123 replies to this topic

[cwilkson]

Ok, here's the schematic for the overkill circuit. I suggest you guys build and test it if you have the parts lying around. If you don't have a 3.3V zener, you can substitute a 300 Ohm resistor in desperation. You wont get the nice regulation filtering, but it will let you play with the rest of it.



Buffers, and buffers, and buffers! OH MY!!!

Opamps 1-3 are simple voltage buffers. Opamp 4 is used as a comparator/switch to yank voltages around.
Opamp 2 with the zener diode and the 200 Ohm resistor form a 3V power supply to drive the pot and the timing capacitor. Not really needed, but it does help reduce VCC noise that gets in and causes jitter. Rp is the pot. Opamp 1 buffers the pot voltage. The output is variable between 0V and 3V. The 330kOhm resistor charges the capacitor from the 3V supply. Opamp 3 buffers the capacitor voltage. Opamp 4 acts as a comparator. During charging, the capactitor voltage is less than the pot voltage, so the comparator output is low and both diodes are off. When the capacitor voltage exceeds the pot voltage, the comparator output goes high, to about ~4 volts. It yanks the capacitor voltage up to 3.3-3.4v through the diode which trips the TIA's paddle input. There is a positive feedback loop from the comparator through the diode and back through the capacitor buffer. So once the circuit is tripped, it trips itself even harder, latching the tripped state. This is done for switching speed and stability. (the feedback theorists are cringing now. good for them!) The second diode is used to pull the comparator's other input high when the comparator goes high. This is to guarantee that the TIA can pull the capacitor voltage below that input and reset the latch if the pot voltage is very low. The 10k Ohm resistor provides isolation between the outputs of opamps 1 and 4 when the diodes are turned on. It's value was chosen so that even if the pot voltage is at the maximum 3V this path will draw more diode current than the 330kOhm path. This guarantees that the inverting comparator input stays marginally lower than the non-inverting input until the TIA tries to discharge the capacitor.

So that's it. Pretty complicated, but I hope it's easy to understand. If anyone else builds it, I want to hear the results. If you have problems, or suggestions for improvements, or if you like it just as it is, or if you have questions about it.

Enjoy!
-Chris

[cwilkson]

A.J. Franzman, on Sat Apr 14, 2007 12:30 AM, said:

Buffering's not a problem but I don't know what a walking ring oscillator is, and I only have a few 74xxx series chips (of any flavor), which do not include a 74LS221 unless I can find one on one of my junk boards.

A ring oscillator is a built with a bunch of delay elements connected in a circle. The classic example uses an odd number of digital inverters. The period of the oscillation is equal to the sum of the individual propagation delays. One way to build a walking ring oscillator is by tying the serial output of a shift register to it's own serial input, load a single "1" into it, and then just let it run. (Think "ROTATE" instruction in assembly language.) If you look at the outputs of a 4-bit version, they have this kind of timing:

	_	   _
___| |_____| |____
	  _	   _
_____| |_____| |____
		_		_
_______| |______| |_
  _	   _		_
_| |_____| |______| |_

-Chris

[SpiceWare]

Jitter free paddles? wow - I'll have to keep an eye on this thread

[cwilkson]

I thought a transconductance amp had been outlawed for this? Due to not matching the original charging characteristics? Or maybe that was just me. Yeah, probably in my own head during the "I dont understand the point" part of this thread. If people are satisfied with that solution, it *is* the ideal approach. In it's essence, a single transistor and you're done. You can make it a little better by adding a resistor. You can make it a lot better with a resistor and an opamp. You can make it a little better than a lot better by adding a second resistor. If I can find an appropriate transistor tonight, I'll see what I can come up with.

-Chris

[supercat]

cwilkson, on Mon Apr 16, 2007 7:01 PM, said:

I thought a transconductance amp had been outlawed for this? Due to not matching the original charging characteristics? Or maybe that was just me. Yeah, probably in my own head during the "I dont understand the point" part of this thread. If people are satisfied with that solution, it *is* the ideal approach. In it's essence, a single transistor and you're done. You can make it a little better by adding a resistor. You can make it a lot better with a resistor and an opamp. You can make it a little better than a lot better by adding a second resistor. If I can find an appropriate transistor tonight, I'll see what I can come up with.

