Resistor sub tolerance

[Govmnt_Lacky]

What is the 'acceptable' percent tolerance for resistor substitution? Been a while since I have had to substitute so....

+/- 5%.......... +/- 10%?

(First with the correct answer gets to point and laugh) 

[EBK]

It will always depend on what the circuit is.  Is there a specific problem you're trying to solve?

[garcho]

+/- 5% resistors are cheap enough, why not? that's my thought anyway. biasing transistors, filters, etc. probably could get by with 10% just fine but troubleshooting is a PITA already so better to not waste time testing every resistor.

[vigilante397]

Trust your ears, but for breadboarding/prototyping purposes I usually say 10%.

[amptramp]

If you buy ±10% resistors, understand that you are getting +5% to +10% and -5% to -10% resistors because the values within ±5% have been culled for sales as the slightly more expensive ±5% units.  The circuit will determine whether you need tighter tolerances on resistors.

[Ice-9]

+/-1% resistors or cheap enough, I can't imagine any reason not to use them. Not exactly critical in small pedal audio circuits but for the price difference it is a no brainer.

[EBK]

If you buy ±10% resistors, understand that you are getting +5% to +10% and -5% to -10% resistors because the values within ±5% have been culled for sales as the slightly more expensive ±5% units.

Is it really cost effective for a modern manufacturer to do that?  I'd imagine instead a QC with random one-per-batch sampling instead, with rejected batches being more thoroughly tested or simply being discarded.  I don't have any specific knowledge of how these things are made though, so I may be completely wrong.

[reddesert]

I figured the question - about substitution - was more like "The schematic calls for a 240K resistor and I only have typical values, is a 220K okay or do I really need to buy a 240K or make a teepee of 220K + 22K?"

Personally I think a 10% discrepancy is generally going to work in effects circuits. It might change the sound a little, but I expect a lot of the time that could be canceled out by a slight tweak of the knobs.
Some examples:

• Pulldown resistor or high-value resistor connecting an input to Vref. 10% shouldn't matter.
• RC tone circuit. Changes the knee by 10% in frequency, a bit less than two semitones. Probably not audible or neutralized by a small twist of tone knob (has anyone ever heard a pedal that sounded great in D but crap in E?)
• Feedback or gain circuit, like an op-amp feedback resistor. Could change gain by 10%. Might not be audible - circuits with adjustable gain often cover a range of 10x to 100x, so a 10% change is like a tiny fraction of pot rotation.
• Bias or threshold of a voltage-sensitive circuit like a clock or compressor. This is where I think it might matter, but you often have a trimmer anyway - think of a BBD clock or the trimmer on an Orange Squeezer. If you adjust the trimmer to be different by 10%, it does affect the sound, but that's why the circuit has a trimmer.

[Rob Strand]

Is it really cost effective for a modern manufacturer to do that?  I'd imagine instead a QC with random one-per-batch sampling instead, with rejected batches being more thoroughly tested or simply being discarded.  I don't have any specific knowledge of how these things are made though, so I may be completely wrong.

I've looked into this a few times.   There's a number of discussions on the internet.  There seems to be cases where there is no evidence of sorting and cases where there has been sorting (eg. the 5% resistors have the 1% resistors removed leading to a "bi-modal" distribution which has two humps with hole in the middle).

From time to time I have measured batches of resistors.  In the last 25 years I've only seen one case where some sort of sorting *might* have occurred.   In this one case there were about 20 to 50 parts of the same value, and there were many different part values, all from the one manufacturer.   Can't remember if they were 5% or 10%.  I did not see two humps I saw only one.  All the values were all off in one direction and none sat in the +/- 1% region.   This still doesn't imply sorting.  It just means the mean is off and still in spec.  Generally I do see 5% resistors which have some parts within 1%, in fact I've had large batches where they were all within 1%.

