Very low noise bandgap reference design

[mithlesh]

low noise bandgap

HI,
Can anyone tell me what structure is best suited for very low noise application. My requirement is < 1nV^2 from 1-20KHz in cmos process with 1.2V as output (1% accuracy) from 3V supply.

Appreciate your fast responses.

Thanks
Mithlesh

 

[rfsystem]

low-noise band-gap -patent -datasheet

Sorry I did not have my notes with me to calculate the required bias current for a bandgap to deliver an equivalent noise resistor of ~60Ohm (1nV/(4*k*T)). But this value seems out of range.

There are two solutione known to me:

1. Sampling the bandgap voltage on a big cap. That is useful for a sampled system where you can accept noises at the clock edges. The voltage is transfered to the cap by a sampled domain or switch cap lowpass filter.

2. Use a noisy bandgap at system startup and calibrate a diode reference. The diode is driven by medium noisy kT current source from a basic bandgap architecture. But because of the low dynamic resistance of the diode the noise is low. After calibration, which simple select a feedback gain to scale the diode reference to the noisy start bandgap, switch to the scaled diode reference.

I hope that is a solution to you! Could you say which application require this level of noise?

So give me some return for my tips

Added after 1 hours 7 minutes:

Sorry I forgot that the calibration should be updated every significant temperature change. That limits the application. The method 1 is more useable here.

 

[elantra]

low noise bandgap reference

possible for using off-chip cap as filter?
pin avaiable?
It's the good way and you can find some low noise LDOs use this approach.

 

[mithlesh]

bandgap reference + fast -patent

Hey rfsystem,

Thanks for the reply but i could not understand the second point fully. Can you please share some document on the two methods?

This is for audio application.

`Let me collect few points first.

Thanks

Added after 4 minutes:

Hey Elantra,

The pin is available but is there no other way?? What structure should be used then? Is there any advantages of current mode operation over voltage mode in terms of output noise?

Thanks