How to tune the devices parameters in process?

zenner zapping

Some analog IC require a precision resistor. Some datasheet says "... are fine-tuned". I want to know how a resistor is fine-tuned. Is is a automated process in fab or must be done manually? To be precise, what I really want to know is how much the cost will increase if I add a fine-tuned resistor in the layout. Does the cost increase with the number of fine-tuned devices?

zener zap

Resistor tunning is always expensive.

You can do by laser (ablatering the resistor) or in steps by shorting or opening resistor sections that you previously put.

Make a seach for zenner zapping, antifuse, fuse tunning

What kind of circuit do you wanna tune?

There are some technologies with resistors of good precision but with fixed values.

current trim fuse

Hey, it's not so expensive - really, the main cost is the die size used for the trim pads, right? And test time, but that's pitifully cheap.


Hughes - a couple notes that may help.

1) To laser trim, your fab must be able to make thin film resistors intended for trim. (Usually any thin film means they intend for laser trim - NiChrome, SiChrome, etc) Poly resistors are not intended for laser trim, and diffusion resistors definately arent!

2) Now your assembly or test house must be able to laser trim at wafersort. Say you make a 10um wide resistor, measuring 8-12k out of fab, and you want a 20k. During trim, the precision voltage is measured while they shoot little holes in a line across the resistor. This narrow part causes the resistance to increase in rough chunks until you reach about 19k or so. Now they cut lengthwise (to make an L shape overall) until the resistor is finely trimmed to 20k.

The other methods use a bunch of resistors in a divider network. For fuse trim, you make your 20k precision resistor a 15k, then you add a 5k, 2.5k, 1.25k, 0.75k on top. These little trim resistors are all shorted out by fuses out of fab, and to trim the whole network to 20k, you blow some fuse links with a hi-current pulse, opening each link and adding the little resistor onto the main resistor.

Fuse trim kinda sucks in my opinion...
1) takes a lot of current which also gets pulsed thru your small resistor (imagine trying to put 7v across a 0.7k - your little resistor leeches 10mA and can be damaged)
2) It is somewhat "Blind". You don't know whether your 0.75k resistor is 0.7 or 0.8 until u zap the fuse. Your trim software needs a learning algorithm, or you'll lose 2-5% yield due to mistrim

Antifuse like zener zap is better in some ways. Zener zap uses a good old npn or pnp like an E-B diode. Top of diode goes to the higher voltage, bottom of diode goes to lower voltage. In normal operation, this reverse-biased diode sits across your 0.75k and doesn't allow any current thru - hence the 0.75k is in the circuit. To short out the trim resistor, you hit the zener with about 10-30v, and about 50-300mA. This causes zener breakdown of the EB junction, heating up the zener until metal flows from the contact and shorts it out. Now it's shorted out and you just trimmed DOWN 0.75kOhm.

Each trim style has it's own approach. I've found that zener zap and fuse trim are often used because they are CHEAP! The fab does not need to make any new process (as long as npn exist) and even at the cheapest assembly houses in china, they can still zap links very easily.


All of this assumes you are trimming a bandgap or something. If you really want a precision resistor you may need thin film & laser trim. If you just want a perfect 4/1 ratio for gain, then many other resistors can give nice matching. Try poly or cap poly (poly2) for pretty good accuracy, or even diffusion for bad accuracy but still nice matching.


Check out the book "art of analog layout" for more info on each type of trim.