"Overnight circuit" -- don't expect too much from a quick look, or even from an hour! It needs to be that slow, or it doesn't work right. Table of results below.
One reason for the slowness is that the 9kHz filter needs an extremely high Q to filter out 7kHz. Only 1/1000 (-60dB) of the 7kHz harmonic comes through the 9kHz filter.
The other filters don't need to be so tight, but there's no point finding the lower resistor values when the 9kHz filter takes so long to come up to full output anyway. Or maybe there was a point: It would have saved me finding the right value for the extra capacitor in at least 2 of the filters. With such a high Q, the filters need to be very accurate.
The other reason it's so slow is because transient simulation isn't very good at simulating filters, so it needs that 29.3kHz source with the consequent limit on how fast the simulation runs. It's 29.3kHz because I wanted to avoid anything that might be a multiple of any of the signals.
Results from +-1v square wave input:
1khz 1.23v
3khz 388mv
5khz 209mv
7khz 125mv
9khz 76mv
All are in phase, if that makes sense.
Precision may be 0.1% at best.
Re-assembling the results:
http://everycircuit.com/circuit/4904379164655616