One of the difficulties of making a good circuit is producing moderately-linear behavior. Trying to have something that would produce a continuous current of the correct magnitude would be the goal, but I don't see any elegant way of doing it. If you're in a prototyping mood, something like the simple emitter-follower I diagrammed above might be easier.

[cwilkson]

supercat, on Tue Apr 17, 2007 2:22 AM, said:

One of the difficulties of making a good circuit is producing moderately-linear behavior. Trying to have something that would produce a continuous current of the correct magnitude would be the goal, but I don't see any elegant way of doing it. If you're in a prototyping mood, something like the simple emitter-follower I diagrammed above might be easier.

[looks back through thread]

I don't see an emitter follower. Do you mean the single-transistor amp you drew with [code] tags to work with PWM? If so, that's almost exactly what I'm talking about. BTW, there's a typo there. You might want to edit it. 2N3906 is what you meant. It's the PNP. Replace the PWM input with the pot and you have a pot-programmable constant current source. But the 2n3906 is not sufficient for that...the betaF is too low. Because the pot resistance is so high, if you want to use a single transistor you really need a FET. Also, the whole point here is to keep the wiper current = 0. So no matter what the pot value (if you were to replace it) you still need a FET. Or you can build a super-current source by adding an opamp to the mix, like this:

The negative feedback loop forces the inverting input to equal the pot voltage. Neglecting base current (~1% error) the output current is equal to (VCC-Vpot)/R. If a linear current source is acceptable, then this is what you want. The problem once again is the common mode input range. The output of the opamp shouldn't go below the 2V trip point of the TIA (allowing for a few hundred mV of margin here). That means that V- can't go below about 2.7V. But most cheap and readily available opamps choose to include GND in the common mode input range instead of the positive supply. So there's a top-end limit too. A good number to carry around for the top of the range is 3.5V for a 5V supply. So....that means that the valid range for the pot voltage (allowing a safety margin at the top) would be between 2.7V and 3V (plus a little bit). If you go below that, the circuit malfuctions because the capacitor can never charge to 2V. If you go above it the opamp itself malfunctions in a way that depends on the amplifier type. You could bias the pot by adding a resistor at each end, but then you only have about half a volt of swing. Not very precise. If you get a rail-rail opamp, things get easy.

-Chris

[supercat]

cwilkson, on Mon Apr 16, 2007 10:26 PM, said:

I don't see an emitter follower.

This one:

http://www.atariage....s...t&p=1277202

The wiper current wouldn't be quite zero, but would be a small fraction of what it normally would; jitter should thus be reduced considerably. Adding a filter cap to the wiper (connect other end to ground) may help further.

[A.J. Franzman]

Tonight I've built two more circuits from this topic, supercat's from post #97, and cwilkson's from post #101. I also tried re-connecting the pot I've been using per the original design but still using micro-clip test leads.

Things I've noticed:
A) The circuit from post #97 works, if the resistor between +Vcc and the top of the pot is removed (connect the pot to Vcc). But, the usable range of the pot is only about 20 degrees instead of 90 in Video Olympics -- probably a bit too fiddly. Also, I'm still seeing an unacceptable amount of drift.

B) The circuit from post #101 works pretty much as-drawn, though I used 1N4148 diodes and a 220 ohm resistor instead of the 1N914s and 200 ohms. But I'm also seeing about 4-6 pixels of drift with a period of about 14 seconds, when the paddle onscreen in Video Olympics Game 1 is about 1/2 to 2/3 off the screen at the bottom. Also, I don't know if this was intentional or not, but my 3.3 volt zener actually runs at about 2.7 volts with such a low current through it. Replacing it with a 3.6 volt zener gives about 3.0 volts, but does not seem to affect the circuit operation at all. I have a 1-watt 3.9 volt zener that I could also try, but I didn't bother because of the hassle of connecting it (leads too fat for the breadboard) and not seeing any difference with the previous change. Finally, I noticed that the polarity of the pot is reversed with this circuit -- the ground wire has to go on the end where the yellow wire is usually found. Control motion range seems identical to the original, but slightly shifted toward the right end (this is the same end where there wasn't much physical travel remaining in the original design, so now there's even less). Perhaps this is due to offset in the op-amps and could be corrected, or even over-corrected to bring the useful range nearer to the middle of the physical range of the pots.