Some good threads,
http://www.eevblog.com/2011/11/08/eevblog-215-gaussian-resistors/
http://paulorenato.com/index.php/109
And,
https://electronics.stackexchange.com/questions/77325/reducing-the-tolerance-of-resistors-manually
https://electronics.stackexchange.com/questions/157620/can-a-10-resistor-be-used-as-a-1-resistor
http://www.kerrywong.com/2013/02/01/100k-5-carbon-film-resistor-value-distribution/
https://electronics.stackexchange.com/questions/157788/resistors-binning-and-weird-distributions

As a side issue,  some people specify 1% in order to achieve better temperature stability (say for test instruments).  However, it's really just an indirect way of specifying metal films -and maybe not the best way to do it.

[vigilante397]

I figured the question - about substitution - was more like "The schematic calls for a 240K resistor and I only have typical values, is a 220K okay or do I really need to buy a 240K or make a teepee of 220K + 22K?"

Personally I think a 10% discrepancy is generally going to work in effects circuits

That was my understanding as well, and I agree.

[highwater]

Is it really cost effective for a modern manufacturer to do that?  I'd imagine instead a QC with random one-per-batch sampling instead, with rejected batches being more thoroughly tested or simply being discarded.  I don't have any specific knowledge of how these things are made though, so I may be completely wrong.

I've looked into this a few times.   There's a number of discussions on the internet.  There seems to be cases where there is no evidence of sorting and cases where there has been sorting (eg. the 5% resistors have the 1% resistors removed leading to a "bi-modal" distribution which has two humps with hole in the middle).

I would imagine that, once upon a time, the tighter-tolerance resistors *were* selected out of batches.

Nowadays, I would expect that they just keep an eye on how well-calibrated the machine is, and move to a different tolerance range when the calibration starts to drift.

--

Back on topic:

What is the 'acceptable' percent tolerance for resistor substitution?

reddesert pretty-much nailed it. In pedal-building, at-least, it is *highly* unlikely that the specific value is that important, but I would add that there *are* cases where two resistors should be matched to each-other.

Say the schematic for a stereo pedal calls for one 24k in each channel, and all you've got is 22k -- the 22k is fine... but if you've got one 24k and one 22k, you'll probably be better-off picking an even more-different value that you *do* have two of -- for instance, a pair of 16ks or 27ks.

Another example would be a graphic EQ - if you use a substitute value for one band but not the one next to it, the two bands may not overlap properly.

Of course, even in cases like that, +/-20% is probably still close-enough for rock-and-roll - we're not going to be using a stereo pedal if both channels are supposed to be the same, we're not using a graphic-EQ to get a flat frequency response, etc.

---

One likely case where it *would* matter would be something like a voltage divider that needs to be adjusted with a trimpot (for instance, to cancel-out the carrier in a ring-modulator)... if you go too far with resistor substitutions, you might have to also use e.g. a 5k trimmer instead of a 1k.

Another would be where a too-small resistor could burn-out or lock-up an IC - like the Iabc pin on an OTA, or the delay-setting pin on a PT2399.

[Rob Strand]

I would imagine that, once upon a time, the tighter-tolerance resistors *were* selected out of batches.

Nowadays, I would expect that they just keep an eye on how well-calibrated the machine is, and move to a different tolerance range when the calibration starts to drift.

That's pretty much how I see it.   I wouldn't be surprised if some manufacturers don't do 100% testing, they just use sampling and process control methods.    The last time I check up on it, SMD resistors are all tweaked, probably with a laser.  Apparently they zap away a good deal of the resistor and the power rating would be somewhat higher if it wasn't zapped.

[BetterOffShred]

I figured the question - about substitution - was more like "The schematic calls for a 240K resistor and I only have typical values, is a 220K okay or do I really need to buy a 240K or make a teepee of 220K + 22K?"

Personally I think a 10% discrepancy is generally going to work in effects circuits

That was my understanding as well, and I agree.

This. 

I think some circuits are far more finicky about values, and then you throw power supply variance into the mix and it gets even more interesting.

That being said, I've followed the 10% rule thus far and it's been smooth sailing.  That isn't to say if I had the correct values the effect wouldn't sound better, but am I going to notice ? Probably not 

[R.G.]

