The Designer's Guide Community Forum

The Designer's Guide Community Forum
https://designers-guide.org/forum/YaBB.pl
Design >> Analog Design >> Bandgap reference question
https://designers-guide.org/forum/YaBB.pl?num=1142391492

Message started by 4analog on Mar 14^th, 2006, 6:58pm

Title: Bandgap reference question
Post by 4analog on Mar 14^th, 2006, 6:58pm

Hi,
I have a question on the bandgap refence. I am refering to the the bandgap reference where the Opamp is used to match the two nodes X & Y and bias the current sources that set the current in both the legs , where v(X) = VBE1 and V(Y) = dVBE + VBE2.

(Razavi, Pg:391, Fig:11.18)

The current is mirrored to generate the PTAT current and then the Vref is obtained by using another Resistor ( R3) and BJT #3, such that Vref = Iptat* R3 + VBE3 .

My question is, why cant we just use the node voltage V(y) as Vref instead of mirroring the PTAT current and using another resitor,BJT ..?

If we can tap the voltage V(Y), then what about the voltage V(X)..? SInce the Opamp forces, both X and Y to be the same, is V(X) also a bandgap voltage..?

Thanks.

Title: Re: Bandgap reference question
Post by vivkr on Mar 15^th, 2006, 12:37am

Hi,

You can do this. See Thomas H Lee's book (Chapter on References), Fig. 9.9 (CMOS Bandgap Reference)

Regards
Vivek

Title: Re: Bandgap reference question
Post by chase.ng on Mar 15^th, 2006, 11:22am

Hi,

Interesting question. After some calculations, the I realized that it is not possible to get a bandgap voltage at Y. The voltage at Y is VY = VBE + VTln(A2/A1). To get 0 TC at 300k, assume a -2mV/K change in VBE, the ratio A2/A1 will be exp(23.2), which is a whopping 1.2e10. So i guess it is not possible anyway to have bandgap voltage at Y. Did I make any mistakess?

Regards,
Chase

Title: Re: Bandgap reference question
Post by 4analog on Mar 15^th, 2006, 11:54am

Lets say, I am not worried about how big the size of the resistor is ...

My question is more interms of the concept....

Let say we can place a big resisitor such that V(y) = VBE2 + dVBE. If we can use V(y) as a bandgap voltage, what about the node V(X)..?? Sicne the Opamp forces these node to be equal, can I also assume that V(X) is also a bandgap voltage..??

What about there temp coeff...?? V(X) = VBE1 whereas V(Y) = dVBE + VBE2...

What will be the temp coeff of V(x)...?? CTAT..?? or constant..?

Thanks

Title: Re: Bandgap reference question
Post by chase.ng on Mar 15^th, 2006, 12:19pm

Hi,

Just want to clarify that A2/A1 is not the ratio of a resistor but the ratio of BJT ratio where n=A2/A1 in Razavi book, A1 is always 1. As for Vx, I am sorry that I guess I do not have the answer. Maybe someone else will be able to help.

Chase

Title: Re: Bandgap reference question
Post by river on Mar 15^th, 2006, 6:02pm

both Vx and Vy are temperature independent.
Vx = Vbe1 = Vt * ln (Ic/Iss)
Because of feedback, Ic is not constant.
It will compensate temp variation of Vt, thus make Vx constant.

Title: Re: Bandgap reference question
Post by RobG on Mar 17^th, 2006, 10:18pm

Chase is correct... Even though v(y) is Vbe + dVbe, the dVbe term cannot be made large enough for v(y) to become a "bandgap." The current ratio required is something like 1e9:1.

Title: Re: Bandgap reference question
Post by qltong on Mar 18^th, 2006, 11:42pm

Vy=Vbe+dVbe=Vbe+Vtln(A2/A1), A2/A1 must be very large for the compensation and voltage on Ry is constant. Usually Vy is not irrelevant with temperature.
When using R3,Vz=Vbe+R3*dVbe/Ry. We can choose R3/Ry large enough to make the compensation.
If you want to use Vy, add a resistor R3 below the two BJTs.

Title: Re: Bandgap reference question
Post by 4analog on Mar 19^th, 2006, 6:08pm

Thank you all for actively participating in this discussion.

I am mainly interested in the temp coefficient of V(X) vs V(y) if they are forced to be equal by the opamp.

I am not worried about the huge ratio of A2/A1 required to make V(y) a bandgap voltage. Lets assume (only for this discussion) that in some way, ( ex: adding R3 below the BJT, as mentioned by qltong) we can acheive V(y) as the bandgap voltage.

Now, since the nodes v(X) and V(y) are forced to be the same, will there temp coeff also be the same..?
The reply by posted by "river", suggesting that temp coeff of v(x) may be the same as V(y) , as Ic is PTAT, sounds reasonable, but I am still not sure..

Thanks

Title: Re: Bandgap reference question
Post by RobG on Mar 19^th, 2006, 9:39pm

Well, obviously, if they are forced to be equal by the op-amp they are "equal" in all ways including temp co. If you can find a way to make V(y) constant over temperature, such as adding a PTAT signal to the base, then you also have v(x) constant.

Title: Re: Bandgap reference question
Post by 4analog on Mar 23^rd, 2006, 2:46pm

chase,RobG and Qltong: You were right... :)

V(Y) cannot be a bandgap reference because of the missing factor for dVBE to make it large enough to acheive the bandgap voltage. Hence, V(y) and so does v(x) are not badgap references though they are equal. There temp coeffs are the same, but not zero...

thanks