how to obtain very large resistors

[foreverloves]

i need a 1M resistor , but you know ,it takes too much silicon area, is there any methods that can realize large linear resistors?

switch capcitor may be a choice, but it may induce noises

if you have any good suggestions, tell me, thank you

 

[jiveshgovil]

i need a 1M resistor , but you know ,it takes too much silicon area, is there any methods that can realize large linear resistors?

switch capcitor may be a choice, but it may induce noises

if you have any good suggestions, tell me, thank you

You can use analog input impedance tunable circuits. There are huge number that are availbale.

 

[foreverloves]

what is input impedance tunable circuit?
I have no idea about that circuit,completely puzzled

 

[JPR]

Depending upon your application, you may be able to use a Tee network to approximate the large resistance.

Say, for example, you need to have an inverting amplifier with a gain of 100, and the input resistor is a 10kohm, and the maximum resistor size you can make is 50kohms. You would have the 50kohm resistor from your voltage to be amplified to the (-) input of the amplifier, and another 10kohm resistor from the (-) input to a common point for the tee network. You could have a 100ohm resistance from the common point to ground, and a 1.9kohm resistor from teh common point to the output of your amplifier.

If you put -1mV into the input, you would need +5mV at the other end of the 50kohm resistor. The 20:1 divider created by the 1.9k and 100ohm resistors would mean that the output of the amplifier would need to provide 20x, or +100mV to produce +5mV at the common point. This is the same gain you would obtain with a 1Megohm resistor in the feedback.

This is not EXACTLY right, since it negelects the current in the 50kohm resistor, but, if you need to be that exact, you could figure out the EXACT gain of the circuit, and adjust the 1.9kohm or 50kohm resistors to get the right gain. (I would guess that the solution given would be within a percent or so, and is probably close to what you would get with just resistor matching anyway...)

There are, of course, plenty of circumstances where the tee network will NOT yield the correct answer, and you will need either an off-chip resistor, or will have to accept the nonlinearities, etc. associated with a MOS resistance.

 

[srivatsan]

I am just curious for the usage of 1M inside a chip? I am not too sure about the power dissipation capabilities.

Am I near-right with my assessment? I generally avoid high values as they are too tough to design... and may not work as intended.

Srivatsan

 

[qutang]

can you use a current mirror?

 

[zhenywu]

i need a 1M resistor , but you know ,it takes too much silicon area, is there any methods that can realize large linear resistors?

switch capcitor may be a choice, but it may induce noises

if you have any good suggestions, tell me, thank you

1M resistor is not too large
many founry offer an extra way to make resistor
about serval hundries/unit

 

[husseinadel]

hello

you can use high poly resistor ,,it will not consume very large area,,,

 

[foreverloves]

I am just curious for the usage of 1M inside a chip? I am not too sure about the power dissipation capabilities.

Am I near-right with my assessment? I generally avoid high values as they are too tough to design... and may not work as intended.

Srivatsan

thank you for your reply
that is also why i am looking for a alternative way to realize large resistors

Added after 1 minutes:

Depending upon your application, you may be able to use a Tee network to approximate the large resistance.

Say, for example, you need to have an inverting amplifier with a gain of 100, and the input resistor is a 10kohm, and the maximum resistor size you can make is 50kohms. You would have the 50kohm resistor from your voltage to be amplified to the (-) input of the amplifier, and another 10kohm resistor from the (-) input to a common point for the tee network. You could have a 100ohm resistance from the common point to ground, and a 1.9kohm resistor from teh common point to the output of your amplifier.

If you put -1mV into the input, you would need +5mV at the other end of the 50kohm resistor. The 20:1 divider created by the 1.9k and 100ohm resistors would mean that the output of the amplifier would need to provide 20x, or +100mV to produce +5mV at the common point. This is the same gain you would obtain with a 1Megohm resistor in the feedback.

This is not EXACTLY right, since it negelects the current in the 50kohm resistor, but, if you need to be that exact, you could figure out the EXACT gain of the circuit, and adjust the 1.9kohm or 50kohm resistors to get the right gain. (I would guess that the solution given would be within a percent or so, and is probably close to what you would get with just resistor matching anyway...)

There are, of course, plenty of circumstances where the tee network will NOT yield the correct answer, and you will need either an off-chip resistor, or will have to accept the nonlinearities, etc. associated with a MOS resistance.

thank you ,that is a special occurance

any general ways?

