Musaeus takes great pride in the sound quality of its recordings, while freely admitting that there is always room for improvement. Before going on to describe some of the ways in which Musaeus seeks to maximise sound quality in its (and others’!) recordings, a brief definition of sound quality and an explanation of why it matters may be in order:
Sound quality is a measure of how much (or how little) the technological means of recording and reproduction interfere with the listener’s enjoyment and absorption of the music.
It can reasonably be taken as axiomatic that sound quality cannot be ‘perfect’. Any means of representing and storing such a complex phenomenon as musical sound can only approximate perfection. But when it comes right down to it, what is of paramount importance in the recording and replay of music is that the emotional content created by the composer and performers must be transmitted to the listener, since that is what music is all about. So the best recording, essentially, is the one that puts the least in between the performers and the listener.
With that in mind, Musaeus has done its best to adapt and optimise technology - not to mention such simple expendients as being careful in its choice of equipment to use - in order to maximise sound quality. A typical Musaeus recording session involves the minimum of equipment: a stereo microphone array (usually two microphones, but sometimes three - read on!), a simple microphone amplifier, an analogue-to-digital converter and a digital recorder - usually a CD recorder for reliability and convenience, though since we are only using it as a ‘bit bucket’ (because the sound quality-critical part of the operation is done by the ADC) a DAT recorder does just as well. If recording levels are set correctly at the session, there will not necessarily be any need to make any changes to the audio in the digital domain and the issued CD will be a bit-perfect copy of the original session disc, give or take edits (which in Musaeus recordings are invariably ‘butt-splice’ digital edits rather than crossfades, by the way).
Perhaps the most important single element in a recording of acoustic music is the recording venue. Musaeus is very fussy about venues and most of our recordings have been made in churches, of which the best examples have nicely controlled reverberation, very low outside noise and altogether a very ‘friendly’ sound. They are also nice environments in which to perform, and it is often forgotten that musicians need some acoustic ‘feedback’ from the environment in order to hear what sounds they are producing - lacking those clues it’s surprisingly difficult to play at one’s best, even on a highly mechanical instrument such as a piano: for singers it’s near impossible. Likewise, groups of performers need to hear each other clearly.
Probably next most important factor in determining final sound quality is the type of microphone used, in conjunction with its positioning relative to the performers and to the venue. There is no completely satisfactory microphone. Classical-music recordings for the last forty years or so have been dominated by capacitor (or ‘condensor’) microphones, of which many types and models are available. The fundamental problem with the capacitor microphone is that it is a pressure-sensing device and as such is not directional. Even in mono recording systems it is useful to have some degree of directionality in a microphone, if only to give more options in placement etc. so that the relative pickup of sound from different directions is controlled, giving a more or less ‘direct’ sound as required. In stereo, directional microphones are mandatory for most types of array - the exception being the various kinds of ‘spaced omnidirectional’ array, which however have their own problems if used ‘raw’.
Most commercial capacitor microphones are in fact directional, but if one looks closely one finds that the way they are made is in fact a bodge - a very sophisticated bodge, but a bodge nevertheless. Basically, a carefully designed rear ‘leak’ is provided so that to an extent they sense the difference in presure between the front and the back face, and since pressure difference (or ‘gradient’) is a vector quantity, unlike pressure, this makes the microphone directional. It’s a well honed system that works quite well but it inevitably results in a directional response which is not very constant with frequency, or - looked at the other way round - a frequency response which is not flat off-axis.
A second problem with capacitor microphones is that because they rely on a stretched membrane to sense air pressure or movement, this membrane acts like a drumskin and is resonant. In some omnidirectional microphones the resonant frequency is so high as to be ultrasonic, which reduces its audible impact, but in most omni and all directional capacitors it is in the audio band and although it is generally highly damped one should not assume that its effects will be insignificant. (This is terribly hard to prove one way or the other, and ‘thought experiments’ related to it tend to hinge on the equally controversial area of the audibility of phase distortion. Phase distortion is most certainly audible in extreme cases - if you were somehow to delay all frequencies below 100Hz by half a second it would obviously be audible - but whether or not it is a real problem in the amounts normally seen in practical electronic and electromechanical systems is definitely a moot point at present.)
