High Speed, High Current Op-amp

Sounds like you need a switched driver rather than an op-amp to generate fast pulses at high current. Try searching for laser diode drivers.

Rise/fall in 10nS = slew rate 150A/uS and certainly no less than 200V/uS - I was about to say I didn't think you would find one when I remembered Apex:

May be out of your price range but their  PA09 does 200v/us, 150 MHz bandiwdth - no pulse diagrams on the data sheet.

They do lots of other parts so take a look: www.apexanalog.com

How well do you need to control the current and pulse shape ?

MK

I'm checking it out now.

Any ringing in the first 10 ns or so would be ok. A slight linear droop on the flat top would also be acceptable. Apex seems really expensive. I have been thinking about paralleling op amp outputs to reach 1.5 amps, have any experience with that?

Aw, c'mon, just hang a transistor outside the op-amp, maybe a nice, big Darlington.  They have the shut-off resistors built-in.  It is not like you need anything tricky, like push-pull.  National Semi has an old app note along the lines of 'CMOS Linear applications' where they show how to use UB-type silicon gate cmos in the feedback path of op-amps.  Useful technique.

I need the negative rail, so it would require push-pull.

As mentioned Apex PA107 and maybe PA119 might be adequate, but they will be pricey.

This'll be quite hard - up to 1.5A in 10ns and nice and clean - means short wires between load and power device, very low impedance supplies.

Check out this (Linear technology APP Note 133)

www.linear.com/docs/40550

Jim Williams is only aiming for a much slower rise time and single quadrant drive but at higher current - note well the section "Layout Effects".

About the simplest way to get close (which oddly I've just done for someone) is to uses a MOSFET with proper MOSFET driver and control the current by having some series resistance and an adjustable;e power supply. My circuit (in the picture) has 1.175R series resistance, 5 low ESR 22uF caps in parallel, and 10R in series with the MSOFET gate - rise time to 2A is about 35nS. I've got another one to build and test so if I get time I'll try without the 10R and take some scope pictures. It would probably work better with the MOSFET closer to the LEDs. I'd like to get the max current up to 5A.

This board is at the end of 3M of cable (single twisted pair with two additional power cores) and uses LVDS TX and RX chips to get the control pulses up the wire. The power regulator at the source end of the cable can do about 4-14V. The load is 4 red LEDs in series (it's a strobe light but I can't tell you what its for.)

If I needed to get current out of the load fast I'd add another MOSFET (P) to short it the load out (and worry about how to stop them both being on at once !).

MK

If you need push-pull, sometimes you can get away with a cascode output stage.  This would be using a transistor for a pull-down and a CCS for the pull-up.  This topology gives us push-pull with only one control terminal, avoiding a 'dead zone' between the times the pull-up and the pull-down activate.

Can you provide more detail about the nature of your load?

This is interesting because the first approach I took was to use a FET driver, but without the FET. I let the FET driver simply drive the load. I used a large wattage resistor to control the load current. But I've noticed the pulse, at least to my load, contains a linear decrease in amplitude from start to finish. It's significant too, like maybe 100 mV out of a 700 mV pulse peak. However, I have since discovered that negative bias on the output pulse waveform is also needed. This complicates the issue. I cannot disclose the nature of the load, but negative bias is necessary.

To keep the circuitry simplified I may have to relax the load current specification.