Atari 2600 Power Noise vs Video output

[CkRtech]

Greetings EEVblog,

In the modern era, it is common practice among those interested in old video game consoles - the Atari 2600 in this case - to modify the video output to something more convenient. For the record, I still have a few 300 ohm RF CRT sets to use - but my issue and questions stem from observing composite video and s-video output. The Atari 2600 went through numerous revisions when it was manufactured, and my unit here is revision 14 (woodgrain, 4-switch).

TL;DR - I think the custom AV chip ("TIA") is rather sensitive to noise, and I am wondering if feeding it cleaner power would help. And I am wondering how I should best do that.

This story started with something I noticed in this unit with two different video mods - diagonal lines of interference in an otherwise fairly decent video output. The lines run from the top right to the bottom left. At once point I noticed I could rotate them and reduce/increase their thickness if I messed with the power switch. I thought, mistakenly, that my problem might be from the switch - and I then later noticed that I could also alter their orientation and thickness by pushing the cartridge around a bit.

An example of these lines can be seen in the purple at the bottom of the first screen in this video, and in the blue sky during the second part. They are much more pronounced in person -

There seemed to be no change after doing a bit of a refresh (new power jack, 7805, electrolytics, a film cap, etc. - some of which were part of an early 80s service bulletin) as well as trying a different video mod.

At some point between that and : doing a cartridge port reflow, adding a 220uf cap to the 7805 and a 220uf cap + .1uf ceramic to the TIA power->ground, the lines cleared up A LOT. They don't dominate the signal the way they used to. Sometimes, they are barely noticeable.

After talking to another Atari guy, we wondered what might be done with the power feed to the TIA chip (could an inductor in series with the power feed to the chip help, etc.)

So I reviewed "EEVblog #594 - How To Measure Power Supply Ripple & Noise" to double check my settings, did some measuring, and ended up here to get some thoughts.

I am ashamed to admit that I did still use the ground clip from a single probe to do the measurement.

I also noticed a change in the intensity of the power signal when going from one screen to another for the game I had hooked up at the time (used composite video to a PVM monitor - standard 75 Ohm termination)

Open to thoughts and opinions.

[David Hess]

At some point between that and : doing a cartridge port reflow, adding a 220uf cap to the 7805 and a 220uf cap + .1uf ceramic to the TIA power->ground, the lines cleared up A LOT. They don't dominate the signal the way they used to. Sometimes, they are barely noticeable.

After talking to another Atari guy, we wondered what might be done with the power feed to the TIA chip (could an inductor in series with the power feed to the chip help, etc.)

There is a good chance that adding LC decoupling to the TIA power would work well.  Adding LC decoupling to the source of the interference if you can find it should further improve the situation.  It might be necessary to use a low value resistor in parallel with the inductor to keep the Q down.

When modifying an existing design, I have sometimes added small solid tantalum capacitors directly across the ground and Vcc lines from pin to pin of the integrated circuits in question with the shortest possible leads.

[CkRtech]

Thanks, David.

What about the value of the inductor?

Could I possibly pull one off a parts board (386 motherbord with some LCs at the keyboard port), lift the +5V pin from the socket, solder the LC + a resistor (10 ohm?) to the lifted +5V pin and get the other side of the two to components to ground as close as possible? (Been skimming this a bit - http://www.analog.com/media/en/training-seminars/tutorials/MT-101.pdf)

I wish I knew the source of the interference (or how to best find it) - I do think that it perhaps changes depending on the cartridges used in the system.

It seems like C204 should be closer to the power side than the ground side, or does it make a difference? The chip's +5V feed that I am trying to filter actually runs under the entire chip prior to via-ing to the other side and connecting to the power pin on the TIA. Is there anything "wrong" with this design? I attached a photo pointing out my questions - wondering if it could also have a noticeable effect on interference in the TIA.

[David Hess]

What about the value of the inductor?

The self resonate frequency of the inductor limits performance and the inductor has to carry the DC current without saturation so the same types of inductors used for switching power supplies are appropriate.  I usually find whatever is in my junk box and make a single winding without close spacing using a wire large enough to make the winding stiff.  An inductance of 10s to 100s of microhenries is typical.

