# How to make a PMOS work as a constant current source?

I have a circuit in which, the PMOS is biased to a voltage source, the gate voltage, V1, of the NMOS below varies based on the functioning of the analog circuitry before. Because of this, the current through the NMOS (I2) also varies. When the current through the NMOS (I2) becomes greater than the current through the PMOS (I1), the next part of analog circuit starts working. For this, I need a constant current I1 (and I can not use a constant current source). But, in my circuit, the current through PMOS (I1) is varying based on the current through the NMOS (I2), and hence, my next analog circuit is not functioning perfectly. How can I make the PMOS work as a constant current source, independent of I2? I used 65nm technology for developing the circuit. Here is my circuit,

• What are you trying to accomplish overall? This looks like it might be an X-Y problem. There may be better ways to achieve your overall goal than with the approach you have chosen. Commented Aug 24, 2023 at 1:55

You could make the current sources super robust, with cascodes and what not, yet you will still see that the current provided by it will vary with respect to the input signal of your NMOS.

Why? because you're most likely simulating these circuits in an open-loop simulation, without any feedback (not even DC) around it.

I'd suggest you do the following:

1. Put a feedback path around your amplifier (e.g. if you want to design this op-amp for a voltage amplifier, you can simply connect the output to the (-) input to bias your amplifier properly and put it in a voltage buffer configuration).

2. Bias the PMOS with a current mirror. It could be a simple ideal current source with one arrow direction connected to GND, while the other connected to another, properly-sized, diode-connected PMOS transistor (a current mirror, basically). Here, you can decide what kind of current you want running through your actual PMOS current source.

3. If you actually need a real current source where to derive your currents from, you can look into a bandgap or a PTAT current source. Typically, most ICs have a dedicated block where the current is created (either with a constant or a PTAT behavior). This block can have current mirrors where you copy your current from into your amplifier (via current mirrors).

Feedback and a large loop gain will suppress the voltage swings at the drain of your NMOS, thus your PMOS current source will provide a predictable current to the NMOS and analyzing the stage with an small-signal approximation will be a good approximation to it. You can improve it using cascodes, regulated cascodes as needed, but start from there, I'd say.

As a good former professor of mine used to say: "You can't force a current through a drain!"

EDIT: To expand on an observation of yours. You seem to think that biasing a current source transistor with a voltage source should be good enough (based on your schematic drawing). While this sounds ok at first, there are much better and robust ways to do so:

• Why is this idea bad? if you found a gate voltage for the PMOS that gives you the current you want, it'll work only for that particular corner and temperature point, i.e. if you test this circuit with another corner and temperature, you'll see that your current will be different, then you'd need to find another suitable bias voltage. Cumbersome, no?

• Instead, why don't we let a diode-connected PMOS biased with a current source decide what its suitable gate voltage must be to be biased at a given current? Then, this circuit will automatically find the right gate voltage to keep the transistor biased at your decided current. In this way, we let the circuit find its own suitable bias voltage according to its environment and PVT variations.

In short, we let the loop of that PMOS diode-connected transistor find its suitable gate voltage for the given current. This is the most robust and efficient way to bias a current, and the reason why you don't usually see current sources biased with voltage dividers or other brute-force voltage sources.

Hope that helps :)

I am assuming that you have a constant voltage available in your system. You can use that and then use a voltage-to-current converter and mirror that current as shown below. For this to work as a constant current source across temperature, you need a resistor that does not vary with temperature and the 2 PMOS transistors have to be matched.

P.S: The size of the PMOS transistor is quite small. If you plan to use this solution, you need to increase the sizes to have good matching.

• Why did you put an opamp there? Why don't we connect the Vref directly to PMOS and mirror the current? Commented Aug 23, 2023 at 9:05
• @SivaSandilya you can also do that. Commented Aug 23, 2023 at 11:18
• If you connect the Vref to the PMOS, the current that you will get will vary across process and temperature. That is why we do this V-to-I conversion using the opamp to minimize the current variation across process and temperature
– sai
Commented Aug 23, 2023 at 13:26