Ring of two current source

Base current error is fine-tune the current source.

Yeah, that's what I mean -- more stuff. And then you need an op-amp floating up there, which stinks if your supplies are wide, and... But sometimes, that's the only good way to do it.

Tim

The OP should define "very temperature stable."

So, I'll go for "as stable as possible."

No problem, just add another current source. It's already got two, so the more the merrier. ;)

Cheers

Phil Hobbs

As long as you use a single supply, and don't use a RRIO or one of those odd Motorola parts whose output swing exceeds its input CM range, that circuit is guaranteed to start up, no?

It's a great way to make 10V from an LM329, say.

Cheers

Phil Hobbs

The obvious version (zener grounded, dual supply opamp) powers up with net positive feedback, which, with a little input offset, could slew either way. The feedback gets negative after the zener starts conducting... in either direction.

Last time I did this, it was with a 741 (or maybe a 709?) opamp and a $50 zener that came in a presentation-quality box with signed certificates. I can get a better, trimmed reference now in a SOT-23.

Speaking of references, we are using a ZYNQ SOC (FPGA with dual ARMs) which has an internal ADC with reference. It's terrible. I guess it doesn't help that the FPGA self-heats to over 100C.

It's easy enough to add an extra diode to make sure that it can only work as intended - it's less embarrassing if you put it in during design rather than development.

Sounds like a 1N829, which has a built-in forward diode and doesn't present that problem.

Probably not. An FPGA plus dual ARMs does rather suggest a digital process, which won't help either.

Sure. You have to use a single supply, but as long as the output isn't R-R, and doesn't exceed the input CM range, it'll start up for sure. No?

But not quieter than an LM329!

No huge surprise there!

Cheers

Phil Hobbs

Yes, I suppose that would help. I have to do some rather unpleasant math and analysis to give a definite answer to that question unfortunately, so I'm working on it.

Numbers like 20 PPM per degree C wouldn't be hard or expensive, for moderate currents, like 10 mA or less.

A few PPM/degC, especially at higher currents, would be more interesting.

Wide bandwidth, low capacitance, high accuracy current sources are interesting, too.

My back of the envelope calculations are showing that the high end of that range at a few mA should be OK. What I would really like is a sink and a source that would track each other well over temperature.

y get

goes

be in

stor

Depending on the performance you actually need, a complementary pair of Wil son current sources with proper monolithic dual transistors - the BCV61C an d BC62C pairs from NXP are at the cheaper end of the range,and would need e mitter resistors, perhaps with a pair of trimming potentiometers to get clo sely matched source and sink currents - could do the job fairly cheaply and with relatively few components.

Linear Technology's

LT3092 would be simpler, but it cost about $2.50 in volume and you'd need t wo.

Element 14 in Australia has some hundreds in stock, so it is easily availab le.

I'd spend the dough on a resistor array and a dual chopamp, then. (The OPA2

Cheers

Phil

Using a single phosphide LED instead of the two diodes makes a more stable current source that's a lot quieter than a zener, and the dropout is only a volt.

Cheers

Phil Hobbs

Thanks. Maybe I'm asking too much, but would it also be possible to make such a setup adjustable as well over a few mA range? From something like a DAC? If you have a chance maybe you could show me the topology you mean. I am not very familiar with "chopamps" unfortunately.

If you can float the load or equivalent, namely put a sense resistor in the low side, it's really easy.

"Sink and source" implies not a voltage reference, but a bipolar analog signal that programs the current.

The problem is still poorly defined.

I thought you'd never ask!

This is cute, although I've never used it. A mosfet-bipolar cascode combines precision with low output capacitance, but has base current error. This (I think) corrects for that.

This has probably already been done before, but I'm experimenting with using a bridge rectifier as a temperature-stable variable resistance. So you have a current source sourcing current in to the top of the bridge (where ground would usually be in a power supply) and a tracking sink pulling it out the other side. Then a signal with some DC bias is applied to the other terminals and the bridge presents some impedance that changes nonlinearly with applied current.

The impedance of the bridge looking into said terminals will vary with temperature as a function of the diode equation. I need to go through the math but if I'm not wrong (probably am!) the bridge structure causes the diode saturation current dependence of the impedance to drop out, leaving only the thermal voltage component. Which is less of a problem.

But to experiment I guess I need a variable current source and sink for the circuit that are as stable as possible themselves.

IIRC Fred Bartoli posted a circuit that does that, by adding a couple of resistors to the single op-amp circuit. Probably not as fast, on account of the RC time constants.

Cheers

Phil Hobbs

Fairly often, I want to make a really linear, accurate fast ramp, like

A bootstrap current source is also good in that case. You can (tediously) add + and - v^2 terms to fine tune the slopes, to partly correct for nonlinear capacitances here and there.

R28 and R36 can bend the curve a little either way.