[sdiy] LM394 vs. MAT02 vs. ?

[Ian Fritz]

One more point:  As has been argued here before, matching may not be all 
that important anyway.  For example, if the mismatch is caused by just a 
difference in device area, then the mismatch just causes an unimportant 
offset, without affecting the temperature compensation.  If the mismatch is 
caused by a materials problem such as dislocations, then the temperature 
compensation could be affected.  This possibly explains why only one pair 
in a hundred or so didn't work in the days Dana was working with ARP.

Better reasons for using expensive pairs is the improved log conformance 
(low Rb) and noise.  The transistors in the CA3083 array have quite low Rb, 
and work very well for NPN converter applications.  For PNPs, the 
high-gain, medium power Zetec ZTX795A or ZTX1151A have good characteristics 
and are readily available.

    Ian



>   I approve of this method, because I recorded the Vbe of about 100 each
>of 2N3904 and 2N3906 a few years ago, and just about all of them were
>within 2mV of each other for the same polarity.  All of them were within
>5mV of each other.  Much easier to just bet on the short odds that a pair
>will work well enough, and if it doesn't..tweak it. ;)
>
>   Yamaha must have pulled this stunt on their NE-series submodules as
>well; they'd have 2SA561s shouldered up against 2SC458s under all that
>epoxy resin.
>
>Crow
>/**/
>
>On Mon, 8 Nov 2004, Dana Scott wrote:
>
> >This is the official ARP method:
> >Reach into a bag of 2n3904's.  Pull one out.
> >Reach into a bag of 2n3906's.  Pull one out.
> >Stick them together.
> >Repeat.