The Designer's Guide Community
Forum
Welcome, Guest. Please Login or Register. Please follow the Forum guidelines.
Mar 28th, 2024, 4:30pm
Pages: 1
Send Topic Print
PSRR and CMRR for fully differential amp (Read 104 times)
Winson
Guest




PSRR and CMRR for fully differential amp
Jan 05th, 2004, 8:52am
 
hi,

Does anyone know how to setup up for simulations measurements for fully differential op amp?

to measure:
1.PSRR
2.CMRR

Thanks.
Back to top
 
 
  IP Logged
Mighty Mouse
Community Member
***
Offline

Here I come to save
the day!

Posts: 75
Fantasyland
Re: PSRR and CMRR for fully differential amp
Reply #1 - Jan 5th, 2004, 9:08am
 
Winson,
   You would use the test bench described in A test bench for differential circuits, which can be found at http://www.designers-guide.com/Analysis/. Then you would measure PSRR and CMRR in the same way you would for a single-ended amplifier. In particular, if you were using Spectre, you would run an XF analysis with the differential output signal being identified to the XF analysis as the output of the circuit. Then you would measure the gain from the differential input to the output (Adm), from the common-mode input to the output (Acm), and from the power supply to the output (Aps). They all can be measured in one XF analysis. Then, if I remember correctly,

PSRR = Aps/Adm
CMRR = Acm/Adm

- MM -
Back to top
 
 
View Profile   IP Logged
wstyeah
New Member
*
Offline



Posts: 3

Re: PSRR and CMRR for fully differential amp
Reply #2 - Jan 5th, 2004, 7:21pm
 
hi mighty mouse..

thank you for your replied.  ;)
but what i am thinking is.. the PSRR of fully differential is
PSRR= (Adm)/(Asp - Asn) ?
same for CMRR,
CMRR = (Adm)/(Acmp -Acmn) ?

Asp is supply gian for positive output terminal and so on.

Back to top
 
 
View Profile   IP Logged
Mighty Mouse
Community Member
***
Offline

Here I come to save
the day!

Posts: 75
Fantasyland
Re: PSRR and CMRR for fully differential amp
Reply #3 - Jan 6th, 2004, 12:18am
 
The beauty of this test bench is that it completely separates the differential and common mode signals at both the input and the output, so there is no need reconstruct the desired signal by addition or subtraction. This approach eliminates a lot of confusion.

Having said that, I believe your definition is the reciprocal of  mine, which I believe is wrong.

- MM -
Back to top
 
 
View Profile   IP Logged
wstyeah
New Member
*
Offline



Posts: 3

Re: PSRR and CMRR for fully differential amp
Reply #4 - Jan 7th, 2004, 7:34am
 
hi mighty mouse..

thanks again  for replying  :)

erm.. do you mind to draw out the configuration for me?
Thanks first.


Warmest Regards,
Winson
Back to top
 
 
View Profile   IP Logged
Mighty Mouse
Community Member
***
Offline

Here I come to save
the day!

Posts: 75
Fantasyland
Re: PSRR and CMRR for fully differential amp
Reply #5 - Jan 7th, 2004, 10:31am
 
Winson,
    You would use the test bench as shown in Figure 3 (or 5) in A Test Bench for Differential Circuits (though you probably do not need the resistors). This requires that you get your hands on an ideal balun. You can find one in analog lib in the more recent versions of Artist. If you cannot find it, you can create one as described in the paper (use the Spectre version rather than the Spice version if you can).  Finally, if you are using Spectre, perform an xf analysis with the differential output signal (d from the output balun) passed into the xf analysis to define the output for the analysis. The xf analysis will directly compute Adm, Acm, and Aps as the transfer function from Vcm, Vcm, and Vdd to the declared output.

- MM -
Back to top
 
 
View Profile   IP Logged
jeffyan
Community Member
***
Offline



Posts: 47

Re: PSRR and CMRR for fully differential amp
Reply #6 - Feb 13th, 2004, 10:51pm
 
For fully differential op amp, to get  the right results of PSRR,CMRR, i think, the mismatchings should be added to  differential pairs,right??

according to the opinion of  Mr.Mouse, PSRR and CMRR
of  fully differential amp are identical to single-ended amp's respectively,right? ???
Back to top
 
 
View Profile   IP Logged
Mighty Mouse
Community Member
***
Offline

Here I come to save
the day!

