Vbe bias generator insights

[jan.didden]

Some insights into the working of and designing a low impedance bias generator:

https://linearaudio.nl/sites/linearaudio.net/files/bias generator studentzone-11-2020.pdf

Jan

 

[alayn91]

Hello,
Thank you for this paper, but when is the new LinearAudio available ?

 

[jony]

Some interesting stuff about Vbe multiplier we can find here:
bjt - VBE Multiplier with Emitter Resistance Cancellation - Electrical Engineering Stack Exchange

 

[Mark Johnson]

Hello,
Thank you for this paper, but when is the new LinearAudio available ?

See post #1093 of the long thread.

 

[alayn91]

OK

 

[as_audio]

See also here, Hawksford in 1984 :
https://www.diyaudio.com/forums/att...ified-diode-bias-circuit-audio-amplifiers-pdf

 

[Bonsai]

Figure 2 in the Hawksford paper is where Self got the trick with the resistor in the standard single transistor Vbe multiplier. I have not tried it as I tend to use a two transistor CFP pair for my EF3 OPS without problems.

 

[steveu]

vbe

Agreed. Interesting these "papers" do not mention that option. If you are making an amp that is more than minimal, then a CFP vbe reference is best, however the bypass cap becomes very important to guarantee stability of the CFP.

With a compensation resistor on the single transistor VBE, the question is where should the bypass cap be connected? If it is connected across the top of the compensation resistor then it defeats the compensation resistor and is frequency dependent unless a large cap is used.

If you are using a compensation resistor, you can select it by simulating the VAS current plus an AC component, and then tweak for the least AC output.

 

[U230421]

Hawksford paper can be summerised to a simple formula to calculate R3:

R3 ≥ ( 25mV/I ) * ( 1 + R2/R1 )

If the 'amplified diode' ('rubber zener') total voltage over this cell is called Vz:
Vz = ( 1 + R2/R1) * Vbe
(this is the original bias voltage in standard circuits)
then
R3 ≥ ( 25mV/I ) * ( Vz/Vbe )
or written otherwise
R3 ≥ ( 25mV/Vbe ) * ( Vz/I )

If Vbe is in the range of {600 - 650}mV, say 625mV, then (25mV/625mV = 1/25)
R3 ≥ Vz / ( 25*I )

One can add one small resistor in the circuit for better temperature tracking, e.g.
Vz = 2V, I = 4mA -> R3 ≥ 20 Ω (22E)


With respect to post #8, the most effective connection of this bypass cap would now require a serious consideration. It is depending on the dynamic impedance of the 'amplified diode'. The static resistance of the example is (2/4m=) 500Ω, but how much does it change by modulation of the AC-signal? If not so much, the right place would be on top of R3.

 

[Bonsai]

For my class AB amps, I usually decouple the Vbe pass transistor collector to emitter with a 10uF ceramic cap and then a 100 nF from the base to the collector of the 1st transistor.

 

[Ultima Thule]

Figure 2 in the Hawksford paper...

Quite similar to the Hagerman's Vbe multiplier, except that R2 is replaced with a diode connected BJT, in the Leach amplifier we see a similar Vbe multiplier to Hagmans but the diode connected BJT is now replaced with a string of 4 diodes in series with resistors on both sides to mitigate the stray capacitance effect as the 4 diode string is connected onto the heatsink.
The Leach amplifier doesn't have that extra R3 resistor though as seen in Hawksford's, and R14 in Hagerman's paper.

Hagerman Vbe multiplier
http://www.hagtech.com/pdf/vbe.pdf
https://www.diyaudio.com/forums/att...s-ab-power-amp-200w8r-400w4r-hagerman_vbe-pdf (backup link)

Leach amplifier
https://leachlegacy.ece.gatech.edu/lowtim/graphics/ckt.pdf

 

[jan.didden]

I think the main idea in the original posted article was that the dynamic impedance of the Vbe bias generator can be made a very small fraction of an ohm.
In that setting, a decoupling cap across it doesn't seem worthwhile.

In 'traditional' Vbe bias generators it may be required to help a less ideal situation though.

Jan

 

[Bonsai]

It’s mainly for stability Jan similar to decoupling the output of a reg. There’s no question that the Zo is low at LF. The CFP loop gain is high and the test BW of the amp is 100’s of KHz

In any case, before I dig a hole for myself, I should dig my sims out since last time I looked at this was 2010

 

[as_audio]

I have read somewhere that this cap is intended to throw
dynamic signal current at the bases of drivers or output
transistors when required, but do not remember where it
was - probably in this forum. Not sure this is true, perhaps
somebody else knows.

 

[Mark Johnson]

One can think about the bias spreader as an impedance between the VAS transistor and its current source load. We want this impedance to be as close to zero as possible ... but of course it's not. A capacitor connected in parallel will lower the impedance, for AC signals (music), and that might be a good thing.

 

[Bonsai]

You’ve said it more clearly than I managed to. Thank you.

 

[steveu]

I think it was LV that first used the idea of a separate current for the VBE circuit, and a single net connection to the VAS. This is particularly useful for MOSFETs where the total Vto can be ~8Volts. By driving the middle of that spread, output saturation is minimized. This is also handy for op-amp front-ends. If you connect the (single ended) VAS to the end of the vbe opposite the added current and boot-strap the CCS, you can reduce saturation to almost nothing. It's sort of another boot-strap that avoids reducing the voltage swing by the VBE voltage.

But I should mention that too much driver swing exasperates rail sticking.

 

[Bonsai]

I took a look at my CFP bias spreader circuit. The output R is about 1.5 Ohms out to 1 Meg. With the 10 uF decoupled ( good qual ceramic with low ESR it drops to about 0.8 ohms after 1-2 kHz.

Neither here nor there in the big scheme of things.

The collector to base cap on the sense tranny is 100 pF. If it’s too large, Ro goes up ( loop gain killed).

 

[U230421]

...a very small fraction of an ohm...

Stupid me, what's the (dynamic) impedance of a (near perfect) voltage source... zero ohm. I do not use these voltage cells anymore (or for cascodes) but a symmetric diamond-alike emitters outward cell.

Hagerman Vbe multiplier
http://www.hagtech.com/pdf/vbe.pdf

Graph B in figure 3 surprises me somewhat, I've never noticed such strong dependency. Is that typical? More references available?

 

[brian92fs]

I took a look at my CFP bias spreader circuit. The output R is about 1.5 Ohms out to 1 Meg. With the 10 uF decoupled ( good qual ceramic with low ESR it drops to about 0.8 ohms after 1-2 kHz.

It seems I've seen electrolytic caps often used in this position. Is a ceramic cap just a way to reduce the footprint a bit further?

The collector to base cap on the sense tranny is 100 pF. If it’s too large, Ro goes up ( loop gain killed).

I was reading your Ovation NX writeup and it looks like you specify a 10µF cap in that design (C14 - with a comment that its a bit oversized). I don't understand this vs your comment above. Could you elaborate further to help me understand?

Thanks Andrew!