In rounding off his behind-the-scenes look at the new Rauch DVT series power amplifiers, Ben Duncan outlines speaker currents, 3D imaging, how to fry eggs and why the space shuttle exploded.

In Part 2 we looked at amplifier power supplies in an energy storage context. In freeing the Rauch DVT from sophisticated protection circuitry, alias V & I limiting, the speaker's current draw is relatively unrestricted, so it's time to back up the consequences.

Speaker Current Demands

The EV Sentry 100 monitors' impedance curve (See last month's Figure 2) gives us a lead into the likely worst case of current draw, in this instance for a nominal 6Ω of load impedance. At 150Hz, the impedance dips to 4.8Ω. At 40V drive (about 200W) this implies I = ⁴⁰/_4.8 Amps. That's just over 8 Amps. But wait, the Sentry 100 monitor is renowned for being better behaved in the impedance department than most. Besides, all the loudspeaker impedance curves published in data sheets are made under steady state conditions, using a swept sine wave. Once we bring in computer tools like Audio Precision or TECRON, begin looking at monitors with complex passive crossover networks and make our measurements with pink noise (instead of single, swept tones) the instantaneous impedance dips turn out to be lower than expected: around ⅓rd to ¼th of the nominal which would be 2 to 2.8Ω for an 8Ω system. Taking the worst case value, this implies that we need to source ⁴⁰/₂ = 20 Amps to ride out an impedance dip at 40V drive, without distortion or clipping.

For some monitors with a nominal 4Ω rating, we can anticipate a pro-rata worst case draw of 40 Amps. This is a good moment to stress that these currents are manifest as short millisecond bursts and are by no means continuous. At the same time, the peak we can draw can certainly be repetitive, when we say an instrument or vocal keeps hitting the frequencies where the impedance dip is greatest.

Turning to the big rock 'n' roll stage, cabinets are normally wired in tandem, but it's usually without the imposition of passive crossovers, which are largely responsible for the anti-social impedance characteristics of elaborate studio monitors and hi-fi speakers. For example a pair of Turbosound TMS-3 cabs rated at 4Ω each on the bass end dip to just 2½Ω at 250 Hz when paralleled. So even when they're driven bridged by a DVT250 at 100v (= 1250W into 4Ω), the current draw is no more excessive than some 4Ω studio monitors, driven at a fraction of the power.

Standing on this high ground, we can now see why monitor and PA systems using active crossovers are, in general less fussy about amplifier credentials and conversely, the extent to which certain studio monitors give their partner a hard time.

To sum up this topic, unless and until the devious temporal relationships between Amps, Volts and Joules between the wall socket and the speaker cone are examined and brought together in an holistic way, then deterministic dogma will continue to fulfil its own prophecies, namely that all good amplifiers are un-improvedly perfect-perfectly mediocre, that is. (That's exactly how I'd have put it... ed)

In the Midnight Hour

For the more detailed and intense listening that goes on in the control room, voltage drops in the power and output conductors have a knock-on effect. This arises because all voltage perturbations are after all signals. Signals derived in some grotesque way from the original audio copy. Thus they represent delayed and distorted feedback, of arbitrary phase, superimposed across the amplifier's power source. To some extent, all modern amplifiers exhibit PSR (Power Supply Rejection) which is expressed as a dB ratio, typically referred to the output terminals.

This effect seems marginal, and perhaps it would be if the distortion products were exclusively the gentle 2nd or 3rd harmonics. But a perceptive set of human ears can wince at binaural information 20dB or more below the signal's noise floor; that's below roughly -100dB for analogue tape, or below -110dB for digital sources. Meanwhile, Figure 1 reminds us that any series of voltage fluctuations imposed on the power rails provide lots of scope for harmonic complexity. Here, for example, four different currents with diverse phase relationships are mutually interacting and thanks to feedback, they're simultaneously trying to compensate for each other. In effect, the amplifier is trying to pull itself up by its on bootstraps and vice versa all at once. The broad outcome is spikey, high order harmonics, sounding like a minor second! Just fractional doses of these are enough to poison the sound with dissonance.

Against this, the size of just a single component for a bass response that's flat to 0.1dB at 20Hz, when driving 2Ω, is 10,000μF. Capacitors of this size (or even approaching it) are apt to exhibit high leakage at operating temperatures, lots of dielectric absorption, audible resonances and bass end distortion or microphony, not to mention a failure rate that's potentially as high as the power devices they're setting out to protect the speakers from. In the event, Dr. Anton took a personal dislike to these unsavoury compromises. The 'industry standard' protection scheme uses a relay to break the output circuit, in conjunction with a circuit which senses + or - DC voltages above a threshold, but ignores all kinds of legitimate music signals, which are ideally AC (Figure 3).

