Current reference with Nwell resistor and poly resistor

[lunren]

In order to design a temperature indepedent current reference with bandgap reference, I want to combine a Nwell resistor(positive temperature coefficient) and
a poly reisitor(nagative temperature coefficient). The simulation results are really good, but I am not sure the real results after type out. Anyone who had such a experience of combination of differenct type of resistors please give me some advice? Or tell me some drawbacks about this combination? Any comments are welcome. Thanks in advance.

Lunren

 

[tekno1]

I have not used the combination you are trying to implement. n-well resistor will have a back bias coefficient which simulation ignores. one has to hand calculate its effects and select the proper size resistors during the layout.

 

[lunren]

According to the spice modle, after hand calculation, the leakage current is less than 1E-9A. The spec of the current reference is 10E-6A, so I think the back bias effect can be ignored.

 

[sunking]

try monter-carlo analysis

 

[jcpu]

From process point of view, this is not a bad idea at all.

Other than back bias and parasitic cap which are always to be watched
using Nwell resisitance.

One more thing to mention, this may very from fab to fab,
but the variation in resistance /per square is generally
(one sigma, including die to die base, and lot to lot base)
~5%. And that may degrade the cancellation of the temp coef.

 

[jodenma]

you know the simu results good cann't see the tape out test results is good . pls take care of the error in layout and technology.

 

[rajkumaru]

Hi Lunren,
I think you can do, whatever you are thinking

 

[Shohdy]

hi,
I have seen this technique before on TSMC0.25u fab, and it worked just fine.
We didn't need to worry about the nwell back voltage at all, the sim results were very close the the testing results.

So I suggest you go on.

Regards,
Shohdy

 

[DenisMark]

I've done so in one of my design, results i obtain only in september. But now it's seems to me it isn't good idea. It may suffer from the reason that parameters of diff. resistor varied during tech.process not simultaineously with parameters of poly. resistor. If after manufacturing u'll get different corners for diff. & poly. resistors u don't get full temp. compensation i think. Try to simulate with different corners models for diff. & poly. resistors. If results will meet specification use such technique.

 

[lewisflame]

The Idea is great,
but different resistor process may produce mismatches that can't be ignored

 

[DenisMark]

In additional one more idea. If connect diode in series with resistors (with proper choice device parameters) than it's possible to correct nonlinearity of bandgap output voltage to get temp. independent current.