How to improve the bandwidth of this PTAT circuit by frequency compensation

Hi All, I am using the circuit from this paper in designing a PTAT circuit
"A Novel Wide-Temperature-Range, 3.9 ppm/C CMOS Bandgap Reference Circuit",by changing the input transistor of opamp from MOSFET to BJT.

This is the PTAT circuit and the detailed schematic of Opamp. Startup circuit is not included in this figure.


When I did the stability analysis in Cadence, by injecting IPROB at the output node of opamp, I got this bode plot


As this figure shows, I tried to put a cap in A node and another one in B node, but they does not help in increasing the BW.

(1) I do not know if my IPROB location is correct or not.The purpose of this BW increment is to deal with high Slew rate VCC raising, eg,1us from 0V to 2.5V.
(2) if it is correct, could anyone tell me where I can try to put the cap or RC to increase the bandwidth?
(3) if the iprobe injection is not correct for the preparation of start up, could anyone show me how to do the frequency compensation? I tried to inject the IPROB in the path of current mirror to negative input side of opamp, but the low frequency gain is 0dB and higher frequency gain is smaller, it seems like that is not a right way to probe the bandwidth of this PTAT circuit.

Thank you!

Hi,
The IPROB location is correct, you are breaking the loop at the symetric point.
But all the compensation you do will reduce the BW of the loop to achieve stability. So, you can try to push the complex poles away and reduce the compensation. But its a simple 2 stage loop and I'm not sure why you get complex poles less than a decade. I guess the bode plot you have plotted is for Cap at B, trying to introduce a Zero in the loop?

Hi yuvan,

Thank you for your advise! The bode plot I got is automatically generated from Cadence, after I put the iprobe at that place. So I do not know if this bode plot is for B node or anywhere.
Could you tell me how can I introduce a zero in this loop?

yuvan said:

Hi,
The IPROB location is correct, you are breaking the loop at the symetric point.
But all the compensation you do will reduce the BW of the loop to achieve stability. So, you can try to push the complex poles away and reduce the compensation. But its a simple 2 stage loop and I'm not sure why you get complex poles less than a decade. I guess the bode plot you have plotted is for Cap at B, trying to introduce a Zero in the loop?

Click to expand...

No I was asking if the BOde plot was simulated with the comp cap at node B? I was thinking introducing a CAp at node B will eventually look like a Zero WRT to loop gain. (Differencing impedance iincreases)

How does the current probe measure the loop gain? Which ratio do you define as loop gain?

yuvan said:

No I was asking if the BOde plot was simulated with the comp cap at node B? I was thinking introducing a CAp at node B will eventually look like a Zero WRT to loop gain. (Differencing impedance iincreases)

Click to expand...

Yes the simulation was with a 1fF cap in node B. It seems like adding the cap there does not help and no additional zero is found by adding that cap. Do you have any other suggestion in adding the zero?

At the mean time, can we say, if BW cannot be improved, this PTAT will oscillate when fast VCC is applied?

Thank you!

The compensation cap connected @ node B is 1fF? If so that seems to be too low to see any effect @ 30MHz BW.

Could you also produce the phase plot?

At the mean time, can we say, if BW cannot be improved, this PTAT will oscillate when fast VCC is applied?

Click to expand...

Why do you think that the circuit will oscillate? When doing the tran simulation, did you confirm that the circuit oscilated?


BR Jerry

When doing the tran sim, if the SR is big enough, the output current is not a DC level but look like sine wave, although the amplitude is not big. Besides, it is very time consuming for the simulator in plotting that kind of curve.

helpmejerry said:

Could you also produce the phase plot?

Why do you think that the circuit will oscillate? When doing the tran simulation, did you confirm that the circuit oscilated?


BR Jerry

Click to expand...

- - - Updated - - -

yuvan said:

The compensation cap connected @ node B is 1fF? If so that seems to be too low to see any effect @ 30MHz BW.

Click to expand...

Thank you yuvan, I found the low BW was caused by low bias current, and in that case, compensation does not help anymore. Thanks for your help!

No, opamp output is not (generally) low impedance & it's the loop which makes it low impedance. So, when you are breaking the loop the output of the amp is no longer low impedance & so Stability analysis will give proper results.