JFET with matched JFET current sink for linearising

Oh shoot I just found this while doing some information gathering. An article by Kampf V.A regarding AMTs "warm stone" solid state 12ax7 substitute.


Looks sort of familiar...

The full circuit is a bit more involved:

I assume D1 emulates grid current, and perhaps prevents unintended fold over distortion? T3 probably conditions the feedback going into T2 in order to achieve the right bias point and "knee", as I call it. D2 and D3 migh be used to round of the transition into cutoff? Pretty neat design! I think the circuit can probably be simplified for use in a "bespoke" circuit (doesnt need to be hot swappable with 12ax7 tubes).

So guess i was right in thinking it was already "a thing" and not an original idea, but at least it's proven somewhat valid for the application 🫠 i hope there is no rigid patent surrounding it l 🤔
I don't see much correspondence with your circuit, neither for the simple nor for the complete circuit. The simple circuit consists of an N-channel and a P-channel JFET tied together at the sources, that shouldn't compensate out even distortion terms.
I guess you need to use some (maybe lots of) imagination to see it. It's not so close to the simple "single ended diferential pair" as I sketched it out. But I still think its performing the same, very general, task of shaping the output through nonlinear source feedback. Although admittedly, T2 is operating by very different means.

I got a bit exited because the Warm Stone shows something fairly close to the direction I was planning on taking this "basic idea" I presented.

On disecting AMTs circuit.
I'm not super acquainted with P channel JFETs but I see them as sort of "inverted" N channel (feel free to tell my how wrong I am😅). With that gross simplification I'm thinking it would act to sort of invert the feedback again after it has been inverted by T3. Would that be a somewhat accurate statement? It would be in line with what I think is fitting for a triode emulation (a harder knee for the overall NFB). I'm thinking in super wide strokes. If this I'm not completely off the deep end, couldn't the "Warm Stone"s feedback be fairly similar to adding some feedback down directly from the output to an N channel JFET in the bottom position like so?


Sorry for a rather discombobulated entry. I just woke up. From another night of wild transistor dreams 😵‍💫
...ways of emulating triode like behaviour...
I've attached this pdf (#25) and a patent (pdf), furthermore this link.

The circuit in #22 contains hard to get components and other difficulties to deal with.
The circuit proposesed in the patent (1998) is even more difficult to realise.
The circuit in the link uses the known simple resistor feedback.


    688.1 KB · Views: 35
  • US5734725 Tube Emulator.pdf
    889 KB · Views: 35
  • Like
Reactions: 1 user
Do you at least have a simulated a distortion spectrum plot that you can post?
Doing distortion measurements of the basic idea (single ended differential amplifier) is nothing I'm interested in at the moment. It's not really where my interest lies regarding this, I don't have the proper spice models, and I'm really putting way to much time and thought into this already😅. I really should be focused on writing my master thesis right now (regarding feed forward methods in active noise cancelling headphones, in case someone is interested 🙃).

Hopefully I get more time to do some actual work with it in due time. Right now, i have been very happy to start a discussion and ventilate some ideas.

Enough writing about myself and my reasons already. Gosh.

Maybe someone else is inspired to work a bit with fleshing out the concept? 🙂
Joined 2006
Paid Member
Booberg, what i see in your cd, is a cascode of depletion jfets where the lower fet (because of the low drain source voltage) is forced to work in the
resistive mood, as a resistor.
Replacing the lower fet with a resistor for comparison most likely will confirm that..
Last edited:
Huh? How can it be both.

Jan Mulder and I called the configuration in the article linked to in post #12 ( http://resolver.tudelft.nl/uuid:c52dd5db-5d96-477f-abc1-881b785b742b , pdf at the bottom of the page) a single-ended differential pair because it is a single-ended circuit that has the same large-signal input-to-output characteristic as a differential pair that is driven on one input and has its other input fixed to a DC level, while also only one of its outputs is used.
Joined 2011
Paid Member
Hey, that's me! Reference 14.

We used that circuit in NMOS dynamic RAM chips in the mid-1980s. All Nchannel FETs were built in the (shared) P-type substrate, which was charge pumped down to about -3V to reduce junction capacitance. So, unlike Marcel's Figure 3, the earlier DRAM version of the circuit did not tie the source to the bulk, on either device, and there was nonidentical body effect. But the second iteration in 1993, the VT extractor, certainly did.



  • bott.png
    15.6 KB · Views: 40
  • Like
Reactions: 1 user
I suppose Jan Mulder added that reference after a comment from a reviewer. I had left the university by then.

Jan also added the introduction. Electronics Letters didn't publish an earlier version without it, so Jan added an introduction with all buzzwords that were then in fashion and sent it to the IEEE.
  • Like
Reactions: 1 user
Ok, so give me some time to cook here. What I really like with this circuit is that it's very easy to customise the manner of it's exponential behaviour.. Marcel called his 'very similar' configuration, roughly: "Single Ended Diferential" Amplifier.. Or perhaps a "SED" amplifier?? In my native swedish, a "sed" is defined as an adopted behaviour, a custom, or manner.
It's definitively a sign! Of lack of sleep probably 😅

After some tweaking I have so far landed on this general layout:

Where the potentiometer does a very good job of varying the response from almost completely "bypassed" (drain-gate shorted) to pretty much symmetrical (common gate) depending on the bias. It really needs to be logarithmic though as it is wildly sensitive around common gate settings.

Still only tested it in Falstad so far, but I have a batch of dual 389 and 489 Jfets on the way so I might do some scope readings in the not too distant future 👍
  • Like
Reactions: 1 user
One of the interesting aspects of your SED concept is that it enables the upper FET bias current to be set independent of the lower FET bias current. Just select the potentiometer value needed to bias the upper FET, and then adjust the wiper to bias the lower FET which is in parallel, for whatever desired sound effect.