Volume 3 No. 1
Apodising Meridian
Marin Colloms Assesses Meridian's 808i.2 Reference Signature CD Player
With its Radical 'Apodising' Filter.
Review By Martin Colloms

  Latest and best in an evolving tradition of fine CD players that dates all the way back to the MCD Pro from 1984, this new Reference Signature component is personally signed by the Meridian founders Alan Boothroyd and Robert (‘Bob’) Stuart (JRS). The £8,495 808i.2 is one of two superficially identical players, both of which can output fixed or variable (for direct-to-power amplifier connection) stereo analogue signals, the i part of the type number indicates that this version is actually a combination of CD player and pre-amp. The player alone, designated 808.2, costs £7,495, while earlier 808 models may be upgraded to the latest spec for £2,500.

However, the headline feature of this new Mk2 Mk2 variant on 2004’s 808 theme is a special digital filter with ‘apodization’ characteristics. Put simply, this changes the ‘shape’ of the mathematical function corresponding to the digitally processed (ie filtered) audio signal. The new filter window, necessarily operating at an upsampled 88.2kHz and requiring a heavy duty 150MIPS processor, avoids the usual replay pre-echo, and further provides part correction of the pre-echo on the recording via a small, lower level notch in response at the appropriate location in the frequency domain. The net result is minimum phase rather than linear-phase output. (See Technology Backgrounder Box.)

The 808i.2 comprises a substantial enclosure with black glass top and heavy duty moulded polymer fascia. Our sample had a black lacquer style finish, but the current catalogue lists matt black, gloss graphite or gloss silver as the alternative finishes. A bulky beast, it measures 435mm by 290mm deep by 80mm high. The back has an IEC mains input and main power switch, several S/PDIF coaxial and RJ45 balanced digital audio in/outputs and Comms connections (for synchronising to a complete Meridian system). The six analogue audio inputs are unbalanced, while both balanced (XLR) and unbalanced (RCA phono) outputs are provided. As a digital pre-amp it has two optical and three RCA inputs in S/PDIF, for up to 96kHz and 24-bit. In a complete Meridian system context, the 808.i2 can provide the digital source and control centre for the DSP7200 loudspeakers, which also incorporate the new filter.

The 808i.2 will play CD, CD-R and CD-R/W discs, and its universal, modular, screened, switch-mode power supply is rated for 110 and 230V 45 to 65 Hz. The large dot matrix display may be adjusted for brightness and contrast, and also muted. This development from the original 808 model incorporates significant improvements in the DSP section, the core computer, the clocks, the analogue conversion, the buffering and the output circuitry.

A ROM drive reads the disc data in multiple passes at high speed, allowing heavy duty error correction that is claimed to be a hundred times deeper than standard CD. The data is buffered in a sequence of FIFO (first in first out) re-clocked memories to deliver a claimed sub-90 picosecond jitter, with spectral content effectively low pass filtered at a low 0.1Hz. The corrected red book 16- bit data is upsampled to a 176.4kHz, 24-bit datastream at 48-bit precision. Precise computation of the apodizing filters is done at this level, claiming minimum phase characteristics and no pre-echo. The DACs are Delta Sigma devices, of the latest type.

The volume control has 100 steps of 1dB each, and is part digital, part analogue to deliver the best resolution and dynamic rage. It may be set to a fixed level to provide a near CD standard fixed output. In pre-amp mode the analogue inputs are digitised and sent to the processor like any other digital signal, an approach that also allows them to be sent to a digital loudspeaker, if used. An audiophile might need to consider the implications here, since in truth the analogue side of this player/pre-amp is somewhat auxiliary to its main digital functions.

When a disc is first loaded, the drive spins up for a while, forward reading the data and producing varying degrees of whirring noises, depending on the disc and how well centred or balanced it is. This does result in some audible vibration in the early phase of replay. Whether it is this vibration, or the effect of this rapid acquisition, the sound quality does not seem to reach its best until the drive settles down to a lower reading rate, once it is well under way and the drive is silent once more.

 

Sound Quality
The first thing to establish was whether the Meridian’s switch mode power supply would unduly affect my reference system adversely. As it happens this dropped the sound quality rating a relatively modest 12%, some of which was recovered by unplugging the reference CD player. This result is especially good for a switch mode supply.

We initially tried the pre-amp function. There was some concern that this might affect the CD section, since in part the CD output and the pre-amp output are one and the same, but subsequent auditioning of the player left no doubts about its potential. However, the CD player and digital inputs have a direct digital route, while analogue sources must first be digitised before reaching the control system proper.

