Amplified Diode with MOSFET output

[acoustixman]

I will assume it is common knowledge that an amplified diode is appropriate for use in producing bias spread voltage with Bipolar outputs since temperature tracking is preferred to reduce bias voltage as outputs heat up to prevent thermal runaway. This of course takes advantage of the fact that as the output Vbe reduces with increasing temperature, the same thing happens in the heatsink-mounted bias transistor and the bias drive is reduced proportionately.

OK, on to the real issue... In the use of [lateral] MOSFETs as [class B/AB*] output devices, they are usually touted as having negligible or even slightly inverse thermal characteristics (with some caveats concerning instantaneous channel current); i.e. it seems widely accepted that the bias 'spread' generally does not require reduction as they heat up, and this seems further justified given the substantially lower transconductance relative to BJTs. Not much is said about whether they require increase. A symmetrical source-follower configuration is presumed here throughout.

*I'm not seeking discussion about whether you believe "class B" means the DRIVE is 1/2 wave (classic view) or the CONDUCTION is 1/2 wave (G.R.Slone's view). I mean to imply the bias voltage itself is well characterized and understood.

Part I

Some schematics show the use of a simple rheostat-mode pot in the VA totem pole to provide the V-spread to bias the gates and eliminate zero-cross (I use the word "crossover" to describe something else entirely) distortion.

Still others promulgate the use of an amplified diode, just like in cases of BJT outputs. I assume this is NOT set up to thermally track the FETs.

Would anyone please rationalize the use of either approach over the other?

PART II

Consider the same question given the use of BJT emitter-follower pre-drivers, assumed to be biased in class A since the outputs are still configured in source-follower mode. The class-A biasing here should keep temperature at a constant value assuming heat transfer is managed appropriately... So thermal tracking should still not be a factor, correct?

PART III, another question

Is bias instability possible with large (with small bypasses assumed present) capacitance values across an amplified diode? If so can this be conveniently resolved with a "miller cap" from C-B on the amplified diode transistor to keep it from getting hyperactive in overseeing the maintenance of charge upon the parallel cap(s)?

Thank you all for any available insights!
HAPPY LISTENING!

[acoustixman]

No takers? I had assumed these would be pretty simple questions for lots of folks here at DIYAudio...

Rephrasing the main question: Is inclusion of an amplified diode circuit to reduce the voltage source impedance in biasing output devices (let's just call it class AB mode) worth the extra complexity or simply an extravagance when L-MOSFETS (which do not require thermal tracking and are being driven by "constant-heat" class-A EF stages) are in use?

Some designs merely have a capacitor-bypassed resistance to derive the bias spread for the output, and others have the transistor/divider circuit in place. Is either case problematic in any way?

My only amplifier to date utilized class-A EF buffer stages driving class-AB L-FET-SF stages, and was a 100% success. I am soon to build my next, and I'm just wondering if there is any benefit to simplifying to a calibrated resistance between the VA polarities. If it helps, my circuit uses a full-symmetry cascode-loaded VA stage.

Can anyone shed some light on this?

Thanks!

[Bigun]

Perhaps it's a design choice with no black and white answer. For Class AB the biass needs to be controlled fairly accurately, but Class A is less critical. I wouldn't be choosing a Vbe multiplier for a Class A circuit myself when a couple of diodes will do just fine. I've no concerns about stability with use of a diode across a Vbe multiplier.

[acoustixman]

Thanks Bigun... I like the term "Vbe multiplier" too come to think of it... You mentioned what is probably one of the greatest truths in this art I suppose, as for it lacking a black-and-white answer. I appreciate the bias precision-vs-class insights as well.

I do plan a zener diode across the multiplier but that's just a safety in case I lose the base drive (I've heard horror stories about losing the $15 output transistors over a failed $0.30 trimpot, so I figure why not add a $0.60 clamp which in theory should never actually conduct)... I didn't expect that this would affect stability since the reverse-biased junction capacitance should be pretty small, but as for the capacitance across the VbeM...

First off, I have designed a bias current of 2.0mA down the VA, with the VbeM in between. The EF drivers are at about 12mA past the FET gates, with 200+ hfe (should bypass less than 60uA from the VbeM).

The VbeM will be divider-biased at about 10%, 0.2mA on the resistors and 1.8mA of collector current...

It will be bypassed by about 100uF (with a 100nF ceramic and a 100pF mica along for the ride). I see you have a 22uF cap across your VbeM in the TGM1.0, so maybe it's not a huge reach, but I'm thinking this is nevertheless a fairly substantial capacitive load upon the VbeM.

Accordingly I intend to place a 300pF [mica] Miller cap on it to reduce its bandwidth and let the transients come from the bypasses, so the VbeM will just basically function for DC and LF bias maintenance without reacting to the transients. Is this a 'sound' approach; i.e., is a miller cap all that is needed to kill off bandwidth in the VbeM? Surely bandwith is not the goal or we would not be using xxuF caps there in the first place?

Thanks!

[kenpeter]

The tempcos shouldn't be equal anyways. You want the bias to droop
with any increase of temperature, thus the temperature servos itself
toward your pre-determined operating point of thermal equilibrium.

I don't see any problem overcompensating, only undercompensating.
Reverse compensating could go to hell in a handbasket right quick...

