JLH Liniac Revisited

[wrenchone]

Attached is a variant of Hood's Liniac circuit from the 70's with some bells and whistles thrown in. The intended application is a line amp with gain. The perfect current source in the simulation schematic will be replaced with a JFET current source. The current source is there to supply extra bias current to the input JFET J1 such that the current it swings to push the output around is a small proportion of the total bias current. The current mirror (Q3, Q4) provides a low-headroom 2 mA current source as load for J1. J2 is a cascode for J1, doing the usual cascode things but also clamping the drain voltage of J1 to a reasonably low value, allowing healthy bias current without undue power dissipation.

I'll be building this up soon, as the simulation is one thing, but the listening experience is quite another.

[wrenchone]

Attached is the simulated distortion profile. The distribuition of harmonics looks promising, if I can trust the simulation.

[Joachim Gerhard]

that looks promising good luck
have you tried to use a bipolar for j2 ?

[wrenchone]

I like a JFET because it floats on top pf the input JFET source without requiring any external drive voltage. I may consider using a depletion mode MOSFET if I can find something suitable. I have some lower voltage Supertex parts I may try, at least a far as having a look at them in my sorting jig to see what the VGS is at 2 mA.

[Joachim Gerhard]

i tried the fet-fet cascode and the fet-bipolar performed better for me but i understand your resoning it is such an elegant oportunity

[wrenchone]

If I did a bipolar cascode, I would drive the base of the bipolar referenced to ground rather than the J1 source so as not to inject extra current into the J1 source resistor/summing node. Not really a cascode after that, though it still satisfies most of the purposes for having it there in this circuit.

[wrenchone]

As an extra comment on the bipolar-fet vs. fet-fet cascode, I take pains to select the cascode fet such that there is enough VGS at the bias current of the lower fet to tone down the parasitic capacitances. I like to see 4-5V across the bottom fet. I don't know if other designers have taken this into account when making a comparison.

Having said that, I realize that a bipolar in general will may pin the drain voltage of the bottom fet much more securely than an average jfet, as the current pumped through the base-emitter voltage of the bipolar (the difference in base current due to the change in emitter current) will cause a much lower shift in the bottom fet drain voltage (assuming decent gain and base spreading resistance) than in the case of the fet cascode element, where the VGS of the top fet must change to accommodate the signal current in the bottom fet. If the cascode fet has low transconductance, the VGS modulation may be significant.

Food for thought, anyway.

[Joachim Gerhard]

excellent explanation, i could not do it better. The Paris-Aksa head amp has the base of the cascode bipolar referenced to ground and it works surprisingly well. -143dB noise with only one high gm fet. distortion is mostly second harmonic in single ended topology.

[wrenchone]

Just for grins. I subbed in a BJT cascode device (a 2N4401, no slouch) referred to ground in place of the floating JFET cascode. It's still early days yet, but the substitution caused a 2X increase in THD (didn't look closely at the harmonic distribution, will putter around more with it later). One explanation for this may be that this is a very low gain amp, so the summing node moves around. Next time I look at the sim, I'll plant a probe there.

For information, in what situation did you see better results with the BJT cascode as compared to a JFET cascode, and were the conclusions based on measurements or listening or both? I'm not getting argumentative here just wanting to know the facts in your case.

For the simple single-ended, single supply circuits I'm investigating, reasonably high gain and unity gain (or thereabouts) circuits are simple - low gains (<10) are a bitch.

[Joachim Gerhard]

i look at it from the impedance standpoint. most bipolars have lower impedance at the emitter than comparable fets at the source. the result is more speed and lower distortion.
maybe the gm of your fet is high enough so that is not the problem.
actually i build a fet-fet cascode mc pre-pre (it was more like SRPP) but was not very impressed by the performane. it worked but it did not put a lasting impression in my mind. what i have and sometimes listen to is a fet-bipolar cascode mc pre-pre.
it sounds good, very transparent and open. nothing of the fat and worm sound that people atribute to fets because they behave like tubes. the bipolar specimen in that design is a 2N4401 too, funny