Matchin transistors

[Joachim Verghese]

On Thu, 12 Sep 1996 johns at oei.com wrote:

>      In preparing to build an ASM-1, I am considering searching for 
>      matching transistors.  I assume this means trying to find a pair of 
>      transistors that show a close match of a V-be curve for an identical 
>      I-b current ramp.  (Is this correct?)

The problem with using a transistor (or diode) as an exponentiating
device is that the transfer function is heavily dependent upon
changes in temperature.

In a typical linear voltage-to-exponential-current converter, the
first transistor is connected as a voltage follower, and the second
as the exponentiator. The voltage follower adds a temperature dependent
offset to the control voltage. This compensates for the 1st order
thermal effects of the exponantiating transistor.

In other words, if the two transistors have identical thermal
responses, then you've eliminated a major part of the converter's
thermal drift.

Ideally, then, the two transistors should be matched for *thermal
tracking*. Using a dual transistor (MAT-0x, SSM22x0) obviously
is an advantage since the two transistors are on the same silicon
substrate. If you choose to use two discrete devices (2N390x...),
there's no easy way to guarantee thermal tracking, but in most
cases you'll achieve good results by just using randomly selected
devices, as long as you provide adequate thermal coupling. Some
sort of heatsink or metal clip should be used to hold the transistors
together.

As far as Vbe-Ic (or Ib) matching goes, this is generally not needed
for exponentiator applications. Any mismatch in these parameters only
affect the tuning offset of the oscillator/filter.

In systems with large numbers of oscillators/filters, it might be
desireable to Vbe-Ic match the transistors, since this allows you
to have less coarse tuning trimmers, which makes calibration easier.
(Remember that a change of 18 mV at the base-emitter changes the
frequency by one whole octave.)

-joachim