Thank you very much for the explanation Adjuster, they come very handy.

Relative results in your simulated circuit seems to match with my lab results, but they should match with my relative and absolute simulated too and vary a little bit from my lab results.

Adjuster said: This is simply not correct. A logarithmic converter is an accepted way of generating a true-rms by first generating a square-law function. That's it!!! I need the logarithmic converter to make the square of a function.

 

I am very sorry but my eyesight is not what it once was! I failed to see the negative signs, against the grid background.

The loss of focus accommodation with age is a wretched nuisance.

It's ok, now I'm more sure there are mistakes there. Did you have more observations? I need to find why it's not working.

 

I've tried it on LTSpice. Again, I do not have a 741 model, but the results don't seem all that far from what's in the paper.
Note my graph end points are chosen to give natural logarithm values of about -5 and +2.5.

I think the result does not look all that far out, compared to the graph in the paper. At least, the slope looks about right 3.107V end-to-end.

I'm afraid that's about as far as I can take it, at least for now.

Good luck with your project.

 

Thank you very much for you help.

I think both circuits are ok but I have to calibrate them. I mean, calculate Saturation current from lab results and using the formula. Once I know saturation current I can use the circuits. What do you think of this idea?

 

The simpler circuit is really only suitable as a demonstration of the principle, as it will be very much affectrd by temperature, so calibration will only hold if the temperature is held constant.

Two of the simple circuits could in principle be used together, with one fed a constant input to provide a compensating reference, but this would give no real saving in complexity compared to the compensated version. The use of a pair of transistors on the same silicon could help to make the compensation work better than could be obtained even with matched discrete devices.

 

I have a problem with my logarithmic amplifier circuit.



Voltimmeter shows a value different from the one obtained in my graph simulation. Voltimeter shows -1.164 as a result and value in the graphic is -1.33, is that normal? In this case they should match, right?

 

Are these results by two methods of simulation / calculation, or is one result a practical measurement?

In any case, even with closely matched transistors on a common piece of silicon, some calibration could be required. You could aim as a minimum an offset and slope correction. A correction for absolute temperature variation affecting the slope would also be helpful, perhaps using thermistors.

If you are using randomly selected discrete transistors the situation will be less predictable. I note that you are also using 741 op-amps, which have relatively poor performance and may degrade your results. (I have tried a simulation with LT1006 amplifiers, and I get 1.333V output for 3V in, but the simulations may disagree for other reasons.)

As has already been pointed out more than once by myself and others, this kind of circuit is capable of only limited accuracy. For best results appropriate components must be used, preferably selected or inherently matched. You could then apply offset and gain adjustment, but if your requirement is for a really accurate result you may in the end be better to consider a commercial device.