RC Tolerance -- audible difference

[r_s_dhar]

Hellow all,

The problem for people like me who are unfortunate to have been born without a silver spoon in the mouth and a storage oscilloscope on dad's table is that we do not understand the "audible" versions of what text book graphs and formulas say.

If you were listening to Symphony 5 of Beethoven over a pair of platonic 3-way active speakers, what would you hear if your crossover circuitry were suddenly switched over to another board made of 10% tolerant RC components? Sallen-Key configurations are component sensitive -- but how much?

Would it be possible to construct acceptable boards with 5% RCs?

[tiroth]

Do you have a SPICE package that can do Monte Carlo analysis? This is the easiest way to see such effects IMHO.

[fdegrove]

Hi,

No idea what exactly it would sound like but using 10% tolerance components without knowing their exact value can't be a good thing.

Essentially, if you're really unlucky you could potentially end up with a 20% difference between left and right channels.

Mostly though, even 10% components are better than advertised but I would feel rather uncomfortable working "blind" like this.

Would it be possible to construct acceptable boards with 5% RCs?

That'll depend on what you want to achieve but quite likely, yes.

Not everyone's as fussy as me.

Cheers,

[faustian bargin]

actually to be really picky, the range is going to be the true manufacturer's tolerance, which may be smaller than the nominal tolerance, and probably will affect the potential 'luck of the draw' difference btwn left and right in that example.

say a component is labeled 10% tolerance. if the mfr is capable of producing units within a 5% tolerance, then to save money on materials he could produce them centered in the lower half of the specified 10% range. the average component value would be -5% of the nominal value, but they would all still technically be within the +/-10% range.

the end result in fdeg's example would be a difference of 10% btwn left and right though.

but this is hypothetical...i don't know if any/all mfr's are sneaky like that, and if so just how sneaky they would be.

if someone has some insight on 'the way things really are' i would be interested to know.

/andrew

[Brett]

Don't assume that 10% rated components will actually be spread all over the tolerance region. I bought a whole lot of caps recently for an RIAA that were 20%, so I got some extra hoping I'd be able to hand pick a few and get the values I needed, which were slightly off the labelled value. ALL of them turned out to be within 1%, and all but 2 (of 20) were withing about 0.5%.

If you want an exact value, you either have to measure it, or if possible buy components with a 1% or better tolerance.

[EC8010]

Originally posted by Brett
Don't assume that 10% rated components will actually be spread all over the tolerance region.

When a manufacturer specifies 10% tolerance, they typically mean that the values of their product have a Gaussian (or normal, or random) distribution, which is a bell-shaped curve centred on the design value. As such, you will be very unlikely to find a component at the tolerance limit because thay are guaranteeing that (typically) 99.5% of the components are within that limit.

Nevertheless, I have measured 20% tolerance capacitors having a much smaller spread, whose median value was skewed away from the specified value, exactly as speculated by faustian bargin, presumably to save on materials costs.

[Havoc]

As long as they come from the same batch you won't be far from 2% of with 10% components. But if you have one side manufacturer A and the other side manufacturer B the difference might be greater. If it is audible is another debate.