Magazine Archive

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

Analogue Equipment Design (Part 3)

Article from Home & Studio Recording, March 1986

Ben Duncan waxes lyrical in his own immortal style.

In this pant, Ben Duncan compares the hidden aspects of transistor power amplifiers, discusses some restrictions in the bottom end, and examines just how the Rauch DVT design addresses the problems encountered.

In the beginning, bi-polar transistor power amps were fragile beasts. Even ten years after the advent of rugged, Silicon bi-polar power devices, burnouts and blowouts were a frequent occurrence, especially if you sought a real 'Slam', or coupled the 'wrong' amp to your favourite bassbin.

Manufacturers of the original, bi-polar power amplifier species have taken one or two steps to remove the most obvious and dramatic bad aspects. In other words, to boost reliability up from ±50% to around 85%. Cheapest of all, an over-enthusiastic protection circuit can shut down a nominal 500W amplifier, to just 75W when it sees a 'difficult' speaker load. This means the maker can scrape by with just two cheap power devices. Meanwhile, the unadvertised power droop into a real speaker load isn't audible so much as an obvious drop in volume, but more as a rough, bumpy, soggy or choked bass sound. This might be enough to make you want to change your bass speakers...

Reputable makers of amplifiers can be identified by their more complex, yet more accurately tuned protectionless circuits. This makes inadvertent, surreptitious power-limiting less likely to occur on most 'normal' programmes, with most speakers. Plus multiple transistors, capable of coping with larger (if not wholly adequate) peak currents. But even without the trade-off of wideband devices (fast output transistors improve bi-polar top-end clarity, but are delicate), the sheer number of bi-polar devices needed to attain a truly unfettered output current (ie. bass sound) into real speakers under real op conditions, is daunting. To do the job properly immediately sets the price at £1500+. Could there be a cheaper way?

V-I Limiting Unveiled

In Part 2, we dipped into serious, pro-power amp power supply design from the angle of energy storage point of view. Alas, in the majority of amps with bi-polar power transistors, power and energy sourcing qualities are routinely tossed out of the window, thanks to V and I (or V-I) limiting. This is shorthand for 'Voltage and Current limiting'. In a nutshell, these limiting functions are the work of the infamous output protection circuitry, needed to restrain the spectacular, kamikaze instincts of bi-polar transistors, whenever they go to bed with a funky, reactive load. Like connecting a real loudspeaker across the outputs...

The DVT series power amps use MOSFET output transistors, the good news being that these aren't subject to the same toppling instincts as bi-polars. Given the good margin in the Rauch's POWERFETS, this means that the V-I limiter can be completely dispensed with. But for us to understand fully how this benefits the sound coming out of the speakers, and reliability, we'll need to outline the reason for V-I constraints.

Figure 1. A Power Amplifier output stage: The bare bones

Figure 1 displays the bare bones of a power amp output stage: TR1 and 2 are bi-polar output devices (several may be wired in parallel, in real life), and C1, C2 are the supply reservoir capacitors. Note also the archetypal, and highly commendable split-rail supply: load current is sourced symmetrically from either rail, via either half (TR1 or TR2) or the speaker, into the centre (0v) line, returning via C1/2 and the power supply's transformer windings.

