[sdiy] Ratio switching!

[Tom Wiltshire]

On 24 May 2007, at 09:47, Louis van Dompselaar wrote:

>
>> If the resistors are selected by hand using a good meter then  
>> surely even resistors sold as 5% would give the reqired accuracy  
>> AT THE TEST TEMPERATURE ?
>>
> Yes, but what are the chances of finding two resistors in separate  
> batches that would give you a 0.01% accurate division?  Even if you  
> get two batches of 500 that are guaranteed to be evenly distributed  
> (so 1 resistor for each 0.01%) you'll end up testing on average 250  
> resistors from one batch against a given resistor from the other.   
> But in practice you might end up with thousands of +1% resistors in  
> one batch and thousands of +2% in the other and never get the  
> division you want.
>
> But all I'm saying is that it's unlikely you'll get 0.01% accuracy  
> on all steps.  If you lower your target, it's very feasable.  The  
> octave switch in the Doepfer VCO is just a bunch of matched 1%  
> resistors on a voltage reference and that works fine.

The Sequential Pro-One's CEM oscillators use the same trick (rather  
than switching the timing caps) and it works fine there too.

I've tested good quality 1% metal film resistors and found that in  
practice most of them are accurate to 0.5% or better, even without  
careful matching. I expect they build to a higher standard and simply  
specify them as 1% to avoid having to throw any out.

As Louis said, 0.01% is a pretty high target, although I'd have  
thought 0.1% was easily achievable even without going to the expense  
of buying 0.1% resistors.

The final thing is that given that resistors only come in certain  
values (E24 or whatever) the accuracy of the ratio is going to depend  
on finding a good pair of resistors to create that ratio. For certain  
musical ratios, this could be an issue.