Re: [sdiy] Matching trannies for triangle VCO

[jhaible at debitel.net]

>Hi all. For my triangle VCO
>(http://hem.bredband.net/bersyn/VCO/VCO%20Buchla%20core.gif) I tested some
>different dual PNP transistors last night. These transistors are connected
>in a form of discrete OTA.
>A friend, who work at Infineon, gave me samples of their BC856S dual smd
>transistor. Measuring the Vbe with the Moog test circuit, the matching
>seemed to be very good. The difference was no more than 0,5 mV for any of
>the ten transistor pairs.
>But when I tried them in the oscillator, the result was very disappointing.
>On the other hand, two pairs of hand-selected BC560C gave almost perfect
>results. A BC857BS, which gave more mismatch on the Moog tester, worked
>better in the oscillator. How can this be?
>Maybe Vbe isn't the most important parameter to mach for in this case?
>Maybe transistors with higher current gain will make the circuit less
>sensitive to mismatch?


The most basic requirement for getting identical collector currents is matching of Vbe.

But there is more.

If you have a low Beta, the two base currents (which both go to *one* side of the mirror)
will unbalance the circuit. (high Beta is what you want - in addition to Vbe matching -
not necessarily matched Beta.) This effect can be avoided with a 3-transistor current
mirror. (Normally built from a matched pair and an unselected 3rd transistor.)

And then there is the collector voltage. The Vce of the two transistors in the mirror
is *very* different in your VCO, which will unbalance the circuit as well. This could
be fixed with a cascode configuration. (I don't know how big this effect is, compared
to the Beta issue.)


>Anyway, the selected BC560Cs yielded every bit as good result as the very
>expensive MAT03. I was able to adjust the triangle for perfevt symmetry at
>25 Hz. This means that the symmetry is perfect in the entire audible range.
>And it is OK at LFO frequencies too. I tried with very low frequencies and
>measured 18 seconds for the down ramp and 12 seconds for the up ramp. OK,
>this is not perfect, but anyone who is familiar with the circuit in
>question should be impressed with these figures, I think.

Absolutely. (Assuming HF tracking is fine, as well.)

I'm loosely following various triangle VCO efforts on the list, like Ian's or yours,
because somewhere in my head I still think of an updated Frequency Shifter
some time (now that I'm making PCB layouts again ...), and as the QVCO core
is closely related to a Tri VCO core, there you have the link.
(Background: The QVCO in my FS tracks *fairly* well to 1V/Oct - like
an early Moog VCO - but I want something better if there will be a new
version.)


JH.