The Designer's Guide Community Forum

The Designer's Guide Community Forum
https://designers-guide.org/forum/YaBB.pl
Design >> Analog Design >> about temperature coefficient of bandgap
https://designers-guide.org/forum/YaBB.pl?num=1323086860

Message started by lhlbluesky_lhl on Dec 5^th, 2011, 4:07am

Title: about temperature coefficient of bandgap
Post by lhlbluesky_lhl on Dec 5^th, 2011, 4:07am

i have designed a bandgap in 0.18um 0.33V process, which is the classic structure (an opamp, two bjt of 1:8, a pmos pair, and a resistor), VREF=VEB+VT*ln8/R1*R2, but the tc is 45ppm for the simulation. i used some methods to improve the tc, such as exp compensation, different type of resistors (different tc, one is positive, one is negative), and so on, but it only improves a little. in some papers, i see the compensated bandgap can have a tc of 5ppm or below, but how to achive this in actual case? or how to do curvature compensation for bandgap in real project design?

thanks all in advance.

Title: Re: about temperature coefficient of bandgap
Post by Alexandar on Dec 5^th, 2011, 5:39am

Use trimming.

Title: Re: about temperature coefficient of bandgap
Post by loose-electron on Dec 5^th, 2011, 9:41am

start putting offsets and mismatches into your simulation. You will see that you will need to trim and adjust the device.

Also, look up the topic "chopper offset cancellation" or "chopper stabilized op-amps" - this is one way to deal with offset issues.

Title: Re: about temperature coefficient of bandgap
Post by RobG on Dec 5^th, 2011, 11:10am

I think he wants to know how to remove the curvature (about 4 mV) from a traditional bandgap response, i.e. he wants a decent curvature correction circuit. There are countless ways that look great in simulation but they have problems over corners and mismatches which are hard to correct with a single temperature trim in silicon. Amazingly, I still see a lot of new patents popping up on this ancient subject!

IMO, the most robust are independent of the bandgap reference itself and are a piecewise linear correction. At room temp they do not change the performance (so the zero temp-co output voltage is still ~1.23 V). However at temps greater than room they add a PTAT voltage, and at temps colder than room they add a CTAT voltage.

An current mode example is published by Rincon-Mora JSSC, Oct 1998. http://ieeexplore.ieee.org/xpl/freeabs_all.jsp?arnumber=720402

You can adapt the PWL correction to the "normal" bandgap. Just inject the PWL current into the 1x pnp emitter. It might help to see how it was done in one of my patents 6,642,699

Title: Re: about temperature coefficient of bandgap
Post by loose-electron on Dec 6^th, 2011, 2:24am

yes he wants a "curvature compensated bandgap" (search IEEE JSSC for those terms) - but the other items in mismatch and offset must be fixed first, because they are much larger than the thermal curvature.

Title: Re: about temperature coefficient of bandgap
Post by lhlbluesky_lhl on Dec 6^th, 2011, 4:17am

thanks all.

Title: Re: about temperature coefficient of bandgap
Post by RobG on Dec 6^th, 2011, 6:56am

loose-electron wrote on Dec 6^th, 2011, 2:24am:

I agree - that is what gets most people. He can search for my CICC paper on that too ;)