DC protection, however basic, is always better than no DC protection, but there are three snags.

Firstly, there are insensitive relays which don't open until the DC fault voltage has exceeded the average speaker's meltdown voltage. That's above about 15 or 20 volts.

Secondly, in another scenario the relay opens and closes in time with heavy DUB in concert. Typically, this is the outcome of music's asymmetry, especially on synth bass. Although the waveform is AC in so far as it is alternating, there's nevertheless a nett imbalance in one or other polarity and to the sensing circuit, this looks like a DC voltage.

Thirdly, a DC fault generally causes the relay to hold on only so long as the fault is present. That's fine if the fault is genuine, in which case it's highly unlikely to 'repair itself.' It's baffling though if the DCP (DC Protection) resets itself, so the music returns abruptly after being triggered by a DC 'thump', pulsed through the system at switch on. On the plus side, if relay operation is fast enough, it can take the edge off the Thump, which saves a few ripped ears, not to mention cones.

-45dB is an average figure for power amplifier PSR, at mid-band frequencies, ie. 1 kHz. Surprise, surprise, the DVT improves on this; rejection is 100 times greater, at least -85dB, and whereas PSR falls off in all real world circuits at HF (>10kHz), the DVT's 100 fold improvement is maintained up to 100kHz.

The low level of small signal content of music is primarily ambient information. This is the raw sensory data from which we can mentally reconstruct the original acoustic space. Therefore, a good stereo recording can generate a nearly solid 3D image.

Naturally, nature's spacial clues are readily masked by spurious high-order harmonic garbage (eg. spikey or raspy components), especially if the interlopers are lacking in any direct musical relationship. This is true if the left and right (or high and low) channels are listening in to each other's error signals. Better known as 'crosstalk', it's a cumulative effect. Unless rigorous anti-crosstalk measures are built into a design from its conception, we can expect a fraction of this high-order rubbish to leak across to the other side throughout any tightly packed audio assembly. In this context, a single power supply shared between two or more channels commonly forms an easy pathway for the cross-leakage of garbage. It follows that a bog standard power amplifier with two isolated power supplies, has a better chance of revealing a vestige of the stereo image's nuances. But this isn't the full story: once the role of power supply behaviour on perceived stereo space is fully understood, it's possible to do just as well with a single transformer, shared between the two channels. As in the DVT 50s.

You can soon check out audible crosstalk on an amplifier. It's done by driving channel 1 into a test load (or small speaker, locked into a 'Dub box'), then listening in to Channel 2; the input of which is muted at the XLR. (A shorting plug with pins 2 and 3 linked is handy here.) Then you can reverse the channels. Is it any better listening in to Channel 1, with Channel 2 driving? Even if so, asymmetric crosstalk is no friend of good stereo recordings.

Entropy and Synchronicity

Output house-keeping exists to protect a power amplifier (and speakers) from itself should it misbehave. But given a well thought out design, should it all be necessary? In a word, Yes. Reliability arises out of preparedness, not in spite of it.

First of all, in the present universe, there's a primeval force at large, known as 'Entropy'. Its task — to disperse, undo and undermine all life, structure and coherence, and replace it with disorder, chaos and nothingness. Physicists and deterministic thinkers have no root explanation for this, though they do recognise it as the '2nd Law of Thermodynamics'. The normal response of musicians is to scream abuse, or intone 'Sod's Law'; 'If anything can possibly go wrong, it will, and at the worst possible moment!'

A curious facet of the Entropic Force is the predisposition for accidents to come not singly, but in twos, threes, or dozens. A German called Kammerer lead the work on this, during the early days of this century. He collected a vast body of data on everyday coincidence, published in 'Das Gestatz der Serie' (The Law of Seriality) in 1919. Later, in 1932, Wolfgang Pauli (an early, visionary Nuclear physicist) and Gustav Jung (the psychologist) coined the word 'Synchronicity', to bring together what Schopenhauer called 'the simultaneous occurence of causally unconnected events' and Kammerer had described as 'cyclic processes which propogate like waves along the time-axis of the space-time continuum.' In the normal course of everyday events, this phenomenon is referred to as 'coincidence' or 'bad luck.' 'Causally unconnected' means events which come together, without one being the cause or effect of the other.

