Active Bias Circuitry

[FredB]

Has anybody tried regulators or other active circuitry for bias control, and if so how did that work out?

[Enzo]

COnsidering that plain old resistors do the job well and reliably, what would be the advantage of a regulator? There is no current draw on the bias supply other than the voltage divider itself, so the load doesnl;t change to affect bias voltage. The entire power transformer does sag a bit under load, so we might see a small sag in the bias on peaks, but not enough to worry about in my view. Bias just isn;t such a critical dimension that we need to track it microscopically.

Having said that, I see no reason why one could not use an adjustable voltage regulator to set the bias voltage in a system. Couldn;t hurt anything.


In my experience, the problems we have with bias will be due to filter cap failure or age, then maybe a bad rectifier diode, followed by cracked solder connections and maybe bad ribbon cable connections. I don't remember having to replace a faulty adjustment pot in recent years, but I am sure it happens. Problem is that the added complexity of an active regulator won;t prevent any of those failures.

[loudthud]

A fixed voltage for bias on power tubes tends to run them cold when the line voltage is low and too hot when the line voltage is high. The unregulated bias supply tracks better. Traynor had an active bias regulator in the YBA200. Bias was set by sensing the cathode current. It was needed because a toroidal output transformer was used and they need tightly balanced bias currents. There were problems with it and power tubes were dying under warranty. I'm not sure what the fix was.

[FredB]

Enzo,

With active circuitry, I'm thinking it might be possible to shape the dynamics of the amp, rather than just let them simply be the consequence of transformer or capacitors used that can only be adjusted by adding serial resistance into the power supply or changing the values of the filter caps.


Thud,

I've read about Toroidal output transformers. For my purposes I only find them useful for power transformers. Regular old EI iron still seems the best to me for output transformers.

Variations in mains voltage seems to make a good case for some kind of a regulator.

[Enzo]

Hmmm. Well, again, there is little variation in the load on the bias supply, so a larger filter cap will remove ripple a little more effectively. It is not like the bias supply reacts to the signal. I mean other than overall sag or the main variations Loudthud mentioned.

Your bias voltage for the tubes is usually drawn from a voltage divider, and in my experience it is much more common to vary the shunt resistance at the bottom of the divider rather than the series resistance. But even at that, your bias voltage ultimately comes from a cap at the division point, so the resistances don't matter much.


I think loudthud made a good case against a regulator rather than for one. I myself think it doesn't matter if things drift around a couple volts, but he made a good argument. In an amp with 120v mains, and for example a 480v B+, that means for every change of 1 volt on the mains, you get a 4v change in your B+. SO if the mains goes up 5v, your B+ climbs 20v. But if your bias supply is regulated to stay right at -50v (or whatever you select) that means when your B+ goes up 20v, your bias does not also rise to compensate, so the amp runs hotter. And going the other way, when B+ falls 20v, your bias does not dip to maintain the same dissipation, so the amp falls cooler. Unless of course you also planned on regulating the B+, which I somehow don't expect.


But... it won't hurt anything for your bias to go north and south a few volts, so get yourself a variable regulator IC, and the few cheap passive parts it needs, and knock one out real quick, and find out if it does anything for you.

[Steve Conner]

There's no point regulating the bias voltage unless you're also planning to regulate the screen voltage. I've done it and it worked fine for me. I used a LM337 for the bias.

[exclamationmark]

It depends what you mean by 'active'. I like to mess about with combination bias on pre-amp tubes (as the title eludes, a combination of fixed bias and cathode bias), so I basically did this with a DAC feeding an inverting op amp buffer attached to the grid leak. I found I needed to give the op-amp a slightly positive voltage rail in order for it to get down to 0V completely in case you just wanted it acting as ground (since it can't swing rail to rail). For higher voltage stuff (such as power tubes) you probably need to stick a discrete regulator of some kind at the grid leak (it really depends on the output tube and what bias voltage you want to control).

You could go several ways about the active part of the circuit though - you could just reference it to a fixed voltage (ie, a precision zener), or you could reference it to something else - either analog or digital. If you went digital, you could have different sorts of feedback arrangements. Most of the arrangements of this sort are analog servo circuits used to keep direct coupled designs in check - they're pretty much an integrator with a huge time constant to keep the output at some fixed voltage, regardless of tube drift - the time constant is large so it doesn't affect transient, audible signals.

