The Lead Feature (Part 2)
Ben Duncan delves even further into the wires that connect it all together.
In Part One, Ben Duncan described the basic electrical properties of all kinds of connecting wires used in studios. In this episode, Ben focuses on cables for audio signal interconnections. What properties should different connecting cables possess for best results? What's wrong with ordinary wire?
In Part One, we looked at the effects of capacitance, inductance and resistance, both in the context of audio signal leads (where all three can be significant) and in respect of mains cables, where current capacity (a product of resistance-per-unit length) is the keynote. Of course, the net resistance of a mains cable can be just as important. Thankfully, the practical side of this topic has already been exhaustively documented in 'Studio Mains Supplies'1. Which leaves us with audio signal interconnections. These divide neatly into two groups:
Low level interconnects convey mic and level signals. Levels and impedances vary, but in all instances the movement of musical information involves a minute power transfer, nearly always less than 10 milliwatts (<0.01W).
High level cables connect amplifiers to speakers. Connections to cans (headsets) fall into the same category. In this class, peak signal voltages are usually somewhat higher than line levels. More important, peak current levels are upwards of 1,000X higher (eg, up to 50 amps). In turn, a lot of power is being transferred, and so resistance becomes a foremost element in correct application.
Pulling together the information presented so far leads to tried and tested, traditional advice on the choice of cable: Line and mic level leads should exhibit low capacitance (pF/m or picoFarads per metre), particularly if both the source ('output') and load ('input') impedance are high (meaning above 600 ohms). The overall amount of capacitance that's acceptable depends on cable length, and on the source and load impedances. These details are charted elsewhere2. The resistance of signal leads is almost always small compared to equipment impedances, enough to be ignored. Likewise the inductance of conventional shielded cables is generally considered low enough to be waved on. It more or less has to be, generally because reducing cable inductance generally means increasing the shunt capacitance by the same amount, which is just as bad, if not worse!
On long cable runs, the cable's shunt capacitance combined with a build-up of series inductance can unleash resonances, liable to perturb the high-end response of dynamic microphones and their associated transformers. A sure sign is when a mic lead is longer than usual, and the mic at one end requires more than the usual amount of EQ. Leads long enough to give problems can be 'cleaned-up' either by empirically 'tuning-out' the resonance with an RC (Resistor + Capacitor) Zobel network3, or else by buffering close to the mic with (for example) an active 'mic splitter'.
High Level Cables should be as short as possible and made adequately stout, so their resistance is much smaller than the loudspeaker's worst-case (lowest) impedance. As a rule of thumb, this is half the nominal rated impedance, ie. for a speaker rated at eight ohms, read four ohms, to be safe. As we saw in part one, excess resistance eats up amplifier watts you've paid for, though the loss of output level won't be very audible unless the wires are so thin they actually get hot! More significantly, excess resistance reduces the numeric damping factor and can also change the tonal balance of the speaker.
When choosing the optimum cable size for monitor interconnects, the cable resistance is just one factor in the equation for parasitic resistance. Others include the amplifier's source impedance as well as the resistance of the amplifier's and the speaker's internal wiring, together with the connectors. In other words, it's all academic if your amplifier's binding posts aren't firmly tightened. Since thick wire isn't unreasonably expensive (set against the cost of even the cheapest monitoring package), a generous rule of thumb to aim for is a cable resistance that's around 1/50th of the minimum speaker impedance. You can then be sure it plays a negligible part in affecting the response, both in the time and frequency domain.
Figure 1 allows you to read off the maximum cable length that's allowable while keeping cable resistance below 1/50th of the speaker's worst case impedance. Simply enter the cable size in use. If the length you're using exceeds the lengths recommended in the table, it certainly won't cost the earth (and not just because two-core cables have no earth...) to improve your monitoring system's damping with a larger diameter cable, commensurate with the length involved. On the other hand, if you're working on a shoestring with budget amplifiers and jack plugs (whose contact resistance can be much higher than binding posts' or XLRs') there's no point in going overboard, as the parasitic resistances will negate ultra-low cable resistance. Looking into the table again, a judicious compromise is to accept double the indicated length of cable, for a given size in mm2.
As low frequency speakers generally 'pull' the highest peak currents, and because the audible effects of inadequate damping factor become more obvious the lower you go, wires solely connected to bass or sub-bass speakers benefit the most from being chunky. When it comes to mid and high frequency drive-units, the traditional recipe becomes less clear cut. On the one hand, fat wires should be used, to minimise resistance and avoid frequency response abberations. Then again, we should be using thin wires for least series inductance, again, to conserve a flat high frequency response.
