Cascoded Ring of Two for Plate Load

[wrenchone]

I've been looking at using a cascoded ring of two as a plate load for one of my preamp projects, as it requires less bias current than a cascoded bipolar current source using an LED reference. The circuit is shown in the attachment. Values are currently set fo 3 mA current. The AC source shown in the schematic was used to check the output resistance of the circuit - I got about 4M. Things may be a little better if I set a smaller step time and tolerance for the simulation.

Any way, food for thought...

[wrenchone]

I sharpened my pencil and did a simulation with tighter tolerance. Also, the voltage source I'm using to jerk around the collector of Q3 turns out to be 20V p-p. This all adds up to 10M of output impedance, rather than 4M.

[wrenchone]

Adding another series transistor in the simulation didn't increase the output impedance one bit. I may try to simulate the LED-based current source to see how well it performs.I have a feeling the performance will be similar.

[EC8010]

What's the purpose of C2?

[wrenchone]

C2 has no purpose in a simulation as all the sources are perfect and squeaky-clean. In the real world it would clean up the feed to the ring of two control transistor. Papa uses this trick on some of his earlier Zen amps. I may just use a JFET current source at the bottom instead, so that the line rejection is improved and the control transistor has a well-defined current against which to work.

Anyway, I was exploring this option since the usual depletion mode mosfets are reputed to not work so well at low current, and I didn't like the idea of ramming 5-10ma into an LED reference just to jazz a 2-3ma current source. I'm trying to limit how much I draw from the B+, as I'm planning on generating it in a somewhat unconventional manner. I also have loads of 2N4403 and MPSW92, as well as a Fairchild TO-126 video driver transistor for higher currents.

[wrenchone]

I tried a simple Jone/SY-type bipolar cascoded current source in simulation using a stack of 3 pieces of 1N4003 instead of the LED for biasing. The bottom transistor was a 2N4403, cascoded with an MPSW92.The diode stack was biased at 5ma, and the current source was set to deliver 3ma. Output impedance was around 9M, very similar to the cascoded ring of two source.

[wrenchone]

The ring of two cascode simulates as having slightly better output impedance than the classic LED biased cascoded bipolar source (10M vs 9M), with far less maintenance bias current (0.5-1ma vs 5 ma). If no one finds an obvious hole in the design, I'm going to try this approach for the 6J6 version of my "Evil Sandman" preamp. No pretty lights, but I can live with that.

[Gordy]

I still don't understand C2!

Either way 9 or 10M seems like a good figure.

[SY]

OTOH, a cascode of depletion mode MOSFETs can give an impedance in the hundreds of megohms, with ridiculously low noise.

[wrenchone]

Yes, but will they work well at 2-5 ma? I have another circuit (see the "Evil Sandman" thread) using a triode-connected high gm pentode that will be running at 20ma current to fully exploit its "goodness". I intend to use depletion mode fets as a load there.

At any rate, I ran simulations for the bipolar current sources to compare the simulated performance of the ring of two vs the LED-biased cascode sources, in order to get a figure of merit. The simulation is only as good as the models, and the some transistors/models were used for both. They both appear to be roughly equivelent, with the ring of two circuit being slightly better. Real-life performance may be different, but I intend to try it and see. I figure it'll be good enough for a 6J6...