Thank you for your answers I will try to reply and give some additional info:
Pockels cell amplifiers have some of the same characteristics you are after.
Also. Is 100V enough? At 1MHz, 20uH has an XL of 125ohms - so you'd need 500V to drive 4A through your minimum inductance. Probably look at a cascode output stage to give you some headroom for the future - also less slew across the output devices.
Numbers given are maximums,so yes the 4-5A is just the maximum current and it won't be able to drive it @1MHz into 20uH.(the sought after output variable is voltage).No probelms with that.
The output is a cascode,miller kills any hopes otherwise.
About the Pockels cells,it is the first time I hear about that effect and I will research the topic,thank you.
That 2p2 capacitor looks strange considering the sheer size of the device. Can't you make the capacitive divider into 2n2 / 22pF or 2n2 / 100 pF (then with 23x gain)?
We have an 8 channel audio amplifier "Sirus Pro MXC8" with a very similar schematic and those amplifiers perform very well.
Regards, Dieter
Yes you are right,the capacitive divider will get bumped so the lower capacitor dominates the op-amp input capacitance(probably 2n2/22p will suffice will have a look in the datasheet)
I expect the output stage will have thermal runaway issues. The idle current in the output stage is a function of the quiescent current of U11 and the temperatures & VBEs of Q13, Q16, Q6 to Q10 and Q1 to Q5. Forward biased diodes in series with R15 & R16 as well as R13 & R14 would help but may require rethinking their values. I might consider temperature compensated current sinks/sources to replace these resistors.
R1 to R10 may also be a bit low to help current sharing among the devices in the output stage.
Q13 & Q16 have to stand off almost the full supply rail to common voltage and thus dissipate extra power for no reason. I would reduce their VCEs to a much lower value. Perhaps use the cascode CB base bias voltage?
In the end, the slew rate of the output stage is established by how much current can be quickly driven into and taken out of the bases of the output devices. The driver stage may need voltages outside of the rails or to accept a lower output voltage swing to speed up the turning off of the output devices. It may be possible to use peaking capacitors, inductors or even T-coils to help.
The idea was the output stage to be biased in a hard class B operation,so no quiescent current as I thought that by having a fast slew rate will make the crossover distortion not that of an issue.Am I right?
Thank you about the cascode CB base bias voltage idea,I was wondering where to return those collectors anyway

I plan on using slightly oversized devices so that deviations in current sharing in the low tens of % will be in spec.(pictured "inner"trannies are at the edge of their spec,so no go).
Will think about including some peaking circuitry to speed up things.
Also I am at a bit of a loss how to evaluate the stability of the circuit considering it's 2 effective feedback loops?Do I just have a look at the global one and if it's stable call it a day?
Your specs look similar to those of MRI gradient amplifiers. Googling for "MRI gradient amplifier" I see many papers, you might find some inspiration there.
Thank you,will have a look in those.
The requirements don't look that unusual. So rather similar to a slightly faster Hifi amplifier.
The only unusual part would be the variable power supply and the idea to get close to the rails. This would normally be things not important for a one off. Who cares about a few 10 W of extra heat in such a case.
4-5 A is not such a high current.
My suggestion would be a more or less normal class AB amplifier, just with slightly fast transistors. No need to go exotic, just the regular better audio ones.
An inductive load is not a problem for voltage driving. The tricky part would be of the coil/ load shows self resonance.
An inductive load may only need a little bit more care with the SOA - here it helps the current is not that high.
Sorry I have been misunderstood the output voltage compliance is of the least concern,so a few volts more or less doesn't matter.
I have started with a standard approach with emitter follower output but the 5*2SC5200 and 5*2SA1943 seem like a pain the ass to drive at that speed and swing(probably should just brute force the problem by throwing current and power at it)
Are there better(read faster) trannies that can withstand 100V while passing 0.8-1A current(pure inductive load case)
My research found nothing faster than the 2SC5200 and the 2SA1943?
My concern is that the output trannies parasitic capacitance and the load inductance resonant frequency can make for a nice high powered oscillator,so probably will just include a isolation(or Q-killing if you look at it that way) resistor.
What output voltage swing are you expecting? <nevermind...I see you want rail-rail>>
BTW, 4-way connections are a bad drafting practice.
My bad on the 4-way connection should have tidied up the schematic before uploading.
Seems the wording about the output voltage swing on my original post was bad so I will edit it.
I will research and revise the circuit and post again,still any ideas and comments are welcome.
P.S If somebody is curious about the physical experiment itself,PM me I will gladly explain.