Audiolab Q-DAC filters

Article Index
Audiolab Q-DAC filters
page 2 Performance with CD
All Pages


Audiolab's M-DAC and less expensive Q-DAC  both have no fewer than seven digital filters. Noel Keywood looks at what they do.

Audiolab's Q-DAC. See our review.

From the moment I heard Chord's newly released DAC64, back in 2002, I realised that perhaps CD as we knew it back then could sound better than it did. Designer Rob Watts told me that his unique Watts Transient Aligned (WTA) filter might not measure "as well" as the traditional industry solution, but it sounded better all the same. Since Rob Watts knows more about digital signal processing than most, I listened intently to DAC64 and was impressed, like so many others. His WTA filter rolls off upper treble and does give a slightly warmer sound from CD than usual, but it was also less aggressive and gave a somehow more solid feel to images, as if they were bigger bodied and more stable in position. In the end I just liked DAC64: it was clean, stable and assured and I could listen to CD for hours without suffering the slightly unsettled feeling I often get from it.
    That was then and now, no less than a decade later, Audiolab have also addressed the issue of digital filtering and given us no fewer than seven filter options, first in their upmarket M-DAC and, recently, in their cut down Q-DAC (£400, UK). Three of Audiolab's filters remind me of DAC64, since they give a very similar measured result and sound. The great thing about them is that they can be used to subtly tune – tame I could say – a system. They also improve on digital as we know it: the changes aren't big, but they're worth having in my view.
    Oh, and as an aside, M-DAC comes from another Brit, John Westlake who, like Rob Watts, is a renowned digital guru; both have had the vision and courage to move outside the design box and present us with something innovative and different. Here's a close look at M-DAC's filter set and what to expect from it. All were measured with a Rohde&Schwarz UPL audio analyser, the same instrument John Westlake uses.
The seven filters are -

1) Sharp roll off
2) Optimal Spectrum
3) Minimum phase

4) Slow roll off

5) Optimal transient
6) Optimal transient XD
7) Optimal transient DD

I've put them into groups so you can more easily understand their role. At top we have the fast filters (fast in sense of short rise time) that give strongest treble and the brightest sound. These are the usual industry choice that will give best measured result in terms of frequency extension: they reach highest in other words, if only by a small margin.
    In the middle sits a single medium speed filter, termed Slow roll-off, that under measurement was the best filter of all – something that surprised me.
    At bottom sit three slow filters that Audiolab interestingly place most emphasis on. They quite radically roll down treble from CD, giving a slightly warm balance. Under measurement they are identical. And all three are different from the others in behaviour, measuring quite badly in terms of alias suppression, although an alternative view (used by Pioneer for example, who use similar filtering) is that they have the most analogue-like behaviour.
    In practice, as John Westlake admits, they throw up uncorrelated digital audio rubbish, generally considered a bad thing, but the alternative view is that their energy spectrum is very analogue like and their sound possibly likewise. Is what we hear from these filters coloured more by the way they roll off higher frequencies than any other property, I have to wonder?
    At the end of the day, if you don't like 'em then you can switch 'em out, but I gravitated to Optimal Transient XD and it seems most others prefer the slow filters too, so as "bad" as they are, they may be good, if you see what I mean. But I have more to say on this.
     At a technical level digital filters have both frequency and time domain behaviour, but traditionally it is only the frequency domain that gets attention, meaning frequency response and rejection of uncorrelated digital data. However, looking at an impulse in the time domain shows that digital filters also impose ringing on a signal, both pre and post ringing, meaning before and after the signal. Audiolab's three slow filters "exhibit no ringing" they say. "The transient nature of the music is preserved". So these filters concentrate on the preservation of transients, at the expense of all else. 
    Is all this true, and which filter is best? I was surprised when closely measuring these filters that they they don't fully match Audiolab's claims. Also – and importantly –  you can expect different sonic results from high resolution digital than low-res CD. There are some separate and distinct issues here – and illuminating ones. Let's take a closer look.


1) Sharp roll off


If you want flat frequency response to a little beyond 20kHz with CD, then use this filter. It will give you the sound you are used to, meaning quite bright in balance, although Audiolab's filter doesn't peak like some. Impulse response in the time domain suffers substantial pre-ringing and post-ringing, which is bad. Rejection of unwanted data is good.
    Subjectively, this is most like conventional CD, offering the same balance, but it has a 'clamped' into place sound stage that lacks air around it.

