Designer’s Guide Community

ACWWong

Community Fellow

Offline

Posts: 539
Oxford, UK

Flicker Noise contribution to DSM switches
May 29^th, 2009, 5:49am

Hi,

I have a question regarding 1/f noise in switches in a delta-sigma modulator (or a SC LPF).
In the paper by Terrovitis and Kundert on www.designer-guide.org, it states (on page 9)
"Flicker noise of the switches is not an issue because the transistors used as switches operate as resistors in the triode region". I am confused by this statement since in my PNOISE simulation of the first integrator in a delta-sigma modulator my maximum noise contributors is 1/f noise (1-100Hz signal bandwidth) from the gnded switch and floating switch connect to the + input of the opamp (refer to figure 8 in the paper).

A few more details about my circuit might help generate comments. I am using a 0.13um process to build a fully differential switch-capacitor switch-opamp delta sigma-modulator (DSM). The DSM is a 3rd order 1-bit distributed feedback structure. My signal bandwidth is 1-125Hz and my oversampling clock is 32kHz, Vdd=1V and total power consumption is about 30-35uA.

I have also build a fully differential SC biquad low pass filter (3dB freq = 125Hz) and the same switches connected to the 1st opamp in the filter are the major noise contributor (1/f noise) as quoted in the noise summary. Hopefully someone has some comments.

Thanks.
aw

Back to top

IP Logged

vivkr

Community Fellow

Offline

Posts: 780

Re: Flicker Noise contribution to DSM switches
Reply #1 - Jun 2^nd, 2009, 12:35am

Hi Alan,

I have not read this paper but I would not expect any 1/f noise contribution from switches, primarily because a basic requirement for the presence of 1/f noise is that there be a DC current flow through the device in question.

As switches do not fulfil this requirement, there ought to be no issue here. In all my experience with PNOISE simulations of switched-circuits, I have never seen the 1/f noise of the switch pop up in the noise summary.

Perhaps there is some error either in the simulator or in the manner in which you are using the switch.

Best regards,

Vivek

Back to top

IP Logged

Berti

Community Fellow

Offline

Posts: 356

Re: Flicker Noise contribution to DSM switches
Reply #2 - Jun 2^nd, 2009, 6:10am

Quote:

...primarily because a basic requirement for the presence of 1/f noise is that there be a DC current flow through the device in question.

Is that right???

I think the main reason flicker noise of the switches is often neglected is because the flicker noise corner is very low due to down-folding of white noise.

Regards

Back to top

IP Logged

ACWWong

Community Fellow

Offline

Posts: 539
Oxford, UK

Re: Flicker Noise contribution to DSM switches
Reply #3 - Jun 2^nd, 2009, 9:04am

Hi Vivkr & Berti,

Thanks for the discussion.

I'm not really used to looking at very low frequency bandwidths... passive mixers i've looked at in the haven't had switching device flicker noise anywhere in the top 20 pnoise contirbutors... so i've generally believed no DC -> no flicker noise. Having said that a colleague just sent me the attached paper....
http://www.imec.be/esscirc/esscirc2001/C01_Presentations/16.pdf

Anyway just wondering if people have used SC circuits on signal bandwidths of 1 to 100 Hz.... is it normal to see the switches fn to be a major noise contributor ?? (the switches seem to switching fine without any DC current, and the ADC is only about 8~9 bit)

cheers
aw

Back to top

IP Logged

HdrChopper

Community Fellow

Offline

Posts: 493

Re: Flicker Noise contribution to DSM switches
Reply #4 - Jun 2^nd, 2009, 6:27pm

Berti wrote on Jun 2^nd, 2009, 6:10am:

Is that right???

I think the main reason flicker noise of the switches is often neglected is because the flicker noise corner is very low due to down-folding of white noise.

Hi Berti,

I think not always the WNoise down-folding is strong enough in order to bring the flicker noise corner to very low frequencies, since that exclusively depends on the sampling capacitor size (folded Wnoise power will be KT/C). So if C is large the flicker corner frequency might not be that low.
In this reference (Design-Oriented Estimation of Thermal Noise in
Switched-Capacitor Circuits, R. Schreier, J. Silva, J. Steensgaard, G. C. Temes, IEEE TRANSACTIONS ON CIRCUITS AND SYSTEMS�I: REGULAR PAPERS, VOL. 52, NO. 11, NOVEMBER 2005) flicker noise is neglected "Since the current flow in these switches consists of short pulses occurring at the clock rate, the 1/f noise (which is caused by the trapping and release of charge carriers occurring at long intervals) has only a negligible effect here".
This somehow confirms the fact that no DC current means negligible 1/f noise.

Hope this helps
Regards
Tosei

Back to top

Keep it simple

IP Logged

vivkr

Community Fellow

Offline

Posts: 780

Re: Flicker Noise contribution to DSM switches
Reply #5 - Jun 2^nd, 2009, 11:59pm

ACWWong wrote on Jun 2^nd, 2009, 9:04am:

Hi Vivkr & Berti,

Thanks for the discussion.

I'm not really used to looking at very low frequency bandwidths... passive mixers i've looked at in the haven't had switching device flicker noise anywhere in the top 20 pnoise contirbutors... so i've generally believed no DC -> no flicker noise. Having said that a colleague just sent me the attached paper....
http://www.imec.be/esscirc/esscirc2001/C01_Presentations/16.pdf

Anyway just wondering if people have used SC circuits on signal bandwidths of 1 to 100 Hz.... is it normal to see the switches fn to be a major noise contributor ?? (the switches seem to switching fine without any DC current, and the ADC is only about 8~9 bit)

cheers
aw

Hi Alan,

Indeed, switched-capacitor circuits are overwhelmingly used for lower speed applications than continuous-time circuits, and this because they cannot operate as fast.
The band from 1-100 Hz is probably entirely dominated by switched-cap circuits. I have made upto 20 bit accurate switched-cap circuits for low-frequencies (< 200 Hz), and never have I seen 1/f noise from switches popping up in the noise contributors.

But I will go through this presentation.

Berti: The classical model for 1/f noise requires that there is DC current flow. As a matter of fact, the formula for the output current noise (1/f) is something like

Inoise_1/f = K_process*Idc^alpha_idc/f^alpha_freq

where the default for alpha_idc and alpha_freq is 1. So if there is no DC current, then this noise should go to zero. Note that no textbook says what happens with AC current flow, or if there is a time-averaged DC current flow (such as when a switched-capacitor circuit is simply charging and discharging itself between 2 fixed states).

There was also some interesting work in the research group lead by Eric Klumpernik (Google for him) where they found out that turning transistors completely OFF and then turning them ON periodically reduces 1/f noise much more than what would be expected by mere duty cycle reduction.

As 1/f noise is still not understood fully, one needs to be a bit careful with it, but so far I have never experienced it as coming from the switches, atleast not so far.

Best regards,

Vivek

Back to top

IP Logged