[sdiy] discrete DAC accuracy

[Magnus Danielson]

From: Neil Johnson <nej22 at hermes.cam.ac.uk>
Subject: Re: [sdiy] discrete DAC accuracy
Date: Tue, 13 Jan 2004 23:00:06 +0000 (GMT)
Message-ID: <Pine.SOL.4.58.0401132245390.3981 at orange.csi.cam.ac.uk>

> Hi,
> 
> > * Usually you are to assume a Gaussian distribution of the values such that
> >   3-sigma (i.e. 99,7% of all cases) keeps within the given values (i.e. +/- 1%)
> >   or whatever. It may not be that you have this distribution, it may be that
> >   there is a "hole" in the middle where selections have happend, but there may
> >   also be other deviations, like a shifted average resistance due to production
> >   variances etc.
> 
> Not sure if this is what you meant, but resistors of a given value at a
> given tolerance will never be outside the upper and lower % limits.  E.g.
> a 1,000 ohm, 5% resistor will never be outside (950,1050) ohms.
> 
> During manufacture the unmarked resistors come out of the oven, cooled to
> known temperature, and then their resistance is measured.  The value
> measured operates a selector which, if the value lies within set limits,
> pulls the resistor into a bin.  If not, it passes on to the next
> selector/bin.
> 
> (this all happens rather fast)
> 
> Once the bin is full, its contents are marked with their value and
> packaged up (bandoliered or boxed).
> 
> Processes are varied to produce the full range of values, but ultimatey it
> is the sorting that picks out the required values.  Tighter tolerances
> just means changing the upper and lower limits on the seletor.
> 
> And if you have enough money, you can even have your very own values if
> you pay enough for the manufacturer to spend the extra time programming a
> selector for your specific needs.
> 
> As to what the *actual* distribution curve looks like, I've no idea.  I
> guess it depends on which value you have, what the current flavour of
> process was being used at the time, on which oven, etc.
> 
> This all works out rather well because of the (almost exact) joining of
> one resistor value's upper tolerance value, and the next value's lower
> tolerance value.  E.g., E12 series (10%):
> 
> 	100 upper = 110
> 	120 lower = 108
> 
> Bit of an overlap, so the manufacturer can choose which bin the overlap
> resistor values (e.g. 109) fall in, perhaps depending on stock levels,
> demand, etc.  So you'll never find a 100 ohm 10% resistor above 110 ohms,
> because otherwise it'll have been picked off into the 120 ohm bin.

Well, yes... you're probably right. For resistors that is. For other stuff it
doesn't work that way simply because there isn't the same distribution of
target-values. The point here is anyway that you can't trust the distribution,
whatever it is supposed to be. Real values of resistors will be much better.

Cheers,
Magnus