Low ohm trimpot reliability

ter

Not a 3-terminal resistor, for a start. To get away with 2 terminals, I'd n eed 0.3 to 1.0 ohm (see my OP). 100 BF862s in parallel, maybe?

Plus I really want this board to be all-analogue.

The tempco of the log conformity of my magic transistors is the major impon derable at this point.

Cheers

Phil Hobbs

I once had to design a fancy bias circuit (for a class AB transformer-output RF amplifier) which measured the parasitic emitter resistance of a duplicate of the output transistors and adjusted the bias to suit. That was the only way I could get the amplifier to be linear enough with the predicted range of (systematic) variation in the emitter resistance. It seemed to work ok, but was very complicated and would have been quite sensitive to temperature differences. Fortunately it was all on a chip.

Chris

No, IIRC a regular hunting rifle. Back in those days we didn't have fundamentalist nut cases running amok (yet). The only terrorism was from the left and that was really bad but they (mostly) targeted the upper class. When the ultra-left guys committed terrorism they didn't use guns.

Else there may be folks in fatigues knocking on your door the next time there are exercises in the neighborhood and there is fuzz in their Ku band Radar :-)

I think he meant prayer beads.

Good thing--sounds like it wouldn't have worked otherwise!

Cheers

Phil Hobbs

Nah, with the right bead in the base they work great. You do have to be more careful at higher current. (Nowadays I have test equipment that goes up to 50 GHz, so I'd be able to see it if it oscillated.)

Cheers

Phil Hobbs

Since beads are lossy, ohmic, at higher frequencies, I'd imagine that they have corresponding Johnson noise.

They do. The way I use them would probably horrify a real microwave guy, bu t to me, a 40-cent transistor plus a three-cent bead gets me a 2-3 GHz tran sistor whose f_T is nearly independent of bias, whose flatband noise is way under a nanovolt, and whose Early voltage is practically infinite.

Useful!

Cheers

Phil Hobbs

Not to mention a beta of 500 vs 50.

Cheers

Phil Hobbs

I don't know if you are bound by not using an opamp, but you could do a programmable resistor

Add a 100 mohm sense resistor and a mosfet in series. Monitor the series combination voltage and the voltage across the sense resistor and close the loop with an opamp

Don't know your requirement for bandwidth, may need a very high speed opamp and the 1 ohm RDS on FET will work against that speed

Cheers

Klaus

Can you say anything about the application? It's okay if you cant.

My guess:

Probably this:

Clifford Heath.

Nah, not gravity waves. it's just a general purpose laser noise canceller f or improving all sorts of measurements. I came up with the idea anout 25 ye ars ago, to help a pal with a measurement problem. It improves most laser m easurements by essentially getting rid of

New Focus put basically my original circuit in a box and has been selling l ots of them ever since. So I'm working on a 10x faster one with at least 10 dB better cancellation.

Cheers

Nah, not gravity waves. it's just a general purpose laser noise canceller f or improving all sorts of measurements. I came up with the idea anout 25 ye ars ago, to help a pal with a measurement problem. It improves most laser m easurements by essentially getting rid of

New Focus put basically my original circuit in a box and has been selling l ots of them ever since. So I'm working on a 10x faster one with at least 10 dB better cancellation.

Cheers

... laser amplitude noise. With a bit of care to avoid second order effects that can partially decorrelate the noise between beams(*), it'll let you do shot-noise limited measurements with lasers as much as 70 dB noisier than that.

That's why the paper I posted is titled "Ultrasensitive Measurements Without Tears".

Cheers

Phil Hobbs (fat fingers again)

(*) such as a stray fringe due to a first surface reflection demodulating the laser's FM noise

Had a look at this, finally--thanks. The compensation voltage there is proportional to the tail current. Since Q1 runs at a fixed 10 uA, it's only Q2 that ever runs any significant current, so ignoring Q1's R_ee'+R_bb'/beta doesn't hurt the compensation accuracy much.

In my situation, both I_C1 and I_C2 can be anything up to 10 mA or so, which means I can't ignore either of them, and so need both degrees of freedom in the compensation.

Cheers

Phil Hobbs