Resistor (and other toleranced passives?) manufacturing selection process

[AlfBaz]

I was reading up on worst case and Gaussian spread simulation and found in the comments of one blog that manufacturers make the one resistor batch, measure them and sort them into separate tolerance bins.

This led me to find an actual picture of the mangled Gaussian distribution in this post here
http://dangerousprototypes.com/2010/07/01/actual-values-of-10-tolerance-resistors/

Is this actually what the do with toleranced passives and at what tolerance do they stop taking the cream out of the middle. In other words does a 1% batch have the 0.1% range taken out and so on up to their tightest tolerance range?

[David Hess]

I was reading up on worst case and Gaussian spread simulation and found in the comments of one blog that manufacturers make the one resistor batch, measure them and sort them into separate tolerance bins.

This led me to find an actual picture of the mangled Gaussian distribution in this post here
http://dangerousprototypes.com/2010/07/01/actual-values-of-10-tolerance-resistors/

Is this actually what the do with toleranced passives and at what tolerance do they stop taking the cream out of the middle. In other words does a 1% batch have the 0.1% range taken out and so on up to their tightest tolerance range?

Not all manufacturers do this.  I have never run across it myself.

[Len]

I was reading up on worst case and Gaussian spread simulation and found in the comments of one blog that manufacturers make the one resistor batch, measure them and sort them into separate tolerance bins.

I've heard about that but never seen the data to back it up.

When Dave tested a bunch of 1% resistors he found that they follow a normal distribution more or less. No evidence of being selected from a larger-tolerance batch or having smaller-tolerance resistors removed.

[Rufus]

I was reading up on worst case and Gaussian spread simulation and found in the comments of one blog that manufacturers make the one resistor batch, measure them and sort them into separate tolerance bins.

I've heard about that but never seen the data to back it up.

When Dave tested a bunch of 1% resistors he found that they follow a normal distribution more or less. No evidence of being selected from a larger-tolerance batch or having smaller-tolerance resistors removed.

Probably because the stability and temperature coefficient of resistors produced by that manufacturing process would be inadequate for 0.1% so selected good ones would not be good enough anyway. If he had tested hi-stab 0.1% resistors also available in 0.05% tolerance I would not be surprised to see a hole in the 0.1% distribution.

If they have a market for 1 million 0.05% resistors and 10 million 0.1% resistors and they make 11 million resistors do the first million that are within 0.05% get selected or do they tighten the limits to pick the best from the whole batch? Do they select all 0.05% resistors and if they get 4 million of them sell 3 million as 0.1%?  The 0.1% distribution might show a hole narrower the 0.05%, it might show a 0.05% wide depression, or something in between, maybe you will get a reel of 0.05% resistors marked as 0.1%.

I would say holes or depressions in the distribution from selection is a possibility but not a rule. It depends on the manufacturing process and market.

[David Hess]

I would say holes or depressions in the distribution from selection is a possibility but not a rule. It depends on the manufacturing process and market.

I would say it depends on how iniquitous the supplier is.

[Zero999]

The article linked by the original poster mentions 10% tolerance resistors and Dave measured 1% tolerance resistors. I suspect 10% tolerance resistors could be the rejects from the 5% process but who makes 10% tolerance resistors these days? Potentiometers and power resistors are the only resistors I know of which are still manufactured with a wide tolerance

[David Hess]

Carbon film resistors are available in 5%, 2%, and 1% tolerances.  Carbon composition resistors are available in 10% and 5% tolerances.  They are the ones I would expect to suffer from this problem.

[jmoreland79]

Is this actually what the do with toleranced passives and at what tolerance do they stop taking the cream out of the middle. In other words does a 1% batch have the 0.1% range taken out and so on up to their tightest tolerance range?

I don't think the production is that precise. 

I imagine there is a target value for a batch, let's say 47 kOhm.  Let's say they sample test 1% of the batch, and the measured resistance values fall between 45,500 Ohm and 48,800 Ohm.  Welp, that's a 47 kOhm ±5% batch.  They try another batch, and this time the measured resistance values fall between 46,900 Ohm 47,250 Ohm.  Alright, that's a 47 kOhm ±1% batch.  They try another, and this time the measured resistance values fall between 49,100 Ohm 52,250 Ohm.  Fine, 51 kOhm ±5% batch.

You would think (hope) the manufacturing process itself on a given day would yield results well within 5% of the target value.  So the variances would more likely come from the individual materials used to make the resistors.  Each yield probably produces hundreds of thousands of units at least, so bin sorting the full yield would be incredibly impractical.

[Precipice]

From what I see of resistors (SMD and small TH), they're all trimmed to value before painting / coating. I don't think there's any concept of off-tolerance, that isn't a failure that gets tossed.
I'm not sure what the process is with carbon composition resistors, though - but I'd not be surprised if they were trimmed too, for devices that aren't utter bucketshop crap.