STFM VIDEO SYNC FAULT

[czietz]

I just stumbled across this thread and only skimmed over the past posts, so please excuse if I missed some information there.

My STFM had the same problem using RGB via a SCART cable. You can see this here:

In my case it seems that my TV gets confused when it finds not only composite-sync but composite-video from the Atari on its SCART input in RGB mode. As you of course know, an Atari STFM outputs composite-video on pin 2 of the monitor connector, while an STF only outputs composite-sync there. Anyway, I modified my STFM to output only composite-sync as well. I made this switchable, in case I'll need the composite-video output in the future.

You can see in the following video that this completely fixes the problem. For the first 12 seconds, I use the (original) composite-video output (so the fault shows), then starting at 0:16 I switch to composite-sync and the picture is 100% stable!

Now, it beats me why this doesn't happen for every STFM. Maybe the original composite-video output is slightly different, so some machines "confuse" the TV more than others.

[exxos]

Thanks for your post. I think it is the TV not syncing correctly, at least not until it gets a image up like the desktop.

Thing is, my STFM had been working fine for years, then one day it started to have that problem and it never went away.

Next time I come across a motherboard which does it, I will put it aside and investigate it more.

[czietz]

As you can see in the second video, the TV does sync correctly when it gets composite-sync instead of composite-video.

Unfortunately I only have this one STFM so I can't do any comparisons. If you find an STFM with that problem, maybe you could compare the composite outputs (which are only used as sync when using RGB) between an affected and an unaffected machine. Maybe due to aging something on the composite-video signal starts to drift away (DC level, amplitude?) to such an extent that the composite-video signal isn't acceptable as sync signal anymore -- at least when the picture is mostly white.

[exxos]

I think my scart lead is wired to use composite for the syncs. I did check all the signals before, but with everything having noise its hard to work out if something is actually wrong or not. I suspect this problem will start to show up more and more as the machines age.

I remember I spent a lot of time moddin the RGB signals, but not sure I actually scoped out the composite signal.. It could be some dried up caps around the RF modulator circuit, I always assumed it was a shifter related fault, but with signals going around the RF circuit aswell. The problem could may well be in that area. Some quality caps around the RF box on the 5V and 12V power rails would be a interesting test to do..

[exxos]

I dug out a STFM with this fault, though this machine does not boot, so I am unable to try much..

But in relation to the composite signal...

The scope on the left is what it is normally, the one on the right is with 60R resistor to gnd. While my STFM doesn't boot, it does "cure" the video sync fault.. but I don't know if it is simply killing the video totally or helping solve the issue because this machine doesn't boot.

nores.png

EDIT1:
I tried a 68R resistor on my working STFM, its not "upset" anything... I don't think loading composite with 68R is a good idea though. I'm trying to see the the "video amp" can be changed to see if the output level can be reduced that way next...

EDIT2:

The Output Stage
The output amplifier normally produces about 2.0 Vpp and is intended to be loaded with 150 Ω as shown in Figure 16.
This provides about 1.0 Vpp into 75 Ω, an industry standard level (RS–343). In some cases, the input to the monitor may
be through a large coupling capacitor. If so, it is necessary to connect a 150 Ω resistor from Pin 9 to ground to provide a low
impedance path to discharge the capacitor. The nominal average voltage at Pin 9 is over 4.0 V. The 150 Ω dc load
causes the current supply to rise another 30 mA (to approximately 60 mA total into Pin 14). Under this (normal)
condition the total device dissipation is about 600 mW. The calculated worst case die temperature rise is 60°C, but the
typical device in a test socket is only slightly warm to the touch at room temperature. The solid copper 20–pin lead
frame in a printed circuit board will be even more effectively cooled.

So loading the pin with 150R seems to be the "correct" way... Not happy about that thought because pulling another 30mA doesn't sound good to me..

EDIT3:
The video amp is located on pin8. A 2.7K resistor down to gnd has the same effect as the 68R on pin9. Looking at the composite signal, it only lowers in voltage a fraction. Though this *could* explain the borderline situation where some boards are ok and some are not.

