The Designer's Guide Community Forum

The Designer's Guide Community Forum
https://designers-guide.org/forum/YaBB.pl
Design >> Analog Design >> bandgap reference voltage
https://designers-guide.org/forum/YaBB.pl?num=1311356849

Message started by circuit_cook on Jul 22^nd, 2011, 10:47am

Title: bandgap reference voltage
Post by circuit_cook on Jul 22^nd, 2011, 10:47am

Hi, I have a question about bandgap voltage, I am porting a bandgap design which is a typical Brokaw bandgap, from 45nm to 40nm, but the bipolar model, (Tnf1, tnf2) changed a lot, and Vbe extraplated at absolute 0T shifted down 10mV from 45nm to 40nm in my simulation. The bandgap voltage output in 45nm is 1.23V, while after I adjust the R ratio (minimum change desired) to achieve zero TC in 40nm, I get 1.20V for 40nm.
But I was told both zero TC and the 1.23V is important to keep the same. Because the bandgap is used as reference in AD,etc.
Is that possible I can get the 1.23V just by adjusting the bandgap IPTAT current, to get the VBE TC like in 45nm? the current change would be too much, and it would require other opa change, since TC(vbe) =f(Vbe0), and Vbe0=Vtln(Ic).
Please give me some advice.

Thank you.
JS

Title: Re: bandgap reference voltage
Post by RobG on Jul 25^th, 2011, 9:37am

Knowing the Serenity prayer might help ;).

Personally, I don't trust models to that level, but assuming the model is correct you are pretty limited in what you can do to change the magnitude of the voltage at the 0 TC point.

In theory increasing the bipolar ptat current will do nothing. If you are seeing differences it is because of non-ideal effects such as the voltage drop across the base resistance. A good bandgap design will not rely on these non-ideal effects as they are usually not modeled well.

Another thing that will change the 0 TC voltage (and also the curvature) is the temp co of the resistor used to generate the "ptat" current.

Since you are coming up 30 mv short I think the best option is to gain it up. This was easy with the original "Brokaw" BG. a little less obvious with other designs.

rg

Title: Re: bandgap reference voltage
Post by harpoon on Jul 25^th, 2011, 4:01pm

A standard brokaw bandgap will have vout=vbe + (1+R1/R2)*Vt*ln(m), m=ratio of bipolars. Adjust your resistors to get the right voltage.

but seriously with 40nm or 45nm designs, you should really be thinking about using 0.6V bandgap references ... I presume you are using CMOS, in which case the current mirrors and headroom is going to be an issue.

there are several papers in JSSC that you can refer to ...

Title: Re: bandgap reference voltage
Post by RobG on Jul 25^th, 2011, 4:13pm

Harpoon, unfortunately, when you change the resistor ratio you also change the temp co.

Title: Re: bandgap reference voltage
Post by harpoon on Jul 25^th, 2011, 4:53pm

RobG ... well spotted ...

Title: Re: bandgap reference voltage
Post by circuit_cook on Jul 31^st, 2011, 7:51pm

Hi,
Thanks Robg and Harpoon for the reply, my understanding is my new bandgap output voltage is fundemantally 1.20v for zero TC.
I will try if I can use some offset in the next stage when I generate vbg/R current.
Thanks again.

JS.

Title: Re: bandgap reference voltage
Post by harpoon on Aug 1^st, 2011, 2:49am

hi circuit_cook,

do let us know how you came to that conclusion ...

... and does 30mV in bandgap voltage really make that much of a difference in 40nm ?

like you, I am trying to learn more about these type of circuits esp at sub-90nm

thanks.

Title: Re: bandgap reference voltage
Post by RobG on Aug 4^th, 2011, 9:56am

circuit_cook wrote on Jul 31^st, 2011, 7:51pm:

If you only need a current you should be able to change the resistor on the vbg to vbg/R converter. Alternatively, you could build a bandgap "current source." See the Bamba paper for a current source (they convert the current back to a voltage, but you get the idea). The nice thing about building a current source directly is that you can include the effect of resistor temp-co.

Good luck!

Rob