3

I thought that I had this pretty much figured out over 30 years ago during my initial schooling, but various answers posted to a support forum of a major semiconductor manufacturer(name withheld to protect the possibly innocent) by staff support engineers (and no countering posts) have caused me to second guess myself. There are other questions concerning resistor tolerance, but none even mention the problem I talk about below.

EDIT: This is concerning the statements that the chip makers support engineers make concerning why there are differences between measurement runs. The phrasing used makes it sound like resistors change randomly within their tolerance rating. When nobody called them on this explanation(which was given over a year ago, which is why I did not post in the thread.), given by two different people, is when I came asking. The chip in question is a type of ADC. The excitement signal is give or take a 125Khz, 2Vpp sine wave with a max current around 30mA.

I always thought a resistor's tolerance rating was the permissible difference between its actual value and its posted value.
Meaning if we took a group of 100 ohm resistors they could vary as follows:
A 5% resistor could have any value that was -5 or +5 ohms from 100.
A 1% resistor could have any value that was -1 or +1 ohms from 100
And a .1% resistor could have any value that was -.1 or +.1 ohms from 100
But once it was manufactured, and discounting abuse its value would not change, due to the tolerance rating. I'm not talking about changes due to temperature, that is based on its TCR. E Nor am I talking about aging, these are measurements taken within an hour(or less) of each other.

What caused this question is that more than once in answering a question they stated that the various different values between measurement runs could be due to a resistors "tolerance". As if a resistor's value changes randomly within its tolerance rating, something I have never heard before. If they had stated it might be due to resistance changes due to temperature, I would have thought nothing of it. Or if the poster had been someone other than a tagged support engineer. And there were no posts questioning these conclusions, which messed with my head even more.

Maybe I slept through that part of class and it's such a common thing no one bothers to talk about it, but I'm pretty sure that a resistor's value does not change if there are no outside forces(e.g. temp) influencing it. Am I correct, or have I some how managed to misunderstand how a resistor works for a LONG time? Or does tolerance have a meaning I somehow just have not encountered before?
I do have medical problems, but I'm fairly sure that they have not degraded my thought processes that much(if they have, I need to stop talking).

GB - AE7OO
  • 1,043
  • 7
  • 19
  • Do you mean "a resistor's" or "some resistors"? Do you need an apostrophe somewhere? Are you talking about variation in a single resistor or between samples from a group of resistors? Can you also clarify what you mean by "measurement runs" and how much variation are you talking about? – Elliot Alderson Mar 27 '20 at 18:37
  • The *outside force* that you are neglecting is usually called *aging*, the part resistance will drift over time. If you are building a precision circuit with precision resistors, you need to account for aging. The manufacturer should list the aging factor for precision resistors. They also sometimes list a factor for soldering. The high heat when soldering the part on a board will permanently shift the resistance slightly. – Mattman944 Mar 27 '20 at 18:51
  • Do note that a 5% tolerance 100 Ω resistor *can* in fact be 106 ohms. Tolerance ratings are stochastic; they guarantee that 99.howevermany9s% will be in that range, but unless you're buying individually tested and binned parts, you may still get something out of the range. – Hearth Mar 27 '20 at 20:21
  • 1
    @Hearth What you say may be true, but I would consider it a contract violation with the vendor. The vendor specifies a tolerance, not a standard deviation. I would want my money back for that resistor. If it happens with 1-in-a-million then that is probably acceptable, if it is 1-in-a-thousand I would look for a new vendor. – Elliot Alderson Mar 27 '20 at 22:38
  • @ElliotAlderson I expect they go for something like five or six sigma at least. – Hearth Mar 27 '20 at 23:26
  • @Mattman944 I guess I need to edit it and make more clear what I mean. I don't see aging as a problem between multiple measurements taken in one day. Years maybe, but hours? – GB - AE7OO Mar 29 '20 at 04:16

3 Answers3

5

A decent manufacturer will specify pretty clearly in the datasheet what is meant. For example, here's the relevant table from one vendor I've used:

enter image description here

As you expected, the resistor tolerance is the limit on the resistance at 25 C, and the variation over temperature is covered by a separate TCR specification.

But remember the resistance can also change due to other environmental factors, such as prolonged operation at high or low temperature, operation at high voltage, mechanical stress, etc. If any of these stresses is applied to your device you could see its value vary from day to day, with the value at 25 C remaining within the specified tolerance limits.

The Photon
  • 126,425
  • 3
  • 159
  • 304
  • I"ve clarified the question a little. If the excitation signal was not at such a low level(2Vpp @ 30mA max, 150Khz) I might have been worried about confusing abuse with random change. I've never seen such a small signal have an effect on a resistor/cap/coil. – GB - AE7OO Mar 29 '20 at 04:37
  • @GB what's the resistor value and power rating? 30 mA through a 1/16 W 50 ohm resistor would be a substantial fraction of its maximum power, for example. Aside from that there could be mechanical stress introduced during soldering that slowly releases over time. – The Photon Mar 29 '20 at 04:58
  • I'm not the original poster, so I don't have a clue. But again, it's what was implied by these two support people. Let me go find the thread and post a quote from it. – GB - AE7OO Mar 29 '20 at 05:01
  • @GB-AE7OO you are the original poster of this question. – jDAQ Mar 29 '20 at 05:21
  • @jDAQ I did ask this question, but I DID NOT ask the question/thread that contains the answers I'm asking about. That question was over a year old and I ran across it during research on a ADC. – GB - AE7OO Mar 29 '20 at 09:09
0

... the various different values between measurement runs could be due to a resistors "tolerance". As if a resistors value changes randomly within it's tolerance rating ...

The resistance of a resistor should not change within minutes (unless you strained it with a lot of of heat or something). They are probably referring to different effects (which I am not even sure can create that fast variations). Tolerance does mean the variation from the specified value, the one in the product description, and the actual one, which you can measure.

That is, if you measure the resistor a few times in the same day all the readings should give you very close readings, and the only differences will be due to equipment precision, calibration or temperature variation.

jDAQ
  • 2,517
  • 1
  • 9
  • 19
  • Which is why I made sure to include "But once it was manufactured, and discounting abuse its value would not change, due to the tolerance rating." – GB - AE7OO Mar 29 '20 at 04:30
  • It shouldn't change in the common situations you described, and if those support guys said so, ask for a clarification. They might have meant that they vary over a production lot. But resistors are not quantum stuff or random processes. – jDAQ Mar 29 '20 at 05:23
  • I did not ask the original question on the support forums. I had a browser crash and I'm trying to find the question, so I can quote specific parts. – GB - AE7OO Mar 29 '20 at 09:11
  • Hi GB - AE7OO, Your comments please. – vu2nan Jun 26 '20 at 10:09
0

First of all it's necessary to understand the need for multiple tolerance bands for a component.

For the supplier (component manufacturer), it would be a means to reduce inventory by categorising and differential pricing. The supplier would aim for a nominal value and would need to get out-of-range components (caused by process variations) within range, by trimming. Likewise trimming process variations would necessitate checking & sorting into various tolerance groups.

For the user (product manufacturer), it would enable choice of narrow- or wide-tolerance parts based on the criticality or otherwise of the nominal value to the application. It goes without saying that narrow-tolerance components would be priced much higher than wide-tolerance ones.

Out of spec components/products could be the result of machine-setting not centred on the tolerance band and repeatability issues in the manufacturing as well as test & measurement processes.

vu2nan
  • 15,929
  • 1
  • 14
  • 42