The Designer's Guide Community Forum

The Designer's Guide Community Forum
https://designers-guide.org/forum/YaBB.pl
Design >> Analog Design >> MOS transistors bulk connection
https://designers-guide.org/forum/YaBB.pl?num=1322948702

Message started by Pashtet on Dec 3^rd, 2011, 1:45pm

Title: MOS transistors bulk connection
Post by Pashtet on Dec 3^rd, 2011, 1:45pm

Dear friends,
as we know in order to eliminate dependence of threshold voltage on source-bulk potential and thus reduce gata-source voltage of MOS transistors, we can connect source and bulk terminals together. So we need less voltage headroom in comparing with the case when all NMOS bulk terminals are on GND and PMOS ones - on VDD.
I wondering if there are any limitations or negative consequences of doing that? May be there are some applications where it important to keep bulk potentials on supply rails as i wrote above ?
Thanks in advance.

Title: Re: MOS transistors bulk connection
Post by sriram on Dec 3^rd, 2011, 5:14pm

Hey,
I think the only problem in not doing this connection is that your substrate now is not connected to ground and is carrying some signal(maybe whatever based on your circuit). This may cause unwanted coupling to other parts of the circuit. This might not show up in simulations, but its definitely a problem while measurement.

Title: Re: MOS transistors bulk connection
Post by loose-electron on Dec 3^rd, 2011, 7:19pm

when you do that you are now in a situation where you are driving the capacitance of the well to substrate region.

that capacitance can be very significant and is not in most transistor models, leading to simulation errors.

you need to define the area of your well, that the FET is in, and then approximate that capacitance in your model

Title: Re: MOS transistors bulk connection
Post by Pashtet on Dec 4^th, 2011, 1:46am

Thanks for your sharing.
Both reasons you have mentioned are serios and difficult to simulate. So your general advice is to avoid doing that.

Title: Re: MOS transistors bulk connection
Post by RobG on Dec 4^th, 2011, 8:26am

Pashtet wrote on Dec 4^th, 2011, 1:46am:

Thanks for your sharing.
Both reasons you have mentioned are serios and difficult to simulate. So your general advice is to avoid doing that.

I'm not sure what sriram was talking about, but I agree with Loose Electron. However, I'm sure he'd agree that the extra capacitance isn't a "serious" enough reason to generally avoid doing it - you just need to make sure you model it and weigh the costs against the advantages because the extra capacitance may (not always) cause more issues than is solved.

Another more practical issue in my experience has been the extra area required. The device needs to be in its own well and the well-to-well spacing is relatively large. This can wind up wasting a lot of space so you have to weigh that (and the capacitance) against the performance improvement.

rg

Title: Re: MOS transistors bulk connection
Post by loose-electron on Dec 4^th, 2011, 12:04pm

I agree with Rob, the area taken up by the isolated well is generally not worth the effort.

As a general rule, if I do not need the slightly better threshold characteristics (virtually never), I leave the N-well tied to the positive power rail and have the PMOS devices share a common N-well.

Title: Re: MOS transistors bulk connection
Post by loose-electron on Dec 4^th, 2011, 12:08pm

RobG wrote on Dec 4^th, 2011, 8:26am:

However, I'm sure he'd agree that the extra capacitance isn't a "serious" enough reason to generally avoid doing it - you just need to make sure you model it and weigh the costs against the advantages because the extra capacitance may (not always) cause more issues than is solved.
rg

Its tough to generalize this one, you need to define the well geometry, determine the capacitance of walls and bottom, and then put it into your model.

Driving that well capacitiance can be significant. Some work I did on an RF PA showed some very extreme differences w/wo the capacitiance. At the end of the day, I talked them into tied well to +V and the big I=Cdv/dt currents got eliminated.

Title: Re: MOS transistors bulk connection
Post by Vladislav D on Dec 4^th, 2011, 2:20pm

If you connect a body to a source you should keep an eye on the drain-well diode because you don't want the diode starts to conduct current. This is especially true if you have shutdown mechanism in a circuit.
I would say that in 99% of circuits we connect body to the rail. You need to have a strong reason to connect body to the source. As an example, a differential pair with high PSRR.

Title: Re: MOS transistors bulk connection
Post by thechopper on Dec 6^th, 2011, 6:29pm

Another example in which it results convenient to connect body to source is in stacked diode connected devices: sometimes the overall voltage drop might be higher than body-drain breakdown and thus having bulk connected to source in each device puts drain-bulk diodes in series, thus larger breakdown characteristics.

Tosei