Temperature compensation for output stages

[Nelson Pass]

I do have a couple of patents related to this that you
can see the first page of at www.passlabs.com

The first one, for what became known as dynamic
bias, bears a bit of resemblance to this chip and some
other later designs, while the optical bias patent
addresses the thermal tracking issue and worked
quite well.

Sonically, it is my experience that these things work
well for Class A output stages (where there is less
need) and not as well in AB stages because they do
in fact introduce some of their own distortion when
stages are turning on and off.

[peranders]

Mr. Pass, did the opto circuit work well? Is this same "Opto-thing" as some japanese firm (JVC?) used long time ago?

What do you think about my crazy microcontroller idea?

Can it be done with low emission I think the idea is good.

[Nelson Pass]

The opto circuit worked very well, but required initial
adjustment.

It's very easy to create a bias circuit which locks the
bias and requires no initial adjustment if you are operating
in Class A mode, just take a look at the Aleph current
source.

Microcontrollers are fine, and we have fooled around with
them on and off (feeble joke). The decision to raise or
lower bias is certainly an easy one, and all the difficulty
comes from sensing properly and then isolating the drive
to the bias circuit which is floating.

Ultimately you want to know the minimum value of the sum
of the currents flowing through both halves of the output
stage. This is the bias value to be adjusted.

[peranders]

My idea was to convert a common Vbe multiplier to a 2-wire tempsensing thing. My idea was that the microcontroller works with 2.5-8 volts and the whole circuit draws only 5-8 mA. I want to create a temp-voltage function which is 10mV/degree plus a certain DC-level. If you assume that the temperature lower the Vgs inckuding drivers with a certain factor (maybe unlinear) I think my idea isn't too bad. If a microcontroller is used you can programme the unlinear curve.

My thoughts now is how to design an output device which is "quite" and stable and uses low current. Maybe I will call Mr. Nelson Pass for a patented solution.

I looked at the A40. Clever to "sense" the current and feed it to the Vbe multiplier. Never seen that before.

[Nelson Pass]

A lot of those were built, and I don't recall any problems
with that bias circuit in the hands of DIYers, but again,
it is a Class A output stage. Don't try this with AB.

[P.Lacombe]

I think that correct temperature compensation is very diffucult to obtain. I agree with Mr Pass, and, moreover I don't imagine what kind of sensor can be used to obtain (real time) the actual temperature of the junction, which can vary quickly, depending on the musical transients.

Because of this, I use lateral mos(*), which exhibits very low temperature coefficients, and don't need in fact any compensation in class AB... Note that incorrect temperature coefficient compensation causes some distortion, well known as "thermal distortion", indeed ;-)

(*) Hitachi 2SK1058/2SJ162 Iq = 150 mA per device.

Regards, P.Lacombe.

[Nelson Pass]

If you want low temperature coefficients from Mosfets,
just bias them at the point where the coefficient is
0, which for the IR power parts is somewhere above an
amp.

[P.Lacombe]

Mr Pass,

The difficulty is that in order to minimize distortion, class AB push-pull must be biased at lower idle currents. The correct value seems to be the point where forward transconductance is about the half of the maximum value possible for the device.

For instance, IRF140 (Tj=125°C) exhibit half transconductance at Id=1A, but null temp coefficient occurs at Id=12A. (approximative values).

2SK1058 (Tj=75°C) exhibit half transconductance at Id=120...150mA, and null temp coefficient at the same value. Try and listen carefully... (in class AB without any feedback...)

Regards, P. Lacombe.

[Nelson Pass]

Well, I didn't say it was easy...

I assume you meant to say "in order to minimize dissipation"
not "distortion"

You can get to lower amperage figures by using higher
voltage and lower current parts.

Myself, I just put some Source resistance in the circuit
and sleep well at night.

[P.Lacombe]

Mr Pass,

It was no error in my post. Correct calculation of the optimal idle current is essential in class AB, in order to obtain minimal distortion : Many many authors have written about this (Douglas Self and others).

High voltage Hexfets exhibit poor linearity and high Rdson, and are inusable for high end audio.

Unlike this, lateral mosfets gives design facility, and really superior sound quality.

Regards, P.Lacombe.