Posts: 75
Fantasyland
Re: PSRR and CMRR for fully differential amp
Reply #7 - Feb 14th, 2004, 3:49pm
 
Yes, generally you need to model mismatch in differential circuits to get any meaningful measurements for PSRR and CMRR.

If you use the balun to separate out the differential and common-mode signals, then the measurement prodedure for measuring differential and common-mode PSRR and CMRR are the same (but of course the results will be different).

- MM -
Back to top
 
 
View Profile   IP Logged
erikwanta
New Member
*
Offline



Posts: 8

Re: PSRR and CMRR for fully differential amp
Reply #8 - Feb 14th, 2004, 11:25pm
 
Run spectre XF analysis and plot the following:

CMRR
db20(1/DATA("/Vic/PLUS" "xf-xf"))

PSRR-
db20(1/DATA("/Vn/PLUS" "xf-xf"))

PSRR+
db20(1/DATA("/Vp/MINUS" "xf-xf"))

Where Vic is the input common mode voltage, Vn is the negative supply, and Vp is the positive supply.
Back to top
 
 
View Profile   IP Logged
jeffyan
Community Member
***
Offline



Posts: 47

Re: PSRR and CMRR for fully differential amp
Reply #9 - Feb 14th, 2004, 11:31pm
 
i don't know whether it is right:
for a differential amp op,the value of CMRR or PSRR, is infinite if the circuit is symmetrical perfectly(if no mismatching added)! because we are concerning the differential signal.
so i think the definition of PSRR&CMRR  are PSRR= (Adm)/(Asp - Asn) and  CMRR = (Adm)/(Acmp - Acmn), as same as Mr wstyeah's.
but they are different  to the results we get from  the new test bench.

need mighty mouse's help!
best regards
jeff.
Back to top
 
 
View Profile   IP Logged
Mighty Mouse
Community Member
***
Offline

Here I come to save
the day!

Posts: 75
Fantasyland
Re: PSRR and CMRR for fully differential amp
Reply #10 - Feb 15th, 2004, 11:59am
 
Jeff,
You are correct, the rejection becomes infinite as the undesired gain (either the common mode gain or the gain from the supplies to the output) goes to zero. Therefore
PSRR+ = Adm / Add
PSRR- = Adm / Ass
CMRR = Adm / Acm

So what this says is that there are two PSRRs, one for the positive supply and the other for the negative supply. You would not combine them unless you had reason to believe that the interferrence from the supplies would be balanced relative to ground, which seems unlikely.

Referring to Figure 5 in http://www.designers-guide.com/Analysis/diff.pdf, if an XP analysis were performed with the output being Vod, then these metrics are calculated using
PSRR+ = Vid / Vdd
PSRR- = Vid / Vss
CMRR = Vid / Vic
(Here I am assuming that Vdd and Vss represent the positive and negative power supplies and that Vxx represents the gain from source xx to the output).

Erik was on the right track with his formulas, but he gave the gain from the supplies to the output, not the rejection ratio.

- MM -
Back to top
 
 
View Profile   IP Logged
richloo
New Member
*
Offline



Posts: 9

Re: PSRR and CMRR for fully differential amp
Reply #11 - Feb 6th, 2005, 11:11pm
 
[quote author=erikwanta  link=1073321544/0#8 date=1076829932] Run spectre XF analysis and plot the following:

CMRR
db20(1/DATA("/Vic/PLUS" "xf-xf"))

PSRR-
db20(1/DATA("/Vn/PLUS" "xf-xf"))

PSRR+
db20(1/DATA("/Vp/MINUS" "xf-xf"))

Where Vic is the input common mode voltage, Vn is the negative supply, and Vp is the positive supply. [/quote]

I've tried the analysis sweep until GHz, but i only get a straight line with -6.4K dB. I wonder this is a setup problem.
can u teach me how to setup this analysis?

Thanks in advance!
Back to top
 
 
View Profile   IP Logged
lurx
New Member
*
Offline



Posts: 1

Re: PSRR and CMRR for fully differential amp
Reply #12 - Sep 4th, 2008, 9:57pm
 
it should be
PSRR+ = Vdd/Vid
PSRR- = Vss/Vid
CMRR =  Vic/Vid
Back to top
 
 
View Profile   IP Logged
Pages: 1
Send Topic Print
Copyright 2002-2024 Designer’s Guide Consulting, Inc. Designer’s Guide® is a registered trademark of Designer’s Guide Consulting, Inc. All rights reserved. Send comments or questions to editor@designers-guide.org. Consider submitting a paper or model.