Variable Coefficient Filters in QS-Series Decks - A Spectral Investigation

[ZosoIV]

Having just purchased a JB940 deck, I was intrigued by the variable sound settings. Basically, these are a set of three different anti-aliasing filters, which push aliasing noise (mirrored representations of frequencies above the Nyquist limit) into the higher regions of the spectrum where humans cannot hear. The standard way of going about this is a brickwall filter, which attempts to quickly lowpass everything above the Nyquist limit (or half of the sampling rate, in this case 44100Hz/2 = 22050Hz). The other filters use noise shaping, apparently.

Since playing around with these seems to have no effect whatsoever on sound quality, even through $300 headphones, I decided to "see" what the deck is actually doing with this noise. To accomplish this, I recorded the analog outs into a firewire soundcard (Behringer) at 96/24. (Remember, these filters don't apply when using the optical outputs). I've attached snapshots of the spectrograms for each filter, taken during the playback of an MD (Dire Straits Greatest Hits, recorded optically with the Type-R codec). Though taken at different points in the song, the point is not to illustrate what we can hear (i.e., below 20kHz), but the regions above (20-48kHz).

I don't have any place to put the images up on the web, so they have been attached here as small jpeg files. To give the viewer some perspective, the frequency range shown is 0Hz-48000Hz, so the frequencies we can hear are represented by the lower "mountain" in each picture. The filters seem really change things above about 30000Hz.

Conclusions? Basically, the filters are certainly doing something, though again, well above where humans can perceive. The standard "brickwall" filter and Filter 3 ("a smooth and resonant sound") seem to show the least ultrasonic noise, while filters 1 and especially 2 ("a well-placed sound" and "a sound that is fresh") seem to “create” a whole bunch of ultrasonic noise. The question remains; does this really affect the frequencies we hear? I’ve heard varying things about HF noise, since it is seen in the noise-shaped output of SACD’s as well. Some claim that this random noise, though unperceivable, causes recordings to feel “warmer” and more “analog-like,” while others contend that high levels of ultrasonics are bad for consumer-grade amps.

I hope others with this deck (or other models with these filters) will find this somewhat interesting. In any case, at least we now know these filters ARE doing something, though I doubt that they make any big difference. The little green light on the filter switch is kind of neat, though