The Designer's Guide Community Forum
https://designers-guide.org/forum/YaBB.pl
Design >> Analog Design >> Bandgap temperature sensor
https://designers-guide.org/forum/YaBB.pl?num=1367033396
Message started by cktdesigner on Apr 26th, 2013, 8:29pm

Title: Bandgap temperature sensor
Post by cktdesigner on Apr 26th, 2013, 8:29pm
Hi all,

I need to implement a temperature sensor based on the circuit shown in the attached schematic. Unfortunately, the current flowing thru M1/M2 is not PTAT. So then how can I implement a temperature sensor based on this architecture? Any guidance will be helpful.

Thanks.

Title: Re: Bandgap temperature sensor
Post by yvkrishna on Apr 27th, 2013, 10:36am
If you strictly need a  PTAT current for temperature sensor then probably your are with a wrong circuit.

and why do you want to use this circuit only?

b.t.w the current tempco  in the fig you posted is still a function of resistor tempco..


Regards,
yvkrishna

Title: Re: Bandgap temperature sensor
Post by raja.cedt on Apr 30th, 2013, 1:04pm
hello,
If you don't have tough spec on temp accuracy remove R1,R2 and sense current trough M3, but finally you need dump into some resister which changes coefficient.
Please refer A CMOS temperature sensor with a 3s inaccuracy of ±0.1°C from -55°C to 125°C

Thanks,
Raj.

Title: Re: Bandgap temperature sensor
Post by loose-electron on May 2nd, 2013, 11:50am
lot of process variation in temp sensors unless they are designed to remove process variance isssues. Research the papers written on the subject.

Title: Re: Bandgap temperature sensor
Post by Dan Clement on May 3rd, 2013, 5:34am
Just feed a diode with current. Then you can quantize it directly.

The accuracy is not good but can easily be calibrated.

I agree with loose electron, research it a little. Every CMOS book has a temp sensor section.

Title: Re: Bandgap temperature sensor
Post by RobG on May 6th, 2013, 8:56am
Not sure what the process problems are - if you remove R1 and R2 the output voltage will be pretty much process independent PTAT. Opamp offset and mismatches between the resistors will dominate the errors.

Also be aware that you have to compare this voltage to something - generally you will do an A/D conversion so the reference that you use for it may be a worse source of error. Or it may give you either a trimmed Vbe or a PTAT votlage you can use since the reference voltage is Vbe+PTAT.

Title: Re: Bandgap temperature sensor
Post by cktdesigner on May 10th, 2013, 9:55pm
It is not possible to remove R1, R2 as they are involved in generating the PTAT part of the bandgap output voltage as seen in the equation below.

Title: Re: Bandgap temperature sensor
Post by RobG on May 11th, 2013, 7:38am

cktdesigner wrote on May 10th, 2013, 9:55pm:
It is not possible to remove R1, R2 as they are involved in generating the PTAT part of the bandgap output voltage as seen in the equation below.

Give it a try... Note your formula reduces to Vout = (R3/R0)*ln(N)*kT/q when R1 goes to infinity (you also need to get rid of R2 or the assumption that the current in each leg is equal will be wrong). The circuit was originally designed without those resistors - in fact it was the basis for nearly all references before Banba published a paper putting them to remove the temp-co (which is something you generally don't want to do).

You really need to read some more papers on this - I would focus on the ones that either use a PTAT reference based on the voltage difference between two bipolars with different current densities (like the one I just described) or one based on Vbe. The former is sensitive to mismatches, and the latter may need calibration depending on how it is used. There have also been some papers by Kofi Makinwa's group has published several papers recently that I think use another technique that may be better but I haven't studied them.

Title: Re: Bandgap temperature sensor
Post by nrk1 on May 12th, 2013, 7:42am

RobG wrote on May 11th, 2013, 7:38am:

cktdesigner wrote on May 10th, 2013, 9:55pm:
It is not possible to remove R1, R2 as they are involved in generating the PTAT part of the bandgap output voltage as seen in the equation below.

Give it a try... Note your formula reduces to Vout = (R3/R0)*ln(N)*kT/q when R1 goes to infinity  


RobG is right. If you need only a PTAT dependence(of current), remove R1,2. If you need a bandgap reference voltage, you need R1,2. If you need both, you'll need two separate cells like this as far as I can tell.

With R1,2 removed, the temperature dependence of current won't be strictly PTAT due to temp. variations of R0 and deviations from exponentials of transistors. If you push this current into another resistor similar to R0, the former is not an issue.

Title: Re: Bandgap temperature sensor
Post by Kevin Aylward on Jul 13th, 2013, 7:59am
This is a standard sub bandgap voltage reference, so I guess the poster has a requirement for sub 1V operation voltage reference. If the process actually has semi decent NPN bipolar transistors, and many “CMOS” processes actually do, then simply mirroring Q1 or Q2 to another bipolar transistor will produce a PTAT current output.