The upshot is that accidental circumstances are rarely simple. Being prepared for the worst may mean it will never happen, so glorious and elaborate protection circuits may seem like a waste of time, in so far as they never have an opportunity to prove themselves. Or it may mean that some other highly improbable disaster arises; Sod's Law is a cosmic joker, forever seeking out the weak links. A prime example occurs in nuclear power plants, where new and unexpected combinations of bad circumstances/human error are forever testing the supposedly 100% fail-safe electronic systems.

The way around these quirks is preparedness, not by brute force (like chained, subsidiary protection circuits, ad nauseam), but by holistic design. This means avoiding interdependance, by drawing up a truth table which reveals the internal contradictions of the system, like surreptitious feedback loops: 'If three things all go wrong at once, something happens which defeats the protection'. This table is built up out of a catalogue of all the nasty things that amplifiers have suffered in the name of keeping the show going, since the dawn of rock'n' roll.

Output Housekeeping: DC protection examined

The primary safeguard at power amplifier outputs is 'DC protection'. But why should this be necessary? Well, reverting to our favourite output stage (Figure 1 in Part 3, H&SR March 86), note that the live output terminal (+) is precariously balanced between the two DC supply rails, courtesy of the output transistors. In a healthy power amplifier, this direct-coupled output scheme is nothing to worry about: the output terminal ideally sits centered at 0v under quiescent (ie. tickover) conditions. That's when there's no music at the output.

Direct coupled outputs are, in all but one respect, an excellent thing. It's part of a plot to reduce all analogue electronics to just transistors and resistors, ie. logic circuits and chips. The aim is much the same: the abandonment of reactive coupling components. They're a royal pain, as it's always difficult and unduly expensive to make really good ones, especially when they're large (necessarily so in a power output role), and it's impractical to fully correct their defects with feedback. Overall, with direct coupled outputs, power amplifiers end up smaller and cheaper (or the same price, but with fewer corners to cut). Given that there's one component's less scope for corner-cutting, the direct coupled amplifier's sound quality can only benefit. So much has already happened: since the original Crown DC100 laboratory amplifier arrived in the UK in 1967. Power amps without direct coupled outputs have become a dying breed, and since the mid '70s, they've all but disappeared.

The one snag with direct coupling is the infamous 'DC fault': if any one component goes AWOL, it's likely that the output will rest at a value other than zero volts. Variations in component tolerances and drift at high temperatures with ageing have a similar outcome, but generally, it's on a smaller scale, called a 'DC offset'. Here, the output rises from 0v, to perhaps +250mV, (or, just as likely, -250mV). Either way, a speaker cone rated to move its full distance with 40v AC applied is being offset just (^0.25/₄₀ x 100%) = 0.6% of the full excursion. Unless we choose to plug cans into the output, this is nothing to 'fry eggs on.' But the consequences of a full scale DC fault (as opposed to a DC Offset, which is assumed to be small) are less than savoury. A common failure mode in Bipolar output devices is for them to short circuit internally, whenever their Safe Operating Area is exceeded. The obvious result is that the output terminal rises to the supply voltage, say ±60 volts, and the drive units soon get declared clinically dead. Even if the amplifier's V-I limiting keeps the devices alive and inside the Safe Operating Area, there's a classic Bi-polar drive-circuit topology which guarantees the same, full voltage fault if just one of the supply rail fuses blows, falls out or is casually removed 'for inspection', when the amplifier is switched on: the output unbalances - it 'goes DC'. The same is not true of well designed MOSFET amplifiers. Even with one supply rail fully disconnected, the output remains close to the origin. Nevertheless, the potential and worst case failure mode remains: a shorted output device is enough to melt-down the biggest, most macho monitors in a few seconds. Before examining DC protection circuits which sense and forestall the bloodshed, let's return to AC output coupling, just to put any difficulties into perspective. In the '60s, the original transistor power amps operated from a single supply rail of about +50 volts, and a large electrolytic stood inline with the output terminals. For a modern amplifier with symmetrical + and - rails, the probability of this one blocking DC and not exploding is a disappointing ⁵⁰/₅₀. So, let's use non-polarised electrolytics. But that doubles the size of the output capacitor, already embarrassingly large.

