Resistor's tolerance

Hi,
i begin this post with a big THANK YOU to all the nice people of this board, i'm learning a lot here simply lurking and reading the endless Meta-threads.
My probably silly question is this:
Looking at some DIY project's BOM's, Gyraf's ones in particular (Jakob, you rule!), i've noticed that resistor's tolerances are often 1%, then looking at Farnell, Mouser etc. i see that there exist tolerances as low as 0.001%.
Is there any noticeable performance gain in using tolerances lower than specified in the project?

Thanks a lot again

niccolo' gallio

Hi niccolo,
Welcome to the lab.
I doubt that using tighter tolerance than specified will yield better performance, except from your bank manager.

peter

.
What Jaako said... and if you need one to be exact, you can always hook the ohm meter to them and measure the actual resistance. If you are matching 2 resistors, like for 2 channels of stepped gain switches etc., 1% spec will usually (theoretically) give you more matches than 5%, as the resistors will vary across a smaller range.
Paul

the tolerance can mean more than the value. in many cases it means a difference in temp coef. the better the tol, the better the temp

[quote author="CJ"]the tolerance can mean more than the value. in many cases it means a difference in temp coef. the better the tol, the better the temp[/quote]

CJ's point is a good one, although again few circuit locations will be sensitive enough to the exact value, self-heating shifts, etc. to make a difference.

Usually premium resistors have, as well, low voltage coefficient and low excess noise.

Some circuits will need well-matched resistors in pairs or sometimes higher multiples for best common-mode rejection and/or power supply rejection. But usually their absolute value is not critical, so as suggested you can get a bunch and hand-match with a DVM.

At one old employer the manufacturing and purchasing people pushed back hard when I specified 0.1% tolerance parts in an auto amp frontend, but the precedent was set and they used them in other products thereafter.

Most mono stuff will work fine with 20% tolerance parts.

Heck, transistors and chips are +/-50%!

We use the resistors to get the transistors and chips more in line with what we want, but 20% "errors" are often insignificant.

We used a LOT of 20% resistors in the old days.

When you go stereo, 10% difference between left and right may be audible. Fortunately, by then, 10% resistors were common. When you have a many-stage stereo system with +/-10% errors in each stage, you might get lucky or un-lucky. The resistors that have the most effect on gain tended to be 5%, at least in better gear. I remember when 2% and 1% resistors were VERY expensive and rarely stocked, many pro-gear amps had factory gain-trim parts.

Mostly, 5% is plenty good. Often a detailed design analysis will call for a resistor "between 3K and 10K" with no particular advantage to any value in that range.

2% and 1% are now low-price by DIY economics (few parts, much labor) so they are routinely used.

But tolerance is not the only thing you get. Yes, you could sort through a box of 20% resistors and find one that was within 1% of the desired value. But the old pressed-coal resistors would not stay stable, drifting with heat and age and even with how hard you bent the leads. I've seen old 10% resistors drift 100% in a decade. (Note that the amp had been "working fine" with a resistor drifted 70% or 80% high in value!) The 1% resistors have to be more stable to meet specs.

> It doesn't mean that a 5% will be off 5%. Most likely it's in the 1-2% range, ... Think of the bell curve you see everywhere

These days, they make resistors the right value, and few are 4.99% off.

Back in the very old days, they didn't know what they would get, 10K or 100K. They measured them and marked them, using very wide tolerance so every resistor could be sold.

Then things got worse. You could be sure that a 20% resistor would NOT be within 10% of the marked value: if it were, it was pulled out and sold as a 10% part for an extra penny. So 10K 20% really meant 8K to 9K or 11K to 12K, but never 9K to 11K.

Nice one. Or two. :thumb:

Also,

See that, if you were to solder together 100 ea. 1 ohms resistors, you would probably get a 100.00 ohm resistor.

Of course in series. Who ya kiddin? :wink:

[quote author="CJ"]Nice one. Or two. :thumb:

Also,

See that, if you were to solder together 100 ea. 1 ohms resistors, you would probably get a 100.00 ohm resistor.

Of course in series. Who ya kiddin? :wink:[/quote]

And something of an inductor. Or maybe a nice necklace :razz: