Bandgap: 10 ppm/Celsius

[crystalballs]

bandgap reference ppm

Hi all,

I have to design a bandgap circuit (curvature compensated I guess...) to reach the performance of:

10 ppm/^C.

Usually, I calculate the "ppm/^C" as follows:

PPM = 1000000 * Vbg(T) / Vbg(@25^C) [ppm/^C]

Please, do you have some advices about the bandgap architecture (i.e. the circuit) to use?

Thanks in advance.
CBs.

 

[rfsystem]

how to calculate ppm of bandgap

There are some approaches having about 0.1% from -25...125°C

look for the following patent

US7411380

also Stefan Marinca have a good bandgap reference in his newest patent

US7543253

 

[crystalballs]

10 ppm/celsius to ppm/k

Hi,

0.1%/Celsius is equivalent to 1000ppm/Celsuis.
That's, to much for me...
Thanks anyway!
CBs

 

[rfsystem]

calculation of ppm/celsius

NO!

it is 6.6ppm/K in the best case

take a look for your self....

[/img]

 

[amriths04]

ppm calculation band gap reference

it depends on your technology. Initially follow some simple bandgap structures and keep improving it till achieve your spec.

 

[palmeiras]

Please, take a look at the following papers:

(a) A CMOS piecewise curvature-compensated voltage reference
Lai xinquan, Xu Ziyou, Li Yanming, et all. MIcroelectronics Journal. Elsevier.

(b) A single Trim CMOS Bandgap reference with a 3 sigma inaccuracy of +-0.15% from -40 to 125. ISSC 2010.
Guang Ge, Cheng Zhang, et all.

 

[love_analog]

Depends upon process, modelling etc

But typically you would need some sort of curvature compensation.

A typical trimmed BGap would have ~100ppm variation.

Note, I don't understand 10ppm/C. does that mean 10ppm*100C = 1000ppm. That is pretty easy to achieve. or does it mean 10ppm over the temp range. I am guessing the latter.

 

[neoflash]

Depends upon process, modelling etc

But typically you would need some sort of curvature compensation.

A typical trimmed BGap would have ~100ppm variation.

Note, I don't understand 10ppm/C. does that mean 10ppm*100C = 1000ppm. That is pretty easy to achieve. or does it mean 10ppm over the temp range. I am guessing the latter.

1000ppm = 0.1%. This is a good result for a bandgap over all temperature range.

 

[snafflekid]

crystalballs, I think your ppm/°C calculation is off.

10e6•[1/Vbg(25°C)]•[ΔVbg]/[Δ°C]