4mA current regulator

A constant current diode is just a JFET with the gate tied directly to the source. If you have a good collection of JFET's, you can likely find one where the I(sub)DSS is 4 mA. There is usually quite a spread in this current so picking a particular JFET should be done based on this spread. For example, the I(sub)DSS of a 2N4416 is typically between 5 and 15 mA. Some common JFET's and their corresponding I(sub)DSS ranges are shown in the table below.

Having used many PCB Piezotronics parts before, I can suggest that you buy their charge amplifier if you need the ultimate in performance. If you can wait a while for mail order, Mouser stocks the 1N5311 diode which is spec'ed to be within the range of 3.24 to 3.96 mA.

Dr. Barry L. Ornitz snipped-for-privacy@live.com

JFET Manufacturer I(sub)DSS Minimum I(sub)DSS Maximum MPF-102 On Semi

Note: P-channel or N-channel determines the "direction" of the constant current diode.

Same idea but (a lot) better accuracy...

| VCC | | | | | I(sink) .-. | V | | | | | | '-' |/ +-----+---------| | | |>

| |/ | +---| | |> | | .-. | | | LM385 z | | -or- A '-' REF1112 | | | | === === GND GND

(created by AACircuit v1.28.6 beta 04/19/05

Simple, cheap, source or sink, very nice.

Except with a TC... proportional to absolute °K... still usable in a lot of non-critical applications. ...Jim Thompson

and this,

Yep, in the old days we relied on a LM317 for this task, even tho NSC's spec said the "minimum load current" was 5mA. Most parts were lower, 3.5mA typ, but it was clearly "out of spec" for 4mA uses such as yours.

These days, distributors would do well stock Linear Technology's new LT3092 to serve the voracious current-source community. It has a 0.5mA minimum-load-current spec, and it has an intrinsic 1% accuracy spec (not counting resistor accuracy).

For 0.5mA to 200mA current sources, this could be your baby.

But, oops! if high currents are your requirement, it doesn't come in a TO-220 power case. For that, grab their LT3080ET, it's ready for a heat-sink clip! But it's possible it may need a parallel output capacitor to prevent RF oscillation. It has the same 0.5mA minimum-load-current, and 1% accuracy.

dist: DigiKey

case: TO-220 SOT-223 LT3092EST -- $3.73 LT3080ET $4.38 $4.20

"Jim Thompson" wrote in message news: snipped-for-privacy@4ax.com...

That can be tuned out with an additional diode and resistor.

I've read a green LED (old style ~2V) balances out the transistor's tempco?

Grant.

k)

Nice. I (think) this might be one less part and even one step better (accuracy, drift), but it's 3:20 a.m. so it could be broken and stupid ....

| VCC | | | | | I(sink) .-. | V | | | R1 | | | '-' |/ +---------| Q1 | |>. .-----' | ^ | TLV431 / \

Yep, that's why it's there. Strangely, many LEDs of all colors sport

-2mV/=BAC tempcos, neatly offsetting a BJT's Vbe drift.

-- Cheers, James Arthur

Using a 1.2V bandgap gives you a sink current that's nominally proportional to absolute temperature (PTAT). That's often useful, e.g. in stabilizing the transconductance of undegenerated differential pairs--but mostly inside ICs.

Using a LED is usually superior, since (a) you get first order temperature compensation (i.e. not PTAT), and (b) it's much, much quieter than a bandgap.

Cheers

Phil Hobbs

If it's less you can always stick two or more in parallel.

Since the absolute DC accuracy of constant current diodes (basically a jfet at IDss) is rather poor, but the effective impedance is really high i recommend a Wilson current mirror (positive side) circuit instead; much easier to get parts.

See...

I don't see anything about the OP's circuit that would require absolute accuracy. ...Jim Thompson

Hi Depending on the dynamic range of the voltage, you could put a resistor from the output lead to ground to supply a minimum current. If the voltage range was high enough, you could even use a second LM317 as the minimum current. Sink enough extra current in the second regulator to keep the first one in range. My reading of the spec is that