About the tolerance of the components

[Ghitza]

Probably it is a stupid topic but I see in a lot of schematics resistors, caps and others that have tolerance from 0.xx till 20%.
If in a circuit the capacitors are rated at 5-10-20% tolerance, is there a problem if I replace them with others with 1% tolerance or lower?
The same question applies for resistors and other components.
Kind Regards!

[james_s]

No, you can always replace a part with large tolerances with one that has lower tolerances, it never hurts to make something more accurate than it needs to be. It's going the other direction that's likely to cause you problems.

[AaronD]

Don't know how common this really is, but I heard from an older college professor about "binning" parts from the same assembly line into different tolerances.  The ones that are within the tightest tolerance are picked out first, the ones that fail that go on to the next looser tolerance, and so on.  The result is that a 5% binned resistor is guaranteed to NOT be within 2% or whatever the next tighter tolerance is.

Of course, there are other factors at play, like temperature stability, so that practice is probably not universal.  But it *is* something to think about.

[Benta]

Don't know how common this really is, but I heard from an older college professor about "binning" parts from the same assembly line into different tolerances.  The ones that are within the tightest tolerance are picked out first, the ones that fail that go on to the next looser tolerance, and so on.  The result is that a 5% binned resistor is guaranteed to NOT be within 2% or whatever the next tighter tolerance is.

Of course, there are other factors at play, like temperature stability, so that practice is probably not universal.  But it *is* something to think about.

That used to be true, but I doubt that it's the case today. If you look at the construction of resistors, 5% parts are different from 1% parts, and usually belong to different product series. The same goes for caps.
The only case where I've seen binning today is with LEDs, but that's different. The LEDs from one run are practically identical and will be binned to color/brightness or temperature/brightness according to the datasheet bins, meaning that when you get a reel of parts they'll all be in the same bin.

[David Hess]

Don't know how common this really is, but I heard from an older college professor about "binning" parts from the same assembly line into different tolerances.  The ones that are within the tightest tolerance are picked out first, the ones that fail that go on to the next looser tolerance, and so on.  The result is that a 5% binned resistor is guaranteed to NOT be within 2% or whatever the next tighter tolerance is.

That is the exception rather than the rule.  I have never encountered it myself.

[TimFox]

As an example of production tolerance:  with conventional 3 or 3.75 W wire wound leaded resistors of similar appearance from Vishay/Dale, I have found empirically that both the 5% and 1% parts “usually” fall in the same 1% band, which is not surprising since the automated winding process should be well-controlled.  I can’t guarantee that, of course.

[ambrosia heart]

If cap is used for saftey reason, don‘t
change any spec., follwing service manual.

For other reason, go ahead.
1% more expensive  than 10%-20%.

[Jwillis]

The tolerance is how close to the advertised value the component should be .
Precision and tolerance are two different things. usually you would choose component attributes  based on application .
Depending on the power , a 20% value difference on a power supplies filter caps won't make much difference to ripple  . But you may need a narrow tolerance for an oscillator or band filter .  A current sense resistor needs a better tolerance and low ppm so the measurement is accurate and doesn't drift. But in most other applications a standard 1% or 5% will work just fine without the added expense .
Theirs also other factors when choosing components . Understanding what those other factors are and how they will effect  your project can be just as important as the value .

[gf]

I wonder what kind of deviation is actually meant if the datasheet does not tell explicicly?
Is it the maximum deviation, standard deviation, 2 sigma, 3 sigma, etc. ?
For instance, if I order 100000 components with 1% tolerance, is it guaranteed that all of them are within 1%, or do I still have to expect outliers?

[srb1954]

Probably it is a stupid topic but I see in a lot of schematics resistors, caps and others that have tolerance from 0.xx till 20%.
If in a circuit the capacitors are rated at 5-10-20% tolerance, is there a problem if I replace them with others with 1% tolerance or lower?
The same question applies for resistors and other components.
Kind Regards!

Generally one can chose a closer tolerance part without ill effect but there are a few cases where you need to be careful with your choice of components. Sometimes parameters other than component value and tolerance were considered important to the circuit function by the circuit designer but are not necessarily recorded on the schematic.

For example, an SMT resistor specified as 5% tolerance will probably use thick film construction, which would have a better pulse power rating than an equivalent sized  thin film 1% resistor. In this case the substitution of a closer tolerance part could have deleterious effects on the circuit operation.

[TimFox]

A common safety phrase:  "All other things being equal...".
Other parameters besides tolerance include power rating, parasitic reactance, etc.

