SIT-Preamp

T4 act as a current generator and T5 as a voltage generator.
Yes i like your amp, it is a current amplifier or as i prefer to call it "transconductance amplifier". (see my site also) :)

Other little V-FETs are 2SK63 (equiv. to 2SK79) 2SK69/2SJ19, 2SK78/2SJ24 and 2SK75 as medium power, but all they are very scarce (i have 2sk79 only).

Thanks for your advice about KP903A but at this moment i need a P-CH type.
Is there a similar jfet as KP 903A in a P-channel type?

Francesco

Yes, T4 is a current source like in Zen amps, and T5 acts as a shunt regulator (powerful Zener diode), which shares fixed current with the amplification stage. This is good for power supply decoupling and makes big polypropylene cap not needed. I guess 2SJ162 could be even better than 2SJ200, since currents are close to the thermally stable point of this lateral MOSFET.
Unfortunately, KP903A has no p-channel analog, but an idea of making 100mA cascoded input stage for Pass Aleph J is so attractive, that I consider possibility of using 2SJ200 at its output instead of IRFP240, and this will open way to 100W-400W two-stage power amp design.
 
I plan trying also a version of SIT preamp with CCS load and NFB, with a non-standard solution for volume control. It should be fast and should have small output resistance. NFB increases with decreasing volume.
 

Attachments

  • SIT preamp with CCS load and NFB.JPG
    SIT preamp with CCS load and NFB.JPG
    75.4 KB · Views: 1,496
Last edited:
How did this project turn out with the SIT's?

Are you still using them, the curves look a lot like the devices NP is having made.

A finish of this project was not successful, I have found that another "crazy" preamp configuration sounds a bit better.
I am using now, KP903A j-FET at the input, cascoded by KP926A SIT over it, and a SIT-based (with some gate-source resistance) current source (160mA) as a load to this stage, and shunt PS to this stage. Total power dissipation, 320mA x 100V = 32W, the design looks like a power amp.
Although looks simple, the last approach needs SIT matching and correct choice of PS voltage, according to their individual properties.
 
Because SIT T3 of post #22 must drive all load current swing,
it is not operating on a constant current load line for best Mu.
Better to actively modulate current source T4 by watching the
voltage drop across R6 22R, and attempt to hold SIT current
to the constant instead...
 
Last edited:
Something like this perhaps?

Let Mu happen upon a constant current loadline, and the other N
transistor (specific type unimportant) dumps away load currents.
SIT gain not degenerated by R1, as R1 is held to constant voltage.
J2 function also improves PSRR by actively cascoding SIT J1...

Remove the SIT, and you will see a ring of two current source...
I suppose Q1 could also be a MOSFET, if the threshold is known.
 

Attachments

  • Taylor.gif
    Taylor.gif
    9.7 KB · Views: 1,176
Last edited:
Because SIT T3 of post #22 must drive all load current swing,
it is not operating on a constant current load line for best Mu.
Better to actively modulate current source T4 by watching the
voltage drop across R6 22R, and attempt to hold SIT current
to the constant instead...

Hello, Kenpeter
I have got the best listening impressions with the following schematics.
"Imperfection" of the current source is compensated by the shunt PS (I mean efficiency of transients reproduction).
Other feature of the "imperfect" current source, is that amplification factor is adjusted by SIT matching and PS voltage, hence no NFB needed.
 

Attachments

  • CRAZY PREAMP.jpg
    CRAZY PREAMP.jpg
    78 KB · Views: 1,179
Amplification factor of your SIT does absolutely nothing in cascode.
Only the negative feedback from R6 is giving your amp control now.
Entirely opposite of your intentions.

Amplification factor "Mu" is voltage gain at constant current.
This is due to SIT local negative feedback from drain to channel.

Yet you hold drain voltage of VT1 constant, preventing this very
important function that makes Mu. Why even bother with SIT?

You vary current through all three devices, choosing to operate
upon load lines dominated by non-linear factor Gm, rather than Mu.
 
Amplification factor of your SIT does absolutely nothing in cascode.
Only the negative feedback from R6 is giving your amp control now.
Entirely opposite of your intentions.

Speaking about amplification, I mean its dependance on the dynamic resistance of VT3 at its working point. With a sinple resistive load, this stage would have minimum amplification, with "ideal" current source - it would be very high, and NFB would be needed. With SIT load, it is just OK for my purposes, near 16.
This schematics serves excellently as a preamp, with DACT volume control, it is extremely low noise and very fast, 380V/uS. With all my respect to John Curl's Parasound JC2 preamp and its no-cap concept, it looses obviously to this schematics.
Good SE 300B (manufacturer is NEM from Novosibirsk, their similar tube equipment was nominated as a winner at Las-Vegas, High-End cathegory) integrated amp also looses in soundstage and microdetails (I do not care for the bottom end here) against combination of this preamp - RF transistors based modified follower ( listening with PMC EB1i speakers).
At the same time, I would not advertise it for others, since definite luck is needed with SIT properties and their matching. The shunt PS is also a key factor. Not as straightforward as with BJTs.
 
Last edited:
OK then, 1 transistor not wasted. I am still not seeing the other two
utilized in a way that makes use of Mu.

You can be right in some aspects, and of course can design very different thing. I have tried some other modifications, SRPP-like, and did not choose them.
I would be extremely interested in your listening reports, about more perfect schematics. There are many of them, at the tubecad pages.
In all cases, when I loved some schematics theoretically, no cap in the signal chain etc., I did not manage to get a sound. Main tricks relate to noise level, using of shunt PS concept, parts selection, design induced EMI, slew-rate, etc., i.e. to aspects not directly linked with linearity. In spite that this schematics looks as very trivial, it can not be approached sound-wise by varios other preamps laying around.
 
Last edited:
This type of output curves (artificial triode) is the general property of
all transconductance amplifiers (high output imp.) with the voltage FB.
Vbe multiplier is one example. Opamps with the voltage FB will also
have this type of output curves.
Sure, one transistor is the second shortest FB path (loop), (after "emitter"
follower).
 
Hi Vladimir,
I saw the updated schema with the mu-follower.
I found the KP931A, it is also called a SIT.
Can that replace the KP926 (which is expensive now)? Lower current, lower amplification of course.
Albert

No, they can not. KP931, KP934, KP959 are BSITs, different from SITs. They are essentially the same bipolar, only made using JFET technologies. They have conventional pentode I – V characteristics.
 
Thank you, Vladimir,
I understand BSITs are not depletion types.

What JFET power types are there then? You use KP903A, for instance.

I like that some BSITs have an incredible low input capacitance -
the KP901A= <100pF, (insulated-gate field-induced)
the KP902B < 12 pF (planar silicon insulated-gate field )
the KP902A 11 pF (planar insulated-gate field)​

So that KP934 ? a B-SIT (planar field-static induction).

I understand. One has to be careful with the extensions: KP903A = JFET that you have in post #22; but the describing of KP903B = silicon epitaxial planar field, to me then a normal Mosfet

But why it confuses me:
The KP931A = silicon epitaxial planar vertical channel with static induction[/I] looks in some data (800V) and with that static induction :cool: in the description it really induces me to think it is a SIT :rolleyes: Anyway, there are some heading in my direction.

I like the idea of a simple preamp. With a simple depletion device. Mu-loaded if possible.
 
Last edited: