Sallen-Key input Z

Butterworths and Bessels, common in instrumentation apps, have only one stage with Q>1 at 6-8 poles, and that Q is low. And in a measurement system, the filters are usually for alias rejection and noise bw reduction, so a little wiggle in the frequency response curve doesn't matter much. DC accuracy sure does matter.

Something like a higher-order Chebychev starts to have serious Qs. An

I like the transitional Gaussian filters for instrumentation and ARBs.

We just evaluated a 4 GHz Rohde-Schwarz scope and a 6 GHz LeCroy, both

We just decided to get the LeCroy. It's a Windows scope with a gigantic screen and stunningly complex menus, but it does very clever eye diagrams without an external trigger.

I don't know why the don't just deconvolve out the ringing.

Oh yeah, forgot, I had one of them, too. The wife, not the '45.

Adjusting the gains isn't really much of an issue for me. The gain peaking is only a couple of dB in filters that I'd be likely to build, and there's no reason to suppose that the signal spectrum will be at all close to flat, so keeping the filters from clipping becomes more of a system-level problem. Optimizing the filter gains separately isn't a complete solution to the problem anyway.

BTW I looked up Sevastopoulos's US patent 5426393, filed in 1993. (It expires in six months.) He's driving a general filter using the noninverting input of an integrator, and taking the integrator's feedback from the filter output. That'll work fine as long as the filter's phase shift is well behaved at the integrator's corner frequency--in fact it's a nice generalization of the unity-gain Sallen-Key, if not a particularly earth-shaking one.

So yes, with an additional op amp, a resistor, and a capacitor, it's possible to get good DC accuracy. Of course, that integrator resistor is a sneak path for the input signal to leak into the output, and so will in general degrade the stopband attenuation. So in real life you'd have to split it into probably 3 sections, and add two bypass caps.

Cheers

Phil Hobbs

My 3 GHz TDS694C has BNCs as well, and beautiful smooth 125 ps edges. Not as much fancy software, of course, but for an analogue guy, it's pretty suitable.

Cheers

Phil Hobbs

Gasp! ;)

Well, it's actually a division of Teledyne now, and Walter is retired, so it's time to let the grudge go.

--

John Larkin Highland Technology Inc

Precision electronic instrumentation Picosecond-resolution Digital Delay and Pulse generators Custom timing and laser controllers Photonics and fiberoptic TTL data links VME analog, thermocouple, LVDT, synchro, tachometer Multichannel arbitrary waveform generators

The LeCroy will do PCI Express eye diagrams without a trigger. It apparently takes a long record and accordion-folds it on top of itself to make the eye. The clever part is that it rubber-bands the timebase, instead of just folding at some fixed period, to simulate a clock-recovery PLL. That must be interesting software.

We're doing some PCIe-over-cable products and need to optimize the equalization of cables from 0.5 to 7 meters long. That's not easy to do. The 100 MHz PCIe reference clock is a "hint clock" in that it only suggests to the receiver PLLs what the 8b10b bit rate might be; they then lock to the actual data streams. So the refclock isn't a usable eye diagram trigger, and we can't get at the actual PLLs buried inside the chips on either end.

We're kluging passive pickoffs on the PCIe differential signals, which are approximately fast LVDS. LeCroy wants $7000 for a differential probe.

Their oscilloscope is sure ugly. It looks like Darth Vader's bench scope.

I still have mine.

SK

Cute idea. You could use LTspice to recover the clock. ;)

Good probes cost the earth. You should build a universal one that can be moved from scope to scope, without having to use the proprietary interface. pHEMTs have low enough input capacitance that they're pretty easy to protect--a series cap of 0.25 pF running into a bootstrapped pHEMT follower will get you a nice 5:1 probe with very low noise that's immune to practically anything.

Cheers

Phil Hobbs

SK

We could export a long record to MatLab or Octave and do it ourselves, but that would sure slow down iterations of tweaking the equalizations.

I might go into the scope probe business. If we only "sell" a few to ourselves, it would be worth it.

It even don't necessarily need an additional opamp.

For the triple SK filter John posted a few days ago, I was about to post that he could use the first SK opamp exactly to that purpose, thus having only one opamp offset in the chain. BTW, it, being part of the first filter, might go around that so obvious patent.

Hey, Fred, what topology did you have in mind?

This is one possibility.

. Your SK

u are

rument

-K,

we're

, that

s, not

der

ing

nt

ne,

of dB

for

Qs,

But

e of

ve

s.

the

to

o is

ally.

y out in

ry

th

ty

I'd buy one - if I could afford to ... This kind of thing is exactly what this group ought to be good for (and rarely is). The regular recommendations for Williams and Taylor (I've got the second edition - ISBN 0-07-070434-1 - with the ISBN for the third edition 0-07-070430-9 penciled in next to the original number) is another.

It ups the pole order by 1... fine, but (in the simple case of an S-K low-pass following) adds a REAL zero, so you have to know what you are doing on your pole placements. ...Jim Thompson

nt

ed,

are 5

s.

l on

ey

A wife isn't a possession. My wife and I established a close relationship back in 1970, but she had a Ph.D. to get so we didn't finally move in together or get married until 1979.

Running the equations, I'm not sure it's worth the powder to blow it to hell, other than it's a "cutesy" way to gain the third pole.

Any following low-pass S-K filter is going to have unity gain anyway.

But in switched-cap usage I can see the benefit. ...Jim Thompson

My favorite is when an 0402 part disappears from the tweezers, and I'm left wondering which one of several open bins of other 0402 resistors it sailed into.

I don't care if I drink a few 0402s with my coffee -- hey, it's RoHS, right? :-P -- but it's annoying when I have to test every resistor with a DMM before I install it.

-- john

In principle you can wrap it around one of your more favoured topologies, e.g. a leapfrog, and get the DC gain and offset accuracy of a S-K. It puts in a closely-spaced pole-zero pair at some low frequency, which will give rise to some settling whoopdedoos at late times. A more serious drawback is that resistor that bypasses the fancy filter stuff--it'll blow the stopband rejection out of the water unless it's split and bypassed N times.

Cheers

Phil Hobbs

I did some root-solving (by hand)... solved it... discarded it >:-} ...Jim Thompson