OT- Solid state amp biasing

[space_ryerson]

Hi guys, thanks for the replies!

That's not quite it, but close. Where you have a 470 ohm, make that 620. Where you have a 620, make that 160 or 170 ohms. Keep the 500 ohm pot as is, per this:

Doh! I guess the coffee wasn't strong enough when I drew this up. I've corrected the drawing.

You bias amps like these by putting volt-meter across R57 R55 and adjusting for several dozen milli-Volts. <20mV will rasp small signals. >100mV will overheat at idle. 30mV-50mV is a happy-zone. Higher sounds slightly better BUT not that much better and far more risk of disaster.

R.G. originally had me setting R55 and R56 to 15mV, which seems to be ok, but I'm curious to hear why 30mV - 50mV is a happy zone, as opposed to 15mV.

(It had me fooled. Looks like Q13 Q15 are a Darlington, but Q13 carries a LOT of the load. If we biased-up Q15 to a tenth-Amp, Q13 would be at 1.1A, far too hot. So Q15 idles "off", Q13 carries the first Amp, then Q15 takes the rest. Very odd.)

There are a lot of oddities with this amp, which makes me wonder how they arrived at the design they settled on (and kept with for subsequent models).

> pesky to get right at higher volumes

Bias of this amp affects the soft-level (~~1Watt) tone, NOT the high power tone.

Usually.

The amp in question took a bit of a hit during a recent show, which broke the power switch off. While replacing the switch, I rebiased the amp and there's a nasty noise that occurs during fast transients--like if I hit a chord fairly hard with a clean tone. I tried a quick re-biasing, and it got a bit better, but it's still there somewhat. Here's the best example I can conjure right now:

http://trace-on.net/audio/gtr_noise.mp3

It's in a band context, so it's harder to hear, but the noise occurs at 0:02, 0:18, 0:26

While that problem could be the bias, it could very well be something else, so I plan to double check everything else when I take it apart next. Unfortunately, the amp currently lives at my rehearsal space away from my solder desk, so I don't have constant access to this amp. (gotta love being a musician in New York City...)

I replaced all of the electrolytics in the last 5 years.

The proposal in red is the right idea.

Great!

Connected to chassis - no, the bias device _must_ be thermally intimate with the Q13 Q15 devices. Clamped to the transistor cases or very-close on the heatsink in good contact. (TO220 package is nice because it suits a sink; the added cost is now moot.)

Apologies; I worded that wrong. The existing thermistor is attached to the chassis between the transistors (they're mounted to the chassis as well). On a nearly identical Randall I have at home, the thermistor is between Q15 and either Q14 and Q16. It may be possible to mount the ZTX455 between Q13 and Q15. They appear to be next to each other, but it's a mess of wires, so it's hard to tell.

C33 C34 C36: Polarity marks are wrong. (Q8 Q9 Bases will sit a hair Negative, not a bit positive.) Only 0.1V-0.2V, so they won't die soon.... but they have been that way long enough to maybe un-do them.

Let me open up a Randall I have at home to see...yup, C33, C34 and C36 are oriented as drawn on the schematic. I'll replace them, and orient the new ones correctly.

It would not hurt to put four silicon diodes in series around that whole thing as an ultimate fail safe. It won't keep the amp from overheating if the biasing transistor/resistors decide to open, but it may give you some time to shut it down. Mostly, these will do nothing in normal operation. They're kind of a fail-less-unsafe.

Sounds reasonable to me. Is this what you are suggesting?:

[R.G.]

The amp in question took a bit of a hit during a recent show, which broke the power switch off. While replacing the switch, I rebiased the amp and there's a nasty noise that occurs during fast transients--like if I hit a chord fairly hard with a clean tone. I tried a quick re-biasing, and it got a bit better, but it's still there somewhat. Here's the best example I can conjure right now:
...
While that problem could be the bias, it could very well be something else, so I plan to double check everything else when I take it apart next.

I think it's something else. Bias in solid state amps tends to affect low-level distortion disproportionately. Could be breaking into an intermittent blast of parasitic oscillation. This is so common in solid state amps, there is a name for it - the "bottom side fuzzies", since it's often on the side run by a PNP output transistor with slightly worse frequency response. Put a scope on it if you can.

