MOSFET output stage

[Eeyore]

How did you keep it stable?

There's an extra pole and zero. Hence not dominant pole. Works a charm. I've used
that on every amp design since ~ 1990. Sometimes in more than one place in the loop.

Don't know laterals - haven't kept up. Gimme some numbers and I'll go
have a look.

You're a bit late now they've almost come and gone. Arrived ~ 1980 courtesy of
Hitachi. 2SJ56 and 2SK176 were a classic complementary pair but no longer
manufactured. Equivalent types now sourced by Semelab and Exicon.

http://www.profusionplc.com/pro/gex/prodGen.html?prdtyp=lateral mosfet

Graham

 

[Eeyore]

Don't know laterals - haven't kept up. Gimme some numbers and I'll go
have a look.

Had better luck locating the later plastic package devices and they actually show
the relevant transfer characteristic.

http://www.datasheetcatalog.com/datasheets_pdf/2/S/K/1/2SK1058.shtml
http://www.datasheetcatalog.com/datasheets_pdf/2/S/J/1/2SJ162.shtml

Graham

 

[Eeyore]

Oh, I thought there was something new going on. They have exactly the Gm
characteristic I was talking about. Here's the thing. Bipolar Gm is
enormously bigger than mosfet, and you can use that in an output stage.

Oh sure. But look at the Hitachi data I posted a few mins later. Look how linear that
curve is especially beyond the typical 100mA quiescent operating current.. Id vs Vgs The
Exicon data sheet doesn't have the equivalent plot for some reason.

The high gm of bipolars is great until you get to a few mA or tens of mA of Ic when it's
crap and that's where crossover distortion comes from. You just can't get rid of it.

Graham

 

[Eeyore]

Here's the thing. Bipolar Gm is
enormously bigger than mosfet, and you can use that in an output stage.
You do it with local feedback by emitter degeneration - a fraction of an
ohm does it.

Actually NO. A fraction of an ohm isn't enough. You'd have to use about 1 ohm at least
which would be intolerable in practical designs for obvious loss reasons.

There is a cleverer way to do it which blows away all the classic ideas of biasing bipolar
output stages but it's my secret. That basic design was good for 0.008% THD and I wasn't
even trying hard. It borrows on your idea though, just not the same way.

Graham

 

[Rich Grise]

Who do MOSFET sound better than bipolar, as an audio amp output driver?

Because you believe they will.

Cheers!
Rich

 

[Eeyore]

Yup, looking now. The transfer characteristic is the alomst-square-law
curve I was expecting; I don't think you can get anything else from a
mosfet. As for Gm, the 0.4V change in Vgs from -1.2 to -1.6 yields a
drain current change of 0.24A (-0.35 to -0.59A) at 75C. That is a Gm of
0.6!

It averages out including higher currents at about 1S.

You can cross over a bipolar output stage long before you hit that
kind of number.

And you'll still get crossover distortion.

Graham

 

[Eeyore]

Because you believe they will.

Because they CAN. By a country mile. But all people want today is cheap. And
if they want esoteric they want the high distortion of tubes instead of
ultra-low THD of well-engineered mosfet amps.

Graham

 

[GregS]

Because they CAN. By a country mile. But all people want today is cheap. And
if they want esoteric they want the high distortion of tubes instead of
ultra-low THD of well-engineered mosfet amps.

Graham

Seems like most people like the Hafler amp of old. Is the design still around ?

greg

 

[Jorden Verwer]

How about SOA for one you UTTER MORON ?

Fine, fine, but that doesn't have any direct influence on what you'll hear,
because it's a boundary condition.

Do you even know what SOA is ?

Like I said, it's a boundary condition. It can influence the performance of
the circuit, but only indirectly, through other design decisions.

YET MORE INSANE BOLLOCKS

Now you're being the moron (not that I admit to being a moron before). It
seems that either you don't know what you're talking about, or personal
attacks are a hobby of yours. Because, frankly, everything I said was
true...

Fact - MOSFETs have lower offset than BJTs.
Fact - The fidelity of the sound depends on much more than just device
properties.

 

[Eeyore]

0.1 ohm is plenty in a high current stage - it makes all the difference
and brings the Gm down to a pretty flat 9ish - there isn't much to be
gained going lower. The variable resistance it is fighting is 25/(Ic *
1000). By the time you hit 200mA or so that is pretty much
inconsequential - everything above controlled by the external resistor.

Sorry 0.1 is not enough. I don't even go that low for current sharing purposes ! More like 0.15,
0.22 or even 0.33.

You're still missing the point about crossover distortion though. OK for 0.1 % THD maybe but
who's going to buy that ?

Graham

 

[Eeyore]

You get crossover distortion whatever the topology.

The mosfet curves match into each other far far better, plus you're already using more feedback
too. That amp I designed, quite seriously had invisible crossover distortion on an AP analyser's
output.

Graham

 

[Eeyore]

Seems like most people like the Hafler amp of old. Is the design still around ?

I hope not.

Some people like antiques too.

Graham

 

[Eeyore]

Fine, fine, but that doesn't have any direct influence on what you'll hear,
because it's a boundary condition.

