Forums » CircuitLab Support Forum

Could you gurus look at this circuit:

It simulates funny, like the equations are not converging.

Lots of zig-zag parts of what should be smooth curves.

Run the DC sweep to see the mountains where there should be smooth slopes.

Thanks,

A_H

I ran this, and it was taking forever - usually a sign of convergence problems.

A quick look shows the opamp swinging tens of kV. Swapping in the other type of opamp, and tying the supplies to ground/20V lets the simulation run quickly and produces reasonable plate current graphs. (Results consistent with a triode with unity mu?)

I must be missing something about what this circuit is supposed to be doing.

You could not realise this circuit using real opamps because you are taking the input voltage below ground with only a 0V to +20V supply. You seem to be taking the voltage V(in) way below ground and using some feature of the diode leakage currents to then pull the opamp +ve input down.

There's some analysis of the circuit assuming a +ve input voltage here:

but for -ve inputs I am either very dubious about using simulation models in this way or very confused about what the model is supposed to be doing.

@signality: You have to look at the DC sweep voltages - the voltages sweep from negative to positive, with the opamp input seeing 0V- 3 diode drops (below 1V with these Schottky diodes at these currents). The opamp input actually can go below ground only to the extent that D8 has finite leakage (equal to the I_s term), but in this situation the error amp is turning off the bipolar transistor, and there is no practical effect to the amp swinging too low.

But from a simulator standpoint, if the (-) input remains at ground and the (+) input goes slightly below ground, the opamp's 110dB of gain will drive the output voltage negative by 1kV for every 3mV of input differential. So using the clamped opamp is useful in cases like this where the feedback loop effectively opens when the BJT is cut off (or when it saturates a the other extreme).

I should have explained, its a triode simulator, basically taking the plate voltage to the 3/2 power and subtracting mu times the grid voltage. But the finicky log then antilog transforms are tricky to do right in real life and in simulation.

Gonna abandon this design as it has a horrible temp coefficient.

Doh!

Sorry, I'd completely missed that fact that both Grid and Plate voltages were being swept.

I'd wondered about the horrendous temperature dependency but thought maybe you didn't care because it was only a simulation.

You might like to have a look at:

http://www.allaboutcircuits.com/worksheets/nonlin.html

And for more detail: http://www.ti.com/lit/an/snoa641a/snoa641a.pdf

Other stuff ... Fig 16 in:

http://www.analog.com/static/imported-files/data_sheets/SSM2212.pdf and http://www.analog.com/static/imported-files/data_sheets/MAT12.pdf

and for detailed information including temperature compensation:

http://www.electronics.dit.ie/staff/ypanarin/Lecture%20Notes/DT021-4/6LogAntiLogAmplifiers.pdf

:)

Many thanks for the references. I'm moving on to the better and compensated BJT loggers. I thought I was pretty clever doing the 3/2 power with 3 and 2 diodes, but that is just too horrible temp-wise to even wire up once.

I spent many, many hours designing and testing temperature compensated antiloggers for analogue music synthesisers back in the 1970's.

Somewhere you may find a reference to a Fairchild apps note using something like the LM3046 transistor array:

http://www.ti.com/product/lm3046

as an ovened, temperature compensated antilog amp. I think the longtail pair was used to form the antilogger and the three remaining trannies were used to form an on-chip heater plus a diode drop temperature monitor. The whole LM3046 was then wrapped in thermal insulation.

There's a reference to it here:

http://electronotes.netfirms.com/s019.pdf

but I haven't found the original apps note. Think it was published in something like Elektor or Electronics Today International in the early 1980's.

Another way to make arbitrary transfer functions is to use diode function generators. It's described in an out-of print Burr Brown book but is basically a variation on the opamp precision rectifier circuit.