The Designer's Guide Community Forum

The Designer's Guide Community Forum
https://designers-guide.org/forum/YaBB.pl
Design >> Analog Design >> how to minimise tail current 1/f noise in differential opamp.
https://designers-guide.org/forum/YaBB.pl?num=1440627952

Message started by yvkrishna on Aug 26^th, 2015, 3:25pm

Title: how to minimise tail current 1/f noise in differential opamp.
Post by yvkrishna on Aug 26^th, 2015, 3:25pm

Hi,

I am designing an opamp with sine inputs only and interested in differential output noise (pnoise).

I notice that pnoise gives most of noise contributors from tail current source and its associated diodes+mirrors.

since the diff pair is not balanced i.e, it has some finite swing at the inputs we expect this to happen (like mixers)

I tried to res. degenerate the tail current and was successful in improving the total ouput noise.

beyond a point where headroom limits the amount of degeneration usable, what is the best that can be done from the input pair sizing/gm choice ?

Also another observation is that the pnoise ouput of amp reduces with increasing the total tail current, which is not very clear as total noise from tail has only increased here.. ??

Please help me in resolving this .. please point me any references if any.

Thanks,
yvkrishna

Title: Re: how to minimise tail current 1/f noise in differential opamp.
Post by loose-electron on Aug 27^th, 2015, 11:52pm

chopper methods to move the 1/F noise to an out of band frequency?

Title: Re: how to minimise tail current 1/f noise in differential opamp.
Post by yvkrishna on Aug 28^th, 2015, 1:19am

Thanks for the reply loose-electron,

Well this is not a kind of application where I can use chopper.

I am only trying to understand the strategy for sizing input pair and amount of tail current to use here for minimising this tail noise upconversion to outputs.

Also not sure why I see more upconversion with larger input Frequency for the input pair.

Thanks,
yvkrishna

Title: Re: how to minimise tail current 1/f noise in differential opamp.
Post by RobG on Aug 31^st, 2015, 4:54pm

yvkrishna wrote on Aug 28^th, 2015, 1:19am:

I've never tackled that particular mixer-type problem, but I have done mirrors. It sounds like you are doing what I would do. As you probably know 1/f is reduced with larger areas devices, but you can't just make the devices wider because the gm goes up as fast as it goes down. You can make it longer to degenerate but you lose headroom. I played with this one time and found that it was best to use a resistor to degenerate the mirror and make the devices as wide/short as possible. Basically the MOS will be in subthreshold and the resistor drop will be as large as headroom allows.

If the noise is coming from the diode connected device you need a larger device, meaning more current in that device. So it boils down to headroom and power.

Usually this is a problem only when mismatches are considered because in an opamp the differential input is very small so noise is steered equally, thus it's common mode. Is this an opamp or are you running a diff pair open loop?

Title: Re: how to minimise tail current 1/f noise in differential opamp.
Post by yvkrishna on Sep 2^nd, 2015, 4:43am

Thanks for sharing your thoughts on this RobG,

I agree regarding that headroom choices for resistor and mosfet.

This is an opamp (not open loop ) even without mismatches the differential input is in orders of 50mV due to huge swing requirements at output. (with mismatches its expected to be worser here)

Thanks,
yvkrishna

Title: Re: how to minimise tail current 1/f noise in differential opamp.
Post by loose-electron on Sep 2^nd, 2015, 1:03pm

If you just want to get rid of the current source and mirror noise get rid of the current sources and use a resistor instead.

The circuit will start to look like a design from 1965 (a couple of discrete transistors plus R, C) but it will be quieter.

Title: Re: how to minimise tail current 1/f noise in differential opamp.
Post by RobG on Sep 2^nd, 2015, 1:44pm

loose-electron wrote on Sep 2^nd, 2015, 1:03pm:

If you just want to get rid of the current source and mirror noise get rid of the current sources and use a resistor instead.

I was thinking the same thing, but don't know how the 50mV swing will distort things, or if there is input common mode noise which would get converted to current.

More opamp gain would help reduce the input swing, which is root of the problem. You could use two opamps and break up the gain over two stages.