# Cascode vs Folded-Cascode

I'm having a hard time understanding what the real benefit of a folded-cascode is over a simple cascode:

Also, one more question, is there any significance of there being a low-impedance node (source of M1A) at the drain of M1 (Figure (b))? I can see that M1A forms a common-gate stage.

• What is the minimum supply voltage to implement each? Commented Jul 15, 2020 at 20:49
• @SpehroPefhany I am not sure. This is an example from the Grey/Meyer book (5th Ed), there is no supply voltage specified. Commented Jul 15, 2020 at 20:52
• In a folded cascode design R and Vo use GND as a ref, as opposed to -Vss. GND is considered the most quiet trace on a board.
– user105652
Commented Jul 15, 2020 at 20:56
• Spehro nailed it. Assume some Vds value (say 2V) for each transistor, study both circuits, and work out what you need to power them.
– user16324
Commented Jul 15, 2020 at 20:59
• I see. It looks like the folded cascode saves on the V_overdrive/Vdsat of M1 but has an additional V_overdrive of the Ibias source. I presume this Ibias source overdrive will be small enough, thereby allowing the folded cascode to work with lower supply voltages? Is that correct? Commented Jul 15, 2020 at 21:14

There are two uses of a cascode -- one is to protect a low voltage transistor from a high voltage supply; the other is to increase the gain of the circuit.

In the circuit (a), M1 drives into a low impedance node (S of M1A). Thus M1's gain (gm) is not reduced by driving into a high impedance. M1's drain current flows into M1A which passes it to its R load. Since M1A is fed by a current source (M1), its current is defined by M1. Thus the gain is gm(M1)*R. and imperfections caused by output impedance are dramatically reduced. In practice a resistor is seldom used; a current source would be more common.

In (b), M1A's current is IBIAS-ID(M1). Again, if IBIAS is constant, there is a 1:1 relationship between changes in M1's drain current and M1A, it's just that it substracts instead, but the gain is essentially -(same)

• Thanks for the answer. "M1's gain (gm) is not reduced by driving into a high impedance". What exactly do you mean by this? Why do we care if it's driving into a high impedance node or not since it's a single branch. The current has to flow through that branch, there isn't any other alternative from what I see. I usually thought that we just want to make sure that the output impedance of the amplifier has to have a high impedance to maintain high gain. Commented Jul 15, 2020 at 21:41
• There is a third use : to isolate the input signal from the Miller capacitance of the voltage gain stage, thus greatly improving frequency response. M1 (the current amplifier) has output voltage 0V (AC), thus Cgd (the Miller capacitance) isn't multiplied by the stage gain.
– user16324
Commented Jul 15, 2020 at 21:50
• Isolating Miller capacitance is also a benefit - but can be considered as part of improving gain. #1 -- without a cascode, gain of M1 is gm.Reff, where Reff is ROUT of M1 in parallel with R (the load). Generally if the load is actually an R, Reff is >> R and can be neglected, but if the load is a current source, Reff becomes significant. In that case, the cascode increases gain by approximately gm.ROUT. Note that current source may also need a cascode to avoid it being the gain limiter. Commented Jul 17, 2020 at 19:26
• @user_1818839 Doesn't that mean the folded cascode shown in (b) doesn't have the Miller capacitance isolation that a traditional cascode does? Commented Feb 20, 2022 at 16:35
• It still removes the Miller effect because the (voltage) gain of M1 is low -- its gm divided by the gm of M1a. Commented Feb 20, 2022 at 19:05

Driving low impedance node (S of M1A) have same significance in both the circuits. Low impedance nodes swings nominally, hence, causing no trouble in headroom. Also, MA1 forms a common gate stage, it is providing isolation from output to input. So, any fluctuations at output will be attenuated hugely (by the reverse gain of CG stage), which appears at this low impedance node.

Cascode is good .I have used this with discrete components for Audio and for RF .The concept dates back to triode valves .The drawback of needing 2 devices instead of 1 was more relevant in decades gone by .There is still the need for a greater supply voltage than for an orthodox circuit with one device.The folded cascode is better at low voltages .I have used folded cascode with a MPF102 driving a BC557 on a 455KHz IF amp for 12VDC operation .