In part one, we looked briefly at the way in which a cable's reactive components could make it exhibit a constant impedance, almost regardless of length. But these Transmission Line properties and the religious matching of impedances that goes with them (to prevent signal reflections) are normally only considered important at high (RF) frequencies or for wide bandwidth signals (ie. video), or for very long audio lines (ie. trans-Atlantic phone lines), where the cable's length becomes a significant fraction of the wavelength of the frequency(ies) concerned. Even in a large studio, a 20m run of cable is barely 1/750th of the wavelength, at 20kHz. At and below this frequency, the cable knows it's a 'short' one, and mostly behaves as if the sum of the distributed R, L and C were three discrete components, lashed across and/or in series with a notional 'perfect' cable. This lumped model of a cable (cf. distributed) is shown in Figure 2, with typical values for a rather thin, 10m speaker cable. For thicker cables, L is bigger (but not much), while R is much lower.
The irrelevance of Transmission Line behaviour is fortunate, since few if any speakers present anything like a constant impedance, and precise impedance matching would be impossible! So the video engineer's trick - that of soldering shunt capacitors across the line at intervals, to reduce series inductance in fat coaxial cables - won't work in the way expected. At best it may slightly improve the treble response, but at worst it could blow up an amplifier that's not adequately protected against the de-stablising effects of having 100nF+ of capacitance strapped across its output. Is there some other way to gain low inductance simultaneous with low resistance? Yes: by connecting conductors in parallel (Figure 3), both R and L are seen 'n' times in parallel, and the overall value is divided by n.
Just over a decade ago, hi-fi shops began to offer Litz wire, which makes use of this technique. It comprises numerous strands of wire, all mutually insulated, and twisted to further cancel the net inductance. Some users claimed a great improvement in sonic quality, but lacked the adjectives to describe it fully. Others were red faced: either the cable made their hi-fi sound awful, or their amplifier had died outright. In hindsight, the residual R+L (Resistance + Inductance) in ordinary cables buffers (stands-off) the amplifier's output from the cable's and the speaker's shunt capacitance. Without this shielding effect, flimsy amplifiers with inadequate stability margins4 went down in flames when faced with Litz cable, or otherwise burst into vhf oscillation, and sounded horrible. The upshot is that unless L and C are abolished simultaneously, or reduced in step, there's a risk of creating problems with low inductance cables, in the surrounding equipment. If you're tempted to try out Litz (or similar low inductance) speaker wires, be sure to check for the tell-tale signs of ultrasonic oscillation before judging the results4.
Hi-fi accessory makers are continuing to introduce exotic speaker cables, apparently because a large sector of the public are convinced of audible benefits. Since 1977, Litz leads have been joined by a variety of exotic cables, employing widely differing techniques. All of them aim to improve replay resolution by (amongst other things) changing the ratio of R, L and C. Of these, the simplest comprises thin singlecore wires, generally unshielded and widely spaced in a ribbon. This format, designed after extensive research by DNM (a UK hi-fi equipment maker) provides highish R, very low C and high L - the exact opposite of Litz. Curiously, it's also exactly contrary to what classical theory predicts is good and proper for speaker connections.
Modern psychoacoustics knowledge is beginning to close the gap between what classical theory predicts, and what we actually hear. Equally, a complete understanding of how cables affect the transmission of music means re-evaluating Maxwell's original (1869) Theory of Electromagnetism. Briefly cutting swathes through acres of heavy vector calculus, cable geometry sets up a propagation delay, known (in electronics jargon) as Dispersion. So much is well-known and accounted for when cables are looked at as Transmission Lines. But such effects have long been neglected in audio (aside from telecoms). 'They are too small to matter,' insisted the Professor, 'just experimental noise'. With the kind of inductances prevalent in average cables, the delay time is certainly small. Going by classical theory, it's generally below 1μs. On the other hand, it varies across the audio frequency band, so the high frequencies always arrive before the mid frequencies, and the bass last of all. The plot thickens...
For some individuals, sensitivity to dispersion may range down to arrival time differences of 100μs or less. If dispersion is perceivable, it would be most likely to stand out at the leading edge of the attack portion of instruments' envelopes, before fresh reverberant energy swamps it. For such individuals, small differences in arrival time are audible as 'smearing'; a lack of definition. For them, fitting cables which lessen the differences in arrival time is like snapping a lens into focus. The absolute time delay is secondary: what matters is time alignment.
