Just when you thought that it was safe to look into the back pages of H&SR, Ben Duncan returns with an additional few paragraphs on analogue design.

In this final part, we look at the principles of the Rauch Precision error detection system, embodied in the DVT50 and DVT250s, and introduce the concept of the absolute boundaries of power amplifier modelling.

The importance of accurately indicating the onset of equipment or amplifier overload (ie. hard clipping) has been stressed in countless equipment reviews. For a start, clipping rips the listeners' ears. Serious musicians don't, as a rule, set out to do this. Besides, there's a penalty clause; if a live sound system is driven recklessly into clipping, you'll very likely be 'hoist by your own petard' (literally, blown up by your own munitions). Make no mistake: a due and precise awareness that amplifiers are pumping out gross distortion saves a lot of cones: both recones and deadcones.

'Industry standard' power amp output metering (if it exists) is rarely by VU or PPM: for power amp manufacturers, both are expensive, ineffective and out of fashion. More commonly, it's replaced by one or two LEDs, the first as a clip indicator, the second simply to indicate the presence of a signal. This is an unbeatable combination when it comes to looking over your shoulder, and being able to tell at a glance that everything's working satisfactorily. The fun starts when one becomes aware that conventional clip indicators under or over-read by 3dB or more, according to the instantaneous levels of output current and the mains voltage. Needless to say, when one is high, the other is low, which only goes to further emphasise the discrepancy. The crux of this is that simple clip indicators of the genre illustrated in Figure 1 just light up when the signal exceeds a certain voltage, which represents x watts into an 8Ω load. It does not say 'I am clipping'; it only says whether or not it's putting out a voltage level greater than the rated power into a particular impedance; in this case, it's 28 volts RMS, which spells around 100 watts (into 8Ω), or 200 watts into 4Ω.

An Input-output Comparison

The first IOC (Input Output Comparator) LEDs appeared on a well known 'industry standard' amplifier, which first arrived in England late in 1979. Figure 2 outlines the principle. Here, the amplifier (any amplifier) is modelled like a magnifying glass. The sole difference between what's at the input, and what's emanating from the output, is that the latter is exactly X times bigger.

Looking at Figure 2, we can see that the amplifier's output is (for the sake of example) exactly 20.0 times bigger than its input. Then if we sample the output, attenuate it by 20 times (÷20), and subtract it from a sample of the incoming signal, the difference should be zero... if all is well. But if the output from IC.1 is a finite voltage, greater than a few millivolts in either direction (+ or -), then something, somewhere is feeling slightly under the weather, always assuming the IOC components are finely tuned. So, when the amplifier first clips, the output doesn't quite manage to reach to 20 times bigger than the input, but instead, manages perhaps 18 times. This fraction, ¹⁸/₂₀, represents 90% of the anticipated output, and the missing 10% is the error, detected by IC.1.

One snag is that our 'divide and magnify' trick assumes that the amplifier and subtractor (IC.1) performance is invariant with frequency. But of course, they're not: both have a finite slew limit, and both exhibit phase and amplitude deviations at high and low frequencies. The upshot is that the basic error detection scheme isn't terribly accurate, except perhaps between 500Hz and 2kHz!

In the Rauch DVT50s and DVT250s amplifiers, we developed this error detection principle into a precision circuit, by applying lots of lateral thinking. For example, the slew limit is not just a hindrance to accurate clip detection; it's also an error in its own right. If we could abolish it, for the sake of the error detection, the amp wouldn't have it in the first place! So instead of trying to stop it happening, we made the slew limit error itself well defined. The outcome is that the IOC acts as an accurate 'exceeding the slew-limit' indicator to boot, though with a 300^v/_μS slew limit on the DVT250s, it's not often needed.

The uses of an accurate error detect mechanism don't stop here.

Firstly, if the amplifier's output goes DC, one of the error LEDs comes hard on, telling you whether the DC fault is + or —.

Secondly, if there's an input, but no output (because someone has stuffed a screwdriver across the terminals), this again represents an error. The equation for this one is: Input = any number, output = nil. The result is a pair of flashing LEDs. Equally, the DVT error LEDs will light up if there's an output signal, but no input. That's input = nil, output = any number of volts. Leaving aside the influence of UFOs, and the supernatural happenings in certain studios, this signifies oscillation, probably the supersonic type you can't hear. We've never managed to force a DVT250s into producing ultrasonic oscillations, but in the sure knowledge that someone, somewhere will have it happen to them someday, it's reassuring to know that LEDs to indicate this condition exist.

Thirdly, the circuit detects RF interference, if it's sufficient to drive the amp into clipping (by eating up the headroom), or else enough to precipitate slewing distortion. In effect, if the RFI is harmful, then the error LEDs will tell us about it.

Fourthly, if the speaker or speakers hung across the output cause a truly horrific phase shift, thanks to a massive capacitative impedance, the error LED threshold is displaced downwards, consistent with the V-I and thus thermal stress is placed on the POWERFETS.

Finally, the error detect mechanism is the DVT's own integral analyser, meaning the amplifier can be diagnosed on the road by just plugging a scope across the error detect output.

It's only Rock 'n' Roll

The table in Figure 3 charts the development of amplifier specifications over 60 years. For the first 50 of these, specs have broadly reflected our increasing knowledge of psychoacoustics: the minimum that's necessary to fool our hearing. But after the amplifiers of the 70s, which comfortably exceeded 'perfection' (on paper, at least), this process lost direction, in so far as theoretical Nirvana proved disappointing. One response is to dispense with numerical specifications altogether. Another copout is to go chasing after digital Nirvana. Dr. Anton Rauch's own synopsis is that amplifier specifications must be developed to represent reality, which has more dimensions than a printed spec sheet. Exactly the same problem faced physicists in the closing years of the 19th century.

The physical universe had been fully 'explained', but there remained a few perplexing questions. About atoms, for example. 25 years later, in 1924, a whole new cosmos had unfolded: new knowledge about the atomic nucleus had just ripped the lid off Pandora's box, engulfing everything hitherto known about the behaviour of physical objects. And to cope with the wealth of perplexing data, some very daring models had to be postulated. For example, in 1925, Max Born proposed probability waves to come to terms with atomic phenomena. This involves modelling the electron clouds around the Uranium atom (for example) in no less than 276 dimensions! (Challenge — deviation - Ed.)

Along these lines, but sticking with just our 3 three everyday dimensions, for easy visualisation, Figure 4 shows Dr. Anton's Cube. The 12 sides represent the bounds of a hypothetical, perfect amplifier. These describe the upper and lower limits of power, distortion and frequency (or time), relative to what music demands of it. Statistically, all kinds of levels of music occupy the central space, but any extreme, experimental or new music pushes out towards the cube's boundaries. Amplifier performance also occupies the domain within the cube, but never fills the cube, because equipment is never perfect, and besides, the limits of the cube are continually evolving. Once we begin to think of amplifier specifications in terms of filling even just a simple 3D space, it's suddenly easier to get a feel for how the imperfect can masquerade as the perfect, if only for a limited period of time, or relative to certain types of music: just look at the 'shapes' of amplifiers A & B!

In the end, if we can learn how to build good power amplifiers, this means we've tackled the biggest and hardest task of all: after that, building good analogue electronics for the remainder of the signal chain, from mic capsule to mix amp, and tape head to line driver, is relatively easy stuff.

In closing, I'd like to thank everyone at The Tube, at Nomis and The Damned (the first band to let rip with Rauch amps), and countless others for providing lots of positive feedback.