The Designer's Guide Community Forum

The Designer's Guide Community Forum
https://designers-guide.org/forum/YaBB.pl
Design >> Analog Design >> vdsat
https://designers-guide.org/forum/YaBB.pl?num=1223998167

Message started by rajdeep on Oct 14^th, 2008, 8:29am

Title: vdsat
Post by rajdeep on Oct 14^th, 2008, 8:29am

Hi all,

Can anyone please explain to me why large values of vdsat is required for the MOS transistors used
in a current mirror circuit? I heard it ensures strong inversion for the MOS transistors.
But then why strong inversion and is it the only reason?

Any good link or book's name will also be helpful. Please don't give me any link to a discussion thread.
That confused me more :-[

Rajdeep

Title: Re: vdsat
Post by PalmRunner on Oct 14^th, 2008, 9:12am

For me Vdsat is the required drain-source voltage at which Id enters the saturation region and the transistor starts to operate in strong inversion. When Id is saturated it depends only from Vgs and not from Vds, just like an ideal current source. Basically strong inversion is achieved by increasing the overdrive voltage (Vgs - Vt), which also ensures that random Vt variations would have less impact on Id. Therefore, better matching is achieved as well.

Regards!

Title: Re: vdsat
Post by Tlaloc on Oct 14^th, 2008, 8:58pm

The main reason to increase Vdsat of current mirrors is to minimize the effect of channel-length modulation, i.e. λ. If you make L longer, λ is decreased. That also tends to increase Vdsat. However, there is also an independent effect of λ from the bias condition, i.e. decreasing current through a device of fixed size decreases λ.

In short, maximizing Vdsat also maximizes the output resistance. Infinite output resistance makes ideal current sources which is the intent of current mirrors.

Title: Re: vdsat
Post by rajdeep on Oct 15^th, 2008, 7:19am

Hi PalmRunner & Tlaloc,

Thanks for your replies. PalmRunner's point is really something new!! But I'm not sure how much Vt variation ocurs after manufacturing. I mean if Vt variation is large then the circuit that is generating the Vgs will also undergo some changes, which will make Vgs to vary also. So, I'm not sure whether this is the reason. That is my doubt, I may be wrong. May be you were referring to some other vt variation, not due to manufacturing.

What is the dependency of lambda on vdsat? As far as I know vdsat = (vgs-vt)/1.2 in strong inversion. I agree increasing L reduces mismatch, but I do not agree that it also increases vdsat!!

I have heard of some other explanation., I

In strong inversion the mismatch problem is reduced!!

I wonder why??

If the above statement is accepeted then yes, we need to have large vdsat values, as it implies strong inversion.

But why strong inversion....

It seems to be accepatable becoz strong inversion means all the square law equations of MOS (that we typically use for hand calculation) hold more accurately, the more we go to weaker inversion, the models become less reliable. So the simulation results are no longer reliable :-X

Although in many other circuits where volage swing becomes important, we cannot afford to have such large vdsat values. For example, in low power supply cases. But we still maintain a reasonable vdsat value, >= 140mV!! i.e. we still try to maintain strong inversion whenever possible.

Is reliabilty of the simualtion models the only issue???

Rajdeep

Title: Re: vdsat
Post by Berti on Oct 15^th, 2008, 10:42pm

Hi all,

I recommend you to read the following paper:

Peter Kinget and Michiel Steyaert , Impact of transistor mismatch on the speed-accuracy-power trade-off of analog CMOS circuits ; IEEE Custom Integrated Circuits Conference ; 1996 .

Maybe this will help to clarify the confusions in this discussion.

Cheers

Title: Re: vdsat
Post by rajdeep on Oct 17^th, 2008, 7:50am

Hi Berti,

Thanks a lot. This should answer a lot of queries, nevermind give rise to some new ones :)

Thnx!