The other type of microphone occasionally used in classical recording is the ribbon, which behaves quite differently from the capacitor. For a start, it is intrinsically a pressure-gradient sensing microphone and so naturally has a directional response which is described as a ‘figure-8’: maximum sensitivity directly in front and behind (though behind is in reverse polarity) and zero at the sides. Like a capacitor, a ribbon microphone has a mechanical resonance but in this case it is at a very low frequency, as low as 2Hz in some microphones and seldom higher than 50Hz. As a direct result of its directional nature, a ribbon microphone has ‘proximity effect’ which makes the bass more prominent if the microphone is close to the sound source (less than a few feet away). And then all currently available ribbon microphones are significantly noisier than most high quality capacitors. (There is a widespread belief that a ribbon microphone, which is a purely passive device containing no active electronic components, has no self noise and only requires an infinitely quiet microphone amplifier to work perfectly. This is incorrect. Due to its own resistance it produces ‘Johnson noise’ like any other resistor, not to mention various acoustic forms of noise which are present due to the acoustic design.)
Musaeus has used both capacitor and ribbon microphones, in various configurations. The most recent recordings use a novel array consisting of two ribbons and one capacitor, the design of which was inspired by John Watkinson who pointed out in an article in Studio Sound that the sum of a ribbon (figure-8) microphone and a capacitor (omni) is a cardioid response (highest sensitivity to the front, zero to the rear and descreasing smoothly as one goes round from front to rear). The Musaeus array thus uses an omni and a forward-facing ribbon, summed together to give effectively a forward-facing cardioid. This output is then combined with a sideway-facing ribbon in a mid-side matrix to give a classic Blumlein-style coincident stereo pickup. Musaeus has dubbed the arrangement ‘OMS’ for ‘Omni-Mid-Side’.
The results of this can be heard on MZCD101 and MZCD102.
(Note: the MS arrangement is mathematically equivalent to the standard coincident ‘XY’ stereo setup, where the left channel microphone points 45 degrees left of straight ahead and the right channel one 45 degrees right. Either system produces ‘intensity stereo’ according to the principles outlined in the 1930s by Alan Dower Blumlein, and the results over a good stereo loudspeaker system have been found highly satisfactory over the years. It is possible to use any kind of directional microphone in either XY or MS mode, and both cardioid and figure-8 pattern microphones have been used successfully. The OMS system approximates a pair of cardioid microphones in XY after decoding to regular stereo, but by varying the ratio of omni to M microphone the amount of rear pickup can be adjusted to suit circumstances. In general, the decision on whether to use XY or MS is largely up to the taste of the engineer, but MS has one useful advantage particularly in the case of a physically small sound source (a singer standing in front of a piano, say, as opposed to an orchestra or large ensemble), which is that the M microphone at least is working direct on axis and thus makes use of its best frequency response, whereas both left and right microphones of an XY array would be relying on their pickup of a central source at 45 degrees off-axis. At very high frequencies even ribbons deviate somewhat from perfectly flat response off-axis.)
The OMS system has a further advantage over simple crossed ribbons, apart from the adjustable rear pickup: the omni will in general have a much lower noise floor than the ribbons and so adding it in boosts the overal signal-to-noise ratio, by as much as 3dB. Even better, the omni will often have better treble extension and certainly better bass (intrinsically the bass response of an omni extends to DC: even given the practical limits imposed by various factors it can extend below 5Hz within 1dB). This improves the overall frequency response of the system, albeit in mono.
