Matching transistors

[Joachim Verghese]

On Sun, 10 Nov 1996, Sean Costello (aka David Hopper?) wrote:

> The 4075 seems like an awful lot of work to make properly, if you wanted
> to make one from scratch.  There are something like five matched
> transistor pairs, each wrapped in copper - and matched to a high degree
> of precision to boot.

I actually measured Vbe of all the five PNP pairs of four 4075 units
(40 transistors, ouch!), and I was shocked by the degree of matching the
devices had been selected for.

The older 4075s with PNPs marked "ITT21172" especially had well matched
pairs. In fact, so good that my Fluke 75 wasn't accurate enough to
show any mismatch at all, so I had to get a better DVM. The newer 4075s
with 2N5087s showed a worst-case mismatch of around 0.3 mV.

I guess you could use less precise transistor pairs, but CV rejection
performance would suffer. And if you want to make a 4075 from scratch,
then I'd recommend using Elantec EP2015A PNP arrays.

> Are the ARP filters really better sounding than the Moog filters?

Yes and no. I think that if you want pure sound fidelity in the
traditional sense (i.e. low noise and distortion, high CV rejection) then 
ARP's filters (the 4075 especially) are hard to beat. They're right up
there with the CEMs and the cleaner SSMs.

I think most people would agree, though, that a distorted Moog filter
adds more "character" to the sound than most any of the above filters.
And that's why people use analog in the first place - to counter-balance
the squeaky clean digital mass-production sounds.

> I mean, you don't need to have too much precision in the filter for an
> Omni, for pete's sake - it doesn't even track the keyboard.

Very true, but the 4075 was used in practically all ARP products from
1977 onwards.

-joachim