C) The pot I've been using no longer jitters! I don't know if it's just because it had some stiff grease or other contamination that got scraped away by all the use it's been getting, or if using it in voltage mode somehow "restored" the resistive element. Fortunately, I do have a few other jittery pots, and enough parts and cables to build at least one pair of paddle controllers using them. Also, when connecting the pot per the original design using the clip leads, I get no drift or only one pixel of drift. But I find it increasingly hard to believe that all of these circuits share a flaw that would cause the higher amounts of drift that I've been getting; perhaps it's something wrong with my solderless breadboard?

Edited by A.J. Franzman, Mon Apr 16, 2007 10:21 PM.

[supercat]

A.J. Franzman, on Mon Apr 16, 2007 10:47 PM, said:

A) The circuit from post #97 works, if the resistor between +Vcc and the top of the pot is removed (connect the pot to Vcc). But, the usable range of the pot is only about 20 degrees instead of 90 in Video Olympics -- probably a bit too fiddly. Also, I'm still seeing an unacceptable amount of drift.

I guess it's necessary to include resistors on both sides of the pot then. Still, the fact that it uses a couple of dirt-common 3904's instead of any sort of exotic chips should make this a cheap and easy circuit to build.

Quote

Also, when connecting the pot per the original design using the clip leads, I get no drift or only one pixel of drift. But I find it increasingly hard to believe that all of these circuits share a flaw that would cause the higher amounts of drift that I've been getting; perhaps it's something wrong with my solderless breadboard?

That is very strange. It could be that the dynamic behavior of the TIA's inputs can vary somewhat in a fashion that doesn't affect how they behave when using a resistor, but such variance would be likely to affect one or the other circuit much more than the other.

I don't know how convenient it is for you to throw things together, but I'd be curious, if you were to build two of each of my circuits and plugged them all in at once, whether they'd drift together or independently.

Not sure what the best software to test that would be; Video Olympics reads one pot at a time, but I could send you a program to read all four at once if you want.

[A.J. Franzman]

supercat, on Mon Apr 16, 2007 9:10 PM, said:

I don't know how convenient it is for you to throw things together, but I'd be curious, if you were to build two of each of my circuits and plugged them all in at once, whether they'd drift together or independently.

Not sure what the best software to test that would be; Video Olympics reads one pot at a time, but I could send you a program to read all four at once if you want.

I have no way to run a program from a ROM .BIN file on a real console, so unless you want to s-mail your program on a cart, that's no good. If you know of any game that reads the pots in a way that you'd like me to test, you can see if I have it. My complete game collection is listed in the file from the first link in my sig, beginning just over halfway through it; the section entitled, "HAVE (to keep?):".

Edited by A.J. Franzman, Mon Apr 16, 2007 10:32 PM.

[cwilkson]

supercat, on Tue Apr 17, 2007 3:34 AM, said:

http://www.atariage....s...t&p=1277202

The wiper current wouldn't be quite zero, but would be a small fraction of what it normally would; jitter should thus be reduced considerably. Adding a filter cap to the wiper (connect other end to ground) may help further.

Whoops. I missed it. Yeah, that should work. It's basically the same output stage as your original LM311/diode circuit. But you might be limited in range of the delay generated. Because those 2n3904's are going to see a beta of somewhere around 20-40 with the small collector currents you're trying to use. Maybe less. So R1 and the top part of the pot might develop more voltage drop than is expected. Might work better if you switch the pot and R1. And if you do that, you might need to experiment with the value of R1.