Musaeus has used a pair of Coles 4038 ribbon microphones (as designed by the BBC in the early 1950s and still produced, largely unchanged, by Coles Electroacoustics of Hoddesdon, Herts, UK) in conjunction with a single Brüel and Kjær (now DPA Microphones) 4003 omni. The microphone amplifier used with the 4038s was specially designed and built by Richard Black for the purpose, and incorporates a mild frequency response correction for the rolloff of the 4038 in the extreme treble and bass. It has very low noise and gain optimised for the application. The 4003, being a line level microphone, does not need any amplification before being mixed with the (suitably boosted) signal from the M channel 4038, but it does need a power supply and on various occasions Musaeus has used B&K’s own 2812 dedicated power supply and line driver (modified for better sound quality) or a simple unregulated power supply.
(Note: DPA makes two versions of much the same microphone, the 4003 and the 4006. The latter is much more common as it runs on a conventional studio phantom power system and requires no external power supply. The 4003 requires a power supply and apart from the fact the DPA’s own supply is very expensive this means that extra cabling must be used and convenience is compromised. However, the requirements of phantom powering, and the extremely small output transformer used in the 4006, restrict the performance of the phantom powered version and the 4003 seems overall to offer many advantages if one is after ultimate sound quality. The power supply requirements are very simple (130V ± 5V with low ripple and noise: 7mA draw per microphone) and the output of the microphone is high enough, and the output impedance low enough, to drive moderate-length cables direct to many types of studio recorder with no intervening electronics at all - apart from a DC-blocking capacitor, unless you are confident that your recorder can handly 65V DC on its inputs! - in many cases.)
For ADC, Musaeus has for the last three years standardised on the superb models from dCS Ltd of Cambridgeshire, UK - most often the model 902, which is available to rent from several outfits at prices that make buying one unattractive for small companies. These use dCS’s patented ‘Ring DAC’ architecture to give linearity estimated at in excess of 29 bits effective, and are capable of sample rates up to 192kHz (model 904: up to 96kHz on the 902) and word lengths up to 24 bits at the output. Since Musaeus normally records direct to CD which is a 44.1kHz sampling rate, 16-bit medium, we use the internal dithering of the dCS to truncate correctly to 16 bits output and sample at 44.1kHz, avoiding any necessity for later conversion.
Musaeus uses PC-based editing to select from recorded takes, and in addition performs a2i mastering (another Musaeus invention) if appropriate. ‘If appropriate’ is an important rider there: a2i removes ultrasonic distortion, but that distortion will not arise unless there is significant energy present in the recording above about 20kHz. That is certainly the case with many bright-sounding instruments such a trumpet and violin, but close examination of the spectrum of a piano will show that there is next to no output at such a high frequency (at least, not unless the piano has been recorded at ridiculously close range), and human voice also has little output so high up except for consonants which are of a noise-like nature and don’t show up ‘Alias Intermodulation Distortion’ to a significant degree. Accordingly, it is no surprise that no listener has yet found a2i to be consistently beneficial on recordings of voice and piano and so Musaeus has not processed any of these. However, on MZCD101 Wilson Collins’ violin clearly produces harmonics to way in excess of 20kHz so the violin sonata has been subjected to the a2i process. The rest of the disc, and the whole of MZCD102, is pure and unaltered bitstream as output by the dCS 902, as we have verified by digitally subtracting original session take extracts from corresponding passages off the finished, manufactured discs and listening the the stony silence that resulted.
That’s most of the technology - not much, in fact, and in a way the less the better. There is one final tweak which we apply to recordings: it seems (according to reports from inside Philips’ research labs, no less) that there is evidence that many CD players give lower jitter levels in the audio output when the disc is completely opaque. We haven’t tested this, but because it costs nothing and looks nice we have made sure that all recent titles have a completely printed top surface. You can believe that this is significant or not as you please - we refuse to be drawn!
Some related information which you might find of interest can be found on our a2i page a2i.htm and also on our Digital Audio Broadcasting page.