Active Bias Circuitry

[FredB]

or, taking a page out of the Bob Pease book, you could just breadboard some stuff.

I used to sim a lot of tube circuits. Now I have a big ugly piece of plywood with a turretboard and tube sockets and a huge assortment of shunt resistors and controls on it. I can cobble together most any output section in minutes and I've learned a lot more that way than I ever did in spice. Of course, I'm fortunate to have access to some nice scopes and true rms voltmeters so I'm able to experiment and measure rather than speculate and simulate.

I guess it depends on the kind of circuit you're building.

jamie

Yes, it does. I have noticed some circuits absolutely, positively have to be tuned by ear.

There's a couple problems with those "one size fits all" adapter boards - mainly the lack of positions for decoupling capacitors and oscillators. The schmartboard's solve this problem to a degree, in that there are some pads for decoupling caps, but the difficulty of fitting an oscillator is still a problem. Personally I think they're sort of a rip off - If you wanted to, you can simply copy the layout and get them fabbed at one of the numerous chinese services. A lot of places offer 5x5cm boards for about $10 for 10 copies. Of course the wait time is a major inconvenience and most people will pay a premium for it.

I've had similar experiences with those nixie power supplies - I have a feeling that they're essentially the bare minimum for a switchmode converter to even function, and since it's a for a nixie tube, noise is probably irrelevant. I found if you stick a suitably sized resistor so it draws a large quiescent current (compared to transients), you can sort of stop the horrendous noise (presumably from the feedback loop spazzing out). A couple of companies make SMPS's for tube audio, but reviews are scarce - I dunno if I'd trust them.

I've sorted out the Nixie power supply's audio problem, and gave it into a 400V 20mA capacity to boot.

Here is a step transient from 0% load to 110% load:



A properly adjusted compensation loop is a beautiful thing.

Here is the noise spectrum at 50% load. It's kind of funny, the power supply gets quieter as it's loaded down.

[FredB]

or, taking a page out of the Bob Pease book, you could just breadboard some stuff.

I used to sim a lot of tube circuits. Now I have a big ugly piece of plywood with a turretboard and tube sockets and a huge assortment of shunt resistors and controls on it. I can cobble together most any output section in minutes and I've learned a lot more that way than I ever did in spice. Of course, I'm fortunate to have access to some nice scopes and true rms voltmeters so I'm able to experiment and measure rather than speculate and simulate.

I guess it depends on the kind of circuit you're building.

jamie

Yes, it does. I have noticed some circuits absolutely have to be tuned by ear or other actual performance behavior.

There's a couple problems with those "one size fits all" adapter boards - mainly the lack of positions for decoupling capacitors and oscillators. The schmartboard's solve this problem to a degree, in that there are some pads for decoupling caps, but the difficulty of fitting an oscillator is still a problem. Personally I think they're sort of a rip off - If you wanted to, you can simply copy the layout and get them fabbed at one of the numerous chinese services. A lot of places offer 5x5cm boards for about $10 for 10 copies. Of course the wait time is a major inconvenience and most people will pay a premium for it.

I've had similar experiences with those nixie power supplies - I have a feeling that they're essentially the bare minimum for a switchmode converter to even function, and since it's a for a nixie tube, noise is probably irrelevant. I found if you stick a suitably sized resistor so it draws a large quiescent current (compared to transients), you can sort of stop the horrendous noise (presumably from the feedback loop spazzing out). A couple of companies make SMPS's for tube audio, but reviews are scarce - I dunno if I'd trust them.

I've sorted out the Nixie power supply's audio problem, and gave it a 400V 20mA capacity to boot.

Here is a step transient from 0% load to 110% load:



A properly adjusted compensation loop is a beautiful thing.

Here is the noise spectrum at 50% load. It's kind of funny, the power supply gets quieter as it's loaded down.

[R.G.]

It's kind of funny, the power supply gets quieter as it's loaded down.

Bigger load is much like heavier damping.

