Bidirectional current source for high currents

[dino]

Hello!
As the title suggests, I need to redesign a bidirectional current source circuit, with 2.5A peak current into 0.5 ohm load. The previous design was based on a  Howland current pump configuration, with an OPA547 (linear power opamp, 750mA output current). My idea was to put OPA548 (same as OPA547, but with 3A current) instead. The problem lies in the power dissipation, i want to keep it as low as possible, because i will have to put 8 of these circuits in a box.
Another idea was to use a LM3886, but that idea was discarded, due to low efficency at low power outputs.
What would you suggest?

[awallin]

There are a couple of linear TEC drive circuits described here:
http://www.ti.com/general/docs/litabsmultiplefilelist.tsp?literatureNumber=sbea001

my own implementation of this:
http://www.anderswallin.net/2013/08/tec-drive-prototype/

if you need a linear constant-current source then you cannot avoid the U*I power-dissipation in the current-regulating component (op-amp or NPN/PNP transistor in the designs above).
What you can do to minimize power-dissipiation is provide the DC-rails for the op-amp/BJT just barely above what is required to drive your load.
If you have a 0R5 load then at 3 A the voltage across it should be 1.5 V. The transistors used in the app-note have a  low Vce of maybe 300 mV at 3A. so you would need to provide +/- (1.5/2 + 300mV) = +/- 1.05 V supplies to the H-bridge.

Anders

[dino]

Thanks for the answer!
Unfortunately I forgot to mention that my load is grounded, which complicates the subject, elliminating the bridge configuration

OPA548 has a max dropout of about 4V from rails. 4V*2.5A=10W which will have to be dissipated, and that is my main concern here. if I don't find another solution, I guess i'll have to make a push-pull stage like in your schematic.

[fcb]

What are you driving with it?

If indeed it is something like a TEC you could use a MOSFET half-bridge and drive it with PWM, you'd have to make a Volts to PWM converter and the loop-stability would be interesting to develop, but that's not hugely difficult.

If you need something 'analog' and can afford it, why not use the OPA548 and bump your rails.  Your not going to be much more efficient at low levels than the 3886 if you build your own push-pull stage.

Probably better to explain what your driving with the current generators.

[dannyf]

What would you suggest?

A digital power amp. For your spec, using a digital audio amp would work.

[Marco]

Is dynamic cross over distortion important? (ie. does the current have to accurately follow some waveform.) If not I'd just design the current source with opamps and external MOSFETs, with something like OPA548 you just pay a lot of money for it's high voltage capability ... which you don't need because of the low ohm load.

Also I'd make the current source linear and modulate the power supply rails for the MOSFETs with a switcher.

[Marco]

A digital power amp. For your spec, using a digital audio amp would work.

Will those really work well with the feedback loop necessary to turn it into a current source?

[nctnico]

I'm considering something similar by using several LT1210 current feedback opamps in parallel. I haven't tried it yet but the trick to parallelling power opamps is to match the amplification closely, use load sharing resistors and limit the slew rate of the input signal. BTW Apex has some appnotes on parallelling power opamps.

[Marco]

You can just give each it's own shunt and feedback loop ... current sources are trivially paralleled.

[dino]

Is dynamic cross over distortion important? (ie. does the current have to accurately follow some waveform.) If not I'd just design the current source with opamps and external MOSFETs, with something like OPA548 you just pay a lot of money for it's high voltage capability ... which you don't need because of the low ohm load.

This current source will simulate a LEM current sensor. This means that good linearity and low distortion are important. On the other hand, signals are slow, mains frequency.
I think I'll try the opamp+external "output stage" route.

[nctnico]

You can just give each it's own shunt and feedback loop ... current sources are trivially paralleled.

I wouldn't say that too loud. Current sources are not easy to get stable especially if you want a reasonable bandwidth.

[robrenz]

I have the schematics for these if you are interested. These are the equivalent of the Kepko BOP 4 quadrant supplies. It is basically a high power opamp that can do about 20kHz

[dino]

Yes, please! You have a PM 

[robrenz]

My thread on the 2 units I have may contain some info that may be useful if you are going to use those schematics.