The conductors in everyday audio cables are far from perfect - if you look closely enough. In some shielded cables, particularly co-axial types principally made for RF (Radio Frequency) use, the conductor is steel with a thin copper coating. Cables of this type are readily identified: pure copper is soft, steel wires are springy. The reasons for using a steel core are valid enough. It's strong, cheaper than copper and most of the current bypasses it, preferring to flow along the copper surface. Except at (low) audio frequencies, where the skin effect lessens and the current flow spreads into the steel core. Then steel's higher resistance and magnetic character are positively unwelcome. For a start, there's distinct likelihood of introducing distortion if the wire vibrates in the vicinity of a (speaker) magnet.
More often, the conductors in everyday cables are described as 'solid copper'. Since the cost of refining a metal spirals as the purity approaches 100%, it should come as no surprise that 'solid copper' in everyday wires is no more than 99.5% pure. The main impurities are oxidising agents, both sulphides and oxygen itself. Other chemicals migrate into the copper, both from the insulating sheath (plastics like PVC and butyl rubber are less inert than you may imagine), and from the surrounding atmosphere. It follows that conductors become progressively more impure with age.
Because the level of impurities in 'bog standard' wire has a negligible effect on the measured DC conductivity (in milliohms per metre) and even badly corroded specimens don't exhibit easily measurable levels of THD (harmonic distortion) when tested with continuous sine waves, it's natural to conclude that they don't matter.
Manufacturers of exotic electronic equipment requiring high reliability think differently. Looking at the worst case - the potential liability of one imperfect connection in a mainframe or satellite - they specify either wires made either of high (99.95%) purity copper, or going one step further, 'Oxygen Free High Conductivity' (OFHC) copper. Plain wire of this calibre is reasonably priced and readily available in single-core 'Wire-Wrap' format, eg. Tefzel®. Alas, shielded or high-current cables incorporating OFHC are scarce and expensive.
OFHC wires are usually silver plated. Although silver's conductivity is slightly better than copper, and top of the table, its surface corrodes much more readily - remember the black crud that forms overnight on silver cutlery? With this in mind, OFHC wires are insulated with especially inert plastics (eg. PTFE), ones which won't oxidise or pollute the conductor. Single core construction is then presumably necessary to restrict the ingress of atmospheric nasties: if the insulation is tight enough, a continuous gas-tight seal is formed. The same can't be said for ordinary stranded wires, thanks to tiny gaps between the strands. The downside is that cables with single-core wires are easily broken by repeated flexing, making them fairly impractical in the studio for anything but permanent interconnects. So if there's no measurable benefit, why use high purity conductors at all?
Over the past decade, audio researchers experimenting with interconnects made with commercially available OFHC wire found that sonic quality did seem to improve with ascending conductor purity. The question was: why? Kamada, a researcher leading Hitachi's cable division, and a Dutch electro-acoustics engineer, AJ van den Hul were both convinced that the improvement was due to differences in the copper's crystal structure. Conventional copper conductors contain upwards of 150,000 crystalline boundaries per metre. When the wire is drawn, the residual voids and cracks between the boundaries are expanded. It's between these that the major impurity, oxygen, lurks. In OFHC wire, the reduction of oxygen content comes hand in hand with around a third fewer crystal boundaries - say 50,000 per metre.
Van den Hul speculated that the crystal boundaries looked like diodes (ie. nonlinear resistances) to the small voltage changes which ride on the fundamentals to give musical detail. He then went ahead and manufactured a series of cables containing so-called 'non-crystalline' wires. Working in Japan, Kamada followed a similar path. He saw that oxygen content restricted the growth of the copper's crystals during annealing. Using a zone refining process (more usually associated with the production of semiconductor materials), a slab of copper was produced which contained a minimal number of large crystals. Like van den Hul's 'monocrystalline' cable, the resulting wire - dubbed 'Linear Crystal' by Hitachi - had only two to four hundred crystals per metre.
