The Designer's Guide Community Forum

The Designer's Guide Community Forum
https://designers-guide.org/forum/YaBB.pl
Modeling >> Semiconductor Devices >> Noise model (white noise) of MOS transistor in BSIM3
https://designers-guide.org/forum/YaBB.pl?num=1245760480

Message started by vivkr on Jun 23^rd, 2009, 5:34am

Title: Noise model (white noise) of MOS transistor in BSIM3
Post by vivkr on Jun 23^rd, 2009, 5:34am

Hi All,

The textbook formula for thermal noise generated in a MOS channel in strong inversion is

in^2 = 4kT*(2/3)Gm*df, which when referred to the gate of the MOS becomes
vn^2 = 4kT*(2/3)/Gm*df

However, running a reality check with a MOS transistor shows that the factor in the above expression should be more like 0.85 instead of 0.67 (2/3).

Where does the difference come from? Is there a universal formula with a constant factor like 0.85 implemented in BSIM3 which is independent of W, L, biasing conditions (assuming strong inversion still) etc.? Or is the difference due to other factors. Hints are most welcome.

Best regards,

Vivek

Title: Re: Noise model (white noise) of MOS transistor in BSIM3
Post by pancho_hideboo on Jun 23^rd, 2009, 5:51am

If you choose Thermal Noise Model=SPICE2, following equation is used as channel thermal noise.
8*k*T/3*(gm)

However if you choose Thermal Noise Model=BSIM3v3, a different equation is used as channel thermal noise.
According to chapter-8 in BSIM3v3 manual from UCB, channel thermal noise is expressed as a following equation.
8*k*T/3*(gm+gds+gmb)
Here gm=transconductance, gds=channel conductance, gmb=body transconductance.

Do you include effects of gds and gmb ?

In channel thermal noise of bsim3v3 for noimod=1 and 3 of Cadence Spectre, equation is little modified like following.
4*k*T*gamma*(gm+gds+gmb)

where gamma is a new model parameter introduced to model short channel device thermal noise.
The default value for gamma is 2/3.

Is there "gamma" in your bsim3vs model file ? If so, "gamma" is 2/3 ?

For noimod=2 or 4, completely different equations are used as thermal noise.
Your noimod is 1 or 3 ?

Title: Re: Noise model (white noise) of MOS transistor in BSIM3
Post by Frank Wiedmann on Jun 23^rd, 2009, 6:34am

You can find the official BSIM3v3 documentation at http://www-device.eecs.berkeley.edu/~bsim3/get.html. The noise model is described in chapter 8 of the users' manual.

You could also take a look at the documentation of your circuit simulator. In many cases, it also contains information about the available models and their equations.

Title: Re: Noise model (white noise) of MOS transistor in BSIM3
Post by raja.cedt on Jun 23^rd, 2009, 6:45am

hi Vivek,
1.I read some where about this, at some point of the channel length gamma has minimum and for short channel gamma is 1 in some corners.
2.As you said the expression for input referred noise is valid only in long channel (that to saturation), but in short channel there is another expression for week inversion and i am not sure there is some Rg/5 component is also there. I got all these stuff from the following paper.
A. J. Scholten et al., "Noise modeling for RF CMOS circuit simulation," IEEE
Trans. Electron Devices, pp. 618-632, Mar. 2003.
Hope this may help little

thanks,
rajasekhar.

Title: Re: Noise model (white noise) of MOS transistor in BSIM3
Post by vivkr on Jun 24^th, 2009, 1:02am

Dear All,

Thanks for your comments.

Pancho: Thanks for the tip about gmbs and gds. I had forgotten those. gds is not a big issue, but gmbs is something I expected to have as 0 since I had shorted bulk and source together. As the DC operating point summary does not show gmbs, I did an XF analysis to estimate gmbs and gm, and very surprisingly, gmbs was nonzero. As a matter of fact, it was large enough to explain exactly the increased noise level. So, the issue seems to be resolved. I wonder why gmbs is nonzero though.

EDIT: A little thought to the matter suggests that gmbs does not need to be zero for zero bulk-source voltage, since this is basically a pn-junction diode, which will still respond to any AC signal (were there one present).

noimod=3 and I am using Cadence spectre so the SPICE2 model is used.

Best regards,

Vivek

Title: Re: Noise model (white noise) of MOS transistor in BSIM3
Post by pancho_hideboo on Jun 24^th, 2009, 5:17am

Channel Thermal Noise in Cadence Spectre.

Title: Re: Noise model (white noise) of MOS transistor in BSIM3
Post by vivkr on Jun 24^th, 2009, 11:46pm

Just one question comes to my mind.

Why do we need to add gds to the noise model? gds is after all a small-signal resistance and not a physical resistance which would generate thermal noise. Similarly, how about gmbs? I could imagine that this is somehow more logical since the channel resistance is impacted by both the gate (i.e. gm) and by the backgate (i.e. gmbs). But it would be nice to know a physical explanation, especially for gds being in there.

Regards,

Vivek

Title: Re: Noise model (white noise) of MOS transistor in BSIM3
Post by raja.cedt on Jun 25^th, 2009, 1:18am

hi vivkr,
even i don't know why gds will come into the picture but not only for short channel even for long channel transistors also gds will come into picture when it is in triode region...

thanks,
Rajasekhar.

Title: Re: Noise model (white noise) of MOS transistor in BSIM3
Post by Berti on Jun 26^th, 2009, 7:04am

gds is there to make the equation valid in saturation and ohmic region.
(in saturation gds is negligible, while in the deep ohmic region gds dominates)

Regards

Title: Re: Noise model (white noise) of MOS transistor in BSIM3
Post by vivkr on Jun 28^th, 2009, 11:08pm

Berti wrote on Jun 26^th, 2009, 7:04am:

gds is there to make the equation valid in saturation and ohmic region.
(in saturation gds is negligible, while in the deep ohmic region gds dominates)

Regards

Ah yes of course! Forgot completely that one also needs the linear region. Thanks for pointing out Berti.

Regards,

Vivek

Title: Re: Noise model (white noise) of MOS transistor in BSIM3
Post by Asmodeus on Jan 18^th, 2010, 11:39pm

As far as i recall, the thermal noise power were derived for transistors operating in the triode region only when the channel was continuous. These equations are not very valid in saturation region.

Gamma is dependent on the channel length modulation & effective mobility. In earlier technology it was kind of fixed to 2/3 but now its definitely more.

Title: Re: Noise model (white noise) of MOS transistor in BSIM3
Post by aLittleKnowledge on Jul 13^th, 2010, 4:08am

[quote author=vivkr link=1245760480/0#4 date=1245830549]Dear All,

"I wonder why gmbs is nonzero though."

The channel charge is determined by field effects. If you change the potential of the bulk the electric field to the channel must change, and with it the channel charge.
This is the reason for gamma in the earliest MOSFET models - no need for BSIM3 subtleties.