It would not hurt to put four silicon diodes in series around that whole thing as an ultimate fail safe.

Sounds reasonable to me. Is this what you are suggesting?:
[/quote]
Yep, that's it. No matter what happens to the biasing circuit, these things step in and keep it from going open circuit. The devices may (will if they sit too long!) overheat, but more slowly than the detonation that happens if the bias circuit opens entirely.

Here's another trick that doesn't get much play. When you're trying to get an amp to just work at all, jumper the bias. Just connect a wire across the place where you show the four diodes. This will make for bad crossover distortion, but it removes bad biasing as an issue until you can deal with whatever else is going on. Once it's stable, quiet, and performing at power, you can remove the wire and concentrate on biasing.

[space_ryerson]

The amp in question took a bit of a hit during a recent show, which broke the power switch off. While replacing the switch, I rebiased the amp and there's a nasty noise that occurs during fast transients--like if I hit a chord fairly hard with a clean tone. I tried a quick re-biasing, and it got a bit better, but it's still there somewhat. Here's the best example I can conjure right now:
...
While that problem could be the bias, it could very well be something else, so I plan to double check everything else when I take it apart next.

I think it's something else. Bias in solid state amps tends to affect low-level distortion disproportionately. Could be breaking into an intermittent blast of parasitic oscillation. This is so common in solid state amps, there is a name for it - the "bottom side fuzzies", since it's often on the side run by a PNP output transistor with slightly worse frequency response. Put a scope on it if you can.

Hmm. I'll check that out. I may be back asking for advice on how to test with a scope on an amp, since I've never done that before. I suspect something else got knocked loose or a solder joint got broken when it took that hit. There are also bundles of wires to and from the diode bridge, power transistors, speaker outputs, etc., which all will need to be double checked.

Here's another trick that doesn't get much play. When you're trying to get an amp to just work at all, jumper the bias. Just connect a wire across the place where you show the four diodes. This will make for bad crossover distortion, but it removes bad biasing as an issue until you can deal with whatever else is going on. Once it's stable, quiet, and performing at power, you can remove the wire and concentrate on biasing.

Good advice; thanks!

[space_ryerson]

Over the weekend I got time in the rehearsal studio to try to locate where the issue is coming from, while replacing and properly orienting C33, C34 and C46. I re-biased to ~40mV, and touched up any remotely iffy solder joints as well.

Before I could get to jumper-ing the bias, everything was testing normally, and I couldn't find where the issue is coming from, nor could I reproduce it. This isn't much of a surprise, since it usually takes at least 30 minutes of mid to high volume playing for it to crop up.

Last night we had rehearsal, and during the 4 hours of playing, the problem didn't crop up once. While I'm happy that it didn't happen, I'm somewhat frustrated that I couldn't definitively find the culprit.

One of my suspicions is that there are two six pin molex connectors that handle the power from the diode bridge to the pcb, the outs to the power transistors, the speaker outs, etc. as well. The Randall RG's had some questionable design choices when it comes to PCB and grounding (like the channel bleed issue these amps are known for), so I imagine if all of these wires containing dc power, high power signal, etc. are twisted around each other, the culprit could be there. Does that sound like a possibility? I'm thinking of using nylon ties to separate wires handling the DC power from the signal.

I wanted to try the bias mod, but I foolishly bought TO-92 packaged ZTX455's when I originally started this thread years ago, so I'll need to order TO-220 packaged ones, since I don't have any lying around.

[R.G.]

Over the weekend I got time in the rehearsal studio to try to locate where the issue is coming from, while replacing and properly orienting C33, C34 and C46. I re-biased to ~40mV, and touched up any remotely iffy solder joints as well.

Before I could get to jumper-ing the bias, everything was testing normally, and I couldn't find where the issue is coming from, nor could I reproduce it. This isn't much of a surprise, since it usually takes at least 30 minutes of mid to high volume playing for it to crop up.

