Messing about with the filter section in the Apache, with the 6AL5 clipper tube removed, you normally get an "eliptical" filter @>10kHz. That filter (see the previous thread) has an ultimate bandstop of about 18+dB. The advantage of the "eliptical filter" is that you get a really fast slope (drop) above the corner frequency of the filter, as opposed to the more gradual drop of the "butterworth" or "bessel" filter. The "eliptical filter" is formed by making the inductor (in this case) resonate like a notch filter via the use of a shunt capacitor. Otherwise it is a normal 3rd order (three pole) filter, which is nominally an 18db/octave slope.
The downside of the "eliptical filter" is that the ultimate bandstop is not as great as the standard filter. You get the benefit of a fast rolloff near the corner frequency at the expense of higher frequency rolloff. In this application, not a real problem since the ultimate rolloff can be either handled in other stages, or by the inherent roll-off of the xfrmrs in the path, or it can simply be ignored...
What I did is to modify the values in the filter to move the corner frequency to a nominal ~6kHz.
and to also improve the bandstop attenuation considerably[/u] (see the graph, compare to the graph of the stock filter values). The values are somewhat critical, since the circuit relies upon "Q" to work, and I'd suggest that testing them in the rig with a frequency generator (you computer can do that job...) and some way to monitor the output (again freeware on your computer will do a fantastic job) and then tweaking the values, as needed to obtain the desired response is best.
The modeled circuit is with R38 on
ground not lifted on the cathode resistor R35, which is a self bias resistor. The effect of this bias voltage under audio conditions upon the filter is likely to be minimal, but a bypass electrolytic would solve that problem - and without the 6AL5 one does not need a string of resistors there anyhow to bias just one tube...
Anyhow here are the new values:
L31 - 1.5H
(if you select about 1.0H and you get the corner freq around > 8kHz maybe 10kHz. - but this seems to high to me,
if you are going to bandwidth limit inside the rig at all)
Rx - place a 2 meg resistor across C57 (same as across L31) - this adjusts the Q just a tad for the better. Can be left out, but don't lower it below 1 meg no matter what.
R39 - unchanged (do not change)
R38 - unchanged (do not change)
C57 - 100pf
C58 - 700pf
C59 - 500pf
C60 - 0.02 to 0.03ufd
(C60) controls the LF rolloff. the above values show a slight rolloff below 40 Hz. The 0.03ufd cap is shown in the graph response.) Which seems about right to me. It is best
not to drive your mod iron with LF energy that does little or nothing for the output, as it will decrease the headroom by sucking up power for not much. Plus you can always EQ it back in, if you are using an external processing chain, etc.
Where to find a new L31?? Almost any small power xfmr - just test the inductance of the primary and/or secondary, in whole or in partial winding to find the value you need. It's not a critical "audio" component, except to the extent that you need a value pretty close the described value to make the filter work at the right frequency. Since it's purpose is to roll off highs, a little power xfrmr will work fine here. It can be really very tiny, since there's nil power going through the thing. (I haven't seen the one that is in the stock unit...)
My last thought is that one might want to make this switchable. I'd throw in a small relay which could be used to either bypass the filter section (use a resistor in place of L31 that will give you a level drop equal to the insertion loss of the filter, which the model says is 6dB. btw), another option and/or a more gentle option is to use a relay to open C57 - but you'll have to test the relay to see what it's self capacitance is in the open position. I'm sure that ur standard toggle switch has way way too much capacitance for this application. Also check the self capacitance to ground!! (if either's more than a few pf, then bypass is the only option.
Another way to go is to make L31 into two inductors and switch one in/out to change the corner frequency... that's a nice way to go!
All this is untested as of this post, although I hope to try some of this in the not too distant future... your comments and results if you try it will be appreciated.
_-_-WBear2GCR
PS. I've worked out the 1.0H values which gets you a filter right @10kHz., with the improved stopband attenuation, if anyone wants them I will post 'em!