The Designer's Guide Community Forum

The Designer's Guide Community Forum
https://designers-guide.org/forum/YaBB.pl
Design >> Analog Design >> Bandgap Reference: BJT
https://designers-guide.org/forum/YaBB.pl?num=1326079821

Message started by mixed_signal on Jan 8^th, 2012, 7:30pm

Title: Bandgap Reference: BJT
Post by mixed_signal on Jan 8^th, 2012, 7:30pm

Hi,
I m designing a band gap reference circuit and I have some basic doubts.

1. The BJT that I will use is a diode connected with Vbc=0. So, in which region of operation will it work i.e. active or saturation.

2. How this scenerio is handled by SPICE i.e. by eber molls and gummel poon model?

I am using Spectre simulator.

Title: Re: Bandgap Reference: BJT
Post by raja.cedt on Jan 9^th, 2012, 12:28am

hello,
it works in active region, becaz when you have Vbc=0, collector base junction still reverse bias. You need some finite Vbc to drive the transistor into saturation. regarding your 2nd Question most of the times it depends the modeling, so now a days Quite common model is gummel poon model.

Thanks,
Raj.

Title: Re: Bandgap Reference: BJT
Post by mixed_signal on Jan 9^th, 2012, 9:08am

Thank you!

Title: Re: Bandgap Reference: BJT
Post by loose-electron on Jan 9^th, 2012, 11:05am

If you have a quality Bipolar transistor,
why not use one of the classic biplolar BG's?

These devices will be less of an issue with
offsets and mismatch issues.

If this is the lateral PNP available on CMOS, then
the diode approach is what you are stuck with.

Title: Re: Bandgap Reference: BJT
Post by mixed_signal on Jan 9^th, 2012, 8:50pm

Hi loose electron,

i am using substarte/vertical BJT because of its accurate and predictable characteristics. I am not using lateral bipolar because of its low quality junctions.

I think the diode problem is with vertical BJTs and not with lateral BJTs.

Title: Re: Bandgap Reference: BJT
Post by loose-electron on Jan 10^th, 2012, 3:32pm

If you got a Beta around 60-120 thats a good bipolar,'
if its Beta 5-20 thats one of the junky ones.

you got access to C,B,E of the transistor, or is the collector tied to ground?

collector tied is generally a lateral junk PNP avaialable on CMOS.

Title: Re: Bandgap Reference: BJT
Post by Dan Clement on Jan 10^th, 2012, 6:41pm

I've designed quite a few band gaps and I have yet to see one come out as simulated... With either npn or pnp.

Best thing you can hope for is to build in some metal options or trim.

Besides you will get much more error from the offset and noise of your op amp than I think you will from VCB = 0.

Title: Re: Bandgap Reference: BJT
Post by vivkr on Jan 11^th, 2012, 4:59am

Dan Clement wrote on Jan 10^th, 2012, 6:41pm:

Dan,

I would suspect that you didn't get good models. I haven't made so many bandgaps (maybe 2-3 spread over a long period of time), but we made some "corrections" to the manner in which our models were being fitted and then we suddenly started getting good fits whereas we were condemned to having to trim the bandgaps earlier in every single case.

Of course, it also depends on the kind of precision you are trying to get but I would say that 1% accuracy should be possible without trimming if your models are accurate.

Vivek

Title: Re: Bandgap Reference: BJT
Post by RobG on Jan 11^th, 2012, 8:42am

vivkr wrote on Jan 11^th, 2012, 4:59am:

I would suspect that you didn't get good models.

I'm with both Dan and vivkr... the models are not usually precise enough until you build a real bandgap. I suspect part of the problem is that the layout and surroundings don't exactly match between the circuit and device that was modeled, and the modelers also often don't appreciate the precision required. Once you tweak in the model it will work fine up to the point of bipolar IS and resistor Rsheet process variation.

And the bipolar that you can't access the collector is the vertical (aka substrate), not the lateral.

rg

Title: Re: Bandgap Reference: BJT
Post by RobG on Jan 11^th, 2012, 8:59am

mixed_signal wrote on Jan 8^th, 2012, 7:30pm:

I don't understand what your "doubt" is. The simulator will work fine if Vbc = 0 V.

Maybe the problem is with the "saturation" term. Everything was great in the world in the old days when bipolars ruled. All you had to remember was saturation is bad. Do not operate in saturation! Then came MOS and the dummies called the normal operating mode "saturation."

So don't get confused:
Bipolar: Saturation is bad!!! In saturation BC collector diode is forward biased and the device no longer acts like a transistor.

MOSFET: Saturation is good! In this region MOS acts like a transistor!

Title: Re: Bandgap Reference: BJT
Post by loose-electron on Jan 11^th, 2012, 6:07pm

Dan Clement wrote on Jan 10^th, 2012, 6:41pm:

I do agree - which is one of the reasons I suggested going to the old bipolar structures and avoiding the op-amp methods. If its a lateral bipolar with terrible beta, that won't work.

As a general rule a CMOS process BG is using a lateral PNP and
it will need to be trimmed. You can get around
some of the offsets with chopper methods, but the offsets get
gained up so much that generally you got to trim.

Title: Re: Bandgap Reference: BJT
Post by vivkr on Jan 12^th, 2012, 3:04am

one more point I forgot to mention which often causes much grief is not just the modelling of BJTs but that of resistors.

If the resistor tempco is not properly characterized or if you are using dimensions vastly different from those used in model extraction, then you are going to be way off.

Another point where I have seen some careless data extraction from silicon is when there is more than one resistor type available in the process. It may happen that the different resistor types were not all characterized thoroughly on their own and that the results were bunched together (SHOCKING), especially for more expensive measurements like tempco. In that case, you may do all you want and end up tuning your bandgap perfectly at the wrong point. This is quite hard to catch but you may be able to figure out this if you see same tempco quoted for all types of resistors.

Unfortunately, the last of these errors is almost impossible to detect for you as a designer. So, a good solution might be to make a copy of your bandgap with probe pads and suitable strucures for accurate measurement of resistances so as to be able to measure out critical parameters such as resistance, Vbe, tempco and check how well they fit to your simulations. Obviously, you need to do a fair bit of tricky measurements in this case.

Vivek

Title: Re: Bandgap Reference: BJT
Post by RobG on Jan 12^th, 2012, 8:07am

Great point Vivkr, errors in the the temp co of the resistor model as well as its value (dependence on ΔW, ect) will produce errors that are indistinguishable from errors in the bipolar. In fact, it could be a bigger issue at one company since I screamed so much about the bipolar models! Sometimes you folks make me see the obvious a little better.

rg

Title: Re: Bandgap Reference: BJT
Post by RobG on Jan 12^th, 2012, 8:09am

duplicate post...

Title: Re: Bandgap Reference: BJT
Post by loose-electron on Jan 12^th, 2012, 3:03pm

Give this a read:

http://chipdesignmag.com/display.php?articleId=438&issueId=16

available here also:

http://effectiveelectrons.com/whitepapers/Determine%20Foundry-Model%20Problems%20Without%20Touching%20A%20Wafer.pdf

essentially this is a page of what I feel should be common sense and
often is not