[acoustixman]

I received an email that CBS240 had replied here also with a few good points, but I do not see the message upon viewing above...

While I had acknowledged the dynamic thermal issues (more properly the lack thereof) in case of L-FETs being used in SF output stages, and I do acknowledge the purpose of gate coupling (which in my case is done resistively, across the emitters of my class-A EF buffer, using values about 10% of my gate ballast resistors and providing about 12mA of coupling bias current, and the bias bypass caps in question being placed across the bases of the latter; I justify the pure-resistive N-P gate coupling since (A) the coupling resistance is small compared to the recommended ballast, and (B) the bypass cap is usually explicitly justified for BJTs in terms of them being subject to minority carrier recombination inertia, which FETs evidently do not suffer.)...

The question had also been posed as to "why add unnecessary complexity"... This is the very essence of my original question. I do have rationale for using the active Vbe multiplier is a servo mechanism for maintaining bias voltage if the VA bias current should drift, while realizing it is not necessary for thermal concerns. I was wondering if this rationale is sufficient in any designers' opinions to warrant inclusion of the active bias circuit despite the lack of need for a 'thermal servo'...

While absolutely necessary in case of power-BJT outputs, It appears to be 6 of one and 1/2 dozen of the other in case of L-FETs.

[acoustixman]

Originally Posted by kenpeter

I don't see any problem overcompensating, only undercompensating.
Reverse compensating could go to hell in a handbasket right quick...

These statements seem to drive right at the meat of my more critical question; how to approach compensation of the VbeM against the rather large capacitive load that I intend to place upon it. It appears you would say that if in doubt a larger miller cap is appropriate on the VbeM device, true?

Pardon me, what do you mean by "reverse compensating"?? Are you essentially indicating a gross LACK of compensation by this?

Thanks!

[kenpeter]

Thermal negative feedback via the diodes or whatever... not Miller CDom "compensation".
My bad for choosing a word that often means something else.

[keantoken]

Originally Posted by acoustixman

Part I

Some schematics show the use of a simple rheostat-mode pot in the VA totem pole to provide the V-spread to bias the gates and eliminate zero-cross (I use the word "crossover" to describe something else entirely) distortion.

Still others promulgate the use of an amplified diode, just like in cases of BJT outputs. I assume this is NOT set up to thermally track the FETs.

Would anyone please rationalize the use of either approach over the other?

Using a Vbe multiplier is the "safe" way because it voids thermal runaway. However, decreasing bias increases distortion, and so on power peaks the low-signal distortion will increase. So it seems the more hi-fi approach is to try and equalize the bias current vs. temperature curve.

PART II

Consider the same question given the use of BJT emitter-follower pre-drivers, assumed to be biased in class A since the outputs are still configured in source-follower mode. The class-A biasing here should keep temperature at a constant value assuming heat transfer is managed appropriately... So thermal tracking should still not be a factor, correct?

Temperature depends on power dissipation of the devices; this comes as current multiplied by voltage.

Say we have two devices of an arbitrary type. 20V into 8 ohms sine. Rails are 25V, bias is 3A

At max excursion the given device will have 5V across it, and 4.25A. This makes 21.25W. At minimum 1.75A, 45V, giving 78.5W. So we find that the device NOT active actually dissipates more. Add these wattages together, we get 99.75W

When idle, both devices dissipate 3A*25V giving 150W total. So actually, total dissipation decreases when driving a load, which means temperature will also decrease.

BUT, this is for perfect class A. What about egg-shaped class A? The rules could change.

PART III, another question

Is bias instability possible with large (with small bypasses assumed present) capacitance values across an amplified diode? If so can this be conveniently resolved with a "miller cap" from C-B on the amplified diode transistor to keep it from getting hyperactive in overseeing the maintenance of charge upon the parallel cap(s)?

I don't see why there would be a problem. I don't know if a transistor can oscillate in this position (I doubt it, but strange things happen). If there is a problem, it would be because of the transistor seeing nearly zero load impedance through the capacitor. If you're worried you could remove the cap and put the same value cap across the C-B junction of the Vbe multiplier/amplified diode. This way the multiplier has a direct path of negative feedback at AC. (I really doubt it's a problem though, the Vbe multiplier can discharge a parallel cap in a few microseconds. I doubt anything can happen at audio frequencies, and heat transfer is super-slow)

Don't take my words for granted when designing medical equipment, har har...

- keantoken

[kenpeter]

Originally Posted by keantoken

Using a Vbe multiplier is the "safe" way because it voids thermal runaway. However, decreasing bias increases distortion, and so on power peaks the low-signal distortion will increase. So it seems the more hi-fi approach is to try and equalize the bias current vs. temperature curve. - keantoken

You forget the temperature can servo itself, if presented with negative
feedback that isn't too terribly phase shifted. Extra bias will warm the
transistors till thermal equilibrium with the feedback is achieved...
I don't see how trying to drift the input bias in perfect concert with
the output thermals achieves anything but sabotage this process?

Cooling is more effective when outputs are hotter than room temp.
Design temperature to rise and then hold and some safe threshold.
You can't always garauntee what is room temp inside an amplifier?
But you can maybe garauntee what temp in a leaky oven where
you control the heating elements.