Re: Transistors

Jim Thompson wrote:
Winfield wrote:
Jim Thompson wrote:
John Larkin wrote:
Jan Panteltje wrote:
John Larkin wrote...

Yup. THS3062, practically a unique opamp, blinding fast with +-15
supplies. But if it's amplifying a sine wave at, say, 20 volts p-p
out, and the frequency goes up to roughly 12 MHz, it crashes, pulls
tons of power, gets red hot, and phase inverts! If you drop the input
signal way down, it recovers!

Well, John!!!! the data*** specifies input common mode range as
+-13.9V MAX. Page 3 of ths3062.pdf. It is a video amp :-)

We're running these at gains in the 3 to 5 range, and it does this
working inverting or non-inverting, loaded and unloaded, so it's not a
common-mode issue. And at 12 MHz, we're not even slewing a volt per
nanosecond yet.

It crashes and phase inverts in either config! It made a nasty blister
on me poor finger, it did! But the effect doesn't seem to be thermal,
in that freezing it hard doesn't affect the frequency trip threshold
much.

I note the 2nd Harmonic Distortion rises rapidly around 10-20MHz.
Probably an inadequately stabilized multi-loop that is amplitude
dependent.

I think it's likely another matter, one I've struggled
with in high-frequency high-voltage-amplitude push-pull
amplifiers, namely getting the pullup devices completely
off before the pulldown devices turn on. And visa-versa.

Actually, it's generally not possible, or even desirable,
to completely achieve this goal, but one struggles hard
to avoid having too much rail-rail current under these
extreme conditions, because various local parts become
overheated. I fear too that many high-frequency high-
slew-rate amplifier designs don't pay enough attention
to this issue and may experience failures as a result.

I don't think so, not at 12MHz. I'm intimately familiar (not by
choice :-) where multipli-intertwined feedback loops present problems,
and defy analysis/simulation using existing LoopGain techniques.

Dr. Auggie Ochoa (Zarlink) and I are busily trying to solve this
issue.

I'll stick to my theory at 12MHz. Not that I don't
think something like you're describing can happen
with multiple feedback loops and serious rolloffs
and their phase shifts as well as the pure time
delays you like to use in your modeling, etc., but
I haven't heard evidence for multiple intertwined
loops in TI's opamp.

Back to inadequate transistor shutoff, unless the
amplifier output stage circuit is operating class A
this can be a serious issue. I've struggled with it
at 3MHz and 5MHz for modestly-large discrete parts,
and I know it's an issue at much higher frequencies
for the small transistors in ICs. What's required
is separate active perhaps class-A drivers for both
the pullup and pulldown transistors, to insure they
get at least partially turned off. Sadly one does
not often see such pre-driving circuitry where it's
so badly needed in high-power high-frequency ICs.

John can do a simple test to evaluate my theory:
measure the opamp's rail-rail supply current as a
function of high-voltage excursions at the highest
frequencies and then as a function of current load
under full swings in the same condition. Ideally
the opamp should dissipate power only for current
delivered to the load, i.e., small voltage drops
across the output transistors at the high currents,
without any high rail-rail currents in the process.

Looking at the THS3062 data***, I don't see any
plots for opamp supply current vs frequency for
full-scale output voltage swings. To me this is a
bad sign of a serious issue that's been overlooked.
Check it out, John, and let us know.
.