Component tolerance ratings?

[GEOelectronics]

Consider just resistors but this question applies to all components that are marked with tolerance data.

In the early days of solid state as we know it today, transistors were obviously made in large batches, then tested before a part # was assigned. That's why we had thousands of 2N#'s to deal with in he germanium days.

Now that I have access to a multiple decimal place DMM, after testing many hundreds resistors it almost seems like their markings are merely a guideline. Testing a batch of a few hundred identical really ancient carbon resistors have all changed significantly and in the same direction.

Does the marking indicate the testing, sorting or manufacturing accuracy, or perhaps all 3.
A resistor marked 1% upon testing will be clustered around the marked value in a Gaussian Curve. But what about it's tempco and long term aging effects? Are these caveat emptor only?

Really old wire wound resistors of the .1 and .05% class are testing close to the marked value.

If a person were to take an unmarked resistor and test it with the intention of labeling it, would the tolerance be stated as a statistical probability of the testing instrument and procedure, or would it reflect the accurate testing of a large number of samples?

Are .05% components "better" than 1% components or have they just been measured better?

And finally if one were to label a resistor that read 10.0010 Ohm, how would he label it and the tolerance.

Thanks

George Dowell

[David Hess]

In the early days of solid state as we know it today, transistors were obviously made in large batches, then tested before a part # was assigned. That's why we had thousands of 2N#'s to deal with in he germanium days.

This is still the case.  Besides hfe variation, transistors are also grouped into separate part numbers for breakdown voltage for instance and parts which do not meet a stringent specification may be binned as a much more general purpose and poorly specified part like the 2N3055.

Now that I have access to a multiple decimal place DMM, after testing many hundreds resistors it almost seems like their markings are merely a guideline. Testing a batch of a few hundred identical really ancient carbon resistors have all changed significantly and in the same direction.

Does the marking indicate the testing, sorting or manufacturing accuracy, or perhaps all 3.
A resistor marked 1% upon testing will be clustered around the marked value in a Gaussian Curve. But what about it's tempco and long term aging effects? Are these caveat emptor only?

The tolerance includes all of the effects you mention over a given specified time and load.  For instance Vishay makes 2% carbon film resistors with a -250ppm/C temperature coefficient which are suppose to stay within 2% over 1000 hours at maximum dissipation.  Over 8000 hours, they could change 5%.

Better tolerance resistors will use different construction and have a better temperature coefficient.  For instance, 1% carbon film resistors are practically unheard of.

If a person were to take an unmarked resistor and test it with the intention of labeling it, would the tolerance be stated as a statistical probability of the testing instrument and procedure, or would it reflect the accurate testing of a large number of samples?

I would say the later insofar as the distribution is calculated but the temperature coefficient is important also.  See below.

Are .05% components "better" than 1% components or have they just been measured better?

There is a pretty big difference between 1% and 0.05%; they will be different parts.

And finally if one were to label a resistor that read 10.0010 Ohm, how would he label it and the tolerance.

If you measured the temperature coefficient also then that would give some idea of the tolerance.

[Kleinstein]

Carbon resistors are known to drift up over time. So having all the resistors reading high is not by chance, but more or less normal.

Very tight tolerance resistors need to be stable, but one sometimes also gets very stable resistors with higher tolerance, especially those not trimmed to a nominal value. Still they tend to trim the better ones (e.g. lower measured TC).

A similar thing happens with semiconductors: there may be essentially the same part in a cheap grade and in a better tested grade with similar typical ratings but better guaranteed specs. This especially happens with parameters that are a little slow to measure (e.g. LF noise or low leakage currents). So you pay the premium for testing, not for really better part - only very few parts might fail the test.

[GEOelectronics]

"Quote

    Q: Are .05% components "better" than 1% components or have they just been measured better?


A: There is a pretty big difference between 1% and 0.05%; they will be different parts."

For example, you must mean the construction and materials would be different in a fundamental way, eg. carbon vs film vs wire-wound, then the laser trimmer and hermetically sealed variety etc.
That makes sense and it also makes sense that there is no need for precision resistors in a consumer product when a mass produced, cheap specimen "works fine", not to mention being an affordable product.

In a batch of vintage bobbin wound wire resistors I tested, each was hand marked with value and tolerance. Same construction, just tested and graded after the fact as it were. They varied from 1% to .1% and a few were .05%. Whatever instruments they used back then were remarkable, and the parts today still have the same values and tolerance.

It's been an eye opening experience for me to see the fine details of measurement possible in he home lab today.
In many ways I feel like Rip Van Winkle Waking up after a long nap o find the world changed so much.

I look forward to joining the trend.

George Dowell

[hamster_nz]

I think what you are asking is covered here:

http://dangerousprototypes.com/blog/2010/07/01/actual-values-of-10-tolerance-resistors/

(Although some of the links to original sources are dead)

[GEOelectronics]

Yes Hamster_NZ, that link "fills in the gap" quite well.

I test Zeners and TVS's  too, the gap he mentions is present in large samples of those too.\\

George Dowell