Is it possible to make larger resistors from MOSFETS?

[anhnha]

Hi.

I need to create larger resistors from MOSFET (about 1 to 3 MΩ) for my bandgap reference.
I can't use normal resistors because their sizes are too big.
Could you tell me how to do that?
Is it possible to make larger resistors from MOSFETS?
I use TSMC018.

 

[dick_freebird]

You can, provided that you aren't especially fussy about
the value or linearity. Linearity requires that Vds be kept
low (so you might want many segments, such that Vds<
Vgs) and body effect will then come into play, impacting
segment-linearity.

But you see this kind of thing done a lot especially in a
CMOS op amp compensation network, the zero resistor
is often a MOSFET.

 

[anhnha]

Thanks.
I intend to use the resistors for this bandgap.
**broken link removed**

With this bandgap, I can use small values for R1, R2, R3 however the power consumption will be very large.
The resistors from MOS using for bandgap doesn't seem a good idea because non-linearity but I think I have no option.

 

[Dominik Przyborowski]

This resistors require high linearity and positive temperature dependence factor so it's bad idea to replace them with mosfets.

 

[anhnha]

Thanks.
I am somewhat aware about that but is there a solution for this?
I need to design a bandgap with low power consumption, small area.
However, two constrains are conflict.
Larger resistors => Large area (bad) => small power consumption (good)
Small resistor => Small area (good) => large power consumption (bad)
If you have a solution, please let me know.

 

[Dominik Przyborowski]

Hello anhnha. I check one idea by adds some pmosfets in series with resistors. Below my schematic (in ams0.35um) with dimensions and dc OPs.


I found some disadvantages:
1. The reference voltage is more sensitive to temperature than in classic implementation.
2. Vref has strong dependence from R0. You will need to check worst case simulations
3. Probably You supply range decrease (fets in that configuration needs higher supply, but You have to check in your pdk)
4. I spend on it only half hour so I don't guarantee that this is final solution ;-)

//edit: Maybe it will be good solution to add pfet current sources biased from better node than ground or to use simple amplifier stage to control them Vgs.

 

[erikl]

... or to use simple amplifier stage to control them Vgs.

Here you can find a paper on electronically tunable linear MOS resistors in the range of 10⁶ to 10¹¹ Ω. Unfortunately, the additional circuitry for the necessary feedback biasing circuit is rather complex.

 

[anhnha]

Thank you, Dominik Przyborowski.

Your work looks good! :-D
How about your line regulation?
I will try it now.

 

[Dominik Przyborowski]

Thanks erikl. One of my friend looking for high value resistor with good linearity for wide voltage drop for neuro readout.

@anhnha: unfortunately I didn't check it. Now I haven't time for nothing more than writing thesis ;-)

 

[anhnha]

Hi,
If possible, could you explain a bit how the MOSFET resistors work?
(maybe a link)

 

[erikl]

You can get MOS resistors with relatively linear behaviour either by MOSFET operation in the triode region (high Vgs, small Vds, low resistance)or by operation in (a limited part of) the saturation region, where the Ids vs. Vds rise is mainly given by carrier velocity saturation ("Early voltage"), works only at medium Vds voltages, but in a larger Vds range than in the triode region, resulting in relatively high resistance values, which, however, depend strongly on temperature.

See this Ids vs. Vds (= vdd) characteristic:

For more info about MOS resistors see any (design) book on MOSFETs and search for this keyword, or for the a.m. operation regions.