dc filament heater - high impedance const. voltage

[Rod Coleman]

Originally Posted by Bigun

The problem with any kind of shunt regulator is that it is essentially a Class A device and will burn lots of juice - the filament is already hungry enough without adding to that. The other problem is that it offers an exceedingly low a.c. impedance across the filament - the opposite of what I want to achieve. I agree, the TL431 might be a good replacement as a voltage reference.

Yes, more or less you are right. Rod Coleman's filament supply provides a good guideline to the design goals - low d.c. impedance, high. a.c. impedance.

A pure current source is a possibility, but you have to include a sense resistor in series with the filament current. And people using Rod's current source have found that the sound is sensitive to the quality of this resistor. This resistor also dissipates wasted heat and increases the voltage headroom required for the filament supply. I'll give this some thought though.

Gareth, one correction: my Regulator does not present low dc output impedance. There is no deliberate tailoring of the output impedance at all, in fact.

You can use an op-amp for controlling current, but the circuit you have drawn will need to be modified for stability:

An op-amp with a power transistor in its feedback loop will usually require external compensation. A very slow opamp might be OK, but the LT1055 you have shown offers 16V/µs slew rate and 6.5MHz gain-bandwidth product - a nice specification, but adding a power-BJT with its high capacitance will disrupt its feedback loop too much. Think of the result as a kind of electronic tail-chase, where the output of the LT1055 reacts too fast to the feedback sense node for the BJT to keep up with.

Next, beware of output drive requirements. The LT1055, like many opamps is not specified for high-current output. For instance, with +/- 15V supplies it can deliver about 4V into a 100 ohm load (the minimum load resistance specified). For most BJTs, this drive-level will not give low-Vce saturation, and you may find that 3V to 5V of Vce saturation is the best it can do.

A power FET will not show this problem, but these present their own challenges, due to input capacitance - too high for most op-amps to drive directly, obliging us to add series resistance. You then have an RC pole on the output, whose value swings wildly with the capacitance-swings of the FETs (check out the capacitance-vs voltage plots in a typical power FET data sheet).

These problems can be addressed, but whether it still sounds good will be your challenge!

Please don't think I am trying to put you off using opamps in audio. It can be done, but requires much more design-effort than appears at first blush. And when you start to apply one in an audio circuit, please make sure you know the thinking behind Walt Jung's question: "How many inputs does an opamp have?"

[Rod Coleman]

Walt has archived his old articles on opamp audio, if you're interested in the cryptic remark about inputs:

http://waltjung.org/PDFs/WTnT_Op_Amp_Audio_1.pdf

[tomchr]

Yeah... Like with any other feedback loop, the stability has to be verified. Think it through, run sims, then verify in the lab. Standard design procedure. At least for me it is. I sometimes reverse the order of the thinking and the simulations. That leads to some interesting learnings at times...

As Rod mentions above, the simple current source circuit does have its pitfalls. I find that they're possible to design around, though. I have that circuit working in a couple of places where I needed a constant current.

~Tom

[Wavebourn]

Why not to close feedback around opamp by a single cap? You don't need fast feedback by current around the transistor. Why don't use Darlington for less opamp load current? Problems solved.

[Bigun]

What I posted was really a concept, I agree that there are quite a few details to be worked out. In reality, I won't be needing to build one for a few weeks but I really appreciate soaking up the knowledge you guys are willing to contribute.

I have never built with integrated circuit op-amps. It looks like a whole genre of electronics waiting to be explored so I'll do some more reading

Darlington - maybe a CFP would have less voltage drop (and require stabilization too) ?

I also like the FET idea.

[Wavebourn]

Add resistor from source to "-" input, and capacitor from opamp's output to it's "-" input. The rest is fine, if OPAMP can provide you needed output voltage to drive MOSFET.

[tomchr]

Originally Posted by Wavebourn

Add resistor from source to "-" input, and capacitor from opamp's output to it's "-" input. The rest is fine, if OPAMP can provide you needed output voltage to drive MOSFET.

Yep. I used that compensation scheme in one of my high voltage regulators. Works well. I used an LMP7701 opamp.

~Tom

[Bigun]

As soon as I get back in front of a PC (all Mac at home) I'll try those mods.

Whilst we don't appear to need fast feedback perhaps it helps. In a way it's a lazy way to deal with any h.f. hash from the mains that gets through the transformer and any h.f. hash generated by the rectifiers. By ensuring the current is constant even at high frequencies we do better in keeping stuff we don't want out of the filament current. Any filament current we don't want will be impressed on the signal.

Of course there's still that noisy zener. So how about a dollar or so for a quieter reference such as this:

[Wavebourn]

Please define "constant current". What dynamic resistance you consider as the minimal criterion? What dynamic resistance your MOSTEF has with resistor in source, without any additional feedback through opamp? I thought you need opamp only to keep DC current well defined by your reference voltage.

[Bigun]

Originally Posted by Wavebourn

Please define "constant current". What dynamic resistance you consider as the minimal criterion? What dynamic resistance your MOSTEF has with resistor in source, without any additional feedback through opamp? I thought you need opamp only to keep DC current well defined by your reference voltage.

I mean that the current through the filament, when no signal is present, doesn't change over time. What dynamic resistance do we need ? - well I can't say I've put any numbers to it. Let's think about it here.

I want to avoid non-signal fluctuations in filament current from generating any voltage drops across the filament that would result in an objectionable audio-frequency signal appearing at the output. I don't know what level that should be.

Let's say the signal swing at normal listening levels at the grid is +/- 10V ? . And let's say we want to avoid contamination to the level of -80dB which means 1mV (did I calculate this right?) and which across a filament of 6 ohms is 1/6 mA = 167uA.

So we want the current stable to of order 100uA ?