If You Don't Have a SIT, Fake It...

[Salas]

Originally Posted by wrenchone

Salas, the whole point of this exercise was to be able to push up the impedance of the feedback network and yet have some drive available for the mosfet gate. The way you show it, you really need that electrolytic (or a great honkin' film cap), plus a buffer to drive the buffer.

Was just an idea to check where the hiss comes from on those horns. Valid points the drive difficulty and the large input capacitor indeed. So I went on investigating the theoretical THD differences. I run two FFT on LT SPICE. First is for the ~5x gain high impedance feedback network, the second is for the low impedance ~3x gain. Both on 20mA depletion mode Mosfet CCS setting. Simulated SNR -91dB HiImp, -104dB LoImp.

[wrenchone]

That's very nice, but if you screw up the signal source trying to drive the low impedance, you're not exactly ahead of the game. You might try increasing the bias current in the mosfet. That definitely worked for me in terms of reducing the distortion. I'm biasing the input jfet at ~5 ma, which should work nice for the 2SK117.

[wrenchone]

Here's the latest effort - a bit neater that my first. There's two channels and an onboard shunt regulator. The output mosfet and its current source drain load are both cascoded. One channel exhibited high amplitude oscillations at ~350 kHZ (so low?), while the other looked stable. About 3" of "gimmick" capacitor was applied to the feedback resistor of both channels. It stopped the oscillation on the one channel, and did not obviously destabilize the other. Next up are tests to determine what is going on. I'll post schematics when I'm reasonably satisfied.

[vgeorge]

Nice board wrenchone.
Waiting your final schematic.
I did try salas suggestion with lower resistors. Noise was down but distortion was higher.
With less sensitive speakers maybe there is no problem.

[juma]

Hi Wrench,
here is my take on "turning a (MOS)FET into a SIT/triode". Fair Zin, low Zout, BW(-3dB)=1MHz, rail-to-rail swing (almost), high PSRR, Id about 15mA, sounds wonderfull - what more could a man want ?

Built, tested, measured (except for THD) - here is the schematics:

[buzzforb]

Juma,
R2-4, D1, and C1 are biasing network for top fet? Also, are ther two feedback loops R5/R7(schade component) and R6/R7? Very interesting thread from Wrenchone. Love following it.

[juma]

Originally Posted by buzzforb

R2-4, D1, and C1 are biasing network for top fet?

Yes, it's called a gyrator (fake choke - the most appropriate load for the fake triode ). Voltage divider R2/R4 sets the gate of Q1 at about 2.5V above Vcc/2

Originally Posted by buzzforb

R5/R7(schade component)

Right. It sets the gain and Zin

Originally Posted by buzzforb

and R6/R7?

No. Voltage divider R5/R6 sets the Id. R5 is already involved in gain setting so the R6 is the one to be manipulated to set the Id (I find the 15mA to be an optimum value)

[buzzforb]

As always, Juma, a clear and easy to understand explanation. I don't know why less thinking R5/R6 was feedback, didn't click that these were enhancement mode devices and needed a positive voltage at the gate. I think I have seen the gyrator, minus the diode and with cap going from divider to ground, used to bias the top Jfet in an SRPP type circuit. I have been watching this thread as I am considering trying to use LU Jfet in FE for BA output stage and wrenchone always interesting circuits if for no other reason than just learning. Thanks to both of you.

[wrenchone]

Don't look now, but I've got another version of this thing up my sleeve. PSpice and the other simulators are wonderful toys (to geeks, at least).

[Salas]

Originally Posted by vgeorge

Nice board wrenchone.
Waiting your final schematic.
I did try salas suggestion with lower resistors. Noise was down but distortion was higher.
With less sensitive speakers maybe there is no problem.

Hi Vgeorge. Having a suspicion also about the CCS amplifying its degeneration resistor's noise enough for your high sens horn speakers, I looked in to just a resistor current source. So I brought back your input impedance higher for your source and pot to better handle, hoping that if there was hiss contribution by the CCS maybe we can trade what we last time cut by lowering Rf/Ri. If you still got it handy its an easy mod to test. Sims >30mA for M1 and a bit more THD 2ndH in the mix. R7 should be a 2W part.