fast Sallen-Key

I would think the output Z would be an issue for a LPF.

The usual problem with active filters at high frequencies is feed-through limiting out-band rejection. It may even climb back up to virtually no rejection. The Sallen-Key configuration particularly sucks the big lemon ;-)

If Larkin knew his math, an LC design is trivial using a separating amplifier :-) ...Jim Thompson

I'm just pulse shaping, so stopband attenuation isn't a big deal. An approximate Gaussian response would work. And the first stage, the passive single pole, rolls off the HF stuff, and eases off the opamp slew rate requirement, before I hit the 2nd order S-K section. I do need the passive RC first, for other reasons.

I may just do it with an RC, opamp buffer, and an R-L-C 2nd order section. It's not hard to come up with the R-L-C values, even by fiddling, for a filter this simple. The problem is "ill-posed" in that there's no absolute "right" filter response.

I was just curious if anyone had done active filters in the hundreds of MHz.

My signal is radar-like, a big, maybe 600 ps "transmit" pulse followed by small echoes starting 12-14 ns out. I need to stretch the pulses to a few ns, for downstream processing to handle, but have the main-pulse baseline recovered to almost zero before the first small echo appears.

John

Active filters with discrete components become less interesting at high frequencies since the component sizes are small. There is much published on high frequency active filters for chip design, where the intent is to not have external components, even at the expense of greater complexity.

Sallen Key filters don't handle GBW limitations well, leading to Q enhancement. On a chip you would use leap frog, but that is messy in a discrete implementation.

If distortion isn't an issue, I'd go with a buffer versus an op amp in your design. With GBW limitations, the driving point impedance of the op amp becomes an issue. As an example, In a flash converter test setup, it required some god awful expensive hybrid opamp to get decent numbers. The problem was trace back to drive point impedance. The op amp was glitched by the flash converter.

Even an emitter follower looks inductive past the Ftau.

..so make *that* inductance a part of the filter (a "free" part).

I did a design to cut off around 45MHz. I wanted a steep roll-off but PCB parasitics bit me. The actual component values where way of the simulated values. I reverted to a passive elliptic filter when I redesigned the circuit. I still need to try the passive filter though. YMMV.

Yeah, opamp phase shift variations and PCB parasitics may make an active filter more trouble than it's worth. Even simulations would be hard to trust at, say, 200 MHz.

John

IIRC emitter followers look inductive above Ftau/beta

I've done an active all-pass filter at 60 MHz about 25 years ago--it was the usual op amp 0-180 degree phase shifter: +2/-1 amp, single-pole RC lowpass. I used a voltage divider plus diff amp to make it +1/-0.5, with a dual varactor for phase shift adjustment.

But nothing with tight phase and gain requirements.

Cheers

Phil Hobbs

er

se

OK, I said past ftau, you said above ftau. Same idea.

The inductance looking up an emitter or even an op amp is due to the output impedance of the device rising with frequency. However, it is not a reliable value since it is a function of GBW, which varies from part to part. The impedance would look resistive before the corner frequency, then look inductive past the corner frequency.

ut

ole

lter

use

hat

wed

to

lse

rs.

Mostly the same idea. The difference is in the 1/beta (i.e. Hfe)

If the echoes are very small you might not be happy with the solution after you have stretched the pulse to a few nsec. Maybe send a diode into reverse recovery on purpose? That should be a bit snappier.

In case the pulse source is some fast FPGA you could possibly use a delay line trick to make a longer but snappy pulse.