[sdiy] Spice/mathematical model of SSI2164

[Neil Johnson]

Hi,

Liam Wall wrote:
> Previous discussion(1) suggests using the current gain formula from
> the datasheet(2). I would like to have a model that can describe all
> the facets of Fig 11 of the data-sheet --- the Temperature-Compensated
> Exponential Voltage to Current Converter. This is the formula I'm
> using so far
>
> i = i(in) * exp(-qA*v(ctl) / (k*(temper+273.15)))
>
> with qA/k being approx 1160.5. temper is the ngspice variable for the
> device temperature.
>
> But this model cannot explain VCA3 in fig 11, which is compensating
> for the "series emitter resistance".

That circuit works by using the output gain transistors as the log
transistors.  In that sense the simplistic gain block model you have
no longer applies.
While Figure 4 is highly simplified, it does show the main circuit
blocks to help you understand this.  Also note the comment about
keeping the lateral PNP transistors turned off - in this application
we want to use the NPN transistors with their better performance.

> The text says the series emitter resistance should be modelled at
> 0.63R, then discusses that as applied to a transistor. I can follow
> that logic to get to the ~0.75% error as it applies to the Ebers-Moll
> equation. But not too sure how this should be factored into the model
> for SSI2164.

If you are able to model the transistors, then you would insert a 0.63
ohm resistor in series with each of the emitters of Q3 and Q4.

> Is it as simple as: tack a 0.63R resistor on to the output and
> subtract the voltage drop from v(ctl) in the equation above?

No, I don't think so, but happy to be proven wrong!

Also note that the comment "SPICE simulations indicate that..." means
Rossum used the SPICE deck of the actual chip, not some abstract
model.  You won't get that from SSI as it is highly
commercially-sensitive information.  Heck, you don't even get that for
the TL072 (those op-amp models you can download from the TI website,
for example, are just that: a standard op-amp SPICE model, e.g., Boyle
model, with specific values for that device).

Cheers,
Neil