About offset voltage of error amplifier

[jefferson]

Hi all,anyone know how to analysis the offset voltage of EA?
When simulated,it's seemed that little offset voltage exist.But infact it exists.
1.How to estimate it?
2.Is offset voltage varies with Vdd? How?
3.Is offset voltage has any relationship with its gain or bandwidth? How?
4.What would be do to avoid it?

Any reply would be appreciated.

 

[elbadry]

Offset voltage usually has two components:

1. Systematic: This is due to bias mismatches ( Johns and Martin provides good discussion of offset )

2. Random: This is due to mismatches that occur during fabrication. For a properly designed Opamp, this will be the dominant source of offset voltage. Usually, this can be overcome by increasing the area ( W*L) of input pair. The fab should provide mismatch data and/or monte carlo modeles so that you can estimate the offset

I think that it has nothing to do with supply voltage or gain/bandwidth.

 

[jefferson]

Thanks elbadry.

Maybe I should know the approximate offset of an Opamp.
But where can I get Johns and Martin'materails.
THanks.

 

[elbadry]

I think you can find the book on edaboard

 

[sankudey]

Hi,
Offset Voltage is an DC characteristics and comes from non-ideality as described above as MISMATCH.

The related docs are :

1. Grey-Meyer Book: Look in Diff-Amp chapter :



2. The following Posts in This board: (A little Similar)

How to measure Offset in Simulation:


sankudey

 

[bicave]

-increase W/L and W*L to reduce offset voltage
-small i bias reduce offset voltage

to estimate it...? there's no expression or equation.

 

[electronrancher]

no equation? I think you need to go back to your textbooks!

random offset follows 1/sqrt(AREA). The exact value will be C/sqrt(Weff*Leff) where C is a constant for your process including Tox and Weff and Leff are the effective W/L of your device.

So thinner Tox gives better matching with a linear dependence, and larger area gives 1/sqrt() better matching. I usually say 1um square gives about 20mV random offset in a 1/2 um process.

You will have offset due to the diff pair and any sets of matched current mirrors. Cascodes usually aren't too bad. To calculate, you can probably estimate only the first stage. Diff pair will have a delta-Vt based on area, and mirrors will have a delta-I. Convert this delta-I to input referred offset and add to the delta-Vt of the diff pair, and that's probably a reasonable estimation.

 

[qutang]

using dc sweep to see the offset. it at the biggest gm position

 

[chenjia]

run monte carlo simulation or hand calculation