[Moved] Bandgap references -Resistor temperature dependence

[circuitking]

Hi, in Razavi book/some other material, while analyzing bandgap references, they have not considered resistors temperature variation. Is there any specific reason for this?. The resistors in my technology library are temperature dependent, does this mean I should use resistor-less schematics?. Thanks

 

[KlausST]

Hi,

we need more information about the used document and the circuit you refer to.

Possible reasons for not including thermal drift of resistors:
* resistors from the same batch may have identical drift an thus may compensate each other.
* resistor drift does not cause output voltage drift at all
* resistor drift is negligible
...all depend on the circut and technology...

Klaus

 

[dick_freebird]

Thin film resistors can have negligible tempco
and negligible curvature, and are laser trimmable.
This may have just been the context for the Analog
Devices folks who did a lot of the original inventing.

Or, maybe it's a matter of making it bit-sized for
clarity, there's enough going on with just the
transistor nonidealities to keep you busy for a
while. If you feel like expanding the R terms to
include temperature effects, go for it.

 

[circuitking]

we need more information about the used document and the circuit you refer to.

I made this circuit using ideal resistors from analogLib in virtuoso and got the current reference that is independent of temp. After I replace it with the resistors from technology, the current is no more constant.

 

[KlausST]

Hi,

It seems to be "IC-Design" but you did not start this thread in the IC-Design section. Please clarify.
*********

There is no "absolute stability", so what error (magnitude) are you talking about?

and please provide all informations at once. Don´t let us ask for every single detail.

Why don´t you provide your simulation results? They should show where the error is coming from. And why and what magnitiude...

*****

Klaus

 

[circuitking]

Hi,

It seems to be "IC-Design" but you did not start this thread in the IC-Design section. Please clarify.

Klaus

Yeah, It is IC design. Sorry I didn't see that there are two seperate groups

and please provide all informations at once. Don´t let us ask for every single detail.

Why don´t you provide your simulation results? They should show where the error is coming from. And why and what magnitiude...

*****

Klaus

The red one is with Ideal Op-amp from ahdLib and ideal resistors from AnalogLib. Yellow one is when only ideal opamp is replaced with the opamp I designed. The green is with resistor from technology and designed opamp. After looking at the three plots I decided it is the resistor's temperature dependence that is impacting.

 

[dick_freebird]

Yes, you always have to tweak the PTAT emitter
resistor and the "+TC" resistors specifically for the
real tempco.

You can also see that the real resistor has a tempco
nonlinearity which adds to the BJT Rb curvature and
means you can't be dead-flat everywhere, so pick
a region where it matters most (like flat-ish from 0
to 70C, center the flat-top at 25C for test and
marketing pictures, and hope that low and high
temp accuracy remains tolerable).

Assessing the resistor options of the PDK for the
likeliest species, given constraints such as area
and power, is something to do early.

 

[KlausST]

Hi,

I decided it is the resistor's temperature dependence that is impacting.

Which resistor?

I guess R2 value is identical to R3 value. And both should have identical tempco, thus both drifts should almost cancel out.

Klaus

 

[FvM]

Presume "resistor from technology" is not thin film but standard diffused or whatsoever. Then R4 temperature coefficient is sufficient to ruin the circuit performance.

 

[sutapanaki]

How about you plot the output voltage, that's what you are interested in. The output voltage is defined by a ratio of resistors so their temco should cancel, at least to a first order.

 

[dick_freebird]

Diffused resistors will produce more curvature but
you can make a "good enough" bandgap with them.
You just have to be ready to accept more overall
deviation from ideal output. I wouldn't call it "ruined",
just needing ratios adjusted.

Here's a plot showing the "typical" LT1009 reference
(2.5V, bandgap based) curve and "test mask", the
simulations from my design and the results from a
first look packaged part. The key difference is that
Linear Tech used Pbase resistors, and I used SiCr.
The difference in curvature is plain. But certainly
Linear Tech has made and sold plenty of these parts.

 

[FvM]

Referring once more to the initial question, you are quoting a problem schematic from chapter 11, bandgap references. If you read the chapter thoroughly, you'll notice that they discuss resistor TC, but don't perform a systematic analysis.

 

[circuitking]

Presume "resistor from technology" is not thin film but standard diffused or whatsoever. Then R4 temperature coefficient is sufficient to ruin the circuit performance.

I am checking the source current of M4. I didn't connect M5 and R4

--- Updated Oct 20, 2020 ---

Referring once more to the initial question, you are quoting a problem schematic from chapter 11, bandgap references. If you read the chapter thoroughly, you'll notice that they discuss resistor TC, but don't perform a systematic analysis.

It is from 2nd edition chapter 12. you mean to replace resistors with switched capacitors.

--- Updated Oct 20, 2020 ---

Hi,

Which resistor?

I guess R2 value is identical to R3 value. And both should have identical tempco, thus both drifts should almost cancel out.

Klaus

Yes, both R2 and R3 are identical. I replaced R1 and R2(and R3) individually. Green when only R1 (R2/R3 are ideal) is replaced and red when only R2 and R3 (R1 is ideal) are replaced.

--- Updated Oct 20, 2020 ---

Hi,

Which resistor?

I guess R2 value is identical to R3 value. And both should have identical tempco, thus both drifts should almost cancel out.

Klaus

I increased R2 and R3 values and I got a better response now.

--- Updated Oct 20, 2020 ---

I am checking the source current of M4. I didn't connect M5 and R4

--- Updated Oct 20, 2020 ---

It is from 2nd edition chapter 12. you mean to replace resistors with switched capacitors.

--- Updated Oct 20, 2020 ---

Yes, both R2 and R3 are identical. I replaced R1 and R2(and R3) individually. Green when only R1 (R2/R3 are ideal) is replaced and red when only R2 and R3 (R1 is ideal) are replaced.

--- Updated Oct 20, 2020 ---

I increased R2 and R3 values and I got a better response now. Actually I want current that is independent of supply. In general, If I get a response that is independent of temperature, is it guaranteed that I get a response that is independent of supply?