The real answer?

Easy: you have to know what the resistor does in that circuit. 

Fact is, some resistor positions can take anything between 1K and 1M and work pretty much OK. Others, you really have to worry about the long term drift of a 1% type (although, granted, there are many fewer of these last in the pedestrian world of pedal making).

It's shockingly like the song "The Gambler": you got to know when to hold 'em, know when to fold 'em.

That is of course pretty far from the answer the OP wanted.  Probably the most expedient thing to do is guess, or make up some series/parallel combination that's not too far off.

On the select-out-the-middle story was once true, especially when carbon comp and wirewound were the only affordable resistors, and the laser hadn't been invented yet. There was a premium for 5%, another premium for 2%, and 1% was not used for common everyday stuff. We're talking 1960s here. In the 1970s, manufacturing technology got better, distributions tightened up, and 5% became the norm, taking over from the 10% from the 60s. Once intelligent, automated manufacturing really took off, precision resistors got cheap.

[garcho]

100 pieces, 1% tolerance, metal film, 1/4W, 10k, through-hole resistor = a wallet-whopping $1.90 at Mouser.

100 pieces, 5% tolerance, carbon film, 1/4W, 10k, through-hole resistor = a bank-burning $1.20 at Mouser.

other, less prolific values will be more expensive. 100 1/4W 68k or 7.68k or whatever @ 1% will set you back between 3 to 4 bucks, most of which you can get still get 10 for less than a dollar.

100 pieces of 10k @ 0.01% will cost you a cool $667.00

[PRR]

Very early, they put carbon ink on paper. Later they mixed coal dust with clay in a big pot, and couldn't stir it too well. These type of by-luck processes made every possible value near what they wanted, randomly, and needed sorting.

Even when the quality of clay and coal were tightly controlled, we "knew" the 5% were good sorted 10%.

Today carbon-film is made by blowing fire-smoke on ceramic bodies. You know that old cars would belch smog and soot. And the computer in new cars get the soot down to 0.01%. You can computer-control a fire to whatever % soot you want. So you make 100 Ohm resistors with 10% soot for 10 minutes, 1 Meg with 0.1% soot for 0.1 minute. Actually you test and find that you need 0.67% soot at 835 degrees for 13 seconds to get 100K every time, etc.

Laser-trim sure is a thing but costly compared to getting production consistency better than most of your customers need.

[Govmnt_Lacky]

Damn! All these responses and really no answer to my question. Maybe I phrased it wrong so let me put it this way.

If I need an 82K resistor and I only have a 75K. Can I use it? It is within 10%

[EBK]

If I need an 82K resistor and I only have a 75K. Can I use it? It is within 10%

Yes, you can use it.  Will it work?  I have absolutely no idea without seeing a schematic.

[duck_arse]

If I need an 82K resistor and I only have a 75K. Can I use it? It is within 10%

Yes No, you can can't use it.  it Will it won't work?  I have absolutely no idea without seeing a schematic, really.

there, both answers. one of us must be right.

[R.G.]

Damn! All these responses and really no answer to my question. Maybe I phrased it wrong so let me put it this way.

If I need an 82K resistor and I only have a 75K. Can I use it? It is within 10%

Yes, you phrased it wrong.

So - you need an 82K and have only a 75K. Can you use it? Sure. Absolutely, you CAN use it.

Whether it will WORK acceptably or not in the circuit is the whole point of those answers that said "gotta see a schematic" and "know when to hold 'em, know when to fold 'em". There is literally no good answer to that question without either knowing what happens in the actual circuit, or just hooking it up and trying it. Who knows? In the pedal world, you might like it better.

I personally have typed on this topic for longer that it would have taken you to just sub it in and try it to see if the result was OK or not.

As a final thought, subbing resistors like this - that is, a 75K for an 82K, and having it work or not, tells you precisely nothing about whether a +/- one preferred value change in any other resistor in that circuit or any resistor in any other circuit will be OK or not. The circuit dictates the resistor and its tolerance, not the other way round.