EDIT4:
R11 which is 1.2K, with 4.7K in parallel it becomes 950R. So changing R11 for 950R should be the same as adding 4.7K on pin 8. Maybe someone who has this video fault can give it a try and report back ?

[czietz]

Great idea! I just tried -- with my STFM switched back to composite-video out, of course -- and it works! Like you, I didn't replace R11 but just paralleled a second resistor. 2.7k or even 10k fix the sync problem. If such a marginal change in resistor value (1.2k || 10k = 1071 ohms) solves the problem, then this totally explains why after so many years some machines are affected and some aren't.

I use my scope later to measure the voltage on the composite-video pin with and without resistor.

[exxos]

Great idea! I just tried -- with my STFM switched back to composite-video out, of course -- and it works! Like you, I didn't replace R11 but just paralleled a second resistor. 2.7k or even 10k fix the sync problem. If such a marginal change in resistor value (1.2k || 10k = 1071 ohms) solves the problem, then this totally explains why after so many years some machines are affected and some aren't.

I use my scope later to measure the voltage on the composite-video pin with and without resistor.

Awesome

I didn't take much notice of the voltage as it was only slight...

Just tried again, 4.8K is the value where it starts working. Anything higher, the fault is still there. So something like 2.2K would probably be a good value to settle on.

As for voltage, its only slight change, around 0.1V or so, will try and get better resolution on it, hard to tell with all the noise on it

[czietz]

Composite-Video, measured at STFM with SCART cable and TV connected.

Without additional resistor:

NewFile6.png

With 2.7K resistor to ground on Pin 8 of MC1377:

NewFile7.png

Interestingly in my case the resistor to ground at the luma input of the video encoder makes the color burst look nicer and slightly increases the black level in the "back porch" of the video signal to stand out more from the sync level. Maybe on a completely/mostly white screen the TV needs this correct black level reference to sync correctly?

[exxos]

Interestingly in my case the resistor to ground at the luma input of the video encoder makes the color burst look nicer and slightly increases the black level in the "back porch" of the video signal to stand out more from the sync level. Maybe on a completely/mostly white screen the TV needs this correct black level reference to sync correctly?

Looks like the voltage "drop" is pretty much what I am seeing here. Its almost nothing. Probably 0.1V - 0.2V something like that. Such a small voltage is likely why it fails as the machine ages and parts drift out of spec.

I'm not sure its down to the black level. I did try adding resistors to the RGB lines, so the entire RGB voltage was a lot lower. I would assume if the sync was related to the "black" regions it would have allowed it to sync ?

[czietz]

Now it starts to get really interesting. I compared the composite video signal at encoder IC (pin 9, output of the output amp, blue scope trace) and at the video out connector of the Atari (pin 2, yellow scope trace).

Without resistor:

NewFile11.png

With 2.7k:

NewFile12.png

As you can see, on the encoder IC itself, adding the resistor mostly decreases the white level by about 200 mV. Black level and color burst almost stay the same. However, on the actual video output, the white level almost doesn't change but -- as already noted in my previous post -- the black level increases and also the color burst looks better.

I'm starting to suspect the decoupling caps (C16, C21) because there isn't much else between pin 9 of the IC and pin 2 of the video connector. What do you think?

[exxos]

I'm starting to suspect the decoupling caps (C16, C21) because there isn't much else between pin 9 of the IC and pin 2 of the video connector. What do you think?

If you scope out the RGB lines from the shifter to the video connector, you will see all sorts of crazyness

Its possible those electrolytics have gone bad causing the supply rails to become unstable. It could explain the bad ripple in the signals. C16,C21 could indeed by bad also. Leaky capacitors will causes the DC bias to drift all over the place. Try it

[czietz]

The coupling caps (not "decoupling", of course) C16, C21 and also C22 turned out to be just fine: Capacitance still above the nominal value and no significant leakage.