I think using a FET will give a better function.* It'll help you avoid the I*R drop in the pot. I'm going to try that when I get home. If I can find the transistors that I think have.

-Chris

* I hate FETs. I'm a BJT guy....such lovely gms!! But sometimes base current kills you and you have to bend your preferences a bit.

[cwilkson]

A.J. Franzman, on Tue Apr 17, 2007 3:47 AM, said:

A) The circuit from post #97 works, if the resistor between +Vcc and the top of the pot is removed (connect the pot to Vcc). But, the usable range of the pot is only about 20 degrees instead of 90 in Video Olympics -- probably a bit too fiddly. Also, I'm still seeing an unacceptable amount of drift.

I'm starting to think you have really dirty power. Also make extra extra extra sure that the center lead of the pot is short. I know that you can't make it perfect when testing, but pickup on a 1Meg pot is a **killer**! Do you have a o'scope? I'd be interested in seeing what the voltage on your pot output looks like when you set the pot to the center. (with nothing else in the circuit)

Quote

B) The circuit from post #101 works pretty much as-drawn, though I used 1N4148 diodes and a 220 ohm resistor instead of the 1N914s and 200 ohms. But I'm also seeing about 4-6 pixels of drift with a period of about 14 seconds, when the paddle onscreen in Video Olympics Game 1 is about 1/2 to 2/3 off the screen at the bottom. Also, I don't know if this was intentional or not, but my 3.3 volt zener actually runs at about 2.7 volts with such a low current through it. Replacing it with a 3.6 volt zener gives about 3.0 volts, but does not seem to affect the circuit operation at all. I have a 1-watt 3.9 volt zener that I could also try, but I didn't bother because of the hassle of connecting it (leads too fat for the breadboard) and not seeing any difference with the previous change. Finally, I noticed that the polarity of the pot is reversed with this circuit -- the ground wire has to go on the end where the yellow wire is usually found. Control motion range seems identical to the original, but slightly shifted toward the right end (this is the same end where there wasn't much physical travel remaining in the original design, so now there's even less). Perhaps this is due to offset in the op-amps and could be corrected, or even over-corrected to bring the useful range nearer to the middle of the physical range of the pots.

1n914 and 1n4148 are functionally very similar. Maybe even drop in replacements. The 220 Ohm resistor will drop the zener current a bit, and therefore the voltage. But you're right....the actual value isn't super critical. In fact, mine was 3.0V. So yeah, don't bother changing the diode. I should've mentioned the reversed leads on the pot. Sorry. I assume you're still using Video Olympics #33 for testing? and you want to move the range so that for a given paddle position the paddle is farther to the right? If that's true, then I think you can reduce the 330k resistor to fix it. Try a 220k Ohm instead. That should give you another...20(?) degrees of margin at the end of the pots travel. I hesitate to make it any smaller. It might start to affect the other end of the range by too much. Actually, you might be able to get a little benefit from increasing the zener voltage too. But that would also raise the pot voltage making the problem worse. A better approach is to pad the top of the pot with a resistor to the "3V supply". Anyway, I think reducing the 330k to 220k will work for you. Let me know how it compares.

Quote

C) The pot I've been using no longer jitters! I don't know if it's just because it had some stiff grease or other contamination that got scraped away by all the use it's been getting, or if using it in voltage mode somehow "restored" the resistive element. Fortunately, I do have a few other jittery pots, and enough parts and cables to build at least one pair of paddle controllers using them. Also, when connecting the pot per the original design using the clip leads, I get no drift or only one pixel of drift. But I find it increasingly hard to believe that all of these circuits share a flaw that would cause the higher amounts of drift that I've been getting; perhaps it's something wrong with my solderless breadboard?

I had the same "problem" with my paddles. They got better with the repeated use. Maybe that's the real solution for us all. Just use them! They'll get better!!! Heh. I had to track down a filthy pair to do my testing. About your drift problem. Very strange. What if you power the paddle circuit with batteries (use 4 NiCad's=5V) and connect only GND and paddle output to the Atari? I'm stumped on this one. Actually, how old is your breadboard? If the contacts are tarnished at all they could be having a significant effect on that huge pot resistance. You're back to a dirty wiper all over again!