Constant *power* loads are ugly for compensation, though.

[FredB]

Constant *power* loads are ugly for compensation, though.

What do you mean by that?

What kind of load would play well with compensation?

[R.G.]

What do you mean by that?

A constant power load sucks harder when the voltage supplied to it goes down. When its input voltage goes up, it backs off on current use. It can act like a negative resistance in some senses. Active power loads like switching regulators sometimes act this way.

What kind of load would play well with compensation?

A constant resistance load is easiest to compensate for. Constant current loads are kind of a null as far as stability issues caused by loads are concerned.

It's often the case that some combinations of loading and/or crossover networks can destabilize marginally stable power amp designs, although this is mostly an issue with solid state amps. The presence of damped inductors in series and Zobel networks to ground are there to prevent some of these issues. Without the series inductor, for instance, a highly capacitive load can make many solid state power amplifiers oscillate. Many IC opamps have the same issue with capacitive loading.

Tube amps don't have enough loop gain to have the issue in most cases, and the stabilization that has to be done for including an output transformer inside a feedback loop

Feedback-stabilized active power supplies are in a sense a power amplifier designed to put out a fixed, unvarying signal, and have much the same issues for stability, although they're full of special cases.

[FredB]

A constant power load sucks harder when the voltage supplied to it goes down. When its input voltage goes up, it backs off on current use. It can act like a negative resistance in some senses. Active power loads like switching regulators sometimes act this way.


A constant resistance load is easiest to compensate for. Constant current loads are kind of a null as far as stability issues caused by loads are concerned.

It's often the case that some combinations of loading and/or crossover networks can destabilize marginally stable power amp designs, although this is mostly an issue with solid state amps. The presence of damped inductors in series and Zobel networks to ground are there to prevent some of these issues. Without the series inductor, for instance, a highly capacitive load can make many solid state power amplifiers oscillate. Many IC opamps have the same issue with capacitive loading.

Tube amps don't have enough loop gain to have the issue in most cases, and the stabilization that has to be done for including an output transformer inside a feedback loop

Feedback-stabilized active power supplies are in a sense a power amplifier designed to put out a fixed, unvarying signal, and have much the same issues for stability, although they're full of special cases.

OK, I understand that. I've noticed that as the supply voltage goes down to a SMPS, it will draw more current to maintain the output, hence the negative resistance: Lower voltage results in more current.

Is the second part aiming at something about my little power supply?

I know changing the output capacitor will upset the balance of the feedback network. Come to think of it, that might require some more attention about the use of the power supply actually feeding an amp.

The model runs so slow that, simulating it with an attached amp will be impractical. The best I could do is to add some caps resistors, and varying voltage sources to power supply's output to sort of approximate an attached amp.

I was already thinking of adding a serial resistance between the power supply and any following reactive circuitry in order to mitigate its effect on the compensation loop. I guess it's going to need some further testing there.

* * *

Then again the original idea for this was for a tube pedal booster, so if the power supply is used as is without adding more reservoir/filter caps it should work fine for powering a class A preamp without any inductors.

Although for a larger SMPS powering a whole amp, I can see those issues coming into play.

* * *

I found a model for the UC38xx that works. What PFC was usually used with that?

Found it, the UC3854 of course!

[exclamationmark]

Thought I'd revive this thread since I've been tinkering with some mcu powered auto biasing circuitry. I've been encountering some weird issues in simulation when using source/emitter followers to drive the grid leak resistor. Ordinarily, when the phase inverter isn't clipping, everything is fine. However, as soon as I pump any sort of asymmetric square wave through the PI, the source follower immediately freaks out and doesn't provide any sort of constant voltage. I've found it's related to the charging/discharging of the coupling caps from the PI to the power tubes or driver tubes/mosfets - the ratio of source resistor and the grid leak resistor pretty much determines how effective the source follower is at providing a constant voltage. In my simulations I've found that the total source resistance has to be smaller than the grid leak resistor in order for it to actually work. This isn't a huge problem, since I can still set the source resistance rather low and still have a lowish dissipation. Rather, I'm just perplexed as to why/how this even occurs and I just can't quite get my brain around it. Sticking an N channel device at the bottom of the source follower so it can sink current seems to work to a large degree, but the ripple is still inferior than the brute force method of just having the source resistance smaller than the grid leak.

