High power voltage controlled bidirectional current source circuit

I have tested various amplifiers and motor drivers and none of them are capable of driving my 20Ohm 100mH load at 0-> +/-2Adc (max 0 -> +/-100Vdc) in function of a control voltage (from a 16 bits DAC) with a closed-loop bandwidth of 5kHz. The bias and noise must be <= 1mA. I can give away 10V on the output for sensing.

Honestly I can't see any particular challenge for manufacturers of such equipment to make a product with these sorts of specs and yet I have found nothing suitable - so I am trying to make my own.

I'm looking for a very simple circuit topology that I can design in details - again, I was not able to find anything online after much research other than an application in the LT1990 datasheet that unfortunately can't go beyond 36V. I have two 4-quadrant +/- 100Vdc 2Adc lab supplies with voltage+current limiters so I thought about a simple opamp controlling the gate voltage of a MOSFET push pull based on the readings of a low-side sense resistor, but maybe there is something better?

Space, cost, dissipation are not an issue - I am solely interested in the high-speed performance.

• What is the $\frac{\textrm{d}I}{\textrm{d} t}$ you want to achieve in the load itself (just to confirm what I think I may be reading)? – jonk Sep 9 '17 at 11:37
• @jonk: the slope of the tangent at the origin of a 5kHz sine of amplitude 1A. – Mister Mystère Sep 9 '17 at 12:21
• So, I get $\frac{\textrm{d}I}{\textrm{d}t}=\omega=2\pi f\approx 31500 \frac{\textrm{A}}{\textrm{s}}$. If I'm not mistaken. Have you applied that to your load to see where that points? (Or maybe just being up for just a moment during my sleep period has blinded me?) – jonk Sep 9 '17 at 12:32
• I have done an analysis that yielded a maximum voltage of about 90V when the loop works as expected, if that is what you are asking. Otherwise what do you mean? – Mister Mystère Sep 9 '17 at 13:34
• Does your analysis include and confirm my result above? – jonk Sep 9 '17 at 13:53