Could I possibly pull one off a parts board (386 motherbord with some LCs at the keyboard port) ...

They might not support the DC current you need without saturating.

It seems like C204 should be closer to the power side than the ground side, or does it make a difference? The chip's +5V feed that I am trying to filter actually runs under the entire chip prior to via-ing to the other side and connecting to the power pin on the TIA. Is there anything "wrong" with this design? I attached a photo pointing out my questions - wondering if it could also have a noticeable effect on interference in the TIA.

The schematic does not show the board interface so it is not really clear.  Usually small bulk decoupling capacitors are located where power enters the board and small ceramic capacitors are located at the loads which in this case are the ICs.

Going by the schematic, if 1 microfarad C200 is the bulk input decoupling capacitor, then it looks like they added 22 microfarad C204 because they were having problems with interference.  I think C204 *at the TIA chip* is where the LC decoupling network should be placed.

If C204 is worn out, then that would explain the problem.  They used such a large value to get a low enough ESR at high frequencies.  A sold tantalum capacitor would achieve the same performance with 1/4 the capacitance and not wear out.

[CkRtech]

The self resonate frequency of the inductor limits performance and the inductor has to carry the DC current without saturation so the same types of inductors used for switching power supplies are appropriate.  I usually find whatever is in my junk box and make a single winding without close spacing using a wire large enough to make the winding stiff.  An inductance of 10s to 100s of microhenries is typical.

Could I possibly pull one off a parts board (386 motherbord with some LCs at the keyboard port) ...

They might not support the DC current you need without saturating.

Ahhh. Well - I do have a few PC power supplies available for parts such as this one with a handful of various inductors. Seven in the included photo, for instance.

then it looks like they added 22 microfarad C204 because they were having problems with interference.  I think C204 *at the TIA chip* is where the LC decoupling network should be placed.

So C204 was actually replaced prior to my starting this thread, and it seemed to not make a difference. Despite being next to the TIA chip itself, it is not close to the +5v input pin. When you emphasized "at the TIA chip" for C204 being where the LC decoupling network should be placed, would it be OK to:

1: Leave C204 where it is
2: Grab power from the VIA near TIA +5V pin and then run it into the LC decoupling network
3: Solder other side of LC net to lifted +5V pin? (TIA chip is socketed)

I've included a component side photo (stock photo from elsewhere) that shows C204 relative to the TIA chip as well as a crude mock-up.

If you think that will work with any of the inductors from the power supply, I can wire it up as illustrated.

[David Hess]

Ahhh. Well - I do have a few PC power supplies available for parts such as this one with a handful of various inductors. Seven in the included photo, for instance.

All of those will likely work but those toroidal inductors are pretty large.  The bar shaped ones rely on the air gap between the ends to prevent saturation but have a lower inductance and are not shielded; on the other hand, they are effectively being used for exactly this application.  The larger of the two I marked shown below are the ones I would try first.

Rewinding the inductor with a single layer of 24 to 28 gauge insulated wire to get more turns will likely make it more effective.   The limit here is that too many turns will cause the inductor to saturate but I doubt the TIA chip draws enough current for that to be a problem.

then it looks like they added 22 microfarad C204 because they were having problems with interference.  I think C204 *at the TIA chip* is where the LC decoupling network should be placed.

So C204 was actually replaced prior to my starting this thread, and it seemed to not make a difference. Despite being next to the TIA chip itself, it is not close to the +5v input pin. When you emphasized "at the TIA chip" for C204 being where the LC decoupling network should be placed, would it be OK to:

1: Leave C204 where it is
2: Grab power from the VIA near TIA +5V pin and then run it into the LC decoupling network
3: Solder other side of LC net to lifted +5V pin? (TIA chip is socketed)

Yes, that is what I would do provided that C204 is going directly between the TIA's +5V pin and the TIA's ground pin.  If you use that big C204, then a resistor across the inductor will not be necessary.

If the noise is getting into the TIA via its ground pin due to a ground loop, then things get complicated.

I should have said this earlier but that Analog Device's application note is a great resource and shows exactly what is needed.  Analog Devices AN-202 and AN-345 also cover this subject with more detail about ground loops.