Analogue derived pre-amp sound was certainly pleasant, undeniably high fidelity, with low noise and quite good neutrality, but somewhat below the best audiophile standards – a rather bland effect with analogue sources. The sound quality score is a jobbing 25 – better than a number of so called audiophile pre-amps, but also a little worse than some inexpensive passive controls. Feeding LP equalised signals from a Koetsu Red T, I discovered that the unit had no chassis ground terminal, and improvised with a croc clip to the nearest screw head.  Dare one presume this means that Meridian no longer listens to LPs?

There is probably no good reason to use an 808i.2 solely as a disc drive, save as player and control centre for the partnering digital speakers. We hope to have this opportunity in the near future.

The new filter technology promised a sweet, ‘non digital’ sound, and from the start this was delivered in spades, through various top quality pre-amplifiers. As an established analogue replay enthusiast, it seems improbable that I could describe CD replay as beautiful, but that is just what it is. Up ‘til now you could play the master tape, recall the studio sound, and hear the unwanted digital artefacts and what was lacking in CD replay, but the 808i.2 seems to deliver much of the character of master tape when compared with conventional CD reproduction. A digital ‘character’ which we have come to accept as inevitable has apparently been swept away. Not only is the somewhat fatiguing hardness and grain of normal CD effectively removed, but that excision has brought a substantial improvement in clarity, leading to gains in focus, image width and depth, low level resolution, and – a Meridian watchword – transparency.

It seems that the usual ‘digital’ pre-echo not only adds coloration and shifts the tonal balance but also has a masking effect that obscures information. The results of its removal not at all subtle, regardless of what the double blind brigade might say.

To draw a video analogy, consider a high resolution image at 25MP (megapixels), equivalent to the original studio feed. Typical CD replay might be the equivalent of 3MP resolution, which is actually pretty good if you don’t look too closely, but subjectively the Meridian 808i.2 comes in at a far better 8MP. It simply has better detail, depth and gradation.

Using this player leads to a re-exploration of your disc library, finding new meaning in performances, greater understanding of the way orchestras and rock bands play. Vocals are more articulate, and sung with greater communication and conviction.

Again and again you hear substantial improvements in timbre, spatial relationships and inner detail, even on such compressed material as the Kaiser Chiefs Off with Their Heads. A dense mélange like this often trips up hi-fi systems, but the 808i.2 successfully unraveled the complex layering, managing to extract more of the essential vocals out of the mix, and soothing some of the usual (now, by comparison, understood to be a digital artefact) electronic ‘glare’.

Classical solo piano is rendered with a more open and articulate sound, the delicate yet complex harmonic richness better evoking the familiar sound of a live piano. Violins sing more naturally, sounding less congested and wiry, more vibrant and resolved. Even bass lines are firmly articulate and tuneful, and we found the improvement in resolution was achieved throughout the audio spectrum.

You feel that now domestic digital replay quality can be judged fairly for the first time. We have to suppose that Lagadec and others were right all along and that pre-echo is a distortion we have lived with for too long. If we and Meridian are right (ignoring for a moment whether 16 is sufficient bit resolution), surely no one can agree that a 44.1kHz sampling rate alone using conventional digital filters is a wholly transparent audio signal path?

In a top class system the 808i.2 can make a substantially improved contribution to CD replay. By removing the now cruelly exposed conventional filter signature, we hear sound which reaches back closer to the performance than before.

Returning to ‘conventional’ reference players, the filter artefacts are now plainly audible as a clouding, a blocking hardness, a compression of complex harmonic lines, a loss of separation of complex notes and chord groups, a sort of grey veiling and an overriding mild grain. It could be described as a sort of solarized ‘edge’, a touch of ‘glitter’ highlighting the upper mid and treble. Stuart describes it as unwanted ‘sparkle’ around sounds.

If this appears cruel, then it is. When the Meridian was returned, after a suitable interval I was able to acclimatise myself to the older reference players again, but could not forget the Meridian’s particular quality. Doesn’t this all sound very familiar? Have we not agreed with the long recorded litany of complaints about the digital sound described by many analogue die-hards?

So is Stuart’s new player the best ever? For many it will be. It pushes so many of the right sound quality buttons that any enthusiast’s back catalogue will replay with substantially greater beauty, resolution, and rather less listening fatigue. It will seem like discs have been recorded all over again, but properly this time.

For a classical music enthusiast and for listeners without too serious devotion to rhythm, and with a straightforward high resolution replay system not especially chosen for exceptional dynamics, the 808i.2 808i.2 will be the answer to a prayer. In this context it offers a truly state of the art sound, delivering a massive subjective score of 120. Indeed it’s so good that many pre-amplifiers may not fully do it justice.

While rock enthusiasts were also mostly held in thrall, they could still be a little disappointed with its cooler approach to rendering dynamics, and subjectively slowed pace. Jazz is rendered well, if a little downbeat. Some listeners suffered a degree of angst here, so unwilling were they to give up the greater drama impact and swing available from the similarly priced but more forward sounding Naim CDS3, for the transparency and tonal purity of the 808i.2. The score for such users could be nearer 70 – still very good.