-Chris

[cwilkson]

OK, here's the FET follower circuit. If you have a 2N7000 N-channel MOSFET available try it out.
Let me know how the range feels....

[A.J. Franzman]

cwilkson, on Tue Apr 17, 2007 3:03 AM, said:

OK, here's the FET follower circuit. If you have a 2N7000 N-channel MOSFET available try it out.
Let me know how the range feels....

I kinda doubt it, but if there's an NTE equivalent I can check my local Fry's for it.

Re the range, I put masking tape on the knob and body of my test pot and marked it for better precision.

[supercat]

cwilkson, on Tue Apr 17, 2007 2:06 AM, said:

Whoops. I missed it. Yeah, that should work. It's basically the same output stage as your original LM311/diode circuit. But you might be limited in range of the delay generated.

The design philisophy is entirely opposite the LM311 circuit. In that circuit, the goal was to start the cap slow charging, and then fast-charge it the rest of the way after a delay controlled by the pot. In this circuit, the goal is to charge a cap quickly to a voltage set by the pot, and then let it slowly charge the rest of the way.

Quote

Because those 2n3904's are going to see a beta of somewhere around 20-40 with the small collector currents you're trying to use. Maybe less. So R1 and the top part of the pot might develop more voltage drop than is expected. Might work better if you switch the pot and R1. And if you do that, you might need to experiment with the value of R1.

Actually, I think my problem was failing to consider that I'll have two diode drops on the B-E junctions, and that I need to have my output voltage get up to slightly over 2 volts; that's going to require 3.4 volts on the base of the first transistor--obviously a 1M pot on the high side of the pot is going to cause a problem.

Quote

I think using a FET will give a better function.* It'll help you avoid the I*R drop in the pot. I'm going to try that when I get home. If I can find the transistors that I think have.

One annoyance with FETs is that when used in the analog domain it can be hard to make a circuit that isn't heavily reliant upon characteristics that vary between different types of FETs.

[supercat]

cwilkson, on Tue Apr 17, 2007 5:03 AM, said:

OK, here's the FET follower circuit. If you have a 2N7000 N-channel MOSFET available try it out.
Let me know how the range feels....

How does the current/voltage "knee" look on that circuit? On the BJT circuit the knee is set by the emitter resistor. How will it look with the FET?

Certainly if that circuit works well it's as simple as one could hope for.

[supercat]

cwilkson, on Tue Apr 17, 2007 3:19 AM, said:

I know that you can't make it perfect when testing, but pickup on a 1Meg pot is a **killer**! Do you have a o'scope?

Why isn't there a lightbulb emoticon?

I suspect that what's being observed is 60Hz pickup on the high-impedance input to the circuit. To test this, stick a 1uF cap between the pot wiper input to the circuit and ground. Pot behavior would get very "mushy" with the cap installed, but it should kill the wandering.

[cwilkson]

supercat, on Wed Apr 18, 2007 4:07 AM, said:

The design philisophy is entirely opposite the LM311 circuit. In that circuit, the goal was to start the cap slow charging, and then fast-charge it the rest of the way after a delay controlled by the pot. In this circuit, the goal is to charge a cap quickly to a voltage set by the pot, and then let it slowly charge the rest of the way.

My point was the output stage works the same way. Not the overall circuit.

Quote

Actually, I think my problem was failing to consider that I'll have two diode drops on the B-E junctions, and that I need to have my output voltage get up to slightly over 2 volts; that's going to require 3.4 volts on the base of the first transistor--obviously a 1M pot on the high side of the pot is going to cause a problem.

Or....that could be it.

Quote

One annoyance with FETs is that when used in the analog domain it can be hard to make a circuit that isn't heavily reliant upon characteristics that vary between different types of FETs.

Couldn't agree more. I hate FETs. But if you need low DC input current, sometimes you gotta bite the bullet and use one.