Last night we had rehearsal, and during the 4 hours of playing, the problem didn't crop up once. While I'm happy that it didn't happen, I'm somewhat frustrated that I couldn't definitively find the culprit.

Aye, there's the rub. With intermittents, you never know if the last time you even TOUCHED it, you fixed it, or just made it hide better. 

One of my suspicions is that there are two six pin molex connectors that handle the power from the diode bridge to the pcb, the outs to the power transistors, the speaker outs, etc. as well. The Randall RG's had some questionable design choices when it comes to PCB and grounding (like the channel bleed issue these amps are known for), so I imagine if all of these wires containing dc power, high power signal, etc. are twisted around each other, the culprit could be there. Does that sound like a possibility? I'm thinking of using nylon ties to separate wires handling the DC power from the signal.

Could be. Could also be an issue with the connectors getting dirty and loose. We used to have a rule in the power supply design group I worked in that NO power line ever went through less than two pins of any connector. If you didn't have two pins for each one, you had the wrong connector.

I wanted to try the bias mod, but I foolishly bought TO-92 packaged ZTX455's when I originally started this thread years ago, so I'll need to order TO-220 packaged ones, since I don't have any lying around.

The to220 is a shorter thermal path, but you can do much the same thing with the TO-92 by rigging some kind of spring clamp to hold the flat face down in a pool of heat sink goo on the sink, and then putting some insulating material on top so that heat can't get out of the body of the TO-92. This is a bit slower in thermal time response, but works. You do have to be sure the TO92 you have will carry the current and withstand the voltage. Notice that the requirements on a VBE multiplier are minimal - they only need to have a minimum HFE, and carry the biasing current with some safety factor. The voltage across them is minimal except possibly at turn on and turn off. Nearly any TO-220 with good gain is as good as the next until you get into truly picky, hifi situations. So you might get by with any TO-220 you have. Even PNPs can be set up to work in this situation, just flip the circuit over.

[space_ryerson]

Aye, there's the rub. With intermittents, you never know if the last time you even TOUCHED it, you fixed it, or just made it hide better. 

I hear you there. I probably should have waited before touching it up.

Could be. Could also be an issue with the connectors getting dirty and loose. We used to have a rule in the power supply design group I worked in that NO power line ever went through less than two pins of any connector. If you didn't have two pins for each one, you had the wrong connector.

Good point. I'll look into that as well. If I were to redesign this thing from scratch, I'd go with that rule as well!

The to220 is a shorter thermal path, but you can do much the same thing with the TO-92 by rigging some kind of spring clamp to hold the flat face down in a pool of heat sink goo on the sink, and then putting some insulating material on top so that heat can't get out of the body of the TO-92. This is a bit slower in thermal time response, but works. You do have to be sure the TO92 you have will carry the current and withstand the voltage. Notice that the requirements on a VBE multiplier are minimal - they only need to have a minimum HFE, and carry the biasing current with some safety factor. The voltage across them is minimal except possibly at turn on and turn off. Nearly any TO-220 with good gain is as good as the next until you get into truly picky, hifi situations. So you might get by with any TO-220 you have. Even PNPs can be set up to work in this situation, just flip the circuit over.

Hmmm. I'll look through my parts tonight to see what I've got. I might be in luck

By the way, in my many Randall troubles over the years, at one point I was in contact with the amp's designer (Gary Sunda), and he had this volume boost switch drawn onto his personal schematic:


Is that safe? 

[R.G.]

By the way, in my many Randall troubles over the years, at one point I was in contact with the amp's designer (Gary Sunda), and he had this volume boost switch drawn onto his personal schematic:
...
Is that safe? 

Most likely. The amp is using mixed voltage and current feedback to raise its output impedance. The footswitch shorts out the current feedback, leaving only the voltage feedback, and making for a gain increase.

However - you knew there'd be one of those, right?  - this doesn't increase the power the amp can put out, only its gain. So it will get louder only if the input signal it was being fed was enough below the maximum output power at the speaker to let its power come up to(wards) maximum. If it was already near maximum output, the output would get more distorted, but no more power would come out to the speakers.

So yes - it's a gain boost, not a power boost.

[space_ryerson]

Very interesting. Thanks!