Designer’s Guide Community

raja.cedt Senior Fellow Offline Posts: 1516 Germany	CMFB bandwidth May 27^th, 2009, 11:08pm hi, Can any body please tell me how to decide cmfb loop UGB from amplifier specification?And is there any relation between DMFB UGB and this? Thanks, Rajasekhar.
Back to top	IP Logged

wave

Senior Member

Offline

Posts: 117
Silicon Valley

Re: CMFB bandwidth
Reply #1 - May 28^th, 2009, 6:31pm

raja.cedt wrote on May 27^th, 2009, 11:08pm:

hi,
� Can any body please tell me how to decide cmfb loop UGB from amplifier specification?And is there any relation between DMFB UGB and this?

Thanks,
Rajasekhar.
�

Absolutely. �

It depends what freqs you want to correct CM errors.
I've never really found a scenario of when I'd want CM correction BW to be Less than the DM signal BW.

For that reason, and there is usually an extra pole, CM loop stability is often the most challenging part of a diff'l circuit.

GHLM text, 4 or 5th edition has a nice write up on this.

Cheesy

Wave

Back to top

IP Logged

raja.cedt Senior Fellow Offline Posts: 1516 Germany	Re: CMFB bandwidth Reply #2 - May 28^th, 2009, 9:07pm hi, thanks for your reply,i agree with the explanation what you have given but in willy sansen he has given always cmbw>dmbw.But even in my designs also i usually put dmbw based on settling requirements and i didn't care about cmbw.Any how thanks. Thanks, Rajasekhar.
Back to top	IP Logged

aaron_do

Senior Fellow

Offline

Posts: 1398

Re: CMFB bandwidth
Reply #3 - May 28^th, 2009, 9:12pm

Hi,

please correct me if I'm wrong here but I believe that for amplifier design you usually have 2 CMFB loops.

1) The first loop is used to set the output common-mode voltage and I think this is the is the loop for which Rajasekhar wants to know the UGB. I think this loop is usually slow as wave pointed out due to the additional pole. The bandwidth will be related to and slower than the DMFB loop.

2) The other loop exists in fully differential amps due to the tail current in a differential pair. This provides common-mode source degeneration which reduces the common-mode gain. This CMFB will be much faster than the loop which sets the output common-mode voltage.

cheers,
Aaron

Back to top

there is no energy in matter other than that received from the environment - Nikola Tesla

IP Logged

raja.cedt Senior Fellow Offline Posts: 1516 Germany	Re: CMFB bandwidth Reply #4 - May 28^th, 2009, 9:30pm hi Aaron, i didn't understand how two loops will comes?Extra pole in the loop because of tail current source gate cap, and this also embedded in the common mode loop so how come two loops? Thanks, rajasekhar.
Back to top	IP Logged

aaron_do

Senior Fellow

Offline

Posts: 1398

Re: CMFB bandwidth
Reply #5 - May 28^th, 2009, 10:45pm

Hi rajasekhar,

I'm not sure if I understood your question correctly, but I assumed you were talking about an arrangement similar to the one attached. Let me just put the disclaimer out that I am by no means an expert in this so feel free to correct me.

For the outside loop on the left, the DM is negative feedback and the CM is positive feedback. Therefore the CM gain must be less than 1 over all frequencies to maintain stability. The opamp CM gain is made much less than 1 by the CMFB.

The CMFB consists of two separate loops. The first loop is shown explicitly with the CMFB amp. Both the dominant and nondominant poles of the DM loop are included (but not necessarily exactly the same) in this CM loop plus an additional pole due to the CMFB amp. So this loop is slow. It is primarily used to set the output CM level to a desired value.

The second loop is not explicitly shown but consists of the source degenerated tail current of the input differential pair. This loop is fast.

cheers,
Aaron

Back to top

DMCMFB.jpg

there is no energy in matter other than that received from the environment - Nikola Tesla

IP Logged

Berti

Community Fellow

Offline

Posts: 356

Re: CMFB bandwidth
Reply #6 - May 29^th, 2009, 12:01am

Hi everybody,

Interesting discussion, but I don't fully agree.

I think that Rajasekhar was referring to the amplifier only (open-loop, no feedback yet). In this case I agree with Wave that the common-mode bandwidth should be similar to the differential-mode bandwidth. However, this is often difficult to achieve or results in high power consumption. Therefore in my designs the common-mode bandwidth is typically somewhat lower than the differential-mode bandwidth.