Quote

How does the current/voltage "knee" look on that circuit? On the BJT circuit the knee is set by the emitter resistor. How will it look with the FET? Certainly if that circuit works well it's as simple as one could hope for.

I'm not sure what you mean here. Can you be more specific?

-Chris

[cwilkson]

supercat, on Wed Apr 18, 2007 4:14 AM, said:

Why isn't there a lightbulb emoticon?

Lemme think about it.




I dunno. What they really need is one for "smartass".
(use "colon idea colon" for lightbulb....it's in the full list)

Quote

I suspect that what's being observed is 60Hz pickup on the high-impedance input to the circuit. To test this, stick a 1uF cap between the pot wiper input to the circuit and ground. Pot behavior would get very "mushy" with the cap installed, but it should kill the wandering.

I agree. I think there's pickup in the system that's making the pot voltage dirty @60Hz. And because the video (and therefore dumping the pots) is running at *almost* 60 Hz, there's a beat frequency being developed at (60-almost 60)Hz = equals Real Slow. Like with a period of several seconds.

Assuming the oscillator is still running at spec frequency (there's no way, but it'll be pretty close) my math says that should be a period of 12.9 seconds.

A.J. - is that anywhere close to what you're seeing? Pot drift repeating every 13 seconds?

Edit: BTW...if that's right you need a pretty big cap to get even modest filtering. A 0.0053uF cap would give you an attenuation of ~20dB (=1 tenth) when the pot's in the middle. Might be really sluggish going through there. Check it out.

-Chris

Edited by cwilkson, Tue Apr 17, 2007 11:56 PM.

[supercat]

cwilkson, on Wed Apr 18, 2007 12:44 AM, said:

Edit: BTW...if that's right you need a pretty big cap to get even modest filtering. 0.01uF would give you an attenuation of 20dB when the pot's in the middle. Might be really sluggish.

Well, I suggested 1uF to test the notion. If 60Hz pickup is the problem, the solution is to convert the high-impedance signal to a low-impedance one as near to the pot as possible, but that may be easier on a small PC board than on a proto board.

[A.J. Franzman]

cwilkson, on Tue Apr 17, 2007 1:19 AM, said:

I'm starting to think you have really dirty power. Also make extra extra extra sure that the center lead of the pot is short. I know that you can't make it perfect when testing, but pickup on a 1Meg pot is a **killer**! Do you have a o'scope? I'd be interested in seeing what the voltage on your pot output looks like when you set the pot to the center. (with nothing else in the circuit)
<snip>
About your drift problem. Very strange. What if you power the paddle circuit with batteries (use 4 NiCad's=5V) and connect only GND and paddle output to the Atari? I'm stumped on this one. Actually, how old is your breadboard? If the contacts are tarnished at all they could be having a significant effect on that huge pot resistance. You're back to a dirty wiper all over again!

I don't know what I can do to shorten the wiper lead, but I did check the pot's output voltage way back right after I first reported the drift problem and it was steady down to +1 / -0 mV while the paddle onscreen was drifting ~10% of the available travel. I don't think the power itself is "dirty", but I suppose it is possible that I have high EM fields where I've been working. I do not have a working oscilloscope -- just a tiny antique half the size of a shoebox, with no built-in sweep circuit, but it's dead. No room or budget for a working scope.

I can try the battery trick if you think it will reveal anything, though I don't think I have any good NiCd's. I do have some AA NiMH cells for the digital camera. My breadboard's at least 20 years old but I don't think its resistance is that bad. If there's a particular path in a particular circuit that needs to be a very low resistance I can build it and test it for you using my Dick Smith kit-built ESR and low ohms meter which goes down to .01 ohm. Or, I could just take a sample of several contact paths and see if any are out of whack.

