Resistors and tolerances-actual value of those 10k 20%?

[brian]

I don't remember ever seeing much about this sort of topic, but then again, I haven't checked up on much around here for a very long time. Anyway, I'm just finishing up a statistics and probabilities course and some of the topics in the course reminded me of a story which led to this post. Bear with me, please.

I had a friend who told me that he once held a job at a manufacturing facility testing resistors for total impedance. He said that depending on where the actual resistance fell with respect to the expected value, the individual resistors were sorted according to how far from the value they actually measured to. For example, if they were making 10k resistors and he measured a value of 10.5k, it went into a 1% bin and if he measured 12k, it went into the 20% bin, etc...If this is the case, you would not expect any 10k 20% resistors to measure exactly 10k because all of those resistors wind up in the higher tolerance bin.

The wind up is this, if his story is true it makes sense to me because instead of trying to guarantee certain tolerances in the manufacturing process, the manufacturer simply makes the product with natural variations and then sorts the product later. If this is the case, I wouldn't expect any 20% resistors to ever actually measure 10K for example, because if they were actually 10K, they could be sold as 1% tolerance for a higher price. Therefore, if you're buying 20% resistors, should you ever expect any of them to be closer than 11% from the expected value? If the process is as I have illustrated, the answer is certainly no. If things are this way, buying 20% tolerance resistors means you will never be closer than say 11% from the expected value of resistance for a given resistor.

Can anybody support or defeat this with personal experience? I would be curious to know if my idea is correct.

Thanks,

Brian

[Pedal love]

That could be true, but you must remember, the discretion of a manufacturer's straight 20% is normally different for the selling company. Theres normally more filtering of real values by consumer time, otherwise the sellers would not be in business. Try different dealers until you find something you trust.pl

[seanm]

I would agree for 20% because I think enough 20% resistors would be made to meet demand.

However, I could see where they might need more 10% than where produced and had to dip into the 5% bin to meet demand.

[R.G.]

This sequence of events was in fact true at one time, but it is not any longer.

It was especially true of our magic-mojo friends, the carbon composition resistors. Carbon comps are made by stuffing a special recipe of powered carbon particles and binders into a cavity inside the resistor body, where it was compressed into place and the binder cured. If everything was right, you got a normal distribution of resistor values centered on the nominal value of the resistor you wanted.

The initial manufacturing processes turned out parts that were mostly within 20%. To get closer tolerances, they'd sort them and sell at a premium parts that were 10% and 5%. I don't know if 1% parts were ever done this way because carbon comp can drift more than that from soldering.

So with bad-old-days carbon comps, you could be pretty sure that a 10% resistor was NOT within 5% of the nominal value, and if you valued your modestly paid engineering job, you'd design to accept any part within the 10% range, with most of them on the outside edges.

However... manufacturing got better. 20% parts died out as the yields all got within 10%, then better than that with tighter controls. Then came carbon film.

Most of today's carbon film resistors are a film of ... yep, carbon... deposited on the outside surface of a cylindrical ceramic body in a spiral. The spiral lets you get several different resistors out of the same mix of film-gook by controlling the width and fineness of the spiral. It also lets you trim parts by laser if you like. Carbon film makes 5% easy, 2% reasonably easy.  That's why carbon comp died out - the cheaper to make carbon film parts had higher performance, lower drift, less noise, lower voltage coefficient of resistance and generally better everything except vintage tube amp mojo and that neato phenolic smell when they overheat.

Wire wound resistors and later metal film were always the part of choice for high precision. 1% metal films are fairly cheap because 1% is a crude value for metal film. 0.1% and tighter are available - but your boss will fire you if you specify costly high precision parts when a 20-millicent carbon film will do. Unless of course you work for military design teams where using extra gold and platinum earns you a bonus.

So yep, used to be true, at least in some places, not any more.

It may still be true in some capacitor lines, but I doubt it is a regular occurance with modern parts.

On the other hand, the only guarantee you get is what the specifications say. So if a manufacturer provides you parts specified at +/-10% and all of them are between +9.5 and +10%, he has fullfilled his end of the bargain. Anyone who claims to "design" in a fashion other than easter-egging ... oops, creating their own beautiful reality in electronics... had better learn to design with tolerances taken into account if they want to be able to predict what will happen when the "on" switch is flipped.

[brian]

Thanks, R.G. I was hoping to get a good answer from someone like you on the topic. As my stats professor keeps telling us, "Guys, everything is a random variable!" A good thing to keep in mind when designing/building.

[NaBo]

Unless of course you work for military design teams where using extra gold and platinum earns you a bonus.

LOL!!!  So thats where the millions and billions of dollars go.  Gold film, platinum leaded, 0.00001 tolerance resistors!  

[Torchy]

I was an engineer in the RAF.

In B&Q you could buy a panel screw for 20p.

In the RAF it cost £100, but you got a "manually inserted friction dependant panel securing and fastening device".

In its own plastic bag. And no, I'm not joking