MOSPOWER Class AB musings

[Active8]

I was thinking I could set up a current controlled Vbe multiplier to
control MOSFETs idle current. Ib, entering upper left of figure
below, is supplied by an xCCS which provides enough current for the
divider with extra for the NPN to shunt - the amount of which
effects a greater/lesser drop across the divider via its effect on
Vbe.

The log nature of this setup allows finer control over the bias
voltage compared to a resistor.

Assuming a way to sense MOSFET idle current, is this approach even
worth considering?

Not much in the IRF ANs. Anyone know of other ANs covering this?
BTW, that LT1166 is interesting, albeit lacking in THD spec.

TIA

View in Courier font, of course. Did Google Beta get that
straightened out?


| |
| |
V |
I |
b +------+ |
| | |
| | |
.-. | |
| | | ||-+
| | | ||<-
'-' +------||-+
| | |
| | |
| |/ |
+----| +------+
| |> | |
| | | |
| | | .-.
.-. +------||-+ | |
| | | ||-> | |
| | | ||-+ '-'
'-' | | |
| | | |
| | | ===
+------+ | GND
| |
|
created by Andy´s ASCII-Circuit v1.22.310103 Beta www.tech-chat.de

 

[John Larkin]

I was thinking I could set up a current controlled Vbe multiplier to
control MOSFETs idle current. Ib, entering upper left of figure
below, is supplied by an xCCS which provides enough current for the
divider with extra for the NPN to shunt - the amount of which
effects a greater/lesser drop across the divider via its effect on
Vbe.

The log nature of this setup allows finer control over the bias
voltage compared to a resistor.

Assuming a way to sense MOSFET idle current, is this approach even
worth considering?

Not much in the IRF ANs. Anyone know of other ANs covering this?
BTW, that LT1166 is interesting, albeit lacking in THD spec.

TIA

View in Courier font, of course. Did Google Beta get that
straightened out?


| |
| |
V |
I |
b +------+ |
| | |
| | |
.-. | |
| | | ||-+
| | | ||<-
'-' +------||-+
| | |
| | |
| |/ |
+----| +------+
| |> | |
| | | |
| | | .-.
.-. +------||-+ | |
| | | ||-> | |
| | | ||-+ '-'
'-' | | |
| | | |
| | | ===
+------+ | GND
| |
|
created by Andy´s ASCII-Circuit v1.22.310103 Beta www.tech-chat.de

Why a Vbe multiplier? Mosfets don't need the temp tracking like
bipolars do.

I've mused on this, too. How about a simple bypassed resistor, pumped
by a current source? Some side circuit would know when the signal
level was small or zero, measure fet current, and tweak the bias. A
small uP and a dac, maybe. The hard part is measuring the smallish fet
idle currents... maybe a sense resistor in one of the power rails,
shunted by a hefty diode. I like having no source resistors, except
for the current-limit thing.

Or just use a pot!

John

 

[Active8]

Why a Vbe multiplier? Mosfets don't need the temp tracking like
bipolars do.

I wasn't talking about Vbe mult biasing of a bjt CE amp. The Vbe
mult in the circuit has nothing to do with temp tracking.

I've mused on this, too. How about a simple bypassed resistor, pumped
by a current source?

Why the bypass? Maybe I'm not following you.

As for a resistor, as I said, there's no fine control. a 9 mA change
through 5k is 45 V whereas that same change through the Vbe mult
only results in mV changes.

Some side circuit would know when the signal
level was small or zero, measure fet current, and tweak the bias. A
small uP and a dac, maybe. The hard part is measuring the smallish fet
idle currents...

Is 100 mA typical?

maybe a sense resistor in one of the power rails,

Not, both? But why in the drain?

shunted by a hefty diode. I like having no source resistors, except
for the current-limit thing.

How do you limit current without a much greater R than the typical
0.1R source R?

Or just use a pot!

And a robotic arm to turn it?

 

[Mac]

I was thinking I could set up a current controlled Vbe multiplier to
control MOSFETs idle current. Ib, entering upper left of figure
below, is supplied by an xCCS which provides enough current for the
divider with extra for the NPN to shunt - the amount of which
effects a greater/lesser drop across the divider via its effect on
Vbe.

The log nature of this setup allows finer control over the bias
voltage compared to a resistor.

Assuming a way to sense MOSFET idle current, is this approach even
worth considering?

Not much in the IRF ANs. Anyone know of other ANs covering this?
BTW, that LT1166 is interesting, albeit lacking in THD spec.

