham
AncientBrit
Green padded cells are quite homely.
 
Posts: 858
Joined: Mon Mar 26, 2007 10:15 pm
Location: Billericay, UK

Previous

Return to Mechanical NBTV

Who is online

Users browsing this forum: No registered users and 29 guests

cron