About fuse trimming !

[elone]

I need a accuracy resistor in my design, so i want to use fuse trimming. But how to design this trimming format! Should use metal fuse or poly fuse?
where to find some docs about fuse trimming!

 

[Colbhaidh]

Metal fuses are better for processing on Silicon. These are usually top metal - 1. Most foundries support these in their design rules (TSMC, UMC, Chartered Semi).

 

[sunking]

Both are ok, it depends on the foundry.

 

[survivor]

Hi,

Metal fuse is much better than poly if you try to use laser to cut it. If you try to use current across the fuse to burn it out, poly is the only choice.

 

[Hughes]

Is the fuse poly different from the gate poly?

 

[frankon_o2]

how about the temperature coefficient of the resistor? Maybe you could only attain an accurate resistance under typcial temperator if it vary a lot with temperature.

 

[electronrancher]

use metal fuses for accuracy - poly *might* work, but putting 100Ohm of poly in parallel with a 1k resistor is going to be pretty crappy. better to use 0.1ohm of metal in parallel with that 1k.

actually, 1k is an easy bit to trim. something below 200Ohm begins to be difficult. i suggest metal.

oh, and metal fuses blow just fine using current - i'm not sure what survivor saw, but most fuses can be blown with 5v, 250-500mA. This is actally pretty easy on prober.

I have only seen poly fuses used for selection (digital), or digital post-package trim, where they again are used for digital selection. They vary too much to be used for anything except 1/0.

 

[andy2000akimo]

digital post-package trim　．．．how to trim under package ?? use large nmos switch and 300ma for metal fuse/poly fuse ?

 

[dick_freebird]

You should be able to blow a poly fuse with a few tens of mA.
That's easy with a probe pad. Not so nice if you have to have
a programming switch on chip, that'll be the biggest piece of
it. Poly is pretty stable. Some metallizations can form whiskers
and regrow over time. Thin films blow with less energy. If your
interconnect stack has a thin film resistor in it somewhere,
this can also be a viable candidate.

Gate zap is another option, there you only have to handle
voltage which needs smaller, but possibly non-qualified
devices.

Your foundry may have figured this out for you. If it hasn't,
the qual is on you. Presuming this is a product and not a
science project.

For any of these, the real design is inthe sense circuit (and
the programming switch, if any). The element itself is almost
always trivial.

 

[laiza]

Hello, in test we encountered a temperature related Trimming issue. Perfectly trimmed at ambient while having a temp drift in high temp. Do you know where to attack this kind of scenario?

 

[dick_freebird]

You would want both tempco and value trims (trimming tempco will bend
the value and you will need to bring it back in). A two-pass test is needed,
two temps (at least) to determine the tempco and null it, and then the
value trim afterward (both trims within the final pass).