TIA

View in Courier font, of course. Did Google Beta get that
straightened out?


| |
| |
V |
I |
b +------+ |
| | |
| | |
.-. | |
| | | ||-+
| | | ||<-
'-' +------||-+
| | |
| | |
| |/ |
+----| +------+
| |> | |
| | | |
| | | .-.
.-. +------||-+ | |
| | | ||-> | |
| | | ||-+ '-'
'-' | | |
| | | |
| | | ===
+------+ | GND
| |
|
created by Andy´s ASCII-Circuit v1.22.310103 Beta www.tech-chat.de

I started a thread similar to this one about a year ago. The subject was:

"quiescent current in audio MOSFET class AB output stage"

It seemed that the consensus was to just use a vbe multiplier and a trim
pot and be done with it.

Someone also recommended Douglas Self's book the _Audio Power Amplifier
Design Handbook_, and after I read that I didn't want to use MOSFET's for
the amp anyway, and started thinking about doing it with BJT's.

Now I'm so conflicted I might not ever do it. In the mean time I'm working
on a pre-amp.

--Mac

 

[Mac]

As for a resistor, as I said, there's no fine control. a 9 mA change
through 5k is 45 V whereas that same change through the Vbe mult

Of course you mean 4.5 Volts.

--Mac

 

[Ban]

I wasn't talking about Vbe mult biasing of a bjt CE amp. The Vbe
mult in the circuit has nothing to do with temp tracking.

It will change its value with temperature like a multiplied diode junction,
so when it was 4.5V at 25°C it will be only 3.96V at 65°C. A constant
current source with a resistor alone is independent of temperature.

Why the bypass? Maybe I'm not following you.

As for a resistor, as I said, there's no fine control. a 9 mA change
through 5k is 45 V whereas that same change through the Vbe mult
only results in mV changes.


Is 100 mA typical?


Not, both? But why in the drain?

50-70mA is typical, to have some accuracy we would need at least 0.1V
voltage drop, which corresponds to 2 ohm. Now if the max. current is 15A,
there would be 30V loss, that is why the resistor has to be bypassed with a
hefty schottky diode. You would then sample the voltage whenever the output
goes through zero. You need only one resistor, because when the output is
zero, the current in the other arm must be of equal value. If you put the
sense resistor in the source line the varying voltage drop affects the bias
setting.

How do you limit current without a much greater R than the typical
0.1R source R?
And a robotic arm to turn it?

you didn't get it, sorry

 

[Active8]

Of course you mean 4.5 Volts.

5000 * .009 = 45

Best Regards,
Mike

 

[Active8]

I started a thread similar to this one about a year ago. The subject was:

"quiescent current in audio MOSFET class AB output stage"

It seemed that the consensus was to just use a vbe multiplier and a trim
pot and be done with it.
I

Someone also recommended Douglas Self's book the _Audio Power Amplifier
Design Handbook_, and after I read that I didn't want to use MOSFET's for
the amp anyway, and started thinking about doing it with BJT's.

Now I'm so conflicted I might not ever do it. In the mean time I'm working
on a pre-amp.

I started that. Need an amp to close the loop

What I'm seeing - and I just plonked down IRF530 IRF9530 for
convenience - is that when Vgs goes up just 0.07 Vdc, Id almost
doubles.

Best Regards,
Mike

 

[Active8]

50-70mA is typical,

Thank you.

to have some accuracy we would need at least 0.1V
voltage drop, which corresponds to 2 ohm.

LTI1166 data sheet disagrees - works for their app.

Now if the max. current is 15A,
there would be 30V loss, that is why the resistor has to be bypassed with a
hefty schottky diode. You would then sample the voltage whenever the output
goes through zero. You need only one resistor, because when the output is
zero, the current in the other arm must be of equal value. If you put the
sense resistor in the source line the varying voltage drop affects the bias
setting.

Yeah. You'd need resistors on at least half of the sources then.
Thanks.

you didn't get it, sorry

A joke?

 

[Mac]

5000 * .009 = 45

Best Regards,
Mike

You are right, of course. My apologies.

--Mac

 

[gwhite]

Why a Vbe multiplier? Mosfets don't need the temp tracking like
bipolars do.

"A Simple Direct-Coupled Power MOSFET Audio Amplifier Topology Featuring Bias
Stabilization" by William "Bill" Roehr, IEEE Transaction on Consumer
Electronics, Vol. CE-28, No. 4, Nov. 1982, pp.546-552.


"One of the major design problems has been providing a means of stabilizing idle
current level because short-channel FET's exhibit an increase of drain current
with temperature at the desired idle level."


I'm not going to comment on Vbe multipliers in this application, but
short-channel MOS AB amps probably require bias stabilization. I don't design
these amps and never have, so I don't know the SOTA. The article claims long
channel FETs have poorer linearity. I think Roehr may have worked for General
Semiconductor.

 

[Pooh Bear]

John Larkin wrote:

The hard part is measuring the smallish fet idle currents...

Actually, the hard bit is distinguishing idle current from load current !


Graham

 

[Active8]

John Larkin wrote:



Actually, the hard bit is distinguishing idle current from load current !

I'm ok with that. I just need to level shift a drop WRT the rail up
to where I can use the measurement to control the idle current.