0.01% resistors

Farnell has got a few parts at a slightly higher prices - 10k, 1k, 100R.

Try contacting Susumu directly. We're paying 30 cents for 0.05% resistors.

Once, long ago, someone decided to use 0.01% resistors. We found the rpcing in US dollars at $1 to $1.50 Delivery data, now this was the 'interesting' part, 6 to 9 months, but NEVER arrived !!!!

It was cheaper to simply put in excessive parts and let a laser automatically cut unwanted ones out. That was assembly was constant, and the tester program did the 'adjustment'

In retrospect, I'd push for the 'design-it-out' principle. Go back to Engineering and beat them about the head and shoulders until they had designed out any problem like that in Manufacturing. ;)

If there's a microprocessor in the loop, make sure that you keep things within range of your ADC (or whatever), and calibrate by changing constants in flash.

It does require tweaking, but it can be automatic tweaking.

The big dollars are not for close tolerance resistors but for resistors that earn that close tolerance number by having commensurate low tempco. Without that you're going to get noise and nasty transients as temperatures change, at best, even if you calibrate out the temperature drift.

Best regards, Spehro Pefhany

The 30-cent Susumu thinfilms seem to stay below 10 PPM/K, and you can get a few-PPM voltage reference for a couple of dollars.

True. I always thought that there's a market for matched resistors on the same chunk of ceramic, like matched transistors. Either the market isn't as big as I think, or it's just not possible to get good enough matching to make it worthwhile.

Farnell stocks a few precision resistor arrays, so your first thought is co rrect. I bought a couple for my low distortion oscillator last year.

These are pretty impressive, unique parts as far as I know.

I have the same requirement (let me know if you find anything!)

There seem to be Vishay "bulk metal foil" types north of $10.

Farnell have a few cheap "0.01%" resistors of unknown origin and specification, very peculiar.

Rhopoint seem to specialise in precision resisistors. There are some manufacturers of lower cost (than Vishay) bulk metal foil precision resistors, as well as other types.

It is very tempting here to use Larkins susumu types. Problem is if you take the worst case datasheet limits on anything bar the metal foils, they always seem incredibly pessimistic. But it seems ridiculous to put say two $10 chip resistors on a board otherwise full of sub-$0.001 parts and a $4 120MHz ARM.

Unfortunately, the ones that are 50x better cost about 50 times more.

Best regards, Spehro Pefhany

They are out there.. matching "typical" +/-0.1ppm/°C, but not cheap (~$20 pair).

It's strange when the resistor part of a BOM cost dwarfs that of the semis.

Best regards, Spehro Pefhany

It's not hard to have $1,000 worth of resistors on a board with semis worth 5-10% of that.

Best regards, Spehro Pefhany

Thanks, that's interesting, and they're relatively inexpensive.

One of the dangers of using parametric search is that I'd missed those due to their rather poor absolute tempco (25ppm/K), but that tempco (typical 8ppm) is matched typically to within +/-2.5%. Not quite as good as the Z-foil types (0.1ppm/K typical matching with

Best regards, Spehro Pefhany

Simplifies the buyers job. Forget everything else, just shop around and negotiate for a couple of resistor types.

I have recently (belatedly) realised that is already the case for ICs.

I wanted to make precision op-amp gain stages (single ended). The LT5400s come in a handful of value combinations, so I got to modify my "parallel-series resistor combination" program to work with the possible networks of 4 resistors.

What combination of 4 resistors come closest to a specified ratio?

I sketched them and got 19 combinations (plus their reciprocals)

"a+b", "a + (b+c)", "(a+b) + c", "a+ (b+c+d)", "(a+b) + (c+d)", "(a+b+c) + (d)", "(a|b) + c", "a + (b|c)", "a + (b + (c|d))", "(a+b) + (c|d)", "a + ((b|c)+d)", "(a+(b|c)) + d", "(a|b) + (c+d)", "((a|b)+c) + d", "((a|b) + (c|d))", "(a) + (b|(c+d))", "(b|(a+c))+d", "a + (b | (c+d)) FB from c-d", "(b | (a+c)) + d FB from a-c"

"|" means "in parallel with" "+" means "in series with"

Then a dumb C program to try them all.

Later extended to 2 and 3 gain stages.

If you use a common-centroid layout, you can get pretty good ratio stability with garden variety ones.

Cheers

Phil Hobbs

Pease on this stuff:-

Irritating that it's so difficult to even come close to the tempco of a well-designed and well-constructed rack mount ratio transformer. (< 1ppb/°C typically)

This is my version:

I think I have one somewhere for voltage dividers, too.

We try to use quad r-packs as often as possible, but the hazard is that they lock you down to values and ratios.