The rationale behind this effort is pretty much so I can be a cheapskate and not buy proper output transformers - I've been playing around with some toroidal power transformers that cost me about 10 bucks and found that they work really well as output transformers for low rp tubes, provided you can keep the DC current offset to less than a couple mA. So far I've been able to cram 2 x negative voltage regulators and 8 fully protected ADC inputs (I've built a test circuit and they can survive 300v+ continuous) on a 5x5 cm board. I'm using a STM32F100 MCU that I've been able to nab for $2.50 in single quantities - the dual 12 bit ADC's and 10 x ADC inputs make this thing amazing value for money for these types of projects (even if the DAC/ADC performance isn't that great, it's sufficient for this project - much cheaper than external ones).

[Steve Conner]

I assume you're talking about a regular class-AB1 power amp where the bias voltage is derived from a source follower buffer, not a class-AB2 where the source followers are driving signal directly onto the grids.

Assuming this, the explanation is that when the power tubes are driven into grid current, the grid current flows back into the bias supply and tries to pump its voltage more negative. The bias supply has to be capable of absorbing this, in other words it has to both sink and source current.

The real challenge in auto biasing a toroidal OT is what to do with DC imbalances caused by asymmetrical clipping of the power stage. These saturate the OT just the same as imbalances in idle current would. The tubes are being driven flat out with a PWM square wave and practically acting as switches, so they'll hardly respond to adjustments from the auto-bias circuit. For hi-fi, not a problem, but for guitar it's a bit of a show-stopper.

We use the STM32s at work.

[Gregg]

I've been encountering some weird issues in simulation when using source/emitter followers to drive the grid leak resistor.

Maybe you should take a look at Marshall's AFD100 schematic and see how they did it.

[exclamationmark]

I totally forgot about the marshall autobias thing... Well, I looked and it appears that they're just doing it the lazy way by sticking a giant RC filter at the end of the voltage regulator, which just eats all the transients. I guess it's faithful to the old bias supplies, but it has the disadvantage of an extremely slow loop response time (which I guess can be a good thing for stability) - an active regulator can easily outperform it in terms of ripple and transient response. Though I wonder how much of the ripple/sag on the bias supply contributes to the tone. I haven't ever scoped the typical fixed bias supply under full load, but I suspect it gets yanked around quite a bit.

Kevin O'Connor has advocated toroidal OPT's for a long time, but really only of the plitron variety, which I suspect has some way of dealing with imbalanced currents by forfeiting it's toirodal-ness. I've probably squeezed a couple of watts out of a 30VA power transformer and haven't seen tell-tale signs of saturation, but once I get an actual power supply I might put that to the test. Another thing that might be troublesome is the fact that because the primary inductance of these things is so low, you need to use low RP tubes, but since these are budget PT's they have relatively high primary winding resistance, which counteracts the first point!

[Gregg]

Well, I looked and it appears that they're just doing it the lazy way by sticking a giant RC filter at the end of the voltage regulator, which just eats all the transients.

I have no problem with that if it works.

Kevin O'Connor has advocated toroidal OPT's for a long time, but really only of the plitron variety, which I suspect has some way of dealing with imbalanced currents by forfeiting it's toirodal-ness.

Plitrons are an overkill for guitar amps. Allegedly they have a tiny air gap to avoid saturation.
I have some experience with toroidal OTs I designed myself and they perform quite well without any special measures taken. A 50W OT (wound on a core usually used for 100W PT) takes easily a 50W sine wave at 50Hz without any saturation even when bias current is mismatched. I haven't done tests using square waves but the OTs sounded good in my high gain amp.