Sub uA battery powered current source.

Can you apply a voltage across the detector and measure current?

I wonder if a transistor with an emitter resistor would work. The more voltage across the resistor, the lower the Johnson current noise. How much voltage drop will the detector have?

Sure, the pre-amp would have to be different... At the moment it's a three wire connection to a photoresistor down the bottom of a ~4' probe. liquid He.

I think it's about 1-2 Meg ohm at 4.2 K. They need to cool to ~2K, which raises the resistance.

(Seems like 1 uA should be low enough.. the grad student might be confused...)

For low noise at low current, bipolars aren't ideal (shot noise). How about the matched MOSFETs from ALD? Cascode current mirror will have good compliance.

There's good references with smallish currents, but mainly (because they're bandgap based) there's minimum current

Why not a power supply and a resistor?

Thinking something like this:

except the single cell battery spozed to be a programmable zener (TLVH431A), and the current mirror shown isn't the cascode that I envisioned, and a capacitor or two would help... and every time I try to edit it, the wires snarl differently.

ALD's matched MOS device determines what R2 value biases the '431, and R3 ~= R1 for bias current compensation, and the op amp can be slow but low-noise and low power.

That's Figure 13.16 in AOE III. I made one of those as my first bench project at the Institute in 1988. But Figure 13.15 on the next page is more interesting, part of our "Designs by the Masters" series. You can simplify it since you don't need the DAC programming.

. ,-- zener ----, . | A2 -in --+----, . +---out +in ---, | . | | Rbias . | | | . | ,------------+ Vminus . | '- -in A1 | . +--- +in out --+---- Vout . | monitor . Iout

This circuit has the advantage that it can work up toward the positive rail, without dramatically lowering the zener bias current. The current might instead only double. Only A1 needs to be a good JFET or MOSFET opamp.

As for the zener, rather than use a two-terminal IC Vbe reference, etc., to get lower e_n, use a 5.6 or 6.2-volt reference zener, a real zener diode, operating in optimum combo field-effect / zener region, see AoE III page 674.

I remember Dave Jones looking for the precision voltage reference in his teardown of a Keithley 5-digit instrument, repeatedly passing over an innocent looking DO-41 glass zener diode, and not realizing that it was the vaunted beast!

Yeah that's what they were using. Battery is noisy, so maybe just some filter. Voltage reference and filter? Anyway I promised a current source.

George H.

Scratch, scratch... OK I'm not getting that. Is it like a high side current source. ref-opamp-pfet?

George H.

I found it in the 'circuit ideas' at the end of some chapter in AoE1 or 2. I've used it several times.

Huh, OK I'll have to redraw, I was trying to think how to use a shunt reference. Aren't you missing the current setting resistor between out of A2 and (+) in of A1?

Hmm where does Vminus come from? (another battery?) I'll probably use a three terminal voltage ref... I've already ordered some LT6654's from DK.

Thanks, George H.

Nah, bipolars are the bomb. You just have to use one whose beta holds up at low current (e.g. a BFT25A), and use enough emitter degeneration.

It's no problem making quiet current sources well down into the nanoamps that way. I have a laser driver product whose noise is pretty nearly unmeasurable: > 30 dB below shot noise at 100 mA--SPICE says > 50 dB, except at very low frequency where op amp noise is important. To get that quiet, you need a low noise transistor and have to be willing to drop several volts across the degeneration resistor.

Cheers

Phil Hobbs

Yes, I'm sorry I couldn't finish the editing to keep those. Have gotten lots of complaints!

Right!

. ,-- zener ----, . | A2 -in --+----, . +---out +in ---, | . | | Rbias . Rset | | . | ,------------+ Vminus . | '- -in A1 | . +--- +in out --+---- Vout . | monitor . Iout

From your opamp's minus power rail.

It's just that those may be more noisy.

Yes, starting with a current sink; a bridge balances an I*R drop on a known resistor against a Zener voltage, both not-near-the-rail. The op amp drives a current mirror; the mirror's output compliance can be nearly the full voltage of the 9V battery.

Where I wanted to put a bandgap reference (diode/TL431), it's a one-cell battery symbol, because there's a quirky symbol set available. Would you believe there's a Shockley diode, but not a Zener symbol?

OK, that makes more sense. I really like the Vref, R, opamp for it's simplicity.

Say I looked up battery noise, there was some Fed. (nasa?) report from 1995 saying NiCd's had the lowest noise. But I could find nothing about LiPo's. Has anyone looked at battery noise? Maybe I could do that some time. They had to use that cool dual amp chain correlation trick to beat down the amp noise, I've always wanted to try that. (Two amps looking at same noise, the amp noise is un-correlated and averages to zero.)

Actually I'm more interested in the much bigger microphonic noise that I saw in the same lab years ago.

George H.

Complaints? Ignore them. They can buy the earlier versions. You know the earlier versions are much better for beginners, are they still in print? And of more importance to me, any guess on the X-chapters?... maybe you could dribble those out in few chapter sections, you might make more money that way. (Maybe not)

OK, but if the noise is less than that of R_set you hardly care. (For the physics lab measurement we'd run with a battery, and series R cooled down to the same temp as the detector. ~4K.)

George H.

Hmm, is that wrong? Does the active voltage reference cancel out the Rset Johnson noise? (at least up to some frequency.)

GH

The current noise density is (4kT/R)^0.5 + e_n(ref)/R. So you want e_n(ref) < (4kTR)^0.5, so if I = 1uA and R=6.2M for 6.2V zener, then e_n

It's certainly likeable, but the unknown load connected directly to an op amp input makes me itch. I like my control knobs away from the playful kittens. Thus, a current mirror.

Hmm I never worried about the load connection. I guess you could put ~1k ohm in series with the (+) input. Or if more worried a pair of LND150's (my new favorite input protection, 'hammer'.)

George H.

My concern was more with output isolation. There's a microsecond or five while an op amp resettles, that could disturb pulse-shapes.