[R.G.]

I've built and tested both a zener regulated bias and a microcontroller bias adjuster.

As Enzo notes, the marginal return on effort is small. I liked the zener regulated one for simplicity and stability. It used a 75V zener purely as a pre-regulator to give a fixed starting point for the actual bias divider. Zeners are OK in low current, low power setups like this. I needed a stable bias supply because it included switch selection of four different bias points for different power tubes to be swapped in without getting out scopes and meters.

The microcontroller adjuster worked too, although it had some human interface issues. I never came up with a good (reliable!) way for the uC to sense bias current with signal flowing in the amp so it had to have a few seconds of silence to rebias well. That's a PITA for the user.

In this one, the uC fed a 0-5V DC signal from a digitally controlled potentiometer into a hacked opamp with transistor booster that would output zero to minus 60V. The uC looked at cathode currents on each tube, then adjusted each tube (it was a dual digipot, two hacked opamps) to be at the bias current selected. Worked.

I never put that into production. In practice, the indictor LED circuit for setting bias on the amp was so fast and easy for the user, we just put that in all of them.

[Steve Conner]

I think Fender's new amps have an auto-bias circuit with a microcontroller, similar to this.

I don't have bean counters to please, so I just put an analog meter on the front panel that reads cathode current.

[FredB]

You all seem to be replying to just what I was wondering about. For mains variation a regulator on the plate supply would be needed to establish a constant level there, although once that is done, then one would have to put the sag back into the PS somehow. That's what the active bias idea is about.

I've worked out a servo amplifier circuit that monitors the power tube current and adjusts the PS output impedance to put an AB push-pull sag dynamic into a single ended amp, and that got me wondering about all kinds of variations and applications of active bias circuitry. If people have traveled this path before, why reinvent the wheel?

I have micro-controller simulators and have made my acquaintance with programming them, although I presently lack the means to make PCBs with fine enough traces for them. So I have yet to actually build any uC circuits. I'm mostly just considering transistors and related stuff at the moment.

RG, Do you have a page on GEOFEX about your uC bias stuff?

[exclamationmark]

You can get some fairly powerful microcontrollers in a dip package - most notably the PIC32 and PIC24/DSPIC series that run at around 40MHz (they can do single cycle MAC's, which make them handy for implementing DSP-esque functions). You don't find that many MCU's faster than 40MHz in a DIP package, except for some oddball chips that are kind of hard to find.

I pretty much avoid TQFP's at any cost, unless I need the extra pins or extra space - The higher clock speed doesn't really play any role in my decision process, as it's generally overkill for 99% of my embedded projects. Plus the month long wait to get my PCB's fabbed is always bad. I much prefer to use plated through protoboards and DIP's with 30 gauge wire. The most expensive part of this endeavour is acquiring a programmer for the chip family you're targeting - other than that they literally require only 1 or 2 external components (excluding voltage regulators) to set up on a breadboard/veroboard.

As for the autobias implementation... I'd assume most commercial manufacturers probably use a FFT method, or at least something that vaguely resembles transforming the current measurements to the frequency domain. You can simply (or not so simply - there are of course problems) just pluck the constant term out and attempt to apply some sort of feedback loop. I wonder they get it to recognizes the tube types, though

[FredB]

You can get some fairly powerful microcontrollers in a dip package - most notably the PIC32 and PIC24/DSPIC series that run at around 40MHz (they can do single cycle MAC's, which make them handy for implementing DSP-esque functions). You don't find that many MCU's faster than 40MHz in a DIP package, except for some oddball chips that are kind of hard to find.

I pretty much avoid TQFP's at any cost, unless I need the extra pins or extra space - The higher clock speed doesn't really play any role in my decision process, as it's generally overkill for 99% of my embedded projects. Plus the month long wait to get my PCB's fabbed is always bad. I much prefer to use plated through protoboards and DIP's with 30 gauge wire. The most expensive part of this endeavour is acquiring a programmer for the chip family you're targeting - other than that they literally require only 1 or 2 external components (excluding voltage regulators) to set up on a breadboard/veroboard.

