[sdiy] Filter Structure Question

[Magnus Danielson]

From: Scott Gravenhorst <music.maker at gte.net>
Subject: Re: [sdiy] Filter Structure Question
Date: Fri, 16 Jul 2004 11:21:39 -0700
Message-ID: <200407161821.i6GILdZ08214 at linux6.lan>

> Magnus Danielson <cfmd at bredband.net> wrote:
> >From: Scott Gravenhorst <music.maker at gte.net>
> >Subject: Re: [sdiy] Filter Structure Question
> >Date: Fri, 16 Jul 2004 10:02:22 -0700
> >Message-ID: <200407161702.i6GH2MZ06562 at linux6.lan>
> >
> >> Magnus Danielson <cfmd at bredband.net> wrote:
> [snip]
> >What you could do is to use one of the 4069UB inverters as voltage-reference 
> >and regulate the powersupply so that it will sitt at ground. The evil way of
> >doing this (if I may be even more evil and subversive and teach you some really
> >evil tricks) is to let one inverter of a 4069UB have both its input and output
> >tied to ground and then have both supply feeds through resistors (which is good
> >for current limiting/over-current protection) and with a cap over the local
> >voltage. This will DRIVE the supply (with feedback) so that the virtual grounds
> >of the other inverters is at close proximity to ground. This is both evil and
> >minimalistic.
> 
> Ok, I think I see this.  We use a single supply, say 5 to 8 volts (I want to
> be able to overdrive the CMOS amps easily).  Then one inverter is used the
> way you described to create a new "ground" line that will be about 2.5 volts
> away from what will then be the rails (+/- 2.5Volts).  ??

Yes, you can do this, but you run into trouble with CV which is not adapted,
also recall that when you run multiple 4069UBs they don't have the same bias
voltage, where as my proposal is done per chip, and then one inverter per 
4069UB will autoadjust the voltagelines so that that 4069UB is correctly 
biased.

> The CV is applied across the MOSFET gates and this new created ground ??

This is indeed your problem when you do it that way. The way I proposed you 
have a common ground for everything, but individualized power.

> So this way, I can use standard 4069UB integrators (not the hipass thing I
> proposed) in a state variable configuration.  

Yeap. Actually the way you proposed was close enought an integrator anyway.

> I think I understand the resistors in supplying power to the 4069UB, but I'm
> not sure why they're needed.  The ground creating inverter will seek a point
> about 1/2 way between the rails, but that inverter will draw only 11 ma
> without the resistors.  Maybe that's too much?  With the resistors, will
> there be power supply coupling between the amplifiers?  How big would these
> resistors have to be (I don't see a total power dissipation spec on the
> datasheet)?  Guess: 10K ?

I imagined a dual PSU situation, with a resistor up to the V+ and a resistor
down to V- of something like around 100 Ohm (I haven't really bothered to
calculate this). Now, when you use one inverter to selfbias it will pull 
additional current on either the V+ or the V- side such that the selfbias point
for the actual chip-voltage approximate the ground being common to the other
circuit. So, this way the individual ground situation has been replaced with
an individualized power-supply for each chip. Also, since this is done per
chip it is also matched for that chip but not necessarilly for other chips.
Beware that the selfbias-point between the inverters of a 4069UB chip is much
better matched than with that of the next 4069UB chip. 

> >> So I've mentally played with the idea of using H11F1s for the voltage
> >> controlled resistors.  This solves the virtual ground problem, but I really
> >> wanted this to be a CMOS filter.  The H11F1 method is close and I may try it
> >> anyway.
> >
> >You could be using another 4069UB, but you know that, right? ;O)
> 
> I suppose so, but I have this love affair with the 4007UB as well, it's just
> a MOSFET array to me and the accessibility of extra electrodes is attractive.

Well, using 4007UB is cheating compared to using _only_ 4069UB, which brings
more of a challenge to the field! ;O)

> >You need to linearize the MOSFET. A pair of resistors will do the trick.
> >That is, if you don't want to be cheapertronics-evil, in which case you just go
> >forward blindfolded and it will be OK.
> 
> Cheapertronics, MML, etc., is all fine, I _want_ this thing to distort, but I
> want the CV to be reasonably useful.  Doesn't have to be exactly linear, just
> not wildly nonlinear. If the 2 resistor gate-drain thing will work, then fine.

I think this is more or less on the money, you want something which is a bit
crude, but behaves sufficiently close to what you can handle and possibly know
that you can't get something out of the crudeness while still not being TOO
hard to handle. Destoying a signal is childsplay, doing it in an interesting
enought maner which allows control and variation is another thing.

> >> And then arrange them in a state variable configuration.  This takes care of
> >> the "source needs to be connected to ground" issue.  I don't have a clue at
> >> this point if 1) it will work and 2) what R should be.  High, I think, like 1
> >> meg or more.  Linearity fixing resistors should be applied to the MOSFET
> >> perhaps or an opamp with a MOSFET in the feedback loop to fix gross CV
> >> nonlinearity.
> >
> >OK. Well, I will think some more about it. We certainly would like a 4069UB
> >clean device and not the WASP hybrid.
> 
> That's what I'm after.  I like the idea of being able to overdrive the 4069UB
> inverters by using a large enough input.  This is said to be one of the
> charming things about the WASP (I know this only from reading, not
> experience).  I just don't want to put bipolar OTAs in this.

Right. I haven't fooled around with the WASP either, so I haven't the foggiest.
I just like the idea of doing something with something which wasn't designed
for it.

Getting an 4069UB into various linear-like things is taking some evil digital
thing (a hex inverter) and having it do nice analogue work. ;O)

Cheers,
Magnus - who just saw a little documentary about Mary Shelley