All too often, speaker manufacturers like us to believe that their drive units are neat 8 or 4 ohm resistors, but as displayed in Figure 2, once loaded up into their enclosure, and when driven over their full range of frequencies, the area of loading which looks purely resistive is limited to the driver's mid-band, at best. That's around 500Hz for a 12" bass/mid driver, or around 5kHz for a tweeter. Elsewhere, the speaker looks progressively more like a large capacitor (Figure 2, curve-region A) or inductive, as in curve-region B. Other than the resonant condition (that's the spiky LF peak in the impedance curve in Figure 2), any driver's impedance is generally capacitative at the bottom-end, and Inductive at the HF end. To ram this point home, we've drawn in a capacitor (C) and an inductor (L) as the load (L) impedance, seen by the amp's output devices, in Figure 1. The next question is what happens to the distribution of volts, amps, and power with these mischievous reactances clapped across the output terminals? Under quiescent conditions (the amp on, but nothing is playing through it), the voltage VCE across the output transistors' collector (c) and emitter (e) terminals is at the rail voltage, namely +60volts. Therefore voltage across the speakers is nil. Moreover, the current draw is nil in either the speaker, or the devices. Now imagine that we unhitch the mute on the console, and hit the amp with a raw kickdrum. With the LF content very likely centring on the 65Hz zone, where the speaker's load looks capacitative (see Figure 2), the power transistor is asked to dissipate an unfair share of the power. Assuming the signal is positive going, TR1 is turned on, while TR2 stays off. So we can concentrate on TR1 dumping joules from C1, into the speaker. Under the ideal conditions of resistive loading, like an 8 ft test resistor, the voltage (VCE) across TR1 falls off, as the current increases, because the load takes it's share of the 60v rail. But for our capacitative bass-end load, the current and voltage got up to 90° out-of-phase. This is tantamount to saying that at one point during the 65Hz cycle, high currents and voltages occur simultaneously. At this point, the leading edge of the wave front, the voltage across CL is nil. That's because a capacitor takes a finite time to charge up. So the instantaneous voltage across TR1 is close to 60v. At the same time, the speaker is sounding a loud note (we hope), so a large current must be flowing. Because TR1 and the speaker are in series, the same current must be flowing in each.

Figure 2. Impedance curve for a Sentry 100A studio monitor

Plugging in some ballpark figures, an 8Ω resistor draws, say, 10 Amps peak, with about 5v across TR1 at full output. This spells a dissipation in TR1 of (5 x 10) = 50 watts. On the other hand, the equivalent capacitive load, the instantaneous voltage across TR1 can be as much as 55 volts. Multiplied by the same current, this equals (55 x 10) = 550 watts, ie. a dead device, and/or a burnt finger! Under repeated kick-drum hits, and even granted that not all capacitative loads present the worst-case, 90° phase angle, it's no surprise that bi-polar transistors need elaborate output shackles, to prevent currents and voltages getting dangerously out of-hand.

For inductive loads, the opposite phase angle applies, ie. -90° under worst case conditions. When 55 volts is slammed across a speaker's voice-coil, the current rises sluggishly. So we can conjecture, say, the voltage across TR1 dropping immediately from 60v to 5v (the other 55 is now across the drive unit), with the current rising slowly to 5 Amps. This represents an instantaneous dissipation of just 25 watts, which is both less than a perfect resistive load, and hardly a problem. However, the stored magnetic energy in the driver's coil sets the scene for Back-EMFs, ie. a whopping counter-voltage. This is particularly true when long wires between amp and speakers add enough extra capacitance and inductance to set up a resonant circuit. In practice, the counter voltage is well damped, and so rarely more than 20 to 30% higher than the supply rail, but as we'll see later, bi-polar transistors are much less tolerant of small over voltages, especially when they're well heated.

Big power transistors

As inferred earlier, the Rauch DVT series are the first rugged Rock 'n' Roll-proof amplifiers without V-I limiting. In part, this is achieved by using MOSFET output transistors. Figure 3 displays the Safe Operating Area (SOA) of a good, rugged bi-polar, called MJ15003 versus the Audio-grade POWERFETs (POWERFET is shorthand for MOSFET power transistor) used in most UK and Japanese MOSFET amps, including the Rauch 250s and 50s.

Figure 3. SOA Curves
Bold lines show SOA of POWERFET transistor for DC, and low frequency switching (10 and 100mS).
Dashed line shows 100mS safe operating for a giant bi-polar transistor.
Under each curve is the Safe Operating Area (SOA); above the line, the device will self-destruct.

Looking closely at the graph, you'll see the scales describe the output (drain) current, through the output device, into the speaker. Meanwhile, the 'drain-source' voltage stretched across the bottom represents the voltage across the output device. In a POWERFET, the drain and source correspond to the bipolar's collector and emitter respectively, so VDS is equivalent to VCE. The curves, themselves describe the bounds of V-I protection, and show how the output device's maximum safe power varies as the current to voltage ratio changes. In particular, whereas a single MOSFET's current draw is restricted to 8 Amps, or less, even at the lowest voltages, its 100mS pulse curve (indicative of low bass performance) is ahead of the giant, strutting bi-polar >40 volts. This is precisely the point above which bi-polar devices become embarrassing. For example, with 80 volts applied, a single POWERFET device can just squeeze 1.7 Amps; about half as much.

