bandgap bode stability

[SwordFish]

Hi,
I have a bandgap and I want to know how can I be sure that it is stable.
I know that the negative feedback must be "stronger" than positive feedback but how can I check this ?
I have 2 bode plots , one for positive loop and one for negative loop.

I have to subtract the dB values of the 2 plots to see if at all freq the delta is higher than zero ?

Is enough to have PM>70degree on the negative loop, or I have to check this also to the positive loop.

In my case the negative loop is at 30dB in negative loop, and at 5dB in positive loop (this are values at low freq).
I use IC5 /Spectre with STB.
The bandgap is classic with PNP and 1 opamp.

Thanks.

 

[dgnani]

Yes you need the differential loop gain to make sure the reference is intrinsically stable. I do not have spectre so I cannot give you specifics using STB analysis. I do it using simply AC analysis with one or two AC sources generating a differential input signal. You can then look at the differential loop gain (calculated using a ratio of voltage differences) to evaluate stability using a Bode plot. If you have positive and negative loop gain their difference is what you need to plot and make sure you have the usual 60+ degree of margin at 0dB. Also make sure the gain does not sneak back up again at high frequency (sweep a large frequency range)

 

[saiaditya]

If you are doing lstb analysis, you can break the loop right at the output of the opamp (rather than in the -ve fb loop). Introduce the lstb source between the output of opamp and the current mirror gate and the final result gives you the effective loop stability