current mirror mismatch with different multiplier

EDAboard.com Forum Index -> Analog IC Design & Layout -> current mirror mismatch with different multiplier

Author

Message

alex.ang1983

Joined: 19 Mar 2009
Posts: 7

05 Jul 2009 12:02 current mirror mismatch with different multiplier

Hi All,

I am not a specialist in mismatch calculations and I am getting crazy to calculate mismatch between mirrors that have differen multiplier.

If I have a current mirror with W1=W2 L1=L2 but M1=1 and M2=N (where M is the multiplier) how can I calculate and take in considerations the different M?

I would really appreciate any kind of help or also docs.

Many thanks in advance
Best regards

erikl

Joined: 09 Sep 2008
Posts: 736
Helped: 139
Location: Germany

05 Jul 2009 12:36 current mirror mismatch with different multiplier

If your M means fingering - and not distributing - transistors, you may consider M mismatch exactly like W mismatch, i.e. you calculate (W*M)/L mismatch.

alex.ang1983

Joined: 19 Mar 2009
Posts: 7

05 Jul 2009 12:45 Re: current mirror mismatch with different multiplier

many thnaks for your answer Erikl!!

Yes with M I mean that I have M transistors in parallel having W=W2 and L=L2.
but Problem is that the 2 MOS have different M because one is the reference and has M=1 and the other is M=3(i.e). So this should be different rispect to a scenario where W1=W2 L1=L2 and M1=M2 (in this case the mismatch variance is =coeff/(W1*L1*M1) )

I hope you can help me
thanks
Best regards

erikl

Joined: 09 Sep 2008
Posts: 736
Helped: 139
Location: Germany

05 Jul 2009 13:09 Re: current mirror mismatch with different multiplier

alex.ang1983 wrote:

in this case the mismatch variance is =coeff/(W1*L1*M1)

Hi Alex, I think your above equ. is correct. For worst case, you should use the lowest M. For a typical estimate (and a best possible centroid positioning - perhaps not quite easy with M2 = 3*M1 Wink

) you could use the geometrical middle value of M1 & M2 :
M = √(M1*M2) .
Cheers, erikl

Bahrouch

Joined: 19 Oct 2008
Posts: 1

05 Nov 2009 9:49 Re: current mirror mismatch with different multiplier

Hi Alex,

Assume there is a reference current source with Wref, Lref, M=1. and multiple (n) current sources with the same W, L and M=1 parameters connected to the reference. Then the mismatch of each current source with respect to the reference would be DeltaI/Iref= A/square(W.L.M=1).

In your case, the drain of the multiple current sources are connected together. (Also the sources and the gates). This mean that the current mismatch would be averaged because of the random behavior of the mismatch.

Therfore:

1) DeltaI_tot= Square ((DeltaI1)^2 + (DeltaI2)^2 +(DeltaI3)^2+....+(DeltaIn)^2)
where DeltaI is with respect to Iref

2) I_tot= nxIref

3) DeltaI_tot/I_tot = square (n)/n . DeltaI/Iref
= (1/n).(A/square(Wref.Lref.M=1)).

So you can use the reference parameters for the mismatch calculation and reduce is by the square of the multiplier.

On the other hand, if the reference has a larger multiplier than the current source, than the mismatch will get worse by the same factor.

I hope i answered you question
Adil

aishrastogi

Joined: 09 Nov 2009
Posts: 4
Location: india

12 Nov 2009 19:24 Re: current mirror mismatch with different multiplier

could u please tell me how i can vary two DC voltages simultaneously in smash
i m trying to plot I-V characteristics of mosfet but how do i vary Vgs and Vds simultaneously.
if i vary just one of them n keep other fixed i get the output else i dont

" .DC Vgs 0 5 1 Vds 0 5 0.1 " doesnt work unlike in other spice softwares

your help would be appreciated
thanks

erikl

Joined: 09 Sep 2008
Posts: 736
Helped: 139
Location: Germany

13 Nov 2009 11:46 Re: current mirror mismatch with different multiplier

Hi aishrastogi,
this is a totally different topic. You should try and ask your question in a new thread with an appropriate title. And you'd better explain "smash".

Google
AdSense
Google Adsense

13 Nov 2009 11:46 Ads

timof

Joined: 21 Feb 2008
Posts: 86
Helped: 6

15 Nov 2009 8:15 Re: current mirror mismatch with different multiplier

alex.ang1983 wrote:

How do you define "a current mismatch between mirrors"? Are you interested in a variance (or standard deviation) of the difference of two currents in a current mirror, or of the current ratio?

For M2=1, the "mismatch" between two devices (i.e. standard deviation of current difference) is larger than a "mismatch" (stdev) of one device by sqrt(2.0)~1.4 (assuming there is no correlation between device characteristics).

For M2>1, this multiplier factor will be less than sqrt(2.) but larger than 1., approaching a value of 1.0 as M2--> infinity.

It's easy to derive rigorous formula, if you tell what parameter are you interested in (current difference or ratio).

There was a similar discussion here:

http://www.edaboard.com/viewtopic.php?p=1170466#1170466

Max
-----------