[james_s]

I wonder what kind of deviation is actually meant if the datasheet does not tell explicicly?
Is it the maximum deviation, standard deviation, 2 sigma, 3 sigma, etc. ?
For instance, if I order 100000 components with 1% tolerance, is it guaranteed that all of them are within 1%, or do I still have to expect outliers?

It's guaranteed that all of them will be within 1%. Many of them are likely to be much closer, but anything that is outside of the tolerance will be rejected.

[mariush]

[wraper]

If cap is used for saftey reason, don‘t
change any spec., follwing service manual.

For other reason, go ahead.
1% more expensive  than 10%-20%.

Nonsense. You can always safely use the same type with tighter tolerance spec. Even if it's a safety critical application.

[wraper]

Don't know how common this really is, but I heard from an older college professor about "binning" parts from the same assembly line into different tolerances.  The ones that are within the tightest tolerance are picked out first, the ones that fail that go on to the next looser tolerance, and so on.  The result is that a 5% binned resistor is guaranteed to NOT be within 2% or whatever the next tighter tolerance is.

It's not quite like that. Even if different tolerance parts are produced at the same time, it does not mean that all of the high spec parts will go into the high spec bin. Factory will bin as many high spec parts as they need, and the rest will go to cheaper lower spec bin. Even though there may be a lot more high spec parts, it does not mean that factory will sell them as such. It all depends on how much demand there is for each bin. Also you should expect that parts like resistors will be binned to 2x tighter spec than specified and it's not usual to find them to be barely within spec.

[golden_labels]

If you are doing a repair, which means you place components one by one by hand, as a last resort you can use tolerance worse than in the datasheet, if its value is right and other parameters are the same. A 5% 1.004 kΩ resistor you have in your drawer is as good as 1% 1kΩ you would obtain, if they do not differ otherwise. Though pay attention to details: different tolerances may have different thermal coefficients or value drift from age. They may also use different technology: for example 5% and 1% resistors are nowadays usually different devices. And your multimeter is having limited accuracy and precision too.

Historically binning resistors and selling the better bin as tighter tolerance was happening. You can still find people mentioning seeing a notch in a distribution of e.g. 5% resistors: instead of a plain Gaussian curve, there is a hole between -1% and 1%. But since production technology changed, this is no longer true. And 5% resistors are often having much tighter tolerance than expected — it’s simply not guaranteed if you buy them in bulk.

[David Hess]

I wonder what kind of deviation is actually meant if the datasheet does not tell explicicly?
Is it the maximum deviation, standard deviation, 2 sigma, 3 sigma, etc. ?
For instance, if I order 100000 components with 1% tolerance, is it guaranteed that all of them are within 1%, or do I still have to expect outliers?

Since the parts are tested, all of them should be well within their tolerance boundary.  This is part of the reason higher tolerance parts cost more; the tests cost more.

[Vovk_Z]

It's guaranteed that all of them will be within 1%. Many of them are likely to be much closer, but anything that is outside of the tolerance will be rejected.

I'm not sure, statistics is not my thing, but if Dave says they are made "6-sigma" then 1 out of 1000 000 can be out of its tolerance. (?)

[TimFox]

Automated test equipment has settings for max and min, so if the manufacturer sets a bit of guard band at each end of the tolerance range, then literally zero should be out of spec.  Since Gaussian noise has an infinite peak-to-peak value, you need to average for a sufficient time to keep sigma small enough for the ppm or ppb reject rate you can tolerate.

[Vovk_Z]

Are resistors being tested?

[wraper]

Are resistors being tested?

Every single part is tested, except connectors.

[golden_labels]

Every single part is tested, except connectors.

Fuses are tested too. If it blows, it was good. If it does not, it’s discarded and resurfaces on eBay.

Aside from testing each part, which may or may not be possible and the extent of the testing may differ, there are ]statistial process control methododologies. Those allow detecting or predicting production process failure without testing each part produced.

[wraper]

Every single part is tested, except connectors.

Fuses are tested too. If it blows, it was good. If it does not, it’s discarded and resurfaces on eBay.

You can test their resistance. No need for destructive test except samples.

[golden_labels]

I thought three will be enough to convey that was a joke. 🤦

[AaronD]

Fuses are tested too. If it blows, it was good. If it does not, it’s discarded and resurfaces on eBay.

Joke or not, it's still a good point about the reliability of parts from eBay and similar sites.