'Linear' and 'Mono' - crystalline wires appear to improve the 'air', 'space' and 'detail' laid down and retrieved in music recording and playback. But how? If real, 'diodic' (diode type) effects should be measurable using DC instrumentation. Although non-linearity can be found, it's usually buried deep below the audio noise floor. A more convincing reason has emerged after Malcolm Hawksford's clearsighted re-evaluation of Maxwell's Theory of Electro-magnetism (1869): The multiple pathways formed by crystal-boundaries and voids aggravate dispersion effects in signal propagation down the wire(s)5. What's more, a re-evaluation of Maxwell's theory suggests that the dispersion time delay is much longer at audio frequencies than predicted by classical interpretation. Moreover, ordinary cables exhibit a 'memory' or storage effect - lasting upwards of 100μs. These effects aren't measurable by steady-state techniques, but are fairly evident to our ears, which can easily resolve small (<1ms) errors in the time domain. Once you know what to listen for... Put simply, ordinary copper conductors contain complex, labyrinthine pathways, which delay and smear the leading edges of music. The same goes for stranded wires: individual conductor's surfaces don't make perfect contact and are liable to oxidise, again creating a chaotic series of pathways and loops. The single 'solid' conductors that DNM popularised avoid this. Equally, you could use permanent mains wiring cable (the grey, flat stuff) though the copper isn't so pure. At the same time, the new theory says that thin conductors exhibit the least time dispersion across the audio range. The optimum size of conductor is around 0.8mm diameter = 0.5mm2.
In recent years, audiophiles have seized on a way of making the best of cable's conflicting behaviour at opposite ends of the audio spectrum. Looking at Figure 4, bi-wiring involves removing the crossover network from the speaker cabinet and mounting it adjacent to the power amplifier. To avoid unnecessary magnetic side-effects, the crossover will require housing in an enclosure that's non-magnetic, ie. not made of steel. As it's carrying a high-level signal, shielding isn't required, so plastic or wood will suffice. Having shifted the crossover, the drive-units are then connected by wholly separate wires, ie. two or four cores for a two-way system. Don't be tempted to use a common 'earth' return, just for the sake of using three-core mains cable. The object is to keep the LF and HF feeds completely separated over most of their length.
Although it appears oddball at first sight, bi-wiring approaches bi-amping. If the inputs to the crossovers' hi- and low-pass filters are also split and each network is driven from a separate amplifier, the outcome is a 'passive' bi-amped system. This arrangement possesses many of the advantages of driving speakers with multiple amplifiers and an active crossover6.
The immediate effect of bi-wiring is clearer sound, because interaction (hence intermodulation) of high level signals is restricted to the length of common cabling between the amplifier's output and the crossover's input, which can be kept to 12"/0.3m or less. At the same time, the wiring can be optimised. Since tweeters (as a rule) don't draw high peak currents (ie. the worst dynamic impedance dip for moving-coil tweeters is rarely much less than the nominal rated impedance), the classic requirement, namely low resistance, can be relaxed without detriment to the bass driver's damping.
It follows that 0.5mm2 wire can be used for the tweeter's wiring, to keep the high frequencies in mutual alignment. At the same time, the bass/mid driver(s) can be connected with the heavy duty cable that's desirable on more pragmatic grounds. Finally, by making the tweeters' cabling around 10 to 15% longer than the LF cables, the low and high frequencies can be (more or less) time aligned. Of course, the excess length shouldn't be coiled-up, but left hanging loose. For three-way speakers, the same idea can be applied pro-rata, using medium (1mm2) wire for the midrange drive-unit, and say, 7% excess length.
References & Further Reading
1. Ben Duncan: Studio Mains Supplies, Part Three, Aug 85, H&SR.
2. Ben Duncan, Matching Interconnects, June 84, H&SR.
3. Ben Duncan, Sound on Stage, Pt 9, E&MM, Aug 83.
4. Ben Duncan, How to Calm Hysterics in Op-Amps, Feb 86, H&SR.
5. Malcolm Hawksford, The Essex Echo, Hi-Fi News & Record Review, Aug 85, Aug 86, Oct 86.
6. Ben Duncan, Monitoring Pt 3. Dec 83, H&SR.
Martin Colloms, Bi-wiring, Hi-Fi News & Record Review, June 86.
John Robert, Exposing Audio Mythology: Speaker Wires, Recording Engineer & Producer (USA), Oct 83.
This is the only part of this series active so far.
Feature by Ben Duncan
Previous article in this issue:
Next article in this issue:
mu:zines is the result of thousands of hours of effort, and will require many thousands more going forward to reach our goals of getting all this content online.
If you value this resource, you can support this project - it really helps!