It TOTALLY proves wrong your assertion "The device properties of BJTs are
superior to those of MOSFETs in all respects"

Like I said, it's a boundary condition. It can influence the performance of
the circuit, but only indirectly, through other design decisions.

Now you're being the moron (not that I admit to being a moron before). It
seems that either you don't know what you're talking about, or personal
attacks are a hobby of yours. Because, frankly, everything I said was
true...

What the **** is this 'offset' you're talking about. Do you mean biasing ?

Fact - MOSFETs have lower offset than BJTs.
Fact - The fidelity of the sound depends on much more than just device
properties.

You're a COMPLETE IDIOT. You need a boundary condition up the backside. It's
clear that you know zilch about high-performance audio, whilst I've been doing
it for 37 years.

Graham

 

[Jorden Verwer]

It TOTALLY proves wrong your assertion "The device properties of BJTs are
superior to those of MOSFETs in all respects"

No, it doesn't.

What the **** is this 'offset' you're talking about. Do you mean biasing ?

No, of course not. If I'd meant biasing I would've said biasing. BTW, I
don't see how one component's biasing can be "better" than another one's -
it's simply a design step that's necessary to make it work. I mean, nobody
would say "this amplifier's frequency compensation is so much nicer than
that one's"...

As for offset, here's one explanation (in the context of opamps):
http://en.wikipedia.org/wiki/Operational_amplifier#DC_imperfections

Note that I never claimed that this is relevant in audio applications - but
it's there.

As for your personal attacks towards me, I should mention that I'm under the
impression that your experience with electronics outside audio applications
is fairly limited, given that you've apparently never heard of the term
offset.

 

[Eeyore]

Take a look at these two graphs I copied from Doug Self's power
amplifier book. They show the voltage gain of the output pair against
operating point (input volts)for a variety of bias conditions. This can
be used to select a best bias. For the bipolars at the top, the fourth
curve up is clearly the best with gain varying from 0.97 down to 0.963
across the range. This is easily tamed, and even a small error doesn't
do much damage.

Now contrast this with the fets (2SK135/2Sj50) below. First there is no
stable flat line - the gain goes on rising all the way out to the 15V
which is the maximum he measured. Secondly there is no decent bias
current that will control the crossover. I guess that again the fourth
from the bottom is as good as it gets but that gives a gain variation
from 0.83 down to 0.77, with a much sharper turnaround into the Gm
doubling region (spiky crossover products result).

http://89.174.169.10/odds/crossover.gif

That is why it is so much easier to control crossover distortion in
bipolars.

Nonsense. I've seen Doug Self's 'blameless amplifier' diagrams. The crossover 'pip' is clearly
visible in all of them. It's quite good but not that good. I can beat that standing on my head.

When I said the crossover of my big Mosfet amp was 'invisible' that IS what I meant. INVISIBLE
and a THD 14% above the AP analyser residual.

For some reason I can't reach your gif btw.

Graham

 

[Eeyore]

too. That amp I designed, quite seriously had invisible crossover distortion on an AP > analyser's
output.

I'm sure it did, but that is the amp, not the output stage. See my next
post.

What good is an amp without an output stage ? Or vice-versa. I'm not interested in how many fairies
can dance on the head of a pin. I'm intereted in real products you can build reliably in quantity
production..

Graham

 

[Eeyore]

No, of course not. If I'd meant biasing I would've said biasing. BTW, I
don't see how one component's biasing can be "better" than another one's -
it's simply a design step that's necessary to make it work. I mean, nobody
would say "this amplifier's frequency compensation is so much nicer than
that one's"...

As for offset, here's one explanation (in the context of opamps):
http://en.wikipedia.org/wiki/Operational_amplifier#DC_imperfections

Note that I never claimed that this is relevant in audio applications - but
it's there.

Oh for Christ's sake grow up. An amp is a closed loop (often DC) servo almost.
Any offset depends on the INPUT transistors you brainless jerk.

You have the tiniest idea what you're talking about. You must be a lecturer to
be this stupid.

Graham

 

[Eeyore]

As for your personal attacks towards me, I should mention that I'm under the
impression that your experience with electronics outside audio applications
is fairly limited,

Radar, video, embedded control, computer graphics, DSP.

given that you've apparently never heard of the term
offset.

Offset is IRRELEVANT to output devices you complete MORON !

Do do you know what a 'closed loop' means ?

Don, can you smack this guy round the chops for me. He must be a recent product
of our educashun system.


Graham

 

[Eeyore]

Sorry, wrong address

http://81.174.169.10/odds/crossover.gif

Exactly. You CANNOT get rid of the crossover distortion.

Graham

 

[Eeyore]

feedback too. That amp I designed, quite seriously had invisible crossover distortion on an >>> AP
analyser's output.
can dance on the head of a pin. I'm intereted in real products you can build reliably in > quantity
production..


We're talking about the difference between fet and bipolar output
stages. It is simple to reduce distortions in amplifiers with either to
negligible proportions, which is why, for the purposes of the chat, it
is necessary to restrict the chat to output stages per se.

I don't recall that being the OP's question.

Care to run a sim of the harmonic spectrum of a bipolar output stage vs lateral mosfet ?

Graham