Linearized Source Follower


Paid Member
2002-07-12 9:55 am
San Diego
Has anyone tried anything like this, to linearize a class-A source follower by indirectly controlling its Vgs by controlling the current in the device?


  • linsf.gif
    4.1 KB · Views: 1,486


2001-02-04 4:23 am
An old, old circuit from Wireless World. It doesn't work the way you think it does. The top device is running class AB. On positive waveforms it is driven further into enhancement and the voltage drop across its drain resistor shuts off the current source. As the waveform drives negative the falling current through the drain resistor makes the current source kick in and deliver the negative portion of the waveform. It will work better with BJTs than FETs. you will need lots of loop feedback to overcome the Vgs error. Even though it is a dynamic bias scheme, it would probably work best with FETs when the idling current is set to about half what you would normally set a single ended class A stage to. In other words, bias it like a push-pull class A stage.

This circuit is interesting, and rather like a Single Ended Push Pull I am presently considering with sliding bias.

However, here's another which more simply achieves the linearising effect you refer to.

It is not original, Self referred to it in a Class AB PP topology in the late eighties.

However, this is a single ended version, and it is very linear. It sounds absolutely wonderful running at around 3A from 50V, and will drive an 8R speaker to around 42Vpp, which is 28W which astonishing impact. I developed an amp using refined versions of this circuit and it is probably my best amp, but oh! the inefficiency is crippling........

There is an asymmetrical distortion (H2, in fact) introduced by the varying Vbe with signal excursion; when Vce is large during the negative half cycle, the Vbe is small, and the step function tends to squash the waveform; alternately sharpening it on the positive half cycle. But the effect is nothing like as pronounced as a straight source follower, and the output could reasonably be called a step function of the input.

Stability can be an issue. A 220R gate resistor is mandatory, and sometimes a 10R resistor inserted into the emitter of the bipolar is required. It is also sensible to decouple the power supply for the resistor from that supplying the mosfet; the two can interact causing intermodulation problems.




  • gh1.gif
    13.9 KB · Views: 1,452


Paid Member
2002-07-12 9:55 am
San Diego
No, it's not original -- it is essentially a folded version of the "hybrid-Sziklai" (?) circuit you drew, AKSA.

Just out of curiosity, how much current do you run in the NPN of your circuit?

The folded version will have more loop gain, because the PNP will buffer the Vgs of the lower NMOS, but then it also probably has some worse stability issues for that very reason.

I think it won't need the power supply decoupling you mention because of that buffering.

As for efficiency -- that's a problem. It is probably going to be worse than the usual class-A output stage. And maybe thus relegated to things like linestage output or headphone amp.

To solve the efficiency problem, maybe it can be run class-B by paralleling it with the complementary PFET / NPN and

On the other hand, being (almost) all N, it might make a nifty tube circuit...

I'll post some results after I simulate it a bit (I just downloaded Switchercad-III) and if it looks promising in the simulator then try to build some version of it.


Paid Member
2002-07-12 9:55 am
San Diego
I meant to say ..

To solve the efficiency problem, maybe it can be run class-B by parallelling it with the complementary PFET / NPN and reducing the controlled current in the (now complementary) source followers, letting the bulk of current handling be taken care of by the current sources.

One thing I never quite understood about MOSFET amplifiers is the gate resistor thing. If I drive the gates from a high impedance, what good do they do? My naive thought is that the gates are driven from high impedance, they do nothing, and if they are voltage driven then gate resistors would actually degrade stability by introducing another pole in the system.
The Pass Patent

Hi Jan,

No, I have not read the Pass patent. However, much of Nelson's work is familiar to me; years ago, I built a Zen and quite liked it and this started me off on this voyage of discovery. I like his minimalism and fundamental grasp of the strengths and weaknesses. He is also a very decent human being......

The amp using this topology I have built I call the Glass Harmony. The input stage, the voltage amplifier, is a 6SL7, and the output stage is two parallel P types while the current source is two parallel N types, IRFP 150s.

Mirlo, I run only 4.2mA through the bipolar by using a 1K resistor on the collector/rail, and a 0R47 resistor in the source to degenerate the OLG a little.

Your comment about the gate resistor is well taken. The pupose of a gate resistor is to block rapid transfer of charge to and from the gate, so that the infamous self-oscillation properties of a mosfet can be 'poled' out, to use your terminology.

