Extending op-amp phase accuracy using active feedback

Yes...

Integrator Design for High-Frequency Active Filter Applications Randall L.Geiger and Glenn R.Bailey IEEE Transactions on Circuits and Systems, Vol. CAS-29, No.9, September 1982

I have a crappy copy that I can scan for you if you'd like. ...Jim Thompson

Thanks Jim, I went to IEEExplore and downloaded their (kinda crappy, but entirely usable) copy.

I have Geiger's VLSI Design Techniques book. Nearly 1000 pages of excellence, that taught me the basics of IC design. More so than my university did.

OK. I stopped paying IEEE due to no useful functions about 10 years ago. I just got another notice... "Wouldn't you like to renew..." ?:-) ...Jim Thompson

-- | James E.Thompson, CTO | mens | | Analog Innovations, Inc. | et | | Analog/Mixed-Signal ASIC's and Discrete Systems | manus | | Phoenix, Arizona 85048 Skype: Contacts Only | | | Voice:(480)460-2350 Fax: Available upon request | Brass Rat | | E-mail Icon at

When I graduated there weren't yet courses in anything but Ge transistors ;-) ...Jim Thompson

I have a benevolent employer who not only pays IEEE dues but also has (limited) access to their library. :-)

I agree that, on a personal level, it doesn't really seem like that great of a value anymore though. The basic membership dues alone should give you *some* access to the library...

I once asked if they had a plan that would allow me access to all the libraries, but otherwise no benefits. No :-(

It's like sending money to PBS. I used to be a full-tilt Gold Card Member, then they degenerated to leftist crap except during the multitude of "gimme weeks". ...Jim Thompson

And indoor plumbing was just being introduced in town :-)

I bought this book after I leaned out of the window a bit far. Way far. One of those cold-call type proposals that I sometimes write when I see a product that I believe is sub-par and can be improved upon. For half a year, nothing. So I thought this one had fizzled. Then a phone call out of the blue and everything happened at warp speed from then on. "So you are sure this works?" ... "Yeah" ... "Really?" ... "Sure" ... some gut cramping set in ... gotta get some books here, fast. So I studied up on IC design and it all looked feasible. Then a major high-tech company waltzed in and said it cannot be done. This did not halp at all. Lo and behold the CEO was (still is) a gung-ho type of guy so he let me have a crack at it despite the six-digit risk. It worked :-)

Somehow the best projects are those were one has to ride really close to the edge.

Be nice! I helped install indoor plumbing at my Grandparent's Thompson house (Riverton, WV, near Spruce Knob, highest point in the state) around 1950, including adding electric pumps to the well _and_ electric sewage pumping, since the only place to build a shower and add a toilet in an 1850 house was in the basement.

Yep. And the most fun. I ride the edge most of the time, taking on stuff for which I have no prior experience... thus no built-in ideas that it can't be done. Like when I did the IBIS modeling for TAOS/Avago... now I have techniques _no_one_else_ knows how to do ;-) ...Jim Thompson

Could have expanded the outhouse ...

Old Chinese proverb: Man saying it cannot be done shall not interrupt man doing it :-)

So Jim... do you use the improved integrator topologies when you're building active filters?

It just so happens I have a bi-quad design (it's kinda a lead-lag design

-- not a bandpass filter) that would be rather happier with less excess phase shift at higher frequencies.

I rarely do any active filters now, except for SED ;-)

But when I need to, I have no problem at all producing 100MHz+ GBW ;-)

Pass me the specs and I'll try my off-the-wall configuration ;-) ...Jim Thompson

=20

=20

"Rather, these second order compesation techniques reduce the phase error a= t lower frequencies by adding an equal but opposite phase shift in the ampl= ifier feedback loop." That's completely wrong and not at all the basis for = the phase correction. Anyone can see the feedback loop has phase lag and no= t lead. Just goes to show that the engineer can be completely clueless and = still succeed in getting a working product.

lower frequencies by adding an equal but opposite phase shift in the amplifier feedback loop." That's completely wrong and not at all the basis for the phase correction. Anyone can see the feedback loop has phase lag and not lead. Just goes to show that the engineer can be completely clueless and still succeed in getting a working product.

