Just over 6 weeks after we filed our first service request on the THS3062 problem, TI got back to us. In the interim, we bugged them at least weekly, and also got involved two distributors, the local TI rep, and the ultimate motivator, TI's vice-president of ethics.
The problem is that, when powered from +-12 volts and outputting a+-10 volt sine wave, at about 12 MHz signal frequency the chip crashes: amplitide falls by about 4:1, the chip draws huge supply currents and gets red hot, and output phase reverses. After a lot of flunkies questioned our layout, bypassing, intelligence, and stuff like that, they finally announced that the amp is intended for use in applications where rapid slews are separated by time in which "internal amplifier nodes are allowed to reach equilibrium."
Jim suggested such, based on small wiggles in the closed-loop frequency graphs on the datasheet. So, why is the problem only hinted at by small wiggles? Why isn't this massive defect noted on page 1, in24-point type?
We're re-doing the pcb layout, an 8-layer VME board with a bazillion parts... uP, two fpga's, dacs, all sorts of stuff. Grrrrr.
This is, incidentally, a function generator that was supposed to output +-10 volts behind 50 ohms, up to 32 MHz, sort of what somebody brought up in another thread nearby. We may keep the +-10 volt spec if the replacement amps hold up, but we are adding power supply hooks to use lower-voltage amps and bail down to about +-5 swing if necessary.