[supercat]

A.J. Franzman, on Wed Apr 18, 2007 1:17 AM, said:

I can try the battery trick if you think it will reveal anything, though I don't think I have any good NiCd's. I do have some AA NiMH cells for the digital camera. My breadboard's at least 20 years old but I don't think its resistance is that bad. If there's a particular path in a particular circuit that needs to be a very low resistance I can build it and test it for you using my Dick Smith kit-built ESR and low ohms meter which goes down to .01 ohm. Or, I could just take a sample of several contact paths and see if any are out of whack.

I wouldn't expect using a battery instead of the 2600 for power would change much; I would expect the real key to be the connection between the pot wiper and the amplifying circuit.

[A.J. Franzman]

supercat, on Wed Apr 18, 2007 4:41 PM, said:

I wouldn't expect using a battery instead of the 2600 for power would change much; I would expect the real key to be the connection between the pot wiper and the amplifying circuit.

Since all the circuits we're playing with here are meant to have low or no wiper current, specifically to eliminate the effect of wiper resistance on circuit operation, how would you expect an additional fraction of an ohm of resistance in the breadboard path of the wiper connection to affect the output?

Certainly you must realize that for a typical paddle controller with severe jitter, the wiper resistance must be varying on the order of 10k ohms or more (possibly including very brief open-circuit transients).

Edited by A.J. Franzman, Wed Apr 18, 2007 7:36 PM.

[cwilkson]

supercat, on Wed Apr 18, 2007 5:55 AM, said:

Well, I suggested 1uF to test the notion. If 60Hz pickup is the problem, the solution is to convert the high-impedance signal to a low-impedance one as near to the pot as possible, but that may be easier on a small PC board than on a proto board.

Got it. I thought it was an intended solution. Though I fear that might be the only practical hope if the leads can't be made shorter. That or a metal enclosure.

A.J. Franzman, on Wed Apr 18, 2007 6:17 AM, said:

I don't know what I can do to shorten the wiper lead, but I did check the pot's output voltage way back right after I first reported the drift problem and it was steady down to +1 / -0 mV while the paddle onscreen was drifting ~10% of the available travel.
<snip>
I do not have a working oscilloscope,

Nothing beats a picture. It's a shame that you don't have a scope. And it's a bigger shame that they're so expensive. Can you borrow some time on one at a local school/TV repair/etc? The problem with volt meters is that they tend to average their readings by design (to get a steady reading). And even if it was showing the signal in real time you can't see that fast.
I wouldn't bother with the batteries. The more we talk about it, the more I'm convinced that it's the same pickup issue I was having, only much much worse. About the shortest you could make that lead is to solder the input of the amp you're using (be it opamp input or transistor base/gate) directly to the center connector of the pot with no wire in between. If you can't shorten that node, it's time to grab the aluminum foil (or a metal project box if you have one).

A.J. Franzman, on Thu Apr 19, 2007 1:32 AM, said:

supercat, on Wed Apr 18, 2007 4:41 PM, said:

I wouldn't expect using a battery instead of the 2600 for power would change much; I would expect the real key to be the connection between the pot wiper and the amplifying circuit.

Since all the circuits we're playing with here are meant to have low or no wiper current, specifically to eliminate the effect of wiper resistance on circuit operation, how would you expect an additional fraction of an ohm of resistance in the breadboard path of the wiper connection to affect the output?

Certainly you must realize that for a typical paddle controller with severe jitter, the wiper resistance must be varying on the order of 10k ohms or more (possibly including very brief open-circuit transients).

I think what he means by "connection" is the node/wire itself is susceptible to pickup (specifically "hum" in this case). The high impedance node acts as an antenna. Generally speaking, the longer it is the more effective it is. (bad in this case)

On the protoboard...if the connections are tarnished, then it's the same exact effect that's happening with the pot wiper. And yes, it can vary from a nice short circuit connection to effectively an open. Granted, that's extreme. But large variation in protoboard resistance (especially in older boards) is standard fare. I don't think that's your problem in this case though. It's gotta be pickup. One simple test you can perform is to grab the wire that's connected to the pot's center lead and see what effect that has on paddle jitter/drift. Don't touch the conductor....just the insulation. (The human body is an *amazing* 60 Hz antenna!)

-Chris