Fabricating resistors and capacitors

[Jassi]

I have been reading up on how chips are manufactured and I am stuck on one important step. How do they fabricate capacitors and resistors on the chip with such low tolerance levels. If anyone has a step by step guide, that would help a lot.

Thanks,
Jassi

 

[RaynorWolfcastle]

Originally posted by: Jassi
I have been reading up on how chips are manufactured and I am stuck on one important step. How do they fabricate capacitors and resistors on the chip with such low tolerance levels. If anyone has a step by step guide, that would help a lot.

Thanks,
Jassi

On-chip passives are definitely NOT low-tolerance. As a matter a fact, it is common for caps, resistors, (and especially inductors) to have tolerances upwards of 10%. I'm not an expert in the actual manufactiruing but I know that the loose tolerances are a problem when designing analog circuits. Digital circuits on the other hand, I imagine are more resilient to process variations.

 

[Mark R]

Some high performance analogue chips absolutely need very close tolerance on their resistors, etc. A/D and D/As particularly. This is less of a problem now with delta-sigma modulators, but before high performance digital filters were available, these devices really did need resistors with better than 0.01% tolerance.

It's done with 'laser trimming'. Essentially, the original resistors are made with a deliberately low value. Then, when the die is undergoing testing, the individual resistors are measured and bit-by-bit burnt out with a laser until the values are exactly right. However, this is a specialist process and is expensive, so is normally reserved for high-performance chips.

 

[TuxDave]

Now when we are talking about tolerance, are we talking about getting close to an absolute value or getting close matching. For high resolution A/D and D/A converters, it's never a wise decision to depend on an absolute value for passive devices. Yes, laser trimming is an option but it's also very costly. The better solution is to allow the circuit to tune itself using some self-calibrating circuitry.

 

[RaynorWolfcastle]

Originally posted by: TuxDave
Now when we are talking about tolerance, are we talking about getting close to an absolute value or getting close matching. For high resolution A/D and D/A converters, it's never a wise decision to depend on an absolute value for passive devices. Yes, laser trimming is an option but it's also very costly. The better solution is to allow the circuit to tune itself using some self-calibrating circuitry.

Exactly, typically you'd use feedback as the main method to desensitize your circuit to process variations. Another method is to use devices like varactors that essentially act as voltage-controlled capacitors to tweak your circuit .

 

[Calin]

The tolerance of different resistors in chips is indeed over 10%. However, in the same chip, the typical resistance error in a single chip is less than 1% (you can have two chips that came one after another from the production line, with a 10% difference in resistance - but if you compare resistances inside the same chip, they are very well matched)

 

[fliguy84]

Originally posted by: Jassi
I have been reading up on how chips are manufactured and I am stuck on one important step. How do they fabricate capacitors and resistors on the chip with such low tolerance levels. If anyone has a step by step guide, that would help a lot.

Thanks,
Jassi

capacitors cant be fabricated on silicon. they need AREA to increase the capacitance. so that's why u see those bad-ass capacitors on power supplies

 

[Calin]

You can have small capacitors (for small voltages) on PCB - just an enlarged line, cut across. However, if you need to stabilize larger currents, you need much more capacity

 

[Special K]

Originally posted by: fliguy84

Originally posted by: Jassi
I have been reading up on how chips are manufactured and I am stuck on one important step. How do they fabricate capacitors and resistors on the chip with such low tolerance levels. If anyone has a step by step guide, that would help a lot.

Thanks,
Jassi

capacitors cant be fabricated on silicon. they need AREA to increase the capacitance. so that's why u see those bad-ass capacitors on power supplies

You can definitely have capacitors (and resistors and inductors) fabricated on silicon. We just talked about it a couple weeks ago in an integrated circuit design class I am taking this semester.

 

[TuxDave]

Originally posted by: SpecialK

Originally posted by: fliguy84

Originally posted by: Jassi
I have been reading up on how chips are manufactured and I am stuck on one important step. How do they fabricate capacitors and resistors on the chip with such low tolerance levels. If anyone has a step by step guide, that would help a lot.

Thanks,
Jassi

capacitors cant be fabricated on silicon. they need AREA to increase the capacitance. so that's why u see those bad-ass capacitors on power supplies

You can definitely have capacitors (and resistors and inductors) fabricated on silicon. We just talked about it a couple weeks ago in an integrated circuit design class I am taking this semester.

Yeah, it's very possible to get capacitors on chip but if you're talking about 10pF capacitors, yeah, that'll kill lots and lots of space. Oh, and if you want relatively good tolerance and linearity on your capacitors, you can give up trying to use gate oxide caps and be stuck with lateral metal to metal caps. But that'll mean you'll be playing with at around 500 of fF and that's still a pretty big chunk of area.

On silicon inductors is another really bad story. The best we can do is make a spiral of metal with a tap in the middle and one on the outside. The problem with that is it will be a relatively low Q inductor.

 

[RaynorWolfcastle]

Originally posted by: TuxDave

Originally posted by: SpecialK

Originally posted by: fliguy84

Originally posted by: Jassi
I have been reading up on how chips are manufactured and I am stuck on one important step. How do they fabricate capacitors and resistors on the chip with such low tolerance levels. If anyone has a step by step guide, that would help a lot.

Thanks,
Jassi

capacitors cant be fabricated on silicon. they need AREA to increase the capacitance. so that's why u see those bad-ass capacitors on power supplies

You can definitely have capacitors (and resistors and inductors) fabricated on silicon. We just talked about it a couple weeks ago in an integrated circuit design class I am taking this semester.

Yeah, it's very possible to get capacitors on chip but if you're talking about 10pF capacitors, yeah, that'll kill lots and lots of space. Oh, and if you want relatively good tolerance and linearity on your capacitors, you can give up trying to use gate oxide caps and be stuck with lateral metal to metal caps. But that'll mean you'll be playing with at around 500 of fF and that's still a pretty big chunk of area.

On silicon inductors is another really bad story. The best we can do is make a spiral of metal with a tap in the middle and one on the outside. The problem with that is it will be a relatively low Q inductor.

Very true, Q-factor is a huge problem for on-chip inductors; I took an RF microelectronics class this semester and we spent a week just on modelling and use of on-chip inductors. At frequencies above 2 GHz or so, the models are horrific because of all the loss mechanisms. What's worse, inductors have a self-resonant frequency above which the inductor starts looking like a capacitor!

As an aside to those interested, in many (analog) RF circuits, the bulk of the die space is used by passive elements, not transistors.
Image
Here is an example of a 5.4 GHz amplifier (something you might find in an 802.11a card). Look at how huge the inductors are compared to everything else!

 

[f95toli]

It is possible to fab caps with relatively large values on Si if you use materials with a large dielectric constant like STO, BSTO, SRO as the dielectric. This can be used for example in microwave circuits. It makes the processing much more complicated but it can be done.
However, as far as I know there is no commerical process that uses these materials, maybe in a few years.

 

[LithographWoker]

Embedded capacitors and resistors have great potentials for high frequency, high density

 

[nellienelson1]

Originally posted by: Mark R
It's done with 'laser trimming'. Essentially, the original resistors are made with a deliberately low value. Then, when the die is undergoing testing, the individual resistors are measured and bit-by-bit burnt out with a laser until the values are exactly right. However, this is a specialist process and is expensive, so is normally reserved for high-performance chips.

yeah the company i worked for last year (goodrich engine controls) in birmingham uk had a little fab effort in a clean room and they had a brace of the laser cutting machines, old tech ( late 70's) but quite cool none the less.