2

I have the following circuit and I'm wondering a little bit about the functionality of the output transistors M12-M19.

schematic

simulate this circuit – Schematic created using CircuitLab

I know that M1,M2,M6 and M3,M4,M5 are two differential pairs and I also think, that M19,M12 and M16,M14 redirect/combine the current from positive and negative input to one current.

What I don't fully understand, though, is the idea behind M13,M15,M17,M18. I believe, that the node connecting all four gates and the drains of M18 and M17 should always have a voltage exactly at the mean of M18's and M17's sources. So I was thinking, that M13 and M15 could be cascodes, with M18 and M17 being a self-biasing network.

Is this pretty much it or is there more to the eight transistors on the right?

Uroc327
  • 131
  • 3

1 Answers1

2

You're correct M16, M17, M18 and M9 are mainly for biasing. Note how they're all "MOS diodes", all in series and between Vdd and ground. They are the biasing for M12 - M14.

At the same time M12, M19 and M14, M16 are current mirrors that push the current difference from the differential pairs in the direction of the output.

The biasing current through M16 - M19 is not constant, M1 and M3 add/subtract current and that influences the voltage at M17 and M18.

M13 and M15 are indeed cascodes even though their gate voltage is not constant. Their gate voltage will not vary much though. M13 and M15 provide a low impedance at their source so that the currents from M2 and M4 can flow into the string M12 - M15.

This cascode structure is similar to a "folded cascode", see this lecture sheet 17 and onwards.

Bimpelrekkie
  • 80,139
  • 2
  • 93
  • 183
  • Thanks for the hint that they're basically mos diodes! What is the advantage of having a dynamic bias voltage the cascodes M15 and M13? And why would one only use the voltages from the positive input to determine this active bias? – Uroc327 Mar 04 '21 at 12:32
  • I also thought about a folded cascode configuration first. But personally, this confused me more, than it helped :D Would it be "semantically correct" to draw M19, M12 and M16, M14 at the left above/below the respective differential stage to show the similarity to a normal single ended amplifier? This would then leave M13 and M15 on the right as cascodes to only combine the output currents of the n and the p input stages regardless of their output voltages and M17 and M18 as active bias for the cascode. Or would this hide or distort some circuit functions? – Uroc327 Mar 04 '21 at 12:33
  • 1
    *What is the advantage of having a dynamic bias voltage the cascodes M15 and M13* For the cascodes themselves, the changing bias makes no difference. What does happen is that the current changes go to the **output**. Note that this circuit has a **current output** so for good performance, you need to terminate the output with a **voltage source** (of about Vdd/2). *And why would one only use the voltages from the positive input to determine this active bias?* The only reason I can think of is that it is **simple**. It is not a robust design, only simple. – Bimpelrekkie Mar 04 '21 at 13:29
  • 1
    *Would it be "semantically correct" to draw M19, M12 and M16, M14 at the left above/below the respective differential stage to show the similarity to a normal single ended amplifier?* Yes and it is often drawn that way just not here. Either is OK, whatever you prefer. I think what you're suggesting is removing M17 and M18 and then using the cascodes (M13, M15) **with a fixed gate voltage** to cascode the outputs of the current mirrors. That is what you see most often in textbooks and a bit more clear than your original schematic. – Bimpelrekkie Mar 04 '21 at 13:31