[sdiy] Re: Leapfrog

[jhaible at debitel.net]

Hi Don,

> Well... you fooled me, too.  I had not seen the NIC positive
> integrator circuit before.  So when I looked at the first stage and I
> immediately thought you had accidently built a lossy integrator and
> forgotten the input resistor.  
> 
> But sure enough, the first stage is just an optimization, and the
> other 'L' stages all really are NIC positive integrators (with R183,
> R187, etc.).


Yes the first and last stages _are_ lossy, because of the input and
output termination resistors of the LC circuit. Funny enough, this means
omitting a resistor in the noninverting integrator, and adding a
resistor (R195 in my drawing) to the inverting integrator.

As a transmission line ideally will also work without ringing when
it's only terminated at one end, one could experiment with a non-
damped integrator for the first (or last stage).

> Very clever!

Thanks - the NIC integrator comes from a textbook with a lot of
obscure filter topologies (must look up name and autor, don't have
it here). First I thought - too bad, you need more than unity
gain around the active part, so this brings trouble. Then when
I found that passive summing (with a loss of 6dB) in combination with
an opamp gain of 2 will do exactly what I need, and provide two inputs,
and all this with just two different resistor values, this is where
my heart really made a jump. (;->)


> Although I would be concerned about low frequency noise performance
> with so many of these in series.

Yes, that was, and still is, my concern. Noise, and all the possible
artefacts from 50 "servo systems". It's really not worth buying
expensive high performance opamps here, when a real inductor just
costs a few cents (and only one L value is required.)

Nevertheless, it was a nice excercise, because it was all new to 
me.

BTW, other than the NIV integrator, I was following another idea:
Using 2 inverting integrators + 1 inverting amp per 2 state
variables (trivial; so far), but now using a CD4069 for this!
That's one cheap CMOS chip for emulating two inductors!
Noise would even more be a problem, but I can think of applications
where the soft clipping of the 4069 (on internal nodes! - see
my last mail to Magnus.) might be desired. 
I've built a speaker emulator / distortion box with 4069's
some years ago, but calculation of the resonant peaks
was increadibly difficult for the single stage, multiple
feedback BPFs I had built then. (The limited gain of
the CMOS devices doesn't help, either ...)
But a ripple filter built from a handfull of 4069's, with
soft clipping an early stages (filtered by following stages),
using leapfrog technology, might be interesting.

JH.


-------------------------------------------------
debitel.net Webmail