Well, it seems like both are equally important, with shunt section maybe even more:
http://www.diyaudio.com/forums/showthread.php?s=&threadid=94506&highlight=
http://www.diyaudio.com/forums/showthread.php?s=&threadid=94506&highlight=
A voltage divider's gain is
g=R2/(R1+R2)=1/(1+R1/R2), with R2 being the lower resistor and R1 the upper.
We can apply an error/nonlinearity factor to each, like
R1*e1 and R2*e2,
then we can rewrite the gain as
g=1/(1+e*R1/R2), with e being e1/e2
As long as e1 and e2 are close to 1 (say, +-10% max), it doesn't matter much which one changes.
But the ratio of the resistors matters, the lower the gain (that is, the bigger the term R1/R2) the more the error dominates the gain equation denominator.
Net result is, as long is the error is low (which we shall assume) it doesn't matter which leg of the divider contributes to it. More important, the effect of any error is worst when the gain is very small and it goes to zero when the gain becomes 1. This is assuming zero drive impedance and infinite load impedance, otherwise these impedance must be factored in (drive in series with R1 and load parallel to R2). In a real world situation one will certainly have to experiment which variant of leg arrangement sounds and/or measures better, and often secondary/overlooked other effect (layout etc) may dominate the performance anyway.
And like (I think) was stated in that other thread, it may be wise to optimize for the mostly used level settings, which is only an option for the DIYer, of course.
- Klaus
g=R2/(R1+R2)=1/(1+R1/R2), with R2 being the lower resistor and R1 the upper.
We can apply an error/nonlinearity factor to each, like
R1*e1 and R2*e2,
then we can rewrite the gain as
g=1/(1+e*R1/R2), with e being e1/e2
As long as e1 and e2 are close to 1 (say, +-10% max), it doesn't matter much which one changes.
But the ratio of the resistors matters, the lower the gain (that is, the bigger the term R1/R2) the more the error dominates the gain equation denominator.
Net result is, as long is the error is low (which we shall assume) it doesn't matter which leg of the divider contributes to it. More important, the effect of any error is worst when the gain is very small and it goes to zero when the gain becomes 1. This is assuming zero drive impedance and infinite load impedance, otherwise these impedance must be factored in (drive in series with R1 and load parallel to R2). In a real world situation one will certainly have to experiment which variant of leg arrangement sounds and/or measures better, and often secondary/overlooked other effect (layout etc) may dominate the performance anyway.
And like (I think) was stated in that other thread, it may be wise to optimize for the mostly used level settings, which is only an option for the DIYer, of course.
- Klaus
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.