Well, one guy wrote the app note and another one worked through the math. Perhaps neither of them really knew, at an intuitive level, how it works. Heck, I don't really have any better intuitive explanation of how it works than the "well, um, the second amp has the same phase shift as the first, so it kinda gets fed back in a way that cancels it out" either. :-) Could you perhaps explain it in a more correct (but intuitive) fashion? -- I don't really like having only "well, clearly the math shows that it works" as the only truly correct answer.

Thanks,

---Joel

lower frequencies by adding an equal but opposite phase shift in the amplifier feedback loop." That's completely wrong and not at all the basis for the phase correction. Anyone can see the feedback loop has phase lag and not lead. Just goes to show that the engineer can be completely clueless and still succeed in getting a working product.

Where is the resident know-it-all when you need a proper explanation? Arms worn out from too much hand-waving ?:-) ...Jim Thompson

or at lower frequencies by adding an equal but opposite phase shift in the = amplifier feedback loop." That's completely wrong and not at all the basis = for the phase correction. Anyone can see the feedback loop has phase lag an= d not lead. Just goes to show that the engineer can be completely clueless = and still succeed in getting a working product.

=20

=20

Think of it as a regulator. The error amplifier in the main gain amp sees a= feedback phase shift that lags Vin, so it must introduce a phase lead to z= ero the error. So it is the error amp in the main amp that introduces the l= ead and not the feedback path. So the idea is introduce a phase lag in the = feedback to force the main amp to produce the compensating phase lead.

rror at lower frequencies by adding an equal but opposite phase shift in th= e amplifier feedback loop." That's completely wrong and not at all the basi= s for the phase correction. Anyone can see the feedback loop has phase lag = and not lead. Just goes to show that the engineer can be completely clueles= s and still succeed in getting a working product.

of

a feedback phase shift that lags Vin, so it must introduce a phase lead to= zero the error. So it is the error amp in the main amp that introduces the= lead and not the feedback path. So the idea is introduce a phase lag in th= e feedback to force the main amp to produce the compensating phase lead.

Ouch, that makes my brain hurt. The main amp sees a 'twice delayed' feedback signal. The feedback amp has 'exactly the same' delay as the main amp, so compensating for it twice the output from the main amp is 'spot on' phase wise. Cute.

I will observe that Kramers-Kronig has not been violated, diddling the phase comes at the cost of diddling the gain.

George H.

error at lower frequencies by adding an equal but opposite phase shift in = the amplifier feedback loop." That's completely wrong and not at all the ba= sis for the phase correction. Anyone can see the feedback loop has phase la= g and not lead. Just goes to show that the engineer can be completely cluel= ess and still succeed in getting a working product.

how

on of

ift

es a feedback phase shift that lags Vin, so it must introduce a phase lead = to zero the error. So it is the error amp in the main amp that introduces t= he lead and not the feedback path. So the idea is introduce a phase lag in = the feedback to force the main amp to produce the compensating phase lead.

If you're into idealities then in the case of a perfect amplifier with a si= ngle feedback loop function, the output is easily shown to be the mathemati= cal inverse of the feedback function. Using the relation between reciprocal= s of transforms of inverse time functions, you know a simple phase delay at= fixed frequency is exp(-j theta) with corresponding reciprocal/inverse exp= ( + j theta), then superimpose that on the reality of an imperfect OA and y= ou have your result.

feedback phase shift that lags Vin, so it must introduce a phase lead to zero the error. So it is the error amp in the main amp that introduces the lead and not the feedback path. So the idea is introduce a phase lag in the feedback to force the main amp to produce the compensating phase lead.

Gotcha, thanks. I'm thinking there's a decent chance the author knows that as well, but certainly the way it's stated in the paper isn't correct as-written.