Leaving the fabled CDS3 aside, against the most of rest of the world’s audiophile players this Meridian does do rhythm, thank you, while its huge recovery of detail adds such meaning to performance that criticising its rhythmic capabilities has less validity than usual. Connecting the 808i.2’s variable output directly to a power amplifier lifted its ‘classical’ best score to a massive 135, an all out record, showing moderately improved dynamic expression and grip thanks to the deletion of an external preamp and cable from the signal path. It worked just fine in this direct connection.

 

Lab Results
There is not much headroom on the pre-amp section, since the output clips at 3.4V (7V balanced) indicating rather small voltage rails for the analogue amplifiers. Given that the 100dB scaled volume control on full modulation CD clips at just past number 87, a lower sensitivity power amplifier is inadvisable.

The frequency response is dead flat 20Hz to 19kHz, and drops like stone at 20kHz with a hint of loss, maybe 0.5dB at the nominal band edge (see graph). The minimum phase impulse response was checked out and unquestionably there is no output before the impulse begins, so no pre-echo. The brick wall filter will now ring after all at a supposedly inaudible high frequency, but after and not before the event – this is the ‘apodizing’ filter in action.

While not state of the art, the test results are more than enough not to prejudice the sound quality as a whole. Full level distortion is 0.003%, channel separation typically 100dB, and channel balance about 0.05dB, all fine results. There was a hint of compression on the full level high frequency IM test, but the 0.0025% of difference tone recorded is still very good. Signal-to-noise ratio was 100dB unweighted, 107dBA. Linearity was excellent with fine level tracking right down to -100dB recorded level, 16-bit.

Output impedance was a load uncritical 50 Ohms, DC output offset was zero, and surface gap concealment about average at 1.5mm gap error. ‘Standby’ has very similar consumption to ‘on’; it’s the rear power switch which shuts it down. It draws 24W/35VA in standby and 27W/45VA when operating – a medium level of consumption for a high performance player.

 

Conclusions
Well made and finished with a comprehensive if bulky tablet-style remote control unit, this player has many versatile facilities and connections, including armchair control of absolute phase and that fine 100dB hybrid volume control.

The lab performance is just fine, with no issues of note save the higher than average ‘fixed’ output setting (requiring caution in A/B comparison with other players) and the modest maximum output level for driving some power amplifiers. The fine results were obtained regardless of the nuisance potential of a switch-mode power supply, a technology which Meridian seems to have under good control.

Good enough for moderate quality auxiliary sources, the on-board analogue pre-amp section does not aspire to audiophile quality nor begin to approach the performance of the CD player – nor do I think we should expect it to. Conversely, connecting the 808i.2 directly to a power amplifier fully exploits the dynamically capable high resolution volume control.

It delivers a sound of such clarity, purity, neutral timbre and immediacy as to leave the competition in  its wake. Direct-to-power-amp mode might handicap sources other than CD, but it delivers the most dynamic sound possible from this player, hinting strongly that the corresponding technology performance with the equivalently equipped DSP loudspeaker should also be promising.

This player is firmly recommended. Despite the mild caveat concerning rhythm and timing, it nevertheless  demonstrates a clear advance in the art of digital replay. I consider it delivers a proof of concept; that custom minimum phase ‘apodized’ filters do help digital replay sound more like the real thing.

 

The System
Marantz CD7, Naim CDS3 CD players. XTC The PRE SE and internal Meridian 808i.2 pre-amps. Conrad Johnson Premiere 350SA power amp. Von Schweikert VR-5 SE Anniversary and Avalon Eidolon Diamond loudspeakers. Transparent Reference XL MM2, Van den Hul First Ultimate, Cardas Golden Reference, Von Schweikert bi-wire cables.

.

Technology Backgrounder
The story behind the unusual digital technology employed in the 808.2i (and also its complementary digital active loudspeakers) could be said to begin with Bob Stuart’s decades long commitment to higher quality sound reproduction. A scholarly series of AES papers and lectures on the thresholds of aural perception could be said to have begun with his pre-university articles on the design of a high quality tape recorder published in Wireless World.

He studied audio electronics and psychoacoustics at Birmingham, then did a  postgraduate MSc at Imperial College, and has maintained an interest in perception ever since, in examining, for example, the required resolution limits for real programme under real room conditions. These results indicate preferred standards for recording dither and for resolution headroom to account for post processing of recordings, and sufficient archive transparency for future generations.

He has generated computer models of human hearing to help guide the design process, and in  recent years, in collaboration with Peter Craven, has investigated the effects of digital anti-aliasing and reconstruction filters.

