About the compensation capacitor in Miller OTA

[threekingtiger]

Question 1:

Considering the classical two stage Miller OTA, if I use MOS capacitor (an MOSFET with D S B terminals connected together) for compensation, which terminal should be connected to the output node, G or B? Is there any influence on the OTA if I exchange the two terminals casually?

Question2:

If I choose to apply the nulling resistor scheme to cancel the RHP zero existed in Miller OTA, and the nulling resistor is made by MOSFET woking in linear region. Is there any influence on the OTA if I exchange the position of the compensation capacitor and the nulling "resistor" casually? In other words, which of them should be connected to the output node of the OTA directly?

 

[erikl]

Question 1:

Considering the classical two stage Miller OTA, if I use MOS capacitor (an MOSFET with D S B terminals connected together) for compensation, which terminal should be connected to the output node, G or B? Is there any influence on the OTA if I exchange the two terminals casually?

1. If you connect D & S to B, this node normally is at a fixed potential, otherwise you have to spend an extra well for this MOSFET. So for such a MOSCAP you'd better use the standard bulk connection.
2. In order to make profit of the high cap/area ratio of a MOSCAP, the voltage difference between G and S+D must create a channel, e.g. must be > Vth. Hence for an nMOSCAP the Gate should be the positive terminal - and the other way round for a pMOSCAP.

Question2:

If I choose to apply the nulling resistor scheme to cancel the RHP zero existed in Miller OTA, and the nulling resistor is made by MOSFET woking in linear region. Is there any influence on the OTA if I exchange the position of the compensation capacitor and the nulling "resistor" casually? In other words, which of them should be connected to the output node of the OTA directly?

Theoretically this shouldn't matter for a series connection. Practically, however, the necessary operating points will decide upon their position: The MOSFET working as resistor in linear region needs a high (and possibly controlled) overdrive voltage at its gate, and the MOSCAP working as capacitor also needs a voltage difference > |Vth|. So - depending on the MOSFETs used (n or p) - and the voltage levels available at the output node of the OTA, at the compensation feedback point, and at the node which controls the gate of the nulling resistor MOSFET, the type of the used MOS devices and their relative position will be decided.

 

[snafflekid]

In some single well processes, it is the nMOS that has the independent body so you will be limited to using that for a moscap. In other processes it is the pMOS you have to use. For an example where the output gain stage is a pMOS, I suggest you use an N moscap and put the DSB side toward the diff amp and G toward the op-amp output.

I suggest this because the DSB side will be close to Vdd and the output less than Vdd most of the time. This moscap operates as a linear cap in accumulation mode except if the output voltage rises above DSB voltage (this should be rare?) Also, the junction capacitances of the body-to-substrate will help add to the Miller cap.

But I think the better solution is to use a metal or poly cap for compensation instead of a moscap.

 

[leebluer]

By measured C_V curve, the reverse based MOS CAP achieve better Temperature coefficient & stable unit capatance. In my view, if isolated MOSCAP available, reversed is better.