[BrianHG]

Looking at the age of the PCB, load of other TTL parts in the system, and some modern knowledge, my choice of action would not to be pulling up the VCC pin and using an inductor.  I would at both sides of that VCC power trace, add 2 ultra-low ESR 330uf caps to the GND shield bar just below the IC.  and, for the video output, I would add a 220pf anywhere to 1000pf cap to that GND bar depending on the IC's output drive current.  When scoping, I would GND my scope pin to that GND bar as well.

There is a lot of TTL lines going in and out of that IC and just a simple small VCC filter on it's power wont do much of anything if all the noise is being conducted through all the digital IO.  I would also add a few more of those ultra-low ESR caps by any other ICs which are the warmest / IE consuming most of the current, and the 1 by the cartridge port.

[CkRtech]

Brian - Could you elaborate a bit about the application of ultra-low ESR filter caps in multiple power locations? If the noise is coming from the digital I/O, how will the power filters help out?

Panasonic FR/FM electrolytics good enough or are you talking like polymer - Nichicon RR5 (6.3v)?

Regarding filtering video output (multiple digital luma signals + color signal) that feeds to a modern board to generate composite and s-video - I have a good number of 0.1uf ceramics on-hand (https://www.digikey.com/products/en?keywords=478-4274-1-nd) that I have had around for <looks at packing date> 10 years... if you they would work.

I'd have to order ultra-low ESR caps, anyway - so I guess I am open to newer. more specific options.

[BrianHG]

You have a problem with a 2 layer board, not knowing where this noise is coming from, and, without a 4 layer PCB with 1 solid ground plane, your options are limited.  You are right that if adjacent IC switching noise feeding into the TIA chip cannot be completely erased.  All I'm trying to do is lower the impedance from VCC to GND as much as possible to give you the best possible results.  However, without an Atari unit here, I cannot help you.  I know my Atari 800 had a pristine video signal, however, it's cpu was mounted in a cast (I think Aluminum') RF shielded brick which was all grounded with the PCB ground screwed in joints all over the place and it had the GTIA chip.

Check to see if that big flat soldered trace at the edge of the PCB is GND.  If so, continue:
That red wire on the 220uf cap you showed is way too long.
I would solder that cap's GND right to the GND buss just above at the edge of the PCB.
Add a 10uf and or 1uf tantalum cap in parallel with the 220uf.
I would also add a thick copper solid core conductor from the TIA chip's GND to that GND bar just above at the edge of the PCB as well.
I would also add a 470pf cap on the video output pins to that GND bar as well.
I would also scope probe the GND to that GND bar as well.
Also, if you are adding RCA/S-Video jacks, their GNDs should go there too.

I would try this first before investing in ultra low ESR caps.
You can use tantalum caps in place of the ultra low ESR caps.  They should be still 10x better than your 220uf cap.
The 220uf cap is still useful for lower frequency noise.

[CkRtech]

Thank you for this!

I know my Atari 800 had a pristine video signal, however, it's cpu was mounted in a cast (I think Aluminum') RF shielded brick which was all grounded with the PCB ground screwed in joints all over the place and it had the GTIA chip.

I've had my Atari 800 open as well, and it definitely has a giant block of metal that sits over the entire motherboard. The power supply is outside of it and connects to the motherboard via a pin header.

Check to see if that big flat soldered trace at the edge of the PCB is GND.

It is. The bar is one part of the metal "lips" around the board (both sides) that kiss the metal shielding halves of a sandwich. My shielding is no longer present (and I do not have it). I included a stock photo from elsewhere to show the board with the racetrack area around the ICs and cart slot.

Add a 10uf and or 1uf tantalum cap in parallel with the 220uf.

Darn. Looks like I need to get some of those. I have ceramics everywhere, but no tants.

I would also add a thick copper solid core conductor from the TIA chip's GND to that GND bar just above at the edge of the PCB as well.

Solid 22 AWG enough?

I would also add a 470pf cap on the video output pins to that GND bar as well.

Same type? Tantalum?

[BrianHG]

Solid 22 AWG enough?

Too thin.  you want to make a very low impedance ground.  Try something like twisting 5 together & add a solder to thicken up.

I would also add a 470pf cap on the video output pins to that GND bar as well.
Same type? Tantalum?

Ceramic down at those low capacitance values.