If these figures weren't so outright tiny, the POWERFET's gain could get interesting. But in the DVT250s, we've paralleled up four devices. This is common procedure, but for a rugged bipolar amp, it's usually necessary to go for six or even eight devices for each half of the output stage. Multiplied four times (for the whole amplifier), this can soon get messy. The prehistoric T03 transistor package (invented in 1960) is time consuming to mount up, and doubly frustrating to test or replace. From this angle, just four easily mounted Flat-pak POWERFETs provide a welcoming 14½ Amps at ±78 volts, whereas 4 x MJ15003 can cope with just 6½ Amps.

But the real drama lies in the instantaneous power sourcing capability into a reactive load, which involves a square term. Thus the DVT can deliver a 1600 watt impulse (per channel), contrast the MJ15003s combined effort just 325 watts. Now let's return to the current ratings at low voltages, say VDS = 5 volts. Four MJ15003s can dissipate 600 watts, whereas four POWERFETs can manage only 160 watts at 5 volts. This power is available into a rather strange impedance (Z optimum) where Z opt = Vmax/Imax, ie. Z opt = 156mΩ (0.15Ω). This is to say that bi-polar amps develop their most unrestricted power into shorted speaker leads (this is the nearest to 0.15Ω that any speaker will reach!), and also, inductive loads. In turn this explains why impressive tests of amp ruggedness, like welding 6" nails on the end of the speaker lead (= approx 150 mΩ!), or driving 10HP electric-motors are all meaningless: the real test for audible power is an output stage which can safely and reliably source Amps at high voltages, of above about 40 volts. Although not perfect, the POWERFETs' combined SOA curve does at least develop the most unrestricted power into 5 or 6Ω, which is much closer to practical speaker impedances.

The Cryogenic Arts

To recap, all bi-polar devices exhibit a limited SOA, and can easily explode when driving real speakers, so protection circuitry is necessary. POWERFETs have around double the SOA, and yet the DVT Amps prove that they can operate unprotected, and with a tenfold increase in reliability, all without any V-I limiting. Having just doubled one parameter (the SOA) this gain in performance doesn't add up. So you'll be reassured to know that there's no cheating involved, just some behind the scenes work on another crucial parameter, called temperature. This is where POWERFETs and bi-polars display radically different behaviour.

Bi-polar transistors have a positive temperature coefficient. Put simply, this means positive feedback: as the transistor guts heat up, resistance drops, and more current is drawn. In turn, this hikes up the temperature still further, and at the same time, the SOA falls off. This goes on until either the V-I limiter steps in and curtails the drive level, and the sound, or the thermal breaker switches off the juice, or the amp starts to melt. One outcome of the disastrous erosion of Safe Operating Area margins at high temperatures is that truly reliable bi-polar amplifiers need really big heatsinks.

POWERFETs have a negative temperature coefficient. Like negative feedback in any case system (whether it's a conveyor belt servo-motor, or an electronic circuit), this always works to create a stable, balanced condition: as the FET heats up, it's resistance rises, so the current draw falls off, and the device cools. This in turn allows the current to increase again (if required). This behaviour has two important ramifications. First, if the heatsink is man enough, the FETs are wholly self-protecting, because they can never exceed their rated maximum temperature, and their resistance is always enough at the highest sustained device temperature, to shut the highest current down to a safe level.

Second, the 'Cryogenic' heat-exchanger in the DVT series amplifiers sets up a thermostatic action: the range of real-life operating temperatures is limited, with thermal negative feedback keeping the devices cycling up and down between 50°c and 70°C, rather like an oven thermostat, set at 60°C. Contrast this to a bi-polar amp, which shoots up to 95°C when thrashed, and stays there. The DVTs' well defined and moderate temperature swing has a big influence on long term reliability, remembering that every 10°C rise in operating temperature halves the longevity of any electronic component. Small wonder the average bi-polar amp expires after one and a half to two years (road use), or 7 to 10 years (gentle studio use)! Indeed, these figures are exactly mirrored by the average 20°C in operating temperatures, meaning a 4-fold difference in lifespan.

