[sdiy] Seeking a very old & specific message from Dec 2006

[Neil Johnson]

Hi Tim,

> The notch filter concerned was this one:
>
> https://www.ti.com/lit/an/snoa680/snoa680.pdf
>
> The rather-too-perfect matching of component values in SPICE can make the simulated notch response really, really narrow, so it can be difficult to see that it is there at all. My suggestion was to replace the op amp follower in the simulation with a voltage-controlled voltage source where the gain can be set directly, to 0.9 say - this should make the notch much easier to see. (If the follower/buffer has gain 'k', the Q of the filter is apparently Q = 1/(4*(1-k)), so it blows up with k exactly 1.)
>
> (And from what I could gather from quick Googling last night, the filter tends to be rather tricky to use in practice, because of its use of such big resistor values and the necessity of getting component values 'just right' in order to have the notch exactly at 60Hz.)

Very high Q notch filters in reality will be far less than the ideal,
both due to component tolerances and also component drift over time.
One way round this is done in the HP 339A where they wrap the notch
filter inside a couple of servo loops locked to the main oscillator.
If you really wanted to notch out mains noise then you would get the
best result somehow locking the notch filter to the local mains
frequency anyway as that is likely to vary during the day.

Cheers,
Neil