Nonetheless, I found the DC bias level at the base of Q1 (the emitter follower that buffers the composite video) to be a little low. When the picture is (mostly) white, what Q1 sees are only short pulses going low from the DC level during and immediately after sync. When the DC bias is to low, the transistor can't properly follow these low-going pulses. This is why the color burst and the sync/black reference looked a bit "squashed" on my scope shots. When you lower the white level by maybe 200 mV by adding the parallel resistor at the encoder IC, the voltage swing of these low-going pulses decreases as well, so they can pass through Q1 unaltered again.

To increase the DC bias, I replaced R13 (nominally 22k) by a 27k resistor. This completely fixed the sync issues! Maybe this problem was exacerbated by the fact that in my STFM the resistor was only 21k, at its lower tolerance limit.

Of course I still don't know the root cause, i.e. why it worked in the first place with the original resistor: Did the voltage swing of the encoder increase? Did the Q1 or its base resistors age? However, the sync problem is fixed now and I'm happy. Maybe you want to try this on your broken STFM as well?

[exxos]

Have you tried changing the coupling caps for some new ones anyway ? Caps will be leaky after all these years. It only needs something slight..

I think this is just a age problem. Silicon goes leaky over time, same with electrolytics. Likely the ST was designed and worked with those value parts so wasn't any reason to think what would happen 30 years down the line. A lot of circuits are only borderline working to start with. something goes out of spec slightly and it plummets into chaos.

It's possible if you changed the transistor for a different one, that could fix the issue aswell. This also means if you change to a different brand of the same transistor it may or may not work. 22K based on 5V pulls 0.227mA.. Granted there is the 10K in there, likely a bias to around 3V or something, so the circuit is being biast will (at a guess) somewhere around 0.1mA. Get a slight bit of leakage in the IC itself, the transistor, the capacitor, or even a slightly out of spec resistor and the circuit goes crazy.

A example I will never forget, is I've built and designed & repaired a lot of audio gear over the years. Circuit has to be spot on everywhere, else you end up with distortions. I had once a small 2watt amplifier demo board which I re-designed and built my own PCB, same parts as the demo board. Didn't work. Did a second design. Didn't work. Built the demo board itself from the gerber files.. and.. didn't work. This confused me for a while. I later concluded the only difference between the PCB I had made and the demo board I purchased, was my board did not have the "green" solder mask. I just coated my board with some regular vanish, left to dry, worked perfect after that. Point being, almost no manufactures take into account any sort of tolerance or drift into their designs. I've seen circuits will 1megohm pull up resistors where you can change the voltage on that pin depending how close your hand is to the PCB.

Anyway, wouldn't upping the value of R13 increase the voltage on pin9 ? The tests so far seem to work when the voltage is lowered on pin 9 ?

Maybe you would try removing C16 and put a 1uF ceramic in there to try ? Can't see why 33uF is needed there. At least it would rule out leakage on the coupling capacitor.

[czietz]

Have you tried changing the coupling caps for some new ones anyway ? Caps will be leaky after all these years. It only needs something slight..

Unfortunately, I don't have any new caps bigger than 10 µF available. The bigger ones I found in the back of my parts drawer are also at least 15 years old, so they could be just as bad. Given that the original caps from my STFM still had the correct capacitance and a very low leakage current, I didn't see any point in exchanging them by other old caps.

Anyway, wouldn't upping the value of R13 increase the voltage on pin9 ? The tests so far seem to work when the voltage is lowered on pin 9 ?

Assuming the cap is still intact, the DC bias of Q1 should not show on pin 9 of the IC. When you lowered the (white level) voltage on pin 9, you also lowered the voltage swing so the emitter follower could still follow.

Maybe you would try removing C16 and put a 1uF ceramic in there to try ? Can't see why 33uF is needed there. At least it would rule out leakage on the coupling capacitor.

Again: 1 µF ceramic cap I would have to order. If you have one available, could you try on your (broken) STFM?

[exxos]

Again: 1 µF ceramic cap I would have to order. If you have one available, could you try on your (broken) STFM?

I will connect it back up shortly and give it a try, ive got a STE plugged in doing some stuff at the moment

Not sure its a totally fair test as I can only see a white screen on power up, Though I can see if the sync fault goes away at least, then swap the capacitor on a working STFM to see if anything changes. Though I am using RGB so changes on the composite line are likely to show any colour difference to me.

