MOSCAP - Mosfet transistor or the NCAP, from the tech lib?

[AMSA84]

Hi guys,

I want to use a MOSCAP and characterize it. I've never used MOSCAP.

I read that we can do a MOSCAP using a MOSFET transistor. However, in my tech (UMC130nm) I've available a MOSCAP in the Capacitor cell section. So I am a bit confused.

I must use a MOSFET transistor to make a MOSCAP or I should use the MOSCAP that exists in the cell from the tech library? The name of the cell is: NCAP_HSL130E.

Regards.

 

[erikl]

You can use both, just as you like. The ready-made NCAP cell will already include the source-to-drain connection, and, probably, the necessary marking layer for the extractor to mark it as a capacitor. I'd suggest to study the NCAP layout.

 

[dick_freebird]

The NCAP may also be a depletion mode device which
would make it better for signals that swing across the
entire rail span (More C-linear). This is not always the
case, but often done when the process makes it a
practical "freebie" - often only requires crossing-up the
well type (N S/D, N gate poly, in Nwell instead of Psub).

You'd have to look at the as-printed layers, or figure out
the mask Booleans, to know.

 

[AMSA84]

Thanks for the reply.

Allow me to ask you guys another thing.

If I use a MOSFET connected in a way to become a MOSCAP I would need to connect the source, drain and bulk together and the capacitor will be between the gate and that S/D/B connection.

Now, taking into account that the capacitor is to be used in the output filter of a power supply, the output voltage around 1V. So I need to feed the MOSCAP with that voltage in order to know how much will be the capacitance (of course depending on the transistor sizes)?

 

[dick_freebird]

Yes, you would want to look at the capacitance and,
perhaps more importantly, the channel resistance at
such a light turn-on. This series R (esp. in a MOSFET
type structure where the bottom plate is only the channel,
and channel resistance lies in series) can make the cap
ineffective, or its effectiveness and corner frequency
quite variable with processing (beyond simple Tox control,
you have (V-VT)^2 and that core difference may be too
small to achieve consistency).

In the stated case you'd prefer a low, zero or negative VT
NMOS cap.

 

[AMSA84]

Okay. So you suggest to use a NMOS Transistor connected as a Capacitor with one of those three options.

Now when connecting the MOSFET as stated above, I must polarize the gate with that 1V (output voltage of the converter) and measure the capacitance using for example S-parameter analysis? The capacitance obviously will depend on the transistor size.

 

[dick_freebird]

If I were you I'd characterize the capacitance on both
sides of your planned bias point unless you believe in
the Small Signal Fairy and her inherent goodness. The
C-V swing ought to be pretty significant about VT(N).
And my point was, capacitance is far from the whole
story when you want to use that cap as either a tuning
(Q) or compensation (C*ESR zero), you want to get a
good modeling handle on the movement of series R with
bias point. You might see things like distortion at high
amplitudes, and high in this case could be just 100-200mV
if you slide on down toward, or below, VT - your R might
swing from kOhm to MOhm and surprise you if you model
it too simply.

And no, I do not necessarily recommend the MOSFET
(standard digital) over the designed MOSCAP. The latter
presumably is designed for the most linearity of capacitance
and lowest series R. It may or may not suit you, the choice
has to depend on your "values".

 

[AMSA84]

Okay freebird. Thanks.

The point here is that the MOSFET Transistor is less reliable. Is that it? I might depend on the application, but as I said this is to do some filtering on the output filter of the converter. Maybe it is not a good decision.

I'll go to the MOSCAP that the technology provides. But here I've a problem. The MOSCAP from the technology has 3 terminal. If I remember if the BULK, GATE and SOURCE or DRAIN and I don't have any clue on how to connect it!

Do you have any experience on using this type of MOSCAP? How can the connections be made?

 

[dick_freebird]

Bulk is probably just your JI substrate handle, there for
modeling purposes (parasitic cap of bottom plate needs
to go somewhere).

MOSFET reliability may be lower when applied as a MOSFET
and you care about every attribute. But as a capacitor
you only care about TDDB reliability and can neglect HCI
(usually the more severe by far) and any other reliability
effects that pertain to short channel high field operation.

GATE +, S/D to GND (along with the BULK, via any rough
substrate-resistance-mesh you might care to guesstimate).

Now, using the cap as a bulk filter does mean that every
fast impulse will be sent directly into the substrate ground.
I would not want to do that on a switcher. A linear, OK -
presumably there is not much going on internally.

 

[AMSA84]

Thanks.

Freebird, so what you'd recommend? To use the NCAP_HSL130E capacitor? (I presume that it's a MOSCAP built by the foundry)

I don't know if you were understanding what I was saying in the previous posts.

 

[AMSA84]

Hi guys,

I have done a simulation with the NCAP using a PSIN (SParameter) source; plotting the Real part and Imaginary part of the ZM I found that the resistance (real part) basically is zero. (fOhms/pOhms)

Why?

I tried to find he model from the tech regarding the NCAP but without success.