Added after 2 minutes:

can you use a current mirror?

current mirror may be a good choice, i will try

 

[icecream]

i'm not sure about this, maybe you don't have enough space on substrate and technology to do this.. but .. you can obtain a large value resistor with deposition of thin film of highly resistance material on the membrane created by way of micromachining techniques, so the rest of the substrate could stay on the lower temperature

 

[JPR]

Again, this is somewhat specific, but perhaps useful:

A very long channel FET with Vgs-Vt just a bit larger than the voltage you plan on placing across the resistor will act as linear with fairly high value. The limitation of having Vgs-Vt larger than Vds will limit to cases where small Vds voltages apply (small voltage across the resistor). This type of resistor will also demonstrate voltage dependance (unless gate voltage and well track source voltage).

A bit less specific in application, but more specific to foundry:
There are some foundry specific high sheet rho resistors available, depending upon the foundry, but I think all of them involve extra processing steps, so inclusion on other foundrys may be limited.

Depending upon the application, you can turn a medium value resistor on for a small percent of time to make it appear (to a slow "observer") to be a large value resistor. For example, if you have a 100kohm resistor that is off for 990nSec, then on for 10nSec out of each 1uSec, it will appear as a 10Megohm resistor if used in a circuit with time constants of several uSec. Of course, to a fast "observer", this will look like exactly what it is, a 100kohm resistor that is switched on and off.

 

[foreverloves]

Again, this is somewhat specific, but perhaps useful:

A very long channel FET with Vgs-Vt just a bit larger than the voltage you plan on placing across the resistor will act as linear with fairly high value. The limitation of having Vgs-Vt larger than Vds will limit to cases where small Vds voltages apply (small voltage across the resistor). This type of resistor will also demonstrate voltage dependance (unless gate voltage and well track source voltage).

A bit less specific in application, but more specific to foundry:
There are some foundry specific high sheet rho resistors available, depending upon the foundry, but I think all of them involve extra processing steps, so inclusion on other foundrys may be limited.

Depending upon the application, you can turn a medium value resistor on for a small percent of time to make it appear (to a slow "observer") to be a large value resistor. For example, if you have a 100kohm resistor that is off for 990nSec, then on for 10nSec out of each 1uSec, it will appear as a 10Megohm resistor if used in a circuit with time constants of several uSec. Of course, to a fast "observer", this will look like exactly what it is, a 100kohm resistor that is switched on and off.

thank you very much for you dedication and patience,now i am interested in the first way of ultilizing long channel FET, can you provide some articles about the high resistance characteristice on long channel FET ,i searched the web but found nothing on it, thank you !!!!!!!!!!!!!!!!!!!!!!!!

 

[lwjbh]

i think current mirror is a good chioce,it can supply very high resistor

 

[Mistico_Pereyra]

hello

you can use high poly resistor ,,it will not consume very large area,,,

Tha's a goog choice, i have done pasive poly resistors (about 100K) or also if your area restriction is very hard, you can use n or p diffusion well resistor (sheet resistance ranges fron 1000-2000 Ω/sq). The disadvantage is that is that some times the resistor has high temperature coefficient.

Active MOS resistors are more complex if you are begining in this topic and it will take you more time in studying the MOS transistor

Look at this PDF

 

[cherrytart]

If you aren't particular about the actual resistance or tolerances, I've seen reverse biased diodes used as mega to giga ohm resistors. Plus or minus an order of magnitude (or more, it's really sloppy)

 

[vladimir1984]

I read that the biggest specific resistance is achieved with poly resistors. They are high ohmic.

 

[oxide]

the size depends on how much missmatch you can tol for your design

 

[safwatonline]

a PMOS in linear region with a large length "20u" will give u few Megs.

 

[horzonbluz]

Use a long DMOS, you will get a big resistor.
Usually its Ron will bigger than the Ron of PMOS.

 

[andy2000a]

1. use Hi-Res resistor .. some process have un-dope poly close 1~3K/square , SRAM Hi-res have 1G~2G/squre and use long snake shape

2. use mos switch , but long channel > 20um
you should know , spice model maybe have error
in gerneal , spice model is fitting for short channel device

3. use circuit , like BJT use common-emitter + Re
and Rin will large (1+beta)*Re
and maybe you can use circuit for improve Resistor
mos current mirror have large Rds