simple current source

[Simon]

OK I've knocked up a simple current source diagram (below) am I far off ? feel free to throw the rotten eggs at me, I'm off to the pub for the evening, I'll clear up the mess in the morning 

[migsantiago]

I have used a LM334 for a 4mA constant current source.

http://www.national.com/ds/LM/LM134.pdf

It has a temperature independent circuit (figure 3).

[Zero999]

How much current?

The LM317 is simpler and more stable than the LM334 which varies with the temperature.

[jimmc]

Not a good circuit, the voltage on the -ve input of the opamp is dependent on the V_GSon of the FET and so the current is ill defined.
Also the output stage of the TL084A is not designed the +ve supply so there is a danger that the FET may remain too hard on.

Since the gate current is effectively zero, source current = drain current, so it would be easier to use a N channel FET with the load from drain to supply and the current sense resistor from the source to ground.
As for as the feedback loop is concerned the FET is now a unity gain voltage follower, so loop gain is independent of the load (unlike your circuit). There is a chance that the TL084 may not like the capacitive load of a large FET, so a stopper resistor between its output and the FET gate would be a wise precaution (say 100R). You may also need a resistor from the TL084 output to ground to ensure the output can go low enough to control the FET.
Of course now -ve input of opamp is connected to current sense resistor  / source.
Current is simply control voltage on +ve input of opamp / current sense resistor.

As others have said, if you want a fixed current source of between 10mA & 1A and can afford a voltage drop of 4.2V minimum, the LM317 is a possibility.
(See page 18 of the Nat. Semi. data sheet).

Jim

[Simon]

I was hoping to keep the voltage drop to a minimum

[desolatordan]

Attach R1 to Vcc instead of the opamp output.

Here's some simple constant current op amp circuits.
http://www.maxim-ic.com/app-notes/index.mvp/id/3869

[Simon]

I found this: http://ww1.microchip.com/downloads/en/AppNotes/00874C.pdf which was quite interesting and ultimately what I would like to do(not neccesarily with a uC), but I can't quite understand how a 20 MHz pic which I think is running at 4 MHz can drive a 50 KHz PWM signal, it would mean just 20 "levels" are possible but maybe with the speed of operation the averages would make up the inbetween values, I think lower values of sensing resistor can be used if an op amp is used to amplify the voltage, I've done this in other projects and it worked well and keeps losses down