For Stuart, MP3 compressed formats may have their place, but not in high fidelity. In a recent presentation at the London AES 9th December 2008, he also commented on the subject of double blind listening tests. His view was that the basis of such tests was flawed because listening to music is a constantly evolving experience, and the knowledge and  understanding of a piece is refined and rememorised every time one hears it, so that listening experiences are additive. You can never go back to the moment before it was last heard and make the same value judgment.

This view, with which I wholly agree, for example disputes that a published AES paper which reported a proof that a conventionally filtered 16-bit/44.1kHz (A/D – D/A) Red Book stage is more than sufficiently transparent to be inaudible as a link in a high quality audio chain. Some references are given below, including that for Michael Gerzon’s paper where he also encouraged researchers to go beyond double blind tests for audibility of signal errors. (Refs 1-4)

Noting that the A/B double blind method operates in the context of short chopped up sections of music created through randomised signal paths, these become more like confusing bursts of noise than the musical performances that are part of our listening culture. Through this loss of meaning, Stuart suggests that our  discrimination is blunted, frequently leading to null results for identifying subtle sound quality differences.

Since the early days of 16-bit linear (if you were lucky) non-oversampled CD players, using thick-film ‘brick wall’ anti-alias filters at 22kHz, the alternative approach of oversampling the data has become the norm, allowing the use of  digital filters with linear phase and low amplitude ringing.

Oversampling is now rife with all kinds of claims and benefits associated. When Lagadec warned of the possible pre-echo problem with commonly used digital filters, he was thinking of certain low-order (eg 64-tap) examples, whose mild but visible pass band ripple was directly associated with potentially audible pre-echo. That warning has since been more recently considered for much lower ripple, higher sampling rate digital filters, where their approach to measured perfection results in an a causal time response, again with pre-echo. Some authorities have discarded the notion of possible resulting audible effects on the basis that the oversample rate places the filter’s pre-ring echo frequencies above audibility. Yet subjective reports have continued of a mild residual hardness and ‘blocked in’ quality for digital replay  when compared with analogue, even though upsampling can often improve tonal quality. Stuart indicates that we may well hear the whole wave front energy including the leading edge, even if some of those components of the leading edge are in theory inaudible of themselves.

With the expert support of Peter Craven, Meridian has the resources to compute advanced filters and carry out careful listening tests. The researchers concluded that pre-echo was audible, altering timbre and masking transparency, even with old 16-bit material.

The new digital filter characteristics involve ‘apodization’, which in simple terms merely involves changes in the ‘shape’ of a mathematical function. An example of apodization is the use of  the Hann window in an FFT analyzer to smooth the discontinuities at the beginning and end of the sampled time record.

Meridian new filter necessarily operates at upsampled 88.2 kHz, requires a powerful 150  MIPS processor, and is designed to avoid the usual replay pre-echo, while also correcting some pre-echo on the recording.

In theory the absence of pre-echo should perhaps be heard as a more natural timbre and perhaps an absence of some masking of lower level detail. The audibility will also depend on the quality of the player as a whole, the transport  and jitter handling systems, and the transparency and neutrality of the subsequent DACs and audio path.

Note that several player manufacturers, right back to the earliest Krell and Wadia designs, have sought to address the specific choice of filter window and the resulting subjective quality issues, with varying degrees of success. Several modern players also have selectable filter windows, though so far without very great import.

 

Further Reading
REF 1: ‘Audibility of a CD-Standard A/DA/A Loop Inserted into High Resolution Audio Playback’. Meyer, Moran (Boston Audio Society). JAES Vol 55 Issue 9 2007.

REF 2: ‘Limitations of Double Blind A/B Listening Tests’ ‘Limitations of Double Blind A/B Listening Tests’ Gerzon (91st AES convention).

REF 3: ‘Antialias Filters and System Transient Response at High Sample Rates’ ‘Antialias Filters and System Transient Response at High Sample Rates’ Craven, JAES Vol 52, No 3, 2004

REF 4: ‘Controlled pre-response antialias filters for use at 96kHz and 192kHz’ ‘Controlled pre-response antialias filters for use at 96kHz and 192kHz’ Craven 114th Convention 2003

REF 5: ‘Coding for High-Resolution Audio Systems’ Stuart, JAES Vol 52, No 3, 2004

REF 6: ‘Predicting the Audibility, Detectability, and Loudness of Errors in Audio Systems’, Stuart, JAES Vol 39, 1991

REF 7: ‘Estimating the Significance of Errors in Audio Systems’ ‘Estimating the Significance of Errors in Audio Systems’ Stuart, Paper 3208 AES Convention 91 (1991) (also Vol 39)

REF 8: ‘Ringing False, Digital Audio’s Ubiquitous filter’ ‘Ringing False, Digital Audio’s Ubiquitous filter’ Keith Howard, Stereophile Jan 2006.

 

SUBSCRIBE!

Click here to subscribe to HIFICRITIC