Quad does not give one exact information in the manual on what the other filters actually do but rather gives one a verbal description in general terms of how they operate and sound. As PS noted in his review, these descriptions are in quite general terms and one needs to listen for oneself. Quad was forthcoming about what they did when I asked. And in practice, I think that people will find experimenting with the filter choices interesting and perhaps useful.
Attention has been focused recently on filtering on account of, first, the “apodising filter” that RH was so enthusiastic about some issues back and more recently MQA, which as I understand incorporates what amounts to apodising filtering. The idea is mathematical but in outline one can think of it as giving up a little frequency response flatness and reducing slightly out of band rejection of undesirable digital products in exchange for a cleaner impulse response and reduced “pre-ringing” and “post-ringing”---sound before or after the actual sound. (Even though this pre and post sound is beyond the usual idea of the audible band, there is some evidence to suggest that it is even so audible at least in extreme cases.) On high-bit-rate PCM, filters can be essentially flat and phase linear in the audible band and their impulse response will be good in terms of what is in the audio band or anywhere near it. In CD digital, one has to give up some of the extreme top and on occasion some of the suppression of out of band products to get a cleaner-looking impulse response. This is related to what Wadia was doing in the 1990s and after. The Wadia players had a response that started to roll within the audible range though it did not roll far down until considerably above the usual limit of one half the sampling frequency. But in exchange as it were, they offered cleaner impulse behavior.
Quad did supply some frequency response and impulse response graphs upon my request. But it is not easy to connect this sort of information up to audible performance directly. Roughly, the “fast” filter is the one with the earliest in-band roll-off at the top but the best-behaved impulse response (to look at) and a lot of out of band rejection. Here we are speaking of being down by what looks to be on Quad’s curves close to 3dB at 20kHz, enough to soften the top a bit, though that may be all to the good with a lot of speakers with rising top ends. The “smooth” filter is maximally flat in the audible range but allows through quite a lot of energy above the nominal stopping frequency of 22.05kHz. And the “wide” filter has minimal pre and post ringing but less out of band suppression. The “narrow” filter when used on CDs played by the unit or USB digital input at CD standard is, as noted, Red Book standard.
So which one is “right”? In trying to decide that, one has a problem because digital audio at CD standard in particular involves a certain compromise from the start. The upper limit, the sampling rate/2 which is 22.05kHz, is too close to the audible band for a transparent filter to be constructed at all easily, or perhaps at all if the filter is to suppress all out of band products to an extreme. Roughly, what happens is that if one wants impulses to be clean—minimal pre and post echo—then one has to give up a bit of frequency response flatness and perhaps some out of band rejection, as Wadia did.
This is a tricky subject, but most people have decided that one really ought to have a higher sampling rate so that a filter could be flat within the sub-20kHz frequencies (and phase linear there) and have a more gentle roll-off than the kind needed to be out of there by half the sampling rate. To suppress out of band products almost completely is of course not going to require such severe filtering near the audible band if half the sampling rate is considerably higher than 22kHz! But for CD one is stuck with the issue. As noted above, the out of band products are not as such a problem—your ears will filter them as will many speakers, though in principle they put some stress on the system,
In tonal terms, small droops up close to 20kHz have rather little effect, and are also consistent with live music (air absorbs frequencies up there and reverberation tends to have almost no energy up there at all), although the amount in the ”fast” filter is audible. It is beginning to seem likely that the improvements wrought by the cleaner impulse response and perhaps some other things outweigh the loss from a reduced output at 20k and also justify letting through some energy that is out of the band where it really represents direct audio information—the stuff above 22kHz is not really part of the audio signal that was originally recorded.
Technical audio people tend to believe in flatness as a matter of religion, and Red Book specifies at the same time the deep suppression of output beyond 22.05kHz. But there is a growing body of anecdotal evidence—and sometimes more than anecdotal perhaps—that better impulse response and phase shift may be more significant in audible terms than a slight top end loss, audible though the latter will be, or out of band energy, which in theory is not directly audible. The reader interested in this general situation might find it enlightening to go back and read people’s reaction to Wadia’s work. One might also think back to years ago when Onkyo introduced a CD player that allowed a lot of out of band energy through (apparently on the grounds that the ear/brain liked some out of band energy but did not much care what it was as long as it was roughly related to the signal). For whatever reason, that CD player actually sounded very good).
So in practice, one is almost inevitably compromising a bit on CD digital, whatever one does.
The thing is, however, that one cannot really judge this in the abstract by listening to material already recorded. One can of course judge what one likes. PS liked the “fast” filter. I thought it made piano music too transient–emphasized, with too much attack, not in the sense of edge but of transient weight as it were. And so it will go—different people will like different things. Perhaps out of habit, I actually preferred my usual (Benchmark DAC1) Red Book CD standard to the Quad’s alternative filters most of the time. Of the Quad alternative filters, I preferred “smooth” almost always. But it is important to realize that such preferences are essentially personal and recording- dependent as well.
All that really counts in universal terms is how a D-to-A arrangement, when combined with an ideal A to D, preserves the signal. One can really only check this completely with a live mike feed and a comparison of the live feed with the A-to-D and D-to-A concatenated. Anything else is just choosing flavors in a context where what is right is not truly knowable and personal preference rules. Still, you may enjoy playing around with the filter choices just to get some understanding that the issue exists.