[exxos]

OK quick test with a 1UF (& 10UF after) ceramic, I would say it 90% solves the issue. It looks like there might be a faint black band at the top of the screen, it goes away with the resistor mod. As to if this will actually show on a real booted up machine I don't know.. Probably if that cap is bad, then all the others are aswell.

[czietz]

Seems like this circuit is really sensitive to changes and every component you replace has the chance of fixing the sync problem.

Given that my resistor solution (changing R13) also solves the problem completely, I doubt I will replace that cap in my STFM for the time being. I've also done some simulations of the circuit that confirm that color burst and black level get distorted in the way I observed it on the scope if either R13 becomes too small or if R12 becomes too large.

[exxos]

Seems like this circuit is really sensitive to changes and every component you replace has the chance of fixing the sync problem.

Given that my resistor solution (changing R13) also solves the problem completely, I doubt I will replace that cap in my STFM for the time being. I've also done some simulations of the circuit that confirm that color burst and black level get distorted in the way I observed it on the scope if either R13 becomes too small or if R12 becomes too large.

I agree.. It proves any variation in the circuit can cause it to start behaving. So in reverse, if any part goes out of spec, the circuit fails. Its possible the caps and resistor can be changed. at least the caps can cure the sync fault, and the resistor (can also cure the sync fault) but if it improves the colour then that could be a worthwhile mod to do anyway. Though I don't use composite video , im RGB all the way

Would the video be possible to record to show the differences ?

[czietz]

Like yours, my SCART cable also uses RGB for the actual video signal and the composite video is only used as sync signal. I have to rewire the cable to see if brightness or colors in the composite signal get affected by the various fixes.

[czietz]

OK, I recreated the sync fault by lowering R13 again and I rewired my SCART cable to use the composite video signal only instead of RGB.

As it turns out, what we perceived "merely" as sync issue with RGB also causes wrong colors within the composite signal. This probably due to the distorted color burst:

IMG_1901.JPG

Your first solution -- the resistor at pin 8 of the IC -- fixes the sync and color but slightly lowers the brightness of the composite video image. That is understandable, as the level of the composite video signal is also decreased a bit. The difference, however, is marginal and only noticeable in direct comparison.

My proposed solution for the sync issue -- increasing R13 -- seems to have no effect whatsoever on colors or brightness of the composite signal. Thus, I'll keep it. Anyway, with RGB -- which is of course the preferred solution for a SCART cable -- all solutions are equivalent.

[exxos]

Thanks for your feedback. Can you try a 1uF ceramic as the decoupling capacitor C16 ? The problem I see, is any changes to the bias network may only be short lived, as the capacitors age more, the issue could come back at some point. If changing the capacitor solves this, then I would be incluned to mod it that way. Though it's also possible all the decoupling caps might need changing.

[czietz]

If I remember to buy such a capacitor the next time I order electronic components, I can give it a try. May take some time, though.

[exxos]

Had a STFM arrive today with a similar fault. Not as bad as in my first video but not far off.

Anyway, Changing the coupling capacitor seemed to solve the black areas on the right of the screen, but created a flickering black bar at the top of the screen. I added the 2.7K on pin8, and that solved all the video issues. BUT, the odd thing is, when the desktop comes up, the video seems to jump upwards 1 pixel

I added 500R into R13 and the video problem is there all the time now, just a flicking black bar at the top of the screen

EDIT1: Seems the de-coupling capacitor value was to low, upping it solved the 1pixel jump thing. All seems stable again now.

[troed]

BUT, the odd thing is, when the desktop comes up, the video seems to jump upwards 1 pixel

Analog or digital monitor? Hard to see how it could happen on the Atari side, but digital screens can pretty much do anything on their side ...

/Troed

[exxos]

BUT, the odd thing is, when the desktop comes up, the video seems to jump upwards 1 pixel

Analog or digital monitor? Hard to see how it could happen on the Atari side, but digital screens can pretty much do anything on their side ...

/Troed

Its a LCD monitor, if you check the first post you can see it all.