Magazine Archive

Home -> Magazines -> Issues -> Articles in this issue -> View

Analogue Equipment Design (Part 1)

for Rock 'n' Roll

Ben Duncan looks at the design philosophy behind a modern power amplifier. If you don't understand it just be seen reading it and it will improve your credibility.


What are the factors that set good sounding studio and stage equipment apart from the run-of-the-mill? Ben Duncan believes that everyday analogue electronics in the studio and backstage can be markedly improved, to everyone's benefit.

Starting this month, we expand on this theme, by delving in public (for the first time ever) into the intimate circuitry, topology and design detail of a new family of high performance power amplifiers, the DVT series, made by Rauch Precision, in Cambridge.

Introduction



The mother of the DVT series is the 250S. Rated at 290W into 8 ohms stereo, or 1.2kW into 4 ohms, in the mono bridge-mode, it's packed into a deceptively small (3U) enclosure. Working in conjunction with Jerry Mead and Dr. Anton at Rauch, we resolved that the refinements of this new amplifier (following on the heels of the Powerblocks, Rauch's interim model) must apply across the board. We saw this proposal as inevitable, after the work of systems designers in other frontiers of sound technology. As a technology matures, it becomes possible for one or two design paths to increasingly fulfill nearly every diverse requirements. This is to say that the intrinsic gap between the best exposition of monitoring, PA, and even audiophile amplifiers, is shrinking fast, if indeed it still exists. Only the scale of power rating differs.

With this in mind, the DVT 250's spinoffs, like the 50s (½ the power) aren't the usual grudging 'dinky toy' concession to the majority, who can't afford or justify the maker's flagship model; even the diminutive 50s, which is 2U high (3½") retains the same mix of features as the big mother.

Figure 1
(Click image for higher resolution version)


Amplifier Structure



Figure 1 is a simplified block diagram, displaying the key circuit elements in the Rauch DVT series amplifiers. It's also broadly instructive of other serious, modern designs.

Input housekeeping deals with interfacing the incoming signal, namely balanced versus unbalanced operation, DC blocking, RF filtration, polarity selection, gain control, and muting at power up/down. After this, the signal is ready to perform its allotted task. To push a lot of power into a low impedance load, (ie. any loudspeaker), our first requirement is plenty of voltage swing, in fact, 5 to 10 times the usual incoming line level voltage. This is provided by the Driver Stage, which is essentially a high voltage 'line' amplifier. It's built out of discrete components, and configured like an IC op-amp. On it's own, the driver stage can provide an output of about 48V RMS (+36 dBu!) but it's current sourcing capabilities, at around 100mA, are nothing to write home about.

The MOSFET output devices need only have a voltage gain of unity (x1); their energies are concentrated into boosting the output current, up to 35 Amps. So far, we have enough voltage and current to drive real loudspeakers. The output housekeeper is the departure lounge for the signal before it's dispatched from the output terminals. At this point, there's DC sensing to protect the speaker, should the output devices do anything naughty! There's also load compensation so the phase of the feedback signal isn't upset enough (by loudspeakers and their crossover networks) to cause instability or gain peaking.

Back at the driver stage, the negative feedback signal (NFB) from the output is tapped off, at the junction of RF and RG. The percentage across RG is inversely proportional to the amplifier's gain. In other words, for a gain of x20, RF (or RG) is adjusted so just 5% (=1/20th) of the NFB voltage is applied to the driver stage's negative input terminal. Associated with these resistors are two vital parts, CH and CB. These set up the bandwidth of the amplifier, gently forcing the gain down to unity at frequencies well above and below the audio band. These components are amongst several that have a profound influence on amplifier sound; something we'll be looking into in greater detail later in the series.

Subsidiary to the main signal flow, but all part of a high performance professional amplifier are the auxiliary circuits, like fan speed control, thermal sensing, power up/down housekeeping, and most crucial of all, the power supplies. Last but not least, the DVT is one of the few amplifiers with true error-reading LEDs, because they retain their accuracy in detecting the exact onset of amplifier clipping, under all conditions.

