Designing bandgap with op-amp

[chunlee]

bandgap and op-amp

hi everyone,

i'm want design a bandgap voltage using op-amp. is it we need to make sure the op-amp is stable before we connect it to the bandgap voltage? how to make sure op-amp is stable?

need to make sure the phase margin of the op-amp more than 45 degree or not?

Thanks.

 

[huojinsi]

Re: bandgap and op-amp

u must do a ac analysis for op-amp,plot amplitude-frequency curve and phase-frequency curve of open loop, insure PM is enough larger than 45 degree !

 

[chunlee]

Re: bandgap and op-amp

thanks huojinsi.

i want to ask one more question.

usually how long the MOSFET length can consider as long-channel? 5*Lmin enough or not?

thanks again.

 

[suria3]

Re: bandgap and op-amp

hi everyone,

i'm want design a bandgap voltage using op-amp. is it we need to make sure the op-amp is stable before we connect it to the bandgap voltage? how to make sure op-amp is stable?

need to make sure the phase margin of the op-amp more than 45 degree or not?

Thanks.

You only able to check stability of the Opamp when you connect to the bandgap. When the opamp is alone (not connected) to any loop, u wont be able to measure PM, coz it is open and it behaves like a comparator actually. So, once you connect to the bandgap, then you break the loop at one point, do AC analysis and plot the gain vs phase, then you will able to check PM. A PM of 60 deg is good.

You may refer to this link in your question regarding the long channel effect. It might help you.

**broken link removed**

 

[chunlee]

Re: bandgap and op-amp

hi suria3,

what you mean break the loop at one point? can you pls explain further?

what point i need to connect the ac input and output node when i want to check the PM? is it i connect the ac input to the op-amp input and measure the ac output voltage at output node of the op-amp?

thanks for your help.

 

[aryajur]

Re: bandgap and op-amp

Usually the opamp gain and phase margins are calculated from the AC response of the opamp itself, because this phase margin is assuming the "worst case" feedback factor β = 1. So you can do the AC response of the opamp itself.
Now it depends on the topology of your Bandgap whether this will be adequate or not. For example if you have a gain stage inside the feedback loop of the opamp then your Phase Margin will get worse since now the feedback factor β > 1. So the way suria suggested to put the opamp in the Bandgap, break the feedback loop at a certain point ( say the input of the opamp , maybe the + terminal may be a good choice) and then put the AC source there and then plot the AC response and calculate the gain and phase margin. This way you account for the loading effects and the feedback factor in the circuit. Although this way is also not totally accurate since you have broken the ciircuit and changed the actual operation, but this may give you agood estimate quickly.
The most accurate way to test stability is the Transient response. Put the whole circuit, smulate the startup with a transient repsonse and you will know if your circuit is stable or not.
Also remember to simulate the transient response at different temperatures, since if the Bandgap is stable at 1 temp does not mean that it will be stable at other temperatures if started also.

 

[suria3]

Re: bandgap and op-amp

hi suria3,

what you mean break the loop at one point? can you pls explain further?

what point i need to connect the ac input and output node when i want to check the PM? is it i connect the ac input to the op-amp input and measure the ac output voltage at output node of the op-amp?

thanks for your help.

Chunlee, Aryajur have given a better explanation on your question on the measurement of phase margin. Anyhow, may be i can explain again. I meant break any point is, let say there is a loop that connecting your one of the input of the opamp to a output of another opamp. So, you break the loop by giving a AC signal to the input of the opamp and plot the gain which is output over input. Normally we give AC to the high input impedance like gate of MOS.

There is another way you can do, where without breaking the loop, you can include the "Middlebrook" measurement as I attached. This is very accurate as it gives the accurate DC point which will tolerate a bit when you break the loop (common method). Give IN point to the input of Opamp and OUT point to other point when u break the loop. It is like including this circuit in the loop.

Hope this helps you.

 

[zenisle]

bandgap and op-amp

pls see some book as razavi or gray

 

[pseudockb]

Re: bandgap and op-amp

Regarding the stability of the opamp in the bandgap, there are two feedback loops in the following diagram. If I am not mistaken, loop A is the negative feedback loop while loop B is the positive feedback loop. I have a couple of questions. Firstly, is it sufficient enough to check the stability of loop A only by plotting a frequency response curve of VoA/ViA? Or should I check the stability of loop A by injecting two small signals at point A and point B and then plotting the frequency response of VoA/(ViA-ViB)? Both curves yield completely different response curves. Could someone tell me which is the more accurate way? Thanks

 

[aryajur]

Re: bandgap and op-amp

I think what you can do is break the opamp input terminals and then input 2 signal sources there say Va and Vb. And then take the output at the broken points as Voa and Vob and then you should plot (Voa-Vob)/(Va-Vb)

 

[suria3]

Re: bandgap and op-amp

I agree with Aryajur, when a opamp is comprised of a dual loop, the stability have to check differentially, i mean the gain and phase have to measure differentially rather than measuring one by one as we do in single loop like, unity gain buffer.

 

[pseudockb]

Re: bandgap and op-amp

Thanks for the help. I will try it out.

 

[pseudockb]

Re: bandgap and op-amp

Hi, I simulated the differential loop gain of the circuit but I observed some strange behaviour. The phase response of the loop gain seems weird, as shown below. I have checked that all my transistors are in saturation region and the opamp has a phase margin of about 70 degree before inserting it in the loop. I also ran a transient response of the bandgap reference by ramping up the power supply and the circuit proves to be stable. Am I misintepreting the frequency response of the loop gain? Thanks.

 

[aryajur]

Re: bandgap and op-amp

I am not sure but I think the dB20 function takes the modulus of the value before calculating the dB20. The Bandgap circuit which you put up before, if you take plot the magnitude and phase of the loop gain you should plot (Vob-Voa)/(Va-Vb) since the circuit has an inversion after the opamp. Sorry I wrote the wrong expression in my last post. When you do that I think the phase plot will come out right.

 

[JT]

Re: bandgap and op-amp

Hi,

You need to assess both loops in one time. The way I do it :

Cut at the amplifier output, inject at the current source nodes, and see what's coming back from the amplifier output. (keep in mind DC bias and avoid AC into the output).