Friday, June 10, 2011

Designing the notch-blend filter for KPAC

So far, no response from KPAC engineer, and station hasn't changed.  So I'm developing a necessarily imperfect solution on my end.  Many high end tuners have a high blend switch, very useful in cases like this.  The Pioneer F-26 doesn't.  But the best possible solution on my end might not be a high blend filter but a "notch-blend" filter that specifically notches the 10.5 kHz buzz into mono where it vanishes.

Here is a great online RLC notch filter calculator.

The difference between notch and notch blend is that there are two series resistances, one for the left and right outputs (which I believe in a simplistic model are simply seen in series and summed), and the LC network bridges them.  At least that's my guess now.

Set for notch at 10500 Hz, and Q of 2 which seems like a good compromise between specificity of filter and ringing.

One possibility is 1.6k total resistance, then you get 0.0047uF capacitor and 0.488H inductor.  Based on this calculation, I found and ordered a Russian Teflon 0.0047uF capacitor on ebay for $2 (plus $7.30 shipping).  Was all set to order 47 mH sheilded RF inductor from Mouser when I decided to check F-26 schematic and specs.

Turns out F-26 has output impedance spec of 2.2k ohms and that looks to be correct.  Each output from the final AF amplifier chip sees a voltage divider with two 4.8k resistors in series.  The output from that is further shunted by 10k ohm pot for the variable outs.  Given some unknown driving impedance of the AF amplifier chip, this could quite reasonably add up to 2.2k ohm impedance, so I'll go with the spec.

It looks like Pioneer is doing several things with this output design.  They are cutting the output voltage in half (and thereby also cutting amplifier noise in half).  They are making it bullet proof so any fool can simply sum the two outputs to mono without increasing distortion.  And they are adding some HF rolloff (aka decoupling) which is determined by the capacitance of the cable connected (making it somewhat tunable).  My calculations show a 240 pF cable (cheap 6' cable) would produce a cutoff frequency of 300kHz, most likely not important.  And a good short cable could have 24pF, giving a cutoff of 3Mhz.  However, a 1200pF cable (and they are out there as audiophile magic cables) would produce a cutoff of 60kHz, which sounds OK until you consider that you must have a cutoff frequency of 10x higher than passband to have less than 0.1dB attenuation anywhere in that passband.  So a 60kHz cutoff starts to get close to 0.1dB attenuation at 6kHz.  That could well be noticeable.  The 10x effect was something I discovered the hard way, after decades of using relatively high resistance passive level controls instead of active preamps.  For some reason, my output pots always sounded a bit dull, but I couldn't believe it, because my calculations said I had cutoffs in the 50-150kHz range)  200kHz bandwidth sounds like over-the-top audiophile madness but isn't.  And then there is also the way successive rolloffs in an audio system add up...

Normally I don't like that sort of output design the Pioneer uses, I like true low impedance outputs, but for an FM tuner, producing lots of HF noise anyway, some of which may be supersonic, 2.2k ohm output impedance is quite possibly desireable.  A feature.  Anyway, the claim has always been that Pioneer has the most transparent midrange.  Indeed the AF amplifier is protected from output loading distortion, which might help that.  The claim has generally not been that the Pioneer has the most transparent highs.  Kenwood tuners often achieve that distinction (and many don't like the resulting sound, bright by comparison with other tuners).  Kenwood was so obsessive about low impedance outputs that my 600T tuner has a separate audio amplifier chip after the variable output pot.

End of digression.  I must therefore design my filter around two 4.4k ohms (the channels in series).

Looks like this combination works:

0.0017uF (1700pf, calculated)
0.135H (133mH, nearest E24 value)
Resistance 4.45k (calculated R)
Q 2
Fc 10500

Closest standard capacitor value is 1800pF, just ordered some Russian teflons.

That changes somethings, actually R at 4.4kohms is probably not exact anyway probably slightly lower, and Q is not critical.  Worst effect is that Fc is lowered to 10210, and that's pretty much determined by L and C.

R=4.42 ohms

So with 1800pf, looks like I I need 0.127H inductor instead of 135 for best result.

No comments:

Post a Comment