Random Number Generator Question

[soggybag]

I just built up Doug Deeper's excellent RNG. Everything seems to be working. The sound is super glitchy, over powering fuzz tone with random pseudo arpeggiation, and sub octave.

My build has a very interesting problem. There is a hum/oscillation that occurs when the unit has been idle for a few seconds. After playing and muting the strings, the unit is dead silent. Then, a few seconds go by and there is sort of schweezip sort of noise that ends in a constant hum/oscillation.

The transistors listed for this are a little odd: 2N4401, PN2222, 2N3904, well the first two anyway. I had a 2N2222 on hand and used that in place of the PN2222, hey they both had four 2's in the name how different could they be?

Anyway, if anyone has any insight into this let me know. I figure, if my build is correct, then the problem probably has something to do with the transistor choices. I can probably swap them out until I find some that are compatible.

[rousejeremy]

I had one that did the same thing from time to time. Try using shielded cable on the input jack to the switch and from the switch to the input of the board.

[PRR]

> a little odd: 2N4401, PN2222, 2N3904

These are all jelly-bean transistors.

The 2N2219 was a milestone in transistor history. Good ratings, good gain, usually clean (low-hiss), very consistent. Also the salesmen handed out 10-packs. Free. With gorgeous gold leads.

2N2222 is 2N2219 in a smaller case.

PN2222 is a 2N2222 in a plastic blob (nobody makes the old TO-5 metal can these days).

2N4401 is, for practical purposes, a large 2N2219/2222. If the '2222 is the 2x4 lumber of electronics, then a '4401 is 2x10 lumber.

2N3904 was another milestone. It isn't very big internally but it was processed cleaner than many transistors of its day. In hi-Z circuits it reliably gave low hiss, millions of phono preamps. Again you are more likely to find it in plastic than in metal.

Many of these old distinctions have faded as processes improved. In 1967 foundries were still fighting Purple Plague and other silicon contaminations, so the '3904's promise of clean low-hiss parts was noteworthy. Now we have millions of transistors in a CPU, that can't happen without dead-clean silicon, and last-year's cast-off foundry can still churn out transistors far-far cleaner than back when these part-numbers were new. So the '4401, specified to be an on/off switch (no claim of noise), today is a favorite in low-hiss LOW impedance mike inputs.

To business.

LINK

> something to do with the transistor choices.

No. I see no reason why the transistors matter at ALL (assuming NPN and gain higher than "dead"). I don't see why three different numbers were used... I bet that's what was laying around the bench. {Hmmm... maybe the Q3 Q4 section was built first, then Q2 and Q1, but by then all the '3904s had been used-up in other schemes.)

> a few seconds go by and there is sort of schweezip sort of noise that ends in a constant hum/oscillation.

When you hit it HARD, it goes "stunned". Like an over-powered car which won't steer while the rear tires are spinning. When you back-off, it recovers.

The hum could be bad shielding against room-hum. But this is a HIGH gain circuit. It has an available gain near a BILLION. The biasing is so "wrong" that I can't picture how it will recover, active or blocked. However ANY signal sneakage from output to input WILL cause oscillations (unless Murphy's Law has the day off). So it needs to be laid out very carefully: MINimum distance between parts, MAXimum distance from output back to input.

It is probably part-dependent (not just any "PN2222" may work with these values). Try reducing the 1Meg to 470K.

[Top Top]

I had the same thing when I built mine. With some swapping/experimenting with transistors, I found a combo that stayed silent when not playing.

[PRR]

Might be time to read the Debugging Thread at the top of the forum.

Not always, but very often, you can learn a lot by DC voltages.

Extra easy hear, because all the emitters better be dead-zero, and all the bases will be so dang-near the same (around 0.6V) that they don't really help.

And I'll assume you've checked for happy battery voltage.

So there isn't much left. Read all four Collector voltages.

With _NO_ input (at all!),



The exact voltages are maybe not critical. Which voltage is WAY off?

> schweezip ... constant hum/oscillation

Oh, yeah. It never settles to a no-signal state. Lift one end of the 10uFd from Q1 to Q2, and one end of the 0.01uFd from Q2 to Q3. That should shut it up. Now measure all four collector voltages.

> swapping... transistors....

...really should be a last resort. Certainly when we paid $7 each, we worked with the resistors to get the transistor happy. True, transistors have become nearly as cheap as resistors... still, an R has one less leg to de/re-solder.

[soggybag]

Thanks PRR, your posts are always so insightful. You really have a way have a way boring through any problem and reducing it to the essentials. I'll take a few measurements tonight and see if I can figure it out.

[Top Top]

I put sockets in so that swapping out the transistors was easy. Part of the reason I did this is because I did not have all of the transistors listed in the design on hand and I built it one night on an impulse...

Not being contrary for the sake of it, just saying what I did, and that it worked for me.

[soggybag]

I put sockets in so that swapping out the transistors was easy. Part of the reason I did this is because I did not have all of the transistors listed in the design on hand and I built it one night on an impulse...

Not being contrary for the sake of it, just saying what I did, and that it worked for me.

I do not doubt there is something to what you are saying. Mr Deeper, is a pretty sharp guy, I'm sure he chose the transistors for a reason.

I put sockets in myself. Working from PRR's train of thought, that the transistors were not critical I changed them all to 2N3904s. The sound was the same, of course I'm still getting the oscillation. So the transistors are not critical to the sound, but do effect the biasing.

I removed the caps and measured the collector voltages as:

Q1: 2.56V
Q2: 0.15V
Q3: 0.61V
Q4: 0.07V

Looks like Q1 is about 75% of the ballpark, Q2 is off by 50%, and Q4 is 70%. Looks like Q2 is off the most.

I'm trying to wrap my head around the numbers PRR came up with: 3.2V, <100mv, .6V and <100mv. My puny computational skills can't quite put it together. I guessing this has something to do with the voltage divider above the collector of each stage biasing the stage.

[PRR]

> My puny computational skills can't quite put it together.

Nor can I. Everything interacts. I asked the idiot computer for numbers, but I don't trust it too much. That does give me a glimmer.

> Q2 is off the most.

Yes, I suspect it should be "bottomed", and with these resistors and any modern transistor that should be less than a tenth volt.

Q1 should probably be in the "active" range, sorta midway between the supply rails.

ALSO Q1's collector voltage forces current through the 1Meg to slam Q2 into a bottomed state. Q1 being low, Q2 isn't being forced hard enuff to bottom.

I think his Q1 was lower hFE than yours. (Note that it runs at 0.125mA, WAY off the bottom of the '4401 datasheet, and lower than many transistor testers aim for.)

Do you have a 3.3Meg to put in place of the 2.2Meg? Or a 1Meg to put in series with 2.2 to make 3.2Meg?

Alsternatively you could reduce the 1Meg (to say 470K) from Q1c to Q2b to force Q2 to bottom better despite the low voltage in Q1 collector. But Q1 collector is already pretty low for what appears to be a "clean" stage (it has high gain and thus will distort, but it should preserve most of the waveshape so that Q2Q3Q4 can take their whacks at the signal).