Title: Re: Bandgap temperature sensor
Post by loose-electron on Aug 2nd, 2013, 7:18pm

RobG wrote on May 6th, 2013, 8:56am:
Not sure what the process problems are - if you remove R1 and R2 the output voltage will be pretty much process independent PTAT. Opamp offset and mismatches between the resistors will dominate the errors.

Also be aware that you have to compare this voltage to something - generally you will do an A/D conversion so the reference that you use for it may be a worse source of error. Or it may give you either a trimmed Vbe or a PTAT votlage you can use since the reference voltage is Vbe+PTAT.


The process variance of just the diodes vs. the bias current makes it a little more complicated.

The implementation I did on this many moons back was in the era of the Pentium I, when thermal monitors first got introduced to microprocessors.

The methodolgy used to get rid of all the calibration nonsense involved a pair of PN junctions with a geometry ration (8:1? 12:1? I forget.) and also a pair of high-low current (that was 10:1) into both of the junctions. With all that, you end up doing a set of difference measurements where most of the process dependent variables fall out of the equation, and you are left with purely a temperature dependent function. The geometry ratios and the current ratios become the control of the output differences, with all the absolutes falling out of the math.

Sorry I don't have more specifics here but I did that design way back around 1992, so the details are a little muddy.

Do a IEEEE JSSC search on this, there was a lot of temp sensor stuff done back there.

Title: Re: Bandgap temperature sensor
Post by RobG on Aug 5th, 2013, 9:23am

loose-electron wrote on Aug 2nd, 2013, 7:18pm:

RobG wrote on May 6th, 2013, 8:56am:
Not sure what the process problems are - if you remove R1 and R2 the output voltage will be pretty much process independent PTAT. Opamp offset and mismatches between the resistors will dominate the errors.

Also be aware that you have to compare this voltage to something - generally you will do an A/D conversion so the reference that you use for it may be a worse source of error. Or it may give you either a trimmed Vbe or a PTAT votlage you can use since the reference voltage is Vbe+PTAT.


The process variance of just the diodes vs. the bias current makes it a little more complicated.

The implementation I did on this many moons back was in the era of the Pentium I, when thermal monitors first got introduced to microprocessors.

The methodolgy used to get rid of all the calibration nonsense involved a pair of PN junctions with a geometry ration (8:1? 12:1? I forget.) and also a pair of high-low current (that was 10:1) into both of the junctions. With all that, you end up doing a set of difference measurements where most of the process dependent variables fall out of the equation, and you are left with purely a temperature dependent function. The geometry ratios and the current ratios become the control of the output differences, with all the absolutes falling out of the math.

Sorry I don't have more specifics here but I did that design way back around 1992, so the details are a little muddy.

Do a IEEEE JSSC search on this, there was a lot of temp sensor stuff done back there.

I've done quite a few since '92 - the voltage difference between two PN junctions is PTAT and independent of process as long as they are biased reasonably (no low/high level injection, base IR drop is negligible, etc). You still need to compare this voltage to a reference (like a bandgap reference) if you want a digital answer. This reference will need to be trimmed to eliminate process dependence.

rg

Title: Re: Bandgap temperature sensor
Post by loose-electron on Aug 5th, 2013, 5:51pm

RobG wrote on Aug 5th, 2013, 9:23am:

loose-electron wrote on Aug 2nd, 2013, 7:18pm:

RobG wrote on May 6th, 2013, 8:56am:
Not sure what the process problems are - if you remove R1 and R2 the output voltage will be pretty much process independent PTAT. Opamp offset and mismatches between the resistors will dominate the errors.

Also be aware that you have to compare this voltage to something - generally you will do an A/D conversion so the reference that you use for it may be a worse source of error. Or it may give you either a trimmed Vbe or a PTAT votlage you can use since the reference voltage is Vbe+PTAT.


The process variance of just the diodes vs. the bias current makes it a little more complicated.

The implementation I did on this many moons back was in the era of the Pentium I, when thermal monitors first got introduced to microprocessors.

The methodolgy used to get rid of all the calibration nonsense involved a pair of PN junctions with a geometry ration (8:1? 12:1? I forget.) and also a pair of high-low current (that was 10:1) into both of the junctions. With all that, you end up doing a set of difference measurements where most of the process dependent variables fall out of the equation, and you are left with purely a temperature dependent function. The geometry ratios and the current ratios become the control of the output differences, with all the absolutes falling out of the math.

Sorry I don't have more specifics here but I did that design way back around 1992, so the details are a little muddy.

Do a IEEEE JSSC search on this, there was a lot of temp sensor stuff done back there.

I've done quite a few since '92 - the voltage difference between two PN junctions is PTAT and independent of process as long as they are biased reasonably (no low/high level injection, base IR drop is negligible, etc). You still need to compare this voltage to a reference (like a bandgap reference) if you want a digital answer. This reference will need to be trimmed to eliminate process dependence.

rg



Thanks Rob!

The Designer's Guide Community Forum » Powered by YaBB 2.2.2!
YaBB © 2000-2008. All Rights Reserved.