From: Patent Tactics, George Brock-Nannestad
> ----- Original Message -----
> From: "Don Cox" <[log in to unmask]>
> > Also the manufacturing tolerances on resistors and capacitors in those
> > days were very wide, so even if somebody designed a circuit to give some
> > desired EQ, the results could be somewhat diferent.
----- to which Steven C. Barr commented:
> The standard tolerance for electronic components has been 20% (no metallic
> band) as far back as I can remember...if not before. You could buy 10%, even
> 5%...if you were willing to pay extra...but few corporations were, unless the
> precision of the product required it... ...stevenc
----- the tolerance just meant that you had to do sorting to obtain the
guaranteed precision. If you had a random distribution of component values
(one hump) and you removed the central portion to get a narrower hump (higher
precision), then the remainder would still be within the originally stated
tolerance, but it would have two humps. For some applications 20 % was fine.
When I did a 3-month's stint at the development laboratory of Bruel & Kjaer
in the 1960s, we used resistors that were precise to 4 1/2 significant
digits. They had been obtained by buying large stocks of low-precision
resistors, burning them in for a week by pulling an appreciable current
through them, and then sorting into these narrow categories by individual
measurement. This permitted the laboratory to compare design calculations to
actual constructions, and anomalies were much easier to trace. That principle
had served them very well in their narrow 10th octave filters. These have to
have many stable stagger-tuned sections. So their professional equipment,
just like that of the General Radio Corporation, was precise.
Just at that period, carbon composition resistors were being replaced by
carbon and metal film, and they could not get the spread in tolerance any
more, every 20% resistor they bought would e.g. be 18% off, within a very
narrow hump. So they gave it up. But then came laser-trimmed resistance
arrays, and the precision came back, both for experimental and production
Imagine the precision needed in resistor ladders for 24 bit resolution!
Right, it is a non-starter from an absolute precision viewpoint, unless you
keep your resistors in an oven at constant temperature.