Figure 2
(Click image for higher resolution version)


Good Connections and Bad Capacitors



The hassles of interconnection standards, namely the lack of universally adopted standards, afflicts everyone who uses pro audio gear, but only a handful of manufacturers have their act together, in catering for everyone's diverse needs. For a refreshing change, we decided to make DVT's easy to connect to the outside world, so we drew up a 'truth table' of the variables: all pro users will want balanced inputs, but sometimes the unbalanced wiring pattern will best suit existing cables, or installation practices. Many PA companies work with pin 3 hot, whereas studios often operate with the mode IEC Standard, which has pin 2 hot. This leads to the unique input switching arrangements, shown in Figure 2. SW1 selects balanced or unbalanced input connections at the XLRs, ie. the inverting (-) input is either driven from pin 3, or grounded via pin 1 respectively. As it stands, the DVT is wired pin 2 hot. This is to say that a positive going voltage applied to pin 2 will result in a positive going wave form at the DVT's output. If you're only using a single power amp, be it mono or stereo, the polarity doesn't matter, but for Bi- or Tri-amped monitors, or installations and PA systems, where many amps are driven in Tandem, mixing pin 2 hot with pin 3 hot equipment leds to spurious phase cancellation, particularly in adjacent speakers or crossover points. The remedy is to standardise. It doesn't matter which standard you choose within your own system but inevitably, if the existing set-up operates as pin 3 hot, any subsequent equipment is best converted. In short, a majority verdict is the pragmatic path. Because polarity is something you'd want to set and then forget, we made it accessible as a pair of jumpers under the cover-plate. Preset for pin 2, these are simply swapped over for pin 3 hot systems.

In recent years there has been a growing awareness of capacitative abberations. The capacitors used to block out DC in every amp on the market are demonstrably imperfect, and their errors can be measured and are audible. The degree of colouration, and loss of resolution depends both on how bad the capacitors are, and how carelessly they're applied in the circuit.

Using servo circuitry, an amplifier can be made without any coupling capacitors or blocking; the servo can adjust to cancel out the relatively small DC voltages anticipated in ideal circumstances. For Rock 'n' Roll, servos need to be approached with caution on two counts. Firstly, the extra active components needed to implement the servo, have a failure rate that's intrinsically higher than the capacitors it replaces. Second, Rock 'n' Roll amps are party to various ad hoc initiation ceremonies, such as having 130 volts DC rammed up the input. Neither of these make the servo concept unworkable, but they push up the cost and complexity of effective servo operation. We therefore decided to skip the servo, and concentrate on demonstrating that a good amplifier topology can be made to sound much better when capacitors are used sparingly, and with due regard for their side-effects and pit-falls.

Looking again at Figure 1, the balanced inputs are coupled via back-to back electrolytics. Because of the double-balancing inherent in the arrangement and matching of C2 to C5, dialectic absorption, distortion on asymmetric pulses, and diodic effects are largely cancelled out. In addition, the maximum sustainable DC voltage is equal for either + or - polarity on either input, and at 4 times the individual rating, the input withstands 140 volts DC. This contrasts sharply with the traditional single electrolytic, which exhibits around 5% dialectic absorption (≈5% distortion), and withstands about 3 volts before it blows up or goes short circuit, if the DC voltage happens to be of the wrong polarity, (a 50/50 chance), the only setback to the Rauch technique is that seriesing raises series inductance, particularly because the capacitors are doubly big to counteract the halving of capacitance when seriesed! The way out of this is to complete the circle, literally. C1 is a parallel plastic-film capacitor. Choosing the right value curtails the electrolytics' rising series inductance just above the audio band, and effectively slugs the major resonances that plague amplifiers when ever long leads are hung on the inputs.

Unhealthy Hertz



This leads us nicely into the unpleasant topic of RF pickup. For a start, frequencies just above the audio band - say 22 to 100kHz - are present to some extent as ultrasonic (sound) energy at the mic(s). Ideally, this should be gently filtered away, but in a recent survey of mixers at Gateway Studios, I found that the HF 'EQ' on several of the consoles tested did just the opposite. For example, when the HF control on one well known home recording console (costing £1400, hint) was set at maximum, the gentle and reasonable 8dB of boost at 8kHz went on and on rising with ascending frequency, eventually reaching a peak of +21 dB at 100kHz. None of this was advertised on the panel, presumably leaving everyone aghast when a 'bit of boost at 8K' blows up the monitors.

Even without this sort of misbehaviour, a lot of studio and PA equipment exhibits marginal stability when real cables are hung on the inputs and outputs. Ideally, the difference between domestic Hi-Fi and Pro audio equipment is that the latter should work stably and predictably under a much wider range of conditions, and or Rock 'n' Roll amplification, this must (apparently) include withstanding a wide margin of error on the frequencies going up the inputs. The unwelcome effects of signals above 20kHz are fourfold:

Significant energy above 20kHz contributes nothing useful (we can't hear it), but still consumes headroom. Therefore, the amplifier crunches up sooner than it needs to.

Well before overload, this same energy is responsible for intermodulation products. That's because the scope for error-correction in any audio circuit falls abruptly as we go up and up in frequency, so distortion rises sharply. In turn, this creates a gamut of audible sum and difference frequencies, which louse up the sound as intermodulation distortion (IMD).