How the common-mode transfer function actually looks like (the poles Aaron mentioned) depends on the implementation of the common-feedback circuit (SC, resistive sensing, DDA etc.). I think that this can be decided before the actual implementation is known.

Finally I think that the CMFB has an influence on the CMRR and PSRR of the amplifier. How much common-mode bandwidth you will trade for lower power consumption will therefore depend on this requirements.

Any comments are welcome!

Cheers

Back to top

IP Logged

raja.cedt

Senior Fellow

Offline

Posts: 1516
Germany

Re: CMFB bandwidth
Reply #7 - May 29^th, 2009, 1:09am

hi berti,
i am talking about the amp which was already in feedback,what i am asking is at least before going to implement CMFB roughly how to decide CMBW?From any spec of the opamp its ok.Fortunately in my designs by default CMFB is having lesser bandwidth (architecture which i am working is pmos triode transistors at the tail current source).

Thanks,
Rajasekhar.

Back to top

IP Logged

aaron_do

Senior Fellow

Offline

Posts: 1398

Re: CMFB bandwidth
Reply #8 - May 29^th, 2009, 1:13am

Hi,

in order to reject in-band common-mode noise I guess it makes sense to have a CM bandwidth at least equal to the DM bandwidth. The noise can come from "common-mode circuits" such as tail currents etc.

Like I was saying before, however, I think you can achieve quite wideband CM rejection due to the tail current. The CMFB loop which consists of the amplifier in feedback is mainly to set the common-mode output level.

Does that sound right?

cheers,
Aaron

Back to top

there is no energy in matter other than that received from the environment - Nikola Tesla

IP Logged

raja.cedt Senior Fellow Offline Posts: 1516 Germany	Re: CMFB bandwidth Reply #9 - May 29^th, 2009, 2:05am hi, how would tail current source contribute noise at the output? Thanks, Rajasekhar.
Back to top	IP Logged

Berti

Community Fellow

Offline

Posts: 356

Re: CMFB bandwidth
Reply #10 - May 29^th, 2009, 2:13am

Quote:

Like I was saying before, however, I think you can achieve quite wideband CM rejection due to the tail current. The CMFB loop which consists of the amplifier in feedback is mainly to set the common-mode output level.

That sounds right to me. If you had a perfect tail current source you wouldn't care much about common-mode variations at the output since the common-mode-to-differential-mode gain is very small.

Cheers

Back to top

IP Logged

aaron_do Senior Fellow Offline Posts: 1398	Re: CMFB bandwidth Reply #11 - May 29^th, 2009, 6:39am Quote: how would tail current source contribute noise at the output? I was thinking that the common-mode noise at the input which comes from the previous stage could come from the tail current of the previous stage... Aaron
Back to top	there is no energy in matter other than that received from the environment - Nikola Tesla IP Logged

HdrChopper

Community Fellow

Offline

Posts: 493

Re: CMFB bandwidth
Reply #12 - May 29^th, 2009, 7:48pm

Hi,

I think CMFB BW has to be decided based on the desired CMRR and PSRR for the opamp as Berti suggested before. This selection has somewhat to do with the DMFB BW since in principle one would look for good CMRR and PSRR inside the DM signal bandwidth.

The tail current - as Aaron suggested - it might be come a noise source as a function of the differential unbalance the opamp might have. The more offset the amplifier has the more poorer CMRR and therefore more output CM variations will get translated into differential signal.

I think it is not totally true when you state "you can achieve quite wideband CM rejection due to the tail current". The tail current souce introduces a zero in the CM response, such zero being a function of the output impendace and capacitance of the current source. IF the output impedance / capacitance of such current source is large enough then such zero moves to lower frequencies. Above the zero frequency the CMRR starts decreasing at 20 dB/dec, degrading the amplifiers rejection capability to CM signals.

Tosei

Back to top

Keep it simple

IP Logged

LazyDesigner Junior Member Offline Posts: 15 South Korea	Re: CMFB bandwidth Reply #13 - Jan 15^th, 2015, 4:52pm Hi Based on above discussion, can I say that if above mentioned differential amplifier process a signal of BW 500MHz and differential mode path UGB is ~5GHz, then can I have BW of CMFB loop ~10MHz if CMFB loop settling time is not a problem? Regards, Pravesh
Back to top	IP Logged