[sdiy] Why do we need the buffers? - gm-C filter question

[JH.]

* Some of the "later" synthesizer chips did this.

* I tried it in simulation, and found much less attenuation of high
frequencies
   in the range way beyond the cuttof frequency in LP filters.
   I think it's pararasitic Cs, and in the resistor dividers prevent HF
coupling
   thru all stages.

* Maybe (maybe!) that's one reason why "first generation" SSM and CEM
   sound better than later ones.

JH.



----- Original Message ----- 
From: "Aaron Lanterman" <lanterma at ece.gatech.edu>
To: <synth-diy at dropmix.xs4all.nl>
Sent: Sunday, February 19, 2006 1:36 AM
Subject: [sdiy] Why do we need the buffers? - gm-C filter question


>
> If you see a four-pole LM13700 type filter, or CA3080 type filter, you
> often see a series four of stages like:
>
>   +----------------------------+
>   |                            |
>   |  |\                        |
>   +--|-\                       |
>      |  \                      |
>      |OTA-----+------[buffer]--+--vout
> vi--|+ /     |
>      | /      |
>      |/    -------
>              cap
>            -------
>               |
>               |
>              gnd
>
> When I look in IC design books, they'll often write what looks to me to be
> an equivalent circuit (assuming you're driving something with a high
> impendence input, say another OTA...)
>
>   +-----------+
>   |           |
>   |  |\       |
>   +--|-\      |
>      |  \     |
>      |OTA-----+-- vout
> vi--|+ /     |
>      | /      |
>      |/    -------
>              cap
>            -------
>               |
>               |
>              gnd
>
> Of course, there's the issue of having to put a resistive divider in front
> of the whole thing to get the input range low; in the first diagram above,
> you wind up with a similar divider after each buffer.
>
> Why not just put one resistive divider at the beginning, cascade four
> stages like what's shown in #3, and then put a buffer on the final output?
> I see stuff like that in integrated circuit design books, where everything
> is just OTAs and caps...
>
> Conjectures:
>
> 1) Maybe the OTAs we typically use, 3080, 13700, discrete 4-BJT, whatever,
> don't have high enough input impedance to avoid having the next stage load
> down the previous one? (One thing that makes me think this is the IC books
> are all CMOS, and a CMOS-based OTA would probably have higher input
> impedance, so maybe they could get away with it)
>
> 2) To keep the input voltage on the next stage low enough to avoid
> distortion, you'd have to keep the current coming out of the previous
> stage pretty small; maybe this gets into noise issues (although constantly
> running through all those resistive dividers and op-amp stages can't be
> good), and also, it may severely limit the range of useful currents at the
> OTA input pin.
>
> Am I right (or partially right and partially wrong) on (1) and/or (2)?
> Other reasons?
>
> - Aaron
>
> --------------------------------------------------------------------------
---
>
> Dr. Aaron Lanterman, Asst. Prof.       Voice:  404-385-2548
> School of Electrical and Comp. Eng.    Fax:    404-894-8363
> Georgia Institute of Technology        E-mail: lanterma at ece.gatech.edu
> Mail Code 0250                         Web:
users.ece.gatech.edu/~lanterma
> Atlanta, GA 30332                      Office: GCATT 334B
>
>