As for the autobias implementation... I'd assume most commercial manufacturers probably use a FFT method, or at least something that vaguely resembles transforming the current measurements to the frequency domain. You can simply (or not so simply - there are of course problems) just pluck the constant term out and attempt to apply some sort of feedback loop. I wonder they get it to recognizes the tube types, though

I see micro-controller perfboards in my future. I was wondering about using one to make a SMPS controller, among other things, because 100khz controllers that have current mode control, and lack subharmonics in the output are a little hard to find, except for the ones that are so small they have to be soldered under a microscope.

[Steve Conner]

I doubt the Fender auto bias uses anything as fancy as a FFT. I think it just sets the bias with the audio muted, when you come out of standby.

At work I deal with BGAs, QFNs, and the like. We just brought out a product that uses a 300MHz DSP chip in a 272-ball BGA. I've had some success prototyping with QFNs using the Schmartboard adaptors. For my own personal projects, I try to avoid micros, and if I need one I'll use something like an Arduino or Raspberry Pi.

Fred I'm pretty sure there's a chip in the UC38xx series that would do what you want. 3842? 3854? I forgot. The LM257x "simple switchers" also can run at 100kHz, and they look pretty stable. I've used them to power some quite demanding analog circuitry.

[FredB]

I've made some little switchers with discrete parts, using two transistor astable multivibriators and 555 timers. After spending quite some time studying SMPS theory and modeling controller chips I found making little boost supplies with discrete parts surprisingly easy. I even hand wound the inductor, using a salvaged transformer from an old pc supply.

I found a old classic controller chip, that works ok, although only open loop. Once I close the feedback loop it makes a horrible sound in the audio spectrum. It turned out to be the burst mode turning on and off under light load conditions. If I could turn the bust mode off, ti would work fine, although the burst mode is hard wired on inside the chip. I've noticed this seems a common thing with the newer more energy efficient controllers tool. I talked with some LT guys and found one that sims well, although it's too damn small to solder.

Anyway, getting back to my discrete controllers, I've found I can create an older style err amplifier and bias down the switching transistor very smoothly, and still get a super transient response. The only trouble with scaling that up to a 100-200 watt flyback is that the larger switcher is going to need a soft start. Current limiting works fine for that, and I think it would be easier to program a MCU to do that then to try and work something out based on comparators.

I spent like a month or two learning how to work out and graph a transfer function for a err amplifier, then I learned all chips stopped using that a long time ago, and are much simpler to tune the transient response. It bring the meaning of the word chagrin to mind.

I guess such is life when one tries to teach them-self how to build switching power supplies. I've got a core for a 200 watt forward converter. One of these days I'll wind it up and see what I can do with it. Learning all that magnetic theory was a chapter in my life too. It's neat stuff.

[R.G.]

I doubt the Fender auto bias uses anything as fancy as a FFT. I think it just sets the bias with the audio muted, when you come out of standby.

That's what I did. I was able to use the $0.50 variety of controller for this. The problems were that (1) it requires the audio to be muted, and (2) tubes drift as they heat up.

It's easy enough to keep the thing muted for a couple of seconds at power on or out of standby, but the issue of letting them warm up enough to get past most of the drift was tough for me to solve. I didn't want to make the user have to do anything explicit, like flipping into and out of standby after the amp was warmed up. And I didn't want to expend the time and money to do FFTs and DSP stuff to tweak the DC currents out of the audio while audio was going on.

The red-light/green-light setup on showing when bias was right was so very well received by our early testers that we just bagged the auto bias. If you think about it, you get probably 99% of the good out of every guitarist being able to set bias pretty accurately whenever they want to by looking for a green light while they turn a control. And since this can be done in seconds, and is mostly the equivalent of setting a bias pot in a shop, it lasts for a long while and doesn't have to be done much, perhaps only when new tubes go in, and certainly not more than once per show.

The value returned for such a small effort was too big to go to the complications of a full controller. That may be a different computation if Fender can afford a full uC development team and use a fraction of the leftover code space for it.

However, I bet they didn't include a 9Vdc output to run pedals. For the life of me, I don't know why amps don't do that as a matter of course today.