High levels of RF energy cause heatsink temperatures to shoot up, and accordingly they're adept at blowing up bi-polar power amplifiers, especially those (otherwise desirable) models with fast output transistors, and an extended bandwidth. The Turner B502 is a classic example of the tradeoff: a sweet and open top-end sound came combined with the occasional unscheduled firework display!

Even assuming a power amplifier can cope with 100s of kHz charging up it's input stage, it's risky to pre-suppose that HF drive units share the same view. Above 20kHz, rising inductance makes it progressively harder to dissipate power in the voice coil, but if you've ever visited a maker or distributor's speaker recone depot, there's plenty of evidence for the contrary viewpoint: that frequencies above 20kHz are doubly destructive. How much this relates to momentary overload and latch-up inside panting, red-faced amplifiers must remain conjectural...

In the Rauch DVT series we sidestepped a large part of the RF/Ultrasonic garbage headache by choosing to work around a MOSFET (rather than Bi-polar) output stage at the outset. With 10 to 100 times the power bandwidth, a carefully designed MOSFET output stage is relatively unlikely to lose control when it comes face-to-face with a few MegaHertz.

Despite this lucky side effect, the fact remains that frequencies above 20kHz are best kept well out of any amplifier. This means filtering the incoming signal before it reaches any components which can resonate, demodulate or become otherwise dazed and confused in the presence of RF.

Over the years, many howlers have been spotted in this regard. For example, a famous, top rank US power amp boasts -12dB per-octave active filtering above 20kHz, yet this is built around the front-end IC's feedback network. As a result, the filtering is next to useless above 200kHz, where the IC runs out of steam and the IC's front end is left wide open to frequencies all and sundry.

Looking at Figure 2, the DVT's input is filtered by R1-R4 and C7-10. Unlike an active filter, this is a passive network, which cannot run out of steam, providing suitable dielectrics are chosen for the capacitors. Then with carefully chosen component values, this 2-step filter attenuates away differential signals above 300kHz, at -12dB per octave, which is double the normal rate. At the same time, the filter becomes progressively disorganised as it approaches the upper end of the audio band, making the initial roll-off gentle enough to be inaudible. This contrasts with digital equipment, where very steep (anti-aliasing) filters which cut in sharply right above 20kHz have a disruptive effect on harmonics around 22-30kHz. Whilst not directly audible, these frequencies shape the leading edge of the percussive envelope.

As for common mode-signals, where RF interference appears in an equal quantity on both sides of the input, the ultimate attenuation rate is -24dB/octave, making the filtering around 1000 times more effective than is usual at radio frequencies. To confirm this, we swept the DVT inputs with a modulated RF signal, all the way up to 400mHz. This is a gruelling test for any sound equipment, but the absence of taxi-cab blah, persistent tweeter burnouts; and hard, fatiguing top-end speaks for itself.

Balanced inputs in the Looking Glass



In H&SR's amplifier reviews we've seen that any conventional, unbalanced output can drive a balanced input to good advantage: potential hum loops are avoided, and provided we use paired cable, hash and hum picked up en-route will vanish without any filtering.

How?

To cut a long story short, any balanced input has two key facets. Firstly, it allows us to make connections without the signal ground source's coming into direct contact with the succeeding signal ground. In this sense, it's rather as if we've lifted the signal grounds (instead of the mains earth connections), in order to break a potential hum loop. Alongside this trick, a balanced input has another vital property: signals appearing in the same phase on the (two) inputs are cancelled out. This is known as common mode rejection (CMR), and at first sight, it appears all rather pointless. That's until we start thinking about the send and return conductors laying side by side. Then an interfering field (be it RFI or hum) will tend to be induced equally into each assuming suitable orientation.

Moreover, if the two conductors could occupy the same solid space, the phase of the induced voltage or current will be identical in each conductor. In other words, the induced, interfering signals will be identical, hence 'common-mode'. The nearest we can get to this is to use a pair of identical wires with a small diameter, and twist them tightly. At high frequencies, wavelengths shorten, so for any given, small dislocation in space, there's a commensurate increase in phase difference. In other words, the twist radius is most influential at RF, and rejection inevitably falls off.

This effect can be overcome to some extent by the starquad configuration (it uses paired pairs to increase the mean coincidence in solid-space of each input's conductor), but failing this, look for a cable with a slight twist. Assuming we've a suitable paired cable to hand, we can now put our balanced input to work: the cable converts interfering fields into common mode dross, and the electronics ignores the lot, seeing only the legitimate differential signal.

