what kind of precision you can get if you use r1/r2?

[liletian]

Hi Guys
What's the normal precision of current technology in mosis of r1/r2? can they reach precision better than 1 percentatge without special layout?
Thanks

 

[HanGu]

You can't get desired ratio if you do not use some layout tricks.But i think these tricks are so easy to implent,why do not you use these?

 

[liletian]

You can't get desired ratio if you do not use some layout tricks.But i think these tricks are so easy to implent,why do not you use these?

what kind of tricks are they? BTW, do you know what kind of ratio I can get if I do not use tricks?
Thanks

 

[JoannesPaulus]

The usual tricks: interdigitate r1 and r2, common centroid. If you have, say, r1=3kohm resistor and r2=2.25kohm resistor use 3 1kohm resistors for r1 and 6 1kohm for r2...

 

[liletian]

The usual tricks: interdigitate r1 and r2, common centroid. If you have, say, r1=3kohm resistor and r2=2.25kohm resistor use 3 1kohm resistors for r1 and 6 1kohm for r2...

got it, what kind of precision can we get in this way?
Thanks

 

[JoannesPaulus]

If you choose the resistor size carefully, you can easily achieve 0.5% on the ratio r1/r2.

 

[liletian]

If you choose the resistor size carefully, you can easily achieve 0.5% on the ratio r1/r2.

can you explain more what is the resistor size? I am a little confusing.
Thanks

 

[JoannesPaulus]

The resistor accuracy depends on its area and perimeter. If you make two 1kohm resistors as a, say, 1umx10um and 2umx20um, the latter will be more accurate. Read your technology manuals and you will find a resistor matching table. That table will tell you what is the size you need to use for the desired accuracy.

 

[liletian]

The resistor accuracy depends on its area and perimeter. If you make two 1kohm resistors as a, say, 1umx10um and 2umx20um, the latter will be more accurate. Read your technology manuals and you will find a resistor matching table. That table will tell you what is the size you need to use for the desired accuracy.

As you said, the resolution actually depends on the size of resistor, the larger the resistor, the better the precision.
On the other hand, large resistor will cause a high noise issue, so there must have a tradeoff.
ok, suppose I use voltage divider to get the desire voltage, let's assume r1,r2=10K, so I will get a thermal noise of 20KxKT. Am I right? Is it actually realistic to use resistor divider to get desired voltages from a fixed voltage for high sensitive application eg. RF?
Is there some research done on it and how actually it is solved in industry like in cell phone? Any research results can be shared?
Thanks

 

[JoannesPaulus]

As you said, the resolution actually depends on the size of resistor, the larger the resistor, the better the precision.
On the other hand, large resistor will cause a high noise issue, so there must have a tradeoff.

The noise is only dependent on R not on the physical size of the resistor... you can have 10kohm as a 1umx10um or as a 2umx20um and it won't change the noise. It will, of course, have a higher parasitic capacitance.

Is it actually realistic to use resistor divider to get desired voltages from a fixed voltage for high sensitive application eg. RF?

It only depends on your application. In principle I see no problem.

 

[rfsystem]

Resistor matching is limited by

1. Poly grain size effects
2. Mask tolerances
3. Etching effects
4. Doping nonuniformity

But first do not underestimate the effect of the contacts. The equivalent length of a contact of 15Ohm/ct to 300Ohm/sq poly resistor is 50nm. But the contacts can have a sigma spread by a value equal to there resistance.

So high accuracy dividers are made by tapped resistors the reference voltages are regulated feeded through tapped feedback.

At some size nonuniformity limit the matching. If you want to fold a long resistor the number of bends does not match.