Variable B+ with Bias Tracking

[MatthewF]

I've thought further on PRR's suggestion of using an inverting amplifier to generate bias voltage as a function of (variable) B+. To this end, I've sketched the attached schematic as a first attempt to visualise the circuit. Not shown is the power supply for the op amp (this should be around +15/-60v, derived from a separate mains transformer). The OPA445 datasheet: http://www.ti.com/lit/ds/symlink/opa445.pdf specs max supply voltage at +/-45v, and details a use case with +60/-12v supply, so hopefully this is reasonable.

R6 (1M pot) would be the front panel control for varying power amp voltages. Trimmer R2 is for bias trim, used when changing output tubes etc. The 9v zener at the input to IC1 is there to protect against over voltage at the op amp input. Resistor values are nominal and could of course be changed per application.

Aside from missing PSU decoupling caps at the op amp supply pins, can anyone spot anythig questionable in the drawing?

 

[joaquins]

I don't like the zenner there. It shouldn't be any voltage across it, neither current, so not a noise problem. Bias is common mode but if we can keep the noise low better, tubes matching is not perfect. The problem I have is that it's allowing up to 9V across the inputs, which is too much, for protecting the opamp better to use 2 diodes in parallel reversed from each other.

What I'd do in a case like this is adding a range limiting at the output. A resistor at the output and two diodes to clamp it to two voltage reference which correspond to the maximum and minimum required bias. Even better if in a case of a failure of the opamp or this circuit as it is you have a fixed voltage to the more negative desired bias so you don't loose your tubes. When I was thinking in designing something related I was thinking in having a  common emitter at the output and the collector clamping the voltage down, with some kind of protection in case of a shorted transistor or loosing the feedback that opens the transistor so the bias network almost turns off the tubes. In my case was probably easier than this since I only wanted the bias to go higher than the limit for the tubes, so you never loose a tube. Here for low voltage you want more current at the anode than recommended at higher voltage.

On the other hand I think you are counteracting the sagging of the supply with this, more current on the anode, less voltage on the B+, less negative bias, more current... Kind of a positive feedback which could be dangerous and will counteract the sagging of the B+ you want to have for keep the tubish sound because of the supply voltage dropping at higher levels.

I think I'd go a different way here. Starting with completely independent variable supplies for B+ and bias, regulating them from the same voltage reference, so you change both at the same time. Then if you want you could sense the B+ current and from there affect the bias voltage and B+ voltage to exaggerate or counteract the sagging effect. I'd try to make some scheme to picture better my idea and add it to this post.

JS

PS: added scheme, most values aren't there, and the ones that are are just for reference. Some caps may be added and the polarity of the opamps inputs may be swapped if looking for a different effect on each variable. I didn't check all possibilities and combinations, maybe the bias is tracking in revers of desired for example. It's just to express the idea described before. Also I used a BJT for the HV reference regulator but probably a mosfet is more suited for the task.

 

[MatthewF]

Thanks Joaquins for such a detailed and enlightning post!

My intention with the 9v zener had been to protect the op amp against input voltages greater than the +ve supply rail. I see now that to achieve that D1 would have to be somewhere to the left of R3.

I like a lot of things about your scheme, particularly the common emitter after Q2 which negates the need for an unusual, expensive op amp with awkward power supply requirements. The protection diode is a good call too.

The cathode current sense is intriguing. Looks like a good way to control the amp's dynamics, but perhaps it's not necessary for the bias supply? I'd imagine that in most fixed bias amps the bias voltage doesn't sag much at all (only if the mains transformer is struggling). Furthermore, it seems that we'd need to decouple the current sense signal to avoid introducing feedback at audio frequencies. Would need to think carefully about the time constant.

For new amp designs, perhaps with a solid state rectifier or entire SS output stage, this could be very very interesting. My only reservation is that by regulating the B+ in an amp like the Fender Deluxe, we're removing the essence of its character (the PSU and output stage dynamics). While the current sense element in your scheme goes some way to reintroducing that character, only experimentation would tell if it provides the same desired response as the valve rectifier and smallish reservoir caps do when strained.

 

[joaquins]

I've posted a answer for that last post but seems to not be there anymore... Lost in the update attempt, I don't remember how it was exactly.

I forgot to mention the need of RC for all of the 4 inputs, the ones from the reference voltage should be matched so they track together, or they could be taken from an already smoothed node, sharing the RC, then two R going to the opamps.

The complete config is hard to predict, I guess a simulation is a way to go for all that.

For what I remember was something like that but quite a bit longer, sorry, I just don't remember what I was thinking at the time.

JS

 

[PRR]

> counteracting the sagging ....less voltage on the B+, less negative bias, more current... Kind of a positive feedback

No.

Figure the gain. The forward gain of the tube is less than 10. The gain of the servo is about 10:1. The overall loop gain is less than unity.

Over the long range, current will go down faster than voltage (3/2 law).

It is stable (assuming the no-gimmicks amp was stable).

Me, I object to a high-price chip. It can be done with 2 or 3 transistors and the existing negative bias supply. I had this drawn-out but that plan is not handy at the moment. I admit that for a one-off, the cost of one chip is moot.

The chip input can be protected by back-back simple (1N914) diodes across the inputs. One input is solid ground, The other input should "never" be any significant voltage away, much less than 0.1V. A 0.5V clamp will do.

 

[clintrubber]

Just curious to the implementation on the new Fender Bassbreaker 45:

"Output level knob takes power from 45 watts to a single watt and anywhere in between, tailoring full-powered Bassman overdrive capabilities for studio, stage and arena – players can dial up many “sweet spots” with various combinations of level and channel settings"

As it seems it does sound good on this amp.

There are also other models which simply use a PPIMV for 'power-reduction', but those don't get all thumbs up.

 

[abbey road d enfer]

On the other hand I think you are counteracting the sagging of the supply with this, more current on the anode, less voltage on the B+, less negative bias, more current... Kind of a positive feedback which could be dangerous and will counteract the sagging of the B+ you want to have for keep the tubish sound because of the supply voltage dropping at higher levels.

Well, in any case, with the B+ adjustment, you lose the sag, because it's a voltage regulator. Making B+ adjustable and maintaining sag is kind of tricky, you need positive feedback.

 

[joaquins]

On the other hand I think you are counteracting the sagging of the supply with this, more current on the anode, less voltage on the B+, less negative bias, more current... Kind of a positive feedback which could be dangerous and will counteract the sagging of the B+ you want to have for keep the tubish sound because of the supply voltage dropping at higher levels.

Well, in any case, with the B+ adjustment, you lose the sag, because it's a voltage regulator. Making B+ adjustable and maintaining sag is kind of tricky, you need positive feedback.

You can regulate and then add degrading resistance after the regulation, you have a constant voltage at the point of the regulator but then some resistance to have that effect. That resistor will get quite hot at higher levels, be sure it can handle it.

A more modern control with a regulator for B+ and bias current with positive and negative feedback sensing all the variables and affecting each other may be a great project and be really flexible in a lot of characteristics to emulate amps from the past or the future, also quite hard to tune and get working properly as expected, software controlled is the way to go, so tweaking the parameters doesn't involve a soldering iron.

JS