[sdiy] moog high pass flter - Transistor Matching

Ian Fritz ijfritz at earthlink.net
Wed Dec 11 04:45:40 CET 2002


Hi Scott --

At 07:54 PM 12/10/2002, Scott Bernardi wrote:

>OK, I see your point, but I still just try to match within half a millivolt or
>so. My 3 1/2 digit DVM works for that.

I agree that half a mV is probably plenty good.  In fact (as we've 
discussed here before) if mismatch comes just from geometric factors then 
matching doesn't even matter -- the output is just displaced by a fixed 
ratio. If it comes from material flaws then it is more likely to matter.

>Probably the best argument for matching as a diff pair is that they will 
>both be
>at the same temperature, even with stray air currents, because you're 
>measuring
>them at the same time.

I agree that you would want to make a final differential check of the 
selected pairs.  I think you could line up a bunch of devices, shield them 
against drafts and measure them quickly, though, and avoid serious 
temperature effects.  Especially if the offset from the reference isn't too 
great.

   Ian




>Ian Fritz wrote:
>
> > Um, well you take one and measure the other n-1 against it.  Then pick the
> > good pairs by comparing the numbers, just as if you had measured them 
> singly.
> >
> > The advantage would be that you are using a differential measurement and
> > therefore get better resolution.  You could in principle use a 3 1/2 digit
> > DVM to determine offsets in the microvolt range.  It would be quite
> > difficult to do this well with the .56 V background you have in a single
> > measurement.
> >
> >    Ian
> >
> > At 08:49 PM 12/9/2002, Scott Bernardi wrote:
> > >Although what you describe would work to compare two Vbe's, I'm not sure
> > >how useful it would be to know just the mismatch between two
> > >transistors.  If you had n transistors to work from, you'd have to
> > >potentially make 2^(n-1) comparisons. Probably not bad if you were only
> > >looking for one matched pair.
> > >If you record the values of the Vbe's, then you make one measurement for
> > >each transistor, and you compare values to see which pairs can go
> > >together.



More information about the Synth-diy mailing list