John Curl's Blowtorch preamplifier part II

[MRupp]

in the context of matching and whether Vds matching is more appropriate than Idss matching.

But matching for Idss gives you matched Vgs and vice versa, and Gm is separate and pretty uniform across devices for a given current, is it not?

[syn08]

Originally posted by MRupp


But matching for Idss gives you matched Vgs and vice versa, and Gm is separate and pretty uniform across devices for a given current, is it not?

It is not.

JFETs have, in a first approximation, two degrees of freedom, usually Idss and Vp (the pinch, or threshold, voltage).

Id=Idss*(1-Vgs/Vp)^2

and transconductance is:

|d(Id)/d(Vgs)|=2*Idss*(Vt-Vgs)

This is how I would match JFETs (and, FWIW, MOSFETs as well) if I would have today an automatic test equipment. It can be done by hand, but it's very tedious and for 99.99% of applications pretty much not required:

For N channel: choose a very small Vds, e.g. +50mV and monitor the drain current while stepping Vgs from Vp (which is < 0) up to zero. In between -Vp and 0, in linear coordinates, the plot Id=f(Vgs) will always have a linear region. Not at to high currents, where Lambda (aka channel modulation) counts and not at to low currents, where the subthreshold conduction cannot be neglected. Identify this linear region and extrapolate the straight line on the graph, until it intersects the Vgs axis. The (negative) value is called "extrapolated threashold" Vpe and has the remarkable property of not depending on the JFET channel geometry variations.

Sort the devices in Vpe classes of your choice. Within each class, sort by Idss using the known method (Vgs=0, and e.g. Vds=5V) or match devices for Idss using your favorite null method.

You now got a population of devices that have both Vp and Idss matched (and hence also matched transconductance).

If you want to expand the matching over a large range of drain currents (which, btw, for JFETs should be always < Idss) you would have to include a third parameter that is Lambda. The equation is:

Ids-l=Ids*(1-Lambda*Vds)

Unfortunately, there's no longer a convenient way to extract the extrapolated threshold. You need to repeat the above procedure at at least two different (but still small) Vds and then use some sort of nonlinear regression to extract both Vpe and Lambda. Complicated and a pretty much useless exercise for any audio applications.

For P channel devices: mirror everything against the vertical (Ids) axis and take a negative small Vds.

[homemodder]

Originally posted by AndrewT
yes, I have.
I find little correlation, although others have posted the opposite view. Maybe we test differently.

I think Vbe and Vgs are far more important, then I select for a close match of gain (hFE) or transconductance (gfs). BUT ALWAYS at operational quiescent currents.
I am not usually able to test at operational Vce nor Vds, due to dissipation limitations, but I doubt our disagreement is due to this lack of ability.


As an aside I used to measure hFE first and then try to select similar Vbe from the close batches. It was a complete waste of my time.
The Vbe matches could come from any hFE batch.

I now test Vbe first. Batch them according to this parameter.
The slight problem with this is that the next amplifier may run at a different current and that initial selection must be rechecked at the new current.

Reading about the arguments for vbe or hfe matching I decided to do a quick test if their was any corrolation between them as you found there not to be. I usually matched for hfe in my projects although I rarely do so because of buying matched pairs. My reciept only states matched hfe and I have no idea how it was done.

I used 2 different Sanyo transistors, 2sc2911 and 2sc2362k. I cannot be 100 percent conclusive because these parts ,2sc2362, were matched by factory but either they went through a lot of trouble or there is a corrolation.

With 2sc2362 pairs that were supplied as matched pairs with hfe +- 5 percent or better matched I find the vbe s to be closely matched too, sometimes even closer matched. All the parts are from the same batch and I cant see parts from different batches used to get first vbe s matched then hfe s.

With the 2sc2911 I find the same corrolation. These were matched to their complementary but taking a few npn s from same batch and matching their hfe s to within 5 percent I find their vbes to be closely matched too.

This is just the opposite from what you said above

Can anyone else try this, maybe with parts from a different manufacturer.

[Wavebourn]

For JFETs I use an ohmmeter, connecting source and gate.
For BJTs I measure Beta. Matching Vbe requires Syn08'th favorite flag.

[PMA]

Originally posted by syn08
[B]

Id=Idss*(1-Vgs/Vp)^2

Oversimplified textbook true. No real part behaves like this.

[PMA]

This is much better. + channel length modulation, that depends on Vds.

[forr]

Janneman
---This Perrot is that the guy aka Hephaistos?---

Hephaistos was Gérard Perrot's pseudo for the articles he published in french magazine "L'Audiophile". A great part of them were about the design of very linear transistorised circuits aimed to avoid thermal distorsion.

[john curl]

Thanks PMA, very interesting formula.

[PMA]

Robert Mauro (copyright 1989 by Prentice-Hall, Inc., Englewood Cliffs, NJ 07632)

[scott wurcer]

I think Schockley 1952 yields the same equations. The excellent Japanese work in the 70's unfortunately uses the atan function for fitting.

This matching stuff can be "make work" If I have a circuit with two FET's and four resistors and force the resistor values and pick FETs for a perfect output or do the visa-versa, what's the difference? Maybe there's a difference large signal maybe not, build and see.