2) Optimal spectrum

This filter is very similar to Sharp roll off, if with fractionally and inconsequentially less bandwidth. Time domain ringing is again substantial, considered poor, but rubbish rejection is good.
    Subjectively, this is a better sounding version of Sharp roll off I found, offering a clear, open sound that seems less constrained than the other filters. It's a nice filter to use, I felt.

3) Minimum phase

Technically interesting, this filter eliminates pre-ringing our measurements show, without affecting frequency response in any significant way. If ringing is subjectively important then this filter should sound obviously different and better, but in my experience it does not. Is time domain ringing audibly important then?
    Subjectively, I could not detect any improvement in transient quality, or any other quality, due to this filter's lack of pre-ringing.

4) Slow roll off.

This is a very interesting filter. Audiolab say it has "significantly less time domain ringing" than filters 1) and 2), but not "no ringing" of the slow, optimum time domain filters, 5), 6), 7). It would appear to be a compromise then.
    However, under measurement it exhibited less time domain ringing than all the other filters, including the Optimal transient set. It also rejected unwanted rubbish very well and retained frequency response quite well too. Under measurement at least this was clearly the best filter of all seven.
    Subjectively, this filter was a disappointment. It seemed to have no special qualities at all, sounding clean but constrained. The sound stage was much like that of the fast filters.

5) Optimal transient

This filter well suppresses pre and post ringing, but it did not match the 'Slow roll-off' filter. It significantly rolls down upper treble and poorly rejects unwanted digital rubbish. Technically – not good.
    Subjectively, the topography of the sound stage changed with this filter. Instead of instruments being displayed in a line between the earpieces, the feeling of a sound stage with instruments on it appeared. This injected a better sense of realism, making the presentation less contrived and mechanical – always a criticism of CD.

6) Optimal transient XD

A variant of Optimal transient, it is identical in behaviour under measurement. It has different "mathematics" to Optimal transient though, designer John Westlake says, and offers an improvement in bass quality.
    Subjectively, Optimal transient XD was a slightly fuller sounding and better sorted version of Optimal transient. Focus was improved and there was a better sense of rhythmic grip and push. This is one nice way to hear CD, loosening a lot of the subjective constraints it suffers.

7) Optimal transient DD

A further variant of Optimal transient, it is again identical in behaviour under measurement but offers more balanced signal current draw on the silicon die and apparently has a different bass quality to XD.
    Subjectivley, Optimal transient DD was difficult to pin down I felt. It didn't quite have the impact of XD, but was only marginally less attractive. Sound staging was as good as that of the other Optimal transient filters.

It is only CD's low sample rate of 44.1kHz that cause the slow Optimum transient filters to intrude into the audio band, audibly suppressing treble. When sample rate is more than doubled to 96kHz even the Optimum transient filter set ceases to affect upper treble, our analysis below (at right) shows. The theoretical upper limit is now 48kHz (half the sampling frequency). Optimal spectrum reaches this limit but no further, whilst Optimal transient extends much higher.


Optimal spectrum and Optimal transient filters with 96kHz sample rate data


The question now is: do these filters still affect the sound? If not, then the audible impact of the Optimum transient filters upon CD is likely more down to their frequency domain response than any other property. Bear in mind here that 96kHz and 192kHz sample rates (multiples of 48kHz used in professional circumstances) were proposed so that anti-alias filter design could be eased, using better damped filters with slower roll-off rates that would give better all-round results than those of CD. So results from the M-DAC here show whether this is a promise fulfilled or not, and what we can expect from high resolution digital in future.
    Listening to these filters with a variety of 24/96 material, Rock and live Classical, showed that as measurement suggested, differences were not easily obvious. The differences heard with CD were repeated, but they were smaller. The Optimum transient filter set was again the most arresting to listen to and again the main difference between them and the others lay in sound stage presentation: it broadened beyond the earpieces of my headphones (Philips Fidelio X1s) as if unconstrained by them. There was more air around images and a better sense of depth, especially with orchestra. Of the three Optimal transient filters XD again seemed the most engaging, putting a bit more bulk and character into instruments, again with better stated rhythm and timing.



Add your comment

Your name:
  The word for verification. Lowercase letters only with no spaces.
Word verification:


Hi-Fi World, Powered by Joomla!; Hosted by Joomla Wired.