It works pretty well, and you are right that buffering the power supply would not be necessary with a cascode. However, such an approach would limit the voltage swing at the drain of the mosfet, which is not good for rail efficiency.


Hi Gromanswe,

I am not sure of the significance of your warning about sellers. :confused:

I am not trying to peddle my wares, good as they are. Nor was I aware I was exhorting anyone to buy anything. The Glass Harmony was an early project of mine, and never did become a commercial product. If anyone wanted one, I would decline, because I deplore it's inefficiency.

Like you, I am attempting to participate in a forum of ideas, in the hope that I may contribute something and perhaps even learn something new.

Your view of trade is interesting, too. In the US if a wealthy man drives an expensive car, people often applaud him and wish him well. In my country, and quite possibly yours, they usually deride him for exploiting others and ripping off society. Therein lies a differentiator between Australia and the US; here, we like everyone to be equal, but sadly, it just cannot happen.

You have quite a lineage of cryptic comments - be aware that riddles do not come over well in this medium and straight talk is always preferred, particularly as there are many language groups.



Nelson Pass

The one and only
Paid Member
2001-03-29 12:38 am
Hugh's diagram is not related to my current source, but
interestingly is very much like the circuit which was part
of the Threshold Stasis amps (later generations).

I have been curious for some time as to the origin of
"Sziklai connection" which describes a complementary
follower of this sort, as I had been taught the term
"conjugate complementary". On Google I learned that
George Sziklai was a distinguished inventor at RCA
and certainly deserves to have something named after

More interestingly, "conjugate complementary" I recall
from a Motorola app note, which might reflect the rivalry
between the two companies designing the first solid
state hi-fidelity audio amplifiers. Both produced classic
circuits, with RCA emphasizing quasi-complementary and
Motorola full complementary.

RCA influenced the HK Citation 12 to say the least, and also
the Phase Linear (and ESS...) products.

Motorola looks to have influenced Daniel Meyer's Tiger
series, and Ampzilla and so on.

As another aside, I'm pretty certain that Hugh's reputation
in the industry is solid enough to offset suspicion of his
commercial interest.

I'm the guy you should watch out for. ;)
Hi Tschrama,

Your point is interesting, but I don't really think there is a problem.

Remember that this circuit relies on the 100% voltage feedback from the mosfet drain back to the bipolar emitter; this is the reference, so anything happening in the collector circuit will merely be accommodated into the feedback loop which still must take the drain of the mosfet as its reference. I call this a Sziklai pair - the European notation since Sziklai was a Hungarian who left Budapest a couple of years before the '56 revolution. In truth, a reasonable analogy is the little man sitting atop the elephant, urging the elephant to move lumber under his control.

No, you want a circuit element between the emitter and the drain which displays negative resistance; increasing current flow, reducing voltage drop. This is, well, tricky..........

In any event, why would you want to remove the H2? Are you aware the recording process strips away some of the harmonics and skews the harmonic spectrum of music, and that injecting just a little H2 'fills' out the music, and makes it sound better? The trick is not to exceed about half a percent, and if it's H2 and a touch of H3, she's apples, and I can assure you, you would prefer it over a surgical, sharp rendition straight from the CD, through a 0.001% amplifier and a pair of highly accurate loudspeakers.

I expect the valve afficionadoes would be concerned with this little heretic gem!

Nelson, people have been watching you for years, and rarely from suspicion....... :D




2001-02-06 6:18 am
heretic gem

Ha ha.. well said, as usual, Hugh.

Hugh and NP, your reputations here are solid as ever, pls
don't ever let some vague innuendo change your contributions to
the audio world, and this forum in particular.
I'd rather give my money to the good guys.

Good on ya, mates!
be so, evrbooda

"Therein lies a differentiator between Australia and the US; here, we like everyone to be equal, but sadly, it just cannot happen.
Sincerely by AKSA"

I have never been outside my land.
This land is my land.

The only australian I ever met was a teacher, in english! :D
in my school here in sweden

I have always and I MEAN always liked "the spirit" around in "your" australian terrain
may be aborginies or newcomers

the success full olympic games in Sidney reinsured
me in my opinion

the audio friends I have met on web
with australian origin hasn't given me anything

sorry not to say thatt from yankeemaan

we are certainly more "on speakin' terms"
you and I
not to mention speakers :)

things may change
The times are achangin'
to put in a quote