Hi,
I have used the LT1010 buffer in a design to be able to drive a 50ohm load with a +-5V signal. The supply used is +-12V.
The schematic of the implementation is very similar to below (the simulation schematic is just missing the decoupling capacitors). 
The problem comes when I try to drive the load with a signal that has relatively fast edges swinging from +5V to -5V. On the rising edge, it works as expected, however on the falling edge at some point the LT1010 buffer enters some kind of limitation, like it's stuck at ~0V for 1us and then at -1.5V for a longer period of time until it eventually recovers. This is very consistent behaviour and goes away if the 50ohm load is removed.
I have noticed that when I remove the bias resistor, the period of the second 'limit' section increases significantly. However, if I reduce the bias resistor to 20ohms it does not improve.
Any ideas what could be the culprit here? I was thinking that I can just replace C1 with R5, and leave R5 open, therefore the buffer is not in the feedback path anymore, but I'm thinking this would also affect the quality of my output.
In the oscillogram below, purple = input, blue = LT1010 input, yellow = LT1010 output. 
You can see the op-amp undershooting because LT1010 limits the output and just saturates at minimum output until LT1010 recovers.
I don't think the effect is thermal because the device sits at 50degC without Rbias or at 70degC with Rbias.
Any feedback is appreciated, thanks!
