cfmd at swipnet.se
Mon Mar 6 21:25:20 CET 2000
From: "urosuros" <urosuros at bits.net>
Subject: Re: vcf5t update
Date: Mon, 6 Mar 2000 03:39:19 +0100
> thank you Magnus I realy got mixed up thinking how it could work with single
> power suply
> > > Also , in old MS50 filter schematic from EFM we have 3046 array in diode
> > > conection. Are you shure Tom that 1N914 will work here ?
> > Notice that the 3046 transistors are wired up to act as diodes, only the
> > base-emitter NP-junction is being operational. The sourcing diodes has
> > replaced by the 100k resistors.
> I know that but I was considering diode maching ( btw are trans in3046
> mached )
> since diode transconductace will depend on temperature ( as in expo
> convertor )
> also should diodes Vt should be mached ?
Ah. I guess this is just a cheapification. It will work, but obviously less
> > > And two question not regarding this filter .
> > > First if transistors in ladder filter are not matched that would decrese
> > > rejection but what else will be affected ?
> > Yes, and it would also decrease the common mode rejection of the input
> > to what ever degree that is usefull.
> > > It seems to me that poles would spread a little ( or a lot ) and that
> > > decrese resonance and overall filter performanca , right ?
> > Yes, the unmodulated poles would not lie on the same place, so the initial
> > part they would be dragged together and only when the feedback would get
> > enougth they would spread out as we are used to. I don't think the really
> > interesting thing is happening in the linear frequency plane (that is pole
> > positions) but rather in what happends in the non-linear range with
> > points of the individual transistors.
> wait a second if we interpretate ladder as four integrators conected
> in series transistor missmach would be equal as integrators having different
> RC and thus their poles would be spread in freq plane ( i was not thinkig
> of s-plane spreading which happens with feedback increase )
Right, but your initial state is that the poles will have diffrent placement
on the real axis before we start to apply the feedback.
> so we would not have resonance overthrow ( or peak if you
> prefer that term ) with feed back increase so moog ladder would sound
> like nothing
I haven't seen exactly what happend when you have large variations in pole
position, but I think for a acceptable range you will eventually get feedback.
Instead of an expanding circle you migth get an expanding elipsis, but even if
you have such, you can have either one or two poles to reach the jw-axis.
With a big feedback you should eventually get basically any system to
> bottom line : if I want to hand mach transistors for ladder how
> well mached should transistors be ( and btw do i need to mach
> anithing else except beta ) and do the perfectly mached transistors
> drasticly improve performance
I don't really know. Maybe one should make some nice simulation.
More information about the Synth-diy