I am working on building a current controller circuit used to drive 8 solenoids (for those interested, the solenoids produce magnetic fields for use in a Zeeman Slower). My problem is that there is persistent current ripple that I cannot seem to remove. Each coil runs a different current ranging from 2 to 8 amps with, ideally, a few percent fluctuation. The circuit consists of 8 channels each of which has a solenoid (1 ohm resistance, 250mH inductance) in series with a mosfet (FDP18N20F) and a sense resistor. The voltage across the sense resistor and a set point voltage are fed into an op-amp (OP275) which drives the mosfet gate to set the desired current. Each channel is powered by a common 13.7V/40A supply (tripplite PR50). I attached a sketch of the circuit.

enter image description here

Each channel works well independently when the others are off. However when they are all running and considerable current is being drawn, the supply voltage begins to ripple resulting in current fluctuations (ripple frenquence is about 300 kHZ). Below is a scope trace showing both the AC coupled signal across the sense resistor (nominal value 1.8V) and the ripple on the 13.7V supply voltage.

enter image description here enter image description here

My confusion is that the supply voltage ripple leads to much larger current fluctuations than expected given the mosfets are biased in saturation. My only explanation is that my control loop is doing more harm than good at removing the disturbance at the ripple frequency. Is this a reasonable explanation? One solution is to limit the feedback at high frequencies by modifying the circuit shown in the following figure where RC is chosen to limit the bandwidth to a few kilohertz.

enter image description here

Any other possible solutions would be greatly appreciated or advice on how to reduce the power supply ripple. Thanks in advance.

  • \$\begingroup\$ How is your bypassing? Do you have 0.1uF bypass caps right at each of the op-amp supply rails to ground? A local bypass on each of the solenoids might be a good idea. Can you put a bypass cap on the + terminal of the op-amp? Seems like if it's just being adjusted statically you could put a relatively large cap there too. \$\endgroup\$ – John D Jun 24 '14 at 23:41
  • \$\begingroup\$ Unrelated to main problem - add a 12V zener reverse biased source-gate on each FET. FET lifetime MUCH enhanced. \$\endgroup\$ – Russell McMahon Jun 25 '14 at 3:33
  • \$\begingroup\$ Bypassing?| Coupling between solenoid fields? | Cap OA-inverting to ground not output (probably) | Reverse biased Schottky mounted as close as possible to FET connected G-S. \$\endgroup\$ – Russell McMahon Jun 25 '14 at 3:36
  • \$\begingroup\$ This may be interesting if not useful (and is quite likely already known to you). - physics.ncsu.edu/jet/theses/pdf/Harris.pdf \$\endgroup\$ – Russell McMahon Jun 25 '14 at 5:52
  • \$\begingroup\$ I expect that it should be "not too hard" [tm] to achieve what you want by means of suggestions made here and ongoing interaction. It certainly looks soluble (and many people are using such equipment). I just realised that I have a friend (postdoc) who did her PhD in related areas and may be a source of practical advice on what they experienced. One way or other it sounds like something interesting enough to help ensure it works :-) \$\endgroup\$ – Russell McMahon Jun 25 '14 at 5:59

Add an additional resistor of about 100 ohms in series with the gate and your circuit is good.

The heavy gate capacitance of your MOSFET is likely causing oscillation. Try 1K and 1nF for the RC with 100 ohms series to the gate.

| improve this answer | |

This is a rolling answer it has possible solutions plus many questions. The questions may be useful in directing what to try and may change in usefulness as you progress.

This is definitely soluble You "just" [tm] need to find out what the main 'driver' is and damp it down. Damping it down MAY in part involve in speeding up response of the FET drivers so that they can track any voltage changes in the main supply that you do not manage to eliminate.

The "ripple" looks more like an oscillatory edge spike.
It's not a certain source, but if one channel rings and modulates the power supply with a current spike which leds to a voltage spike then if this is present on all inductors the opamp has to track it to prevent it causing a current spike in that channel. Decoupling all channels from each other may be useful. Needless to say the opamps should have standard bullet proof decoupling from +12/-12 supply variations.

Are all the channel drivers on the same PCB?
What is the construction method? - breadboard / 2 layer PCB / 4 later wth ground and power planes / ...?

Are all inductors V+ ends commoned without intervening filtering?
and is there decoupling at the psu?.
What are the lead lengths and is there per inductor decoupling and do they share a common connection to supply?

ie here the "circuit diagram" is also really a physical layout diagram potentially including PCB construction, paths across board. wiring connections, lead resistances etc.

(Notionally) A 10A feed for 1uS drops a volt in a 1 uF cap. Placing a say 10 uF ceramic between MOSFET source and inductor positive in each help may provide per inductor decoupling of an order of impedance that is useful. Getting the capacitor to source lead short is probably problematic. If the inductor has a 2 wire connection to pcb then using the V+ return for the V+-source cap is a good start.

In addition to the above caps, placing a resistor in the feed to each inductor to each coil with the cap to source on the coil side should help. 0.1 Ohm will drop 1V at 10A and even higher may be useful. How large is OK depend on headroom available from your supply.

If you temporarily replace your supply with a 12V car battery what effect does it have on the "noise".

| improve this answer | |
  • \$\begingroup\$ To answer your questions: 1) Are all the channel drivers on the same PCB? Yes, they are all on a two layer PCB. 2) Are all inductors V+ ends commoned without intervening filtering and is there decoupling at the psu? Yes, the V+ terminals are directly connected. The psu has 10uf bypass capacitors to ground. \$\endgroup\$ – user2966694 Jun 25 '14 at 19:15

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.