Hello,

I've been trying to replicate the first schematic on http://en.wikibooks.org/wiki/Electronics/Analog_multipliers which shows the use of log and exponent circuits to create an analogue multiplier. The concept is fine, however, when tested using a simulator with the same values for all the resistors and diodes, the output goes into saturation. I've also tried simulating similar schematics from different websites with no success.

Is the diagram incorrect or am I doing something wrong?

 

Let's see your net list.

 

I won't attempt to correct all the errors in that document, but there are several, both in the equations and in the schematic.
Why do you need an analog multiplier? There are analog multiplier ICs available.

 

Squillam, it is very diffult to create log-amps which manage to span many decades of input. The components need to be of excellent quality such as low input bias current (JFET input stage) op amps. So puting random simulation components (the typical lm741 op amp for example) will give you circuits which perform only over very limited range of inputs and which will usually yield inaccurate results. Parameters such as the diode and transistor's Is affect the circuit in ways that it is difficult to build these without either quality or matched components.

These circuits are perhaps not so complicated in theory, but in practise they are quite a bugger if you want to build these from discrete components.

 

Squillam, it is very diffult to create log-amps which manage to span many decades of input. The components need to be of excellent quality such as low input bias current (JFET input stage) op amps. So puting random simulation components (the typical lm741 op amp for example) will give you circuits which perform only over very limited range of inputs and which will usually yield inaccurate results. Parameters such as the diode and transistor's Is affect the circuit in ways that it is difficult to build these without either quality or matched components.

These circuits are perhaps not so complicated in theory, but in practise they are quite a bugger if you want to build these from discrete components.

The circuit doesn't sim as advertised, even with ideal op amps and diodes. I tried it.

 

hmm. The output of exponential and logaritmic circuits depend on parameters in diodes or transistors which make them nonideal. An ideal diode has no reverse leakage current which is a design parameter for log-amps for example. So these circuits may not work at all when simulated with ideal diodes. If the sim simulates ideal diodes for example with Is = 0 then the output will be infinite for all inputs since that parameter appears in a denominator somewhere in the input-output relationship.

You should keep in mind that the circuits you are trying to build are not very practical as they are, imo they are merely concept ones. For example the log-amps in these are of the most simple kind and depend heavily on temperature. They would never be used in practise without alot modification to reduce temperature dependence. Both Vt and Is vary with temperature.

Log-amp circuits are difficult to work with because they rely on parameters which make the components nonideal (Vt and Is). A big flaw also is the fact that these parameters vary alot with each production batch. As an example as you may propably know the parameter which "decides" wether bjt's are well matched or not is not Beta as is often wrongfully stated but the saturation current Is. Who's production variation leads to the need of IC's or specially matched components to realise alot of circuits requiring alot of precision.

You may have noticed how expensive logaritmic IC's are. Creating reliable and accurate log-amps which operate over many decades of input appears to be quite expensive.

 

hmm. The output of exponential and logaritmic circuits depend on parameters in diodes or transistors which make them nonideal. An ideal diode has no reverse leakage current which is a design parameter for log-amps for example. So these circuits may not work at all when simulated with ideal diodes. If the sim simulates ideal diodes for example with Is = 0 then the output will be infinite for all inputs since that parameter appears in a denominator somewhere in the input-output relationship.

You should keep in mind that the circuits you are trying to build are not very practical as they are, imo they are merely concept ones. For example the log-amps in these are of the most simple kind and depend heavily on temperature. They would never be used in practise without alot modification to reduce temperature dependence. Both Vt and Is vary with temperature.

Log-amp circuits are difficult to work with because they rely on parameters which make the components nonideal (Vt and Is). A big flaw also is the fact that these parameters vary alot with each production batch. As an example as you may propably know the parameter which "decides" wether bjt's are well matched or not is not Beta as is often wrongfully stated but the saturation current Is. Who's production variation leads to the need of IC's or specially matched components to realise alot of circuits requiring alot of precision.

You may have noticed how expensive logaritmic IC's are. Creating reliable and accurate log-amps which operate over many decades of input appears to be quite expensive.

True, but log-antilog multipliers were in use before the invention of the Gilbert cell multiplier, and may still be. There are several app notes from companies such as Analog Devices and National, with working schematics. They deal with the temperature and matching issues. Most, if not all of them, use transistor Vbe junctions instead of diodes, but the concept is the same.

 

Art of E has a few.

 

Oh well. I guess the circuit won't work despite the theory being correct.

Thanks to all who've replied. Your responses are much appreciated.