In the CSC patent, the designers have gotten around this laterally, by bravely turning the 20 years of established amplifier topology inside out, so that the existing reservoir capacitors are, at one and the same time, AC output coupling and the DC protection (Figure 2). This is good in so far as the electrolytics are biased with a relatively steady DC voltage, and elegant, because the protection is achieved without addition to the overall parts count.

Workforce 1: Relay Advances

In the early DVT amplifiers, we stepped aside from the worst of relay hang-ups, by carefully defining the protection circuit's sensitivity to DC and asymmetric AC. For example, if the DC sensing is oversensitive, there'll be nuisance tripping, in response to small and otherwise acceptable DC offset voltages, say ±100mV. The presence of asymmetric signals emphasises this effect. At the other extreme, too high a trigger voltage means that for a rising DC fault, a high turn-on threshold equates to a time delay, let alone the added stress on the speaker and relay, as switching is accomplished at altogether higher voltages, currents and powers.

In the end, we set the DC protection's trigger threshold at a comfortable 2½ volts, just 5% of a 250W bass driver's excursion.

More importantly, the relay latches on whenever it's triggered, so it simply can't have a relapse. This means the DVT has to be powered down, to reset, if 'thumped' hard by a console at switch-on. But thumps hard enough to trigger the DVT's DC protection are the result of stray DC voltages or a careless and noisy powering-up sequence. Both are themselves signs that something's amiss. It follows that the DVT's protection makes people stop and think: 'why is this happening?' An occasional spot of sound system investigation pays big dividends in the long run.

Also, the DVT's latching arrangement speeds the opening of the relay: instead of the contacts arcing and bouncing about for 30 to 100mS, the break is fast and incisive enough for the contacts to stay clean and not weld-up, despite moderate current ratings.

Workforce 2: A Shunt Protector.

For the DVT250s, we've taken DC protection a step further, into realms where the dominant failure mode of the protective device leaps in to defend the hallowed equipment which must be protected at all costs: your speakers. It's outrageous that power amps have raped and pillaged speakers for years with impunity. But let's step aside from the temptation to 'cure all' with Americanesque 'brute force' technology.

It all began with lateral thinking; because the dominant failure mode of a relay or transistor is to weld solid when assaulted by high currents, we can make this ominous facet work to our benefit by placing the protective device across the speakers: in other words, a 'shunt protector.' (By contrast, the relay in Figure 3 is a 'series protector.') One encouraging benefit is that audiophiles can no longer moan about the oxidised contacts or extra silicon junctions, in line with the signal. There are certainly no worries about hidden or accumulative degradation in sound quality or detail.

So far, the shunt protection scheme has one snag; it shorts the amplifier's output terminals. Every manufacturer is happy to 'prove' that their beast is short circuit proof, by dropping a 6" nail across the back. But the outcome is often very much more spectacular. The short circuit isn't just momentary, but goes on and on, and when the drive signal persists in overdriving the amp regardless, and the guts were already rather hot, it's called the 'Re-heat mode'.

The obvious answer to this pyrotechnics puzzle is to arrange for the amplifier to smartly switch itself off, once the shunt-protector has operated. In this way, there's no risk of anything overheating. There are, though, two snags. Firstly, when the shunt protector operates, it shorts out the DC across the outputs. So, from the sensing circuit's viewpoint, the DC fault has apparently vanished, leading it to signal the 'All Clear'. Secondly, how can an amplifier, relying on a DC fault voltage and its own DC power to sense and control itself, switch itself off? After a moment's thought, you'll realise that both mechanisms are engaged in defeating their own existence. The outcome is oscillation.

Dr. Anton's patent anti-bootstrap remedy sidesteps the irony neatly: in the DVT250s, the shunt protector latches-on, yet monitors its own output terminals. At the same time, it commands the mains contactor to turn off, and stay off, until someone arrives and switches it off manually. Whereupon it can be reset (if the problem was just an anti-social thump). All the time, it causes a red LED to flash 'Shutdown! Shutdown!', and keep on doing this, long after the amplifier's own supply voltages are down to nought. This same protective instinct comes into play if the cooling fan dies. With the bitter experience of Sod's law on stage, this is important, because the messiest and surest way to completely bugger a Rock 'n' Roll amp is to run it into re-heat. 250°C just isn't fair, to any electronics, however sophisticated.

Further Reading
Manuel Huber: Important Aspects of Power Amplifiers (Studio Sound, November 85)
Electronic Reliability Data (IEE) 1981.