About bandgap reference design

[slchen]

I prepare to design a bandgap reference with the power supply from 3.3V to 2V in 0.35um CMOS technology.
Which kind of OP (PMOS-input or NMOS-input) is good for this application?
And what is the spec. of the OP required?
Thanks.

slchen

 

[|IAngel|]

I'd use P-input for better noise and such. Running as an error amplifier, the higher the gain the better. The higher the gain used, the closer the amp will set diode nodes to and therefore the more accurate and temp. independant the circuit will be. This being said, if you use 2nd order compensation, a simple 1-stage amp should suffice for getting below 10ppm/°C. Since you have the headroom though with respect to supply voltage, I'd suggest a 3 stage amp with 100dB of gain. I picked these numbers simply because they will yield awesome results and with a supply of 2-3.3V, you obviously aren't worried about power consumption.

 

[slchen]

Thanks for your suggestions!
I have other problems about this design!

1. What is the bandwidth of this OP required? as low as possible?

2. Because I must design this bandgap reference in the logic process, I can't use the high-R poly to implement the resistors in the bandgap reference! Therefore, what is your suggestion to implement resistors by using n-well, n-diffusion, p-diffusion, low-R poly?

Thanks!!!
slchen

 

[sunking]

the bandwidth can not be too slow for high speed.
low -R poly is better but the area maybe too much, it just depond on you static current.

 

[chinito]

PMOS input pair OPAMP doesn't have to have high gain as 100dB and more. I've designed BG with opamp gain of 65dB. Too much gain isn't always a good thing. Another reason for PMOS input pair OPAMP is its ability to work with lower input CM voltage which will be the case in BG. That's my 2 cents worth of advice

 

[yuyangjp]

I doesn't think the OPAMP used in the BG need to have high gain too. With high gain, the amp will set diode nodes to closer. but there still have offset voltage,which will be 5mV~10mV. So 40dB gain is large enough.

 

[xuel]

I have done a bandgap with a p & n input opamp, because this structure have better emi and less offset.

 

[dasong]

can use the two stage opa ,the CMIR you need think much!

 

[tanghua_0407]

If the stablity of the opam must be thought? So the one stage opam is best ,then we needn't to thought the stability@

Added after 2 minutes:

If the stability of the opam must be thought ,So the one stage opam is best??