Balanced inputs come in two guises. Transformers were first harnessed in the 20s, and still offer lots of scope for confusion. 'Transformerless' balanced input stages allow us to dispense once and for all with power matching, 600 ohm loading and other conceptual garbage that's often confused with the idea of balancing per se. The means to do this, the long-tailed pair, was developed in the late '40s for analogue computation. Today, this elegant vestige from the tube era is still with us, though it's better known today as a differential input, a base feature of any op-amp. But the two input terminals have to be connected in a special, and carefully balanced fashion to develop any useful degree of CMR, relative to the input cable's 'catch'.

Figure 3
(Click image for higher resolution version)


Practical Common Mode Rejection



There are several ways of configuring op-amps for a balanced input, using one, two, three, or even four op-amps. The strengths and foibles of these circuits have been examined at length in other polemics, and what's more, evaluated in a Rock 'n' Roll environment. The upshot is that unless we seek a high gain, or are anticipating very large common-mode signals, the simplest, one op-amp differential-input (Fig 3) is acceptable. It's also quiet, because there are no more than the usual two inputs contributing noise, rather than four, as in the more complex differential topologies.

Looking at Figure 3, practical CMR depends on how tightly the incoming cable is twisted, how accurately the resistors are matched, and on the intrinsic CMR of the op-amp. A good op-amp, the upgraded like used in the DVT has an intrinsic CMR circa 120dB or more (Figure 4). In practice, small mismatches in the input resistors spoil this; for example, even 1% resistors will typically reduce the CMR by around 60dB, to -60dB. Whereas resistor error is the dominant error at low frequencies, it can at least be overcome by trimming, via R7. By this technique CMR can be lifted back towards the 100dB mark. But at RF, the physical limitations intrinsic in paired cables give rise to CMR losses that can't readily be made good. For example, a standard paired cable degrades effective CMR to around 30dB at 1 mHz. To put matters into perspective, it's important to recognise that a balanced system succeeds provided it excels where it really matters; in and just above the audio band. Any shortcomings above 100kHz need to be seen in perspective against another attenuative mechanism which comes into play; the input filtration.

Figure 4
(Click image for higher resolution version)


Frequency Domain Dimensions for the Input Stage Putting the discussion so far into actual figures, the graphs in Figure 4 outline in two dimensions the DVT's input performance in the frequency domain. In other words, the combined effects of frequency response tailoring and CMR, where applicable. Because we're dealing exclusively with quantities below 0dB, (ie. Attenuation) the graph has been turned upside down for convenience. Beginning with legitimate audio signals, whether the DVT is switched to 'balanced' or 'unbalanced', these are applied differentially to the input amplifier. In this instance, only the input's RF filtration is operative, so curve A applies. Note the accelerating attenuation (upwards) as 1 MHz is approached. In order to cope with the wide range of signal magnitudes, the subtleties of the response around 20kHz are inevitably obscured. So we've tabulated the salient details in the centre of the graph.

Provided the DVT is switched to balanced, and paired cable is used to hook up to the inputs, the CMR action will be operative (curve B), plus the RF filtering. These are combined to show the overall response in curve C. The lower half of the curve B shows how the CMR attenuation would look on it's own, barring any help from the filtration. This dodgy circumstance is alas true of too many amplifiers on the market.

To appreciate the extent to which the DVT's behaviour represents and improvement over the best existing designs, curve D shows overall performance of the well known PSA2, a 1980 US design. At 1kHz, the two curves shadow each other, but the DVT's improved HF rejection becomes obvious as curves C and D fan-out above 200kHz.

Last, in curve E, it's the input filtration's turn to be assessed in lone combat with common-mode garbage. Comparing it with the differential filtering (curve D), the ultimate common-mode roll-off is doubly steep. And so now, here is Alice reading the results: 'Common Toad Rejection 1, Garbage nil'.

Next month's part will be a dissection of the driver stage.



Previous Article in this issue

Patrick Moraz

Next article in this issue

On the Level


Home & Studio Recording - Copyright: Music Maker Publications (UK), Future Publishing.

 

Home & Studio Recording - Dec 1985

Donated & scanned by: Mike Gorman

Topic:

Design, Development & Manufacture

Electronics / Build


Series:

Analogue Equipment Design

This is the only part of this series active so far.


Feature by Ben Duncan

Previous article in this issue:

> Patrick Moraz

Next article in this issue:

> On the Level


Help Support The Things You Love

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!

Donations for June 2021
Issues donated this month: 0

New issues that have been donated or scanned for us this month.

Funds donated this month: £36.00

All donations and support are gratefully appreciated - thank you.

If you're enjoying the site, please consider supporting me to help build this archive...

...with a one time Donation, or a recurring Donation of just £2 a month. It really helps - thank you!
muzines_logo_02

Small Print

Terms of usePrivacy