[sdiy] Re: Leapfrog

[Magnus Danielson]

From: "jhaible" <jhaible at debitel.net>
Subject: Re: [sdiy] Re: Leapfrog
Date: Wed, 7 Jan 2004 09:05:12 +0100
Message-ID: <002001c3d4f5$0f0d1ee0$b576b9d9 at debitel.net>

> > >
> http://www.oldcrows.net/~jhaible/scanner_vibrato/jh_leapfrog_untested.pdf
> >
> > Trouble is... L3 is not being represented by an integrator, that's still a
> > lag filter!
> 
> Really?

Yes! At least if I should consider my quick exercise with pen and paper. I can
redo it with a little more effort put into it. It is also consistent with a
number of other exercises that I did before I saw your schematic (basically a
few trial-and-error attempts which I tend to do just to "feel" the problem).

> I haven't written down the equations for that single stage (just used a
> combined Lag / NIC integrator from the textbook), but the overall frequency
> response is identical (within 0.5dB) to the RLC filter. (Always talking about
> simulation / theory, not about a real circuit.)

Well, I don't doubt that your simulations is pretty much the same, it doesn't
contradict my little theory exercise or vice-versa.

The thing is, for an ideal integrator there is one pole and no zero. The single
pole is sitting at origo in the s-plane, i.e. s = 0. In reality we see a little
leakage, so it is just a tad of to the left, i.e. s = 0.00... something
You would hardly notice in the audio-band anyway (unless you have a high-res
measurement tool, but let's not overdo it).

Naturally we could possibly do with an approximation, or let's say a
compromise, so that we do have a lag-filter, i.e. we have a single pole which
is even in the ideal case way off the origo. However, now comes the discussion
of how much off the mark we can accept, which IMHO is a much more brain-killing
exercise than trying to "do the right thing" if possible (it sometimes take
some time for the problem to mature, so it is easy to make a wrong shortcut
before the problem as such is understood).

Also, making a "sweeped sine" simulation isn't everything. It is actually a
rather poor method for characterizing certain aspects. A "smooth" responce is
not necessarilly a sign of good characteristics, but an "unsmooth" responce is
a good indication that it probably will not be that good. When you look at the
positions of pole and zero locations you can get a better insight, given you
know what you are looking for. The hint here is to keep the way off the "active
area" so this is why I also tend to react here, since it will change the
state-variable design into a state-variable-like thing. The output of this
integrator will not be a real ortogonal state. All of a sudden we might depend
on the scaling and particular design, when we did not expect it. We should be
carefull. Maybe I am a bit oversensitive, but there is a way of thinking which
tends to save me from trouble I see that others get themselfs into, so you may
call it a conservatism...

Cheers,
Magnus