How to measure the gm of a differential pair?

[demodb]

He guys, i have a question. If i have a differentail pair, as shown below, how do you measure its gm. I could attach a resistor to it, but it affects the voltage level. I know how to calculate it, but don't see exactly how to measure it.

 

[aryajur]

Re: simulation gm

Do you want to calculate the gm of a transistor or the whole diffamp? Anyway for the diffamp, connect a resistor to the output with its dc value as very large say 1e18 and its ac value very small say 1e-18 and then calculate the operating point and then do the AC simulation and find the current through the resistor Gm = Iout/Vin

 

[demodb]

Re: simulation gm

i want to measure the gm of the whole diffamp. But what do you mean by setting the DC and AC value of the resistor?

 

[aryajur]

Re: simulation gm

You can set the resistor value so that the DC simulation and the AC simulation takes up different values of it. So set the DC value different from the AC value as described above and so when you run the DC simulation it will be like the resistor is not there and the operating point of the diffamp will be calculated without the resistor. Now to find Gm you need to find out output short circuit current in a small signal analysis, and hence you can find the current through the resistor in a AC simulation and the resistor acts like a short circuit in the AC simulation.

 

[demodb]

Re: simulation gm

thx Aryajur, that helped a lot.

 

[rfle]

simulation gm

Hello all.

In DC, simulation, we can sweep the bias voltage (Vbias), and we can plot the dc I-V characteristics, and gm and gm'... over the large range of sweeping voltage (Vbias). Some papers they mentioned about AC characteristics, they can also plot the gm over input voltage range (it is AC Vinput, not Vbias). Can anyone show me how to do this simulation to get that AC gm plot? Thank you very much

 

[pktiwari]

Re: simulation gm

Hi
It was nice to connect a DC resistance but I think it is best to connect a fixed DC voltage (0V AC) at the output. Give the DC voltage what you expect to have when (if) this amp is used in feedback.

The reason I say this is that otherwise you might end up having some transistors in linear.

Say, if you are planning to use it as unity gain buffer, find gm by setting the voltages from the desired input range.

 

[aryajur]

Re: simulation gm

I agree if you really know what voltage would be there on the output (when you put it in feedback then it is known) then its better to put a voltage source since you are spared the exercise of calculating the impedance loading the diffamp. In case you don't know the output voltage then its better to put a resistor since it simulates the open circuit for the DC case and the short circuit for the AC case as you would normally calculate gm.

 

[pktiwari]

Re: simulation gm

What difference would it make if you have a different loading at DC than any other frequency as you propose by having a large DC resistor.. why not just make the loading same for all frequencies including the DC by having a voltage source..

If you don't know the output voltage put any decent value which keeps everything active. Why risk the op calculation of output voltage to the simulator.. you might end up having a transistor in linear and will find the wrong gm..

I might be wrong and probably you have some subtlety to share..

 

[aryajur]

Re: simulation gm

well actually I guess you are not clear on the resistor values I am suggesting. For Dc simulation you are free to pick a large resistor value as 1Giga ohm or lower/higher, your choice. This is equivalent to an open load. If your transistor goes to linear region with such a large resistance then maybe you may need to revisit your design since leakage currents larger than that can easily happen in IC design. This would be a good check on your circuits robustness also.

 

[pktiwari]

Re: simulation gm

The point I am making is that we are trying to do a measurement of a high voltage gain element which can only be properly biased when operated in feedback OR by killing its gain by high loading..
By putting a 1GOhm resistor you still maintain the DC voltage gain and thus an unreliable bias point..

To highlight this - What bias voltages will you use at the input of diffamp in the simulation? Any systematic offset in the diffamp (atleast due to the base current of load transistors) will put you in trouble..

P.S. I am sure you know all this very well but there is something you are trying to say which I am not getting.

 

[aryajur]

Re: simulation gm

The circuit picture shows a single stage diffamp which should not have any systematic offset. The bias point in a simulation for the output is defined reliably in simulation if both inputs have the same DC value. So the issues you mentioned are not a problem. In fact in this case if you are not planning to use this circuit in feedback putting a voltage source is a messy business since you do not know what is the bias voltage supposed to be with changing corners and temperature.
So well I would prefer to use the resistor method for this circuit. But in general the problem should be evaluated to find out an aceptable method. Maybe it might be better to use a voltage source (when the bias voltage is known) or maybe it might be better to actually put the circuit in feedback as it will be actually used to simulate the gm.
Anyway I think we have beaten around this issue a lot, maybe we should take the discussion offline.