At first sight, the POWERFET's increasing resistance at high temperatures inevitably affects circuit performance. For example, if the FET's resistance reaches 4Ω at high temperature, this would appear to rather limit performance when driving a 4Ω speaker, with half the power being lost in the POWERFET. So far, however, we've neglected the effects of feedback. The DVT's driver module is a power op-amp, with a 20MHz gain-bandwidth product. Dividing 20MHz by 100Hz implies a feedback ratio of 200,000 at bass frequencies. This is to say that the intrinsic open-loop output resistance is divided down by 200,000, with the feedback in place, as in normal operation. So even if the open-loop impedance was as high as 500, the real output impedance is 50 ÷ 200,000Ω = 0.00025Ω, or 0.25 mΩ. Of course, this is a hypothetical figure once wiring resistance is accounted for, but it illustrates why the potentially high resistance condition in hot POWERFETs isn't the end of the world.

The Damping Factor

Particularly in the USA, the damping factor specification of a power amplifier is gospel, thanks to advertising hyperbole. This figure is the ratio of amplifier output impedance (RO), versus the speaker impedance (RL), expressed as a resistive divider, ie. (RO+RL/RO). Plugging in some real life figures, an RO of 30mΩ (taken across the amp's output terminals) and an 8Ω speaker, produces a damping factor of (8 + 0.03/0.03) = 268. This is a moderate figure, typical of pro transistor power-amps. Even higher clamping factors are feasible if we can cut down on wiring loom resistance, and if the driver circuit could cope with a higher level of negative feedback, thereby increasing the feedback ratio. In this respect, rugged bi-polar output transistors are their own worst enemy. Figure 4 shows their limited HF response, relative to a MOSFET device. In turn, this translates into excess phase shift at HF, which acts to limit the feedback ratio, before the bipolar circuit bursts in positive feedback (HF oscillation). Likewise, for tube amps, the damping factor is much lower, because the output transformer further complicates the application of large doses of impedance-reducing feedback. Therefore tube amp damping factors hang out in the 50 to 100 region, contrast 300 to 1000 for well designed transistor amplifier, using MOSFETs, or delicate, high speed bi-polar outputs.

Figure 4. Bi-polar vs POWERFET transistors: relative frequency response

The damping factor is widely supposed to influence the bass-end qualities of an amplifier, in direct relationship to the speaker's impedance. For example, if we plug a 4Ω speaker into the above equation, the damping factor apparently falls to just 134.

Put this point to a speaker designer, and you'll get laughed at: from their angle, provided the source impedance seen by the speaker is less than 1/10th to 1/50th of the speaker's minimum impedance, then there's no way the speaker's performance will be influenced. Even if we assume zero-resistance connecting leads, the concept of the amplifier damping the loudspeaker motor is highly dubious. A leading speaker company like ElectroVoice figure in the amp/lead impedance into the Thiele bass cabinet equation, only when it's in excess of 0.25Ω. So why have a damping factor figure? Well, when negative feedback was first applied to tube amps in the late 40s and early 50s, there were some spectacular, even meaningful gains, as the damping picked itself up from a fractional 1.8 (a 10Ω transformer drives an 8Ω speaker), and became an incredible 50, or 100. This sort of change would certainly tighten up the bass, but beyond this point, damping factor is a red herring, at least when we look at the speaker aspect.

The essential issue is 'interface intermodulation distortion'. This horrific word was coined by Matti Otala, a perceptive audio designer from Finland. Here, we're looking at the fraction of the speaker's input energy being reflected back up the amplifier's back-end. Being delayed and out-of-sync with the main feedback signal, the result is a blurred, distorted after-image. The significance of this effect is that damping factor acts to counter it, but assumes importance across the whole audio band. The emphasis is important, in that it contradicts the assumption that damping factor should be piled-on high below 500Hz (for a tight bass) despite the inevitable trade-off, namely less feedback is available to tidy up the remainder, above 500Hz. Moreover, stressing the feedback ratio at LF achieves nothing useful, because human sensitivity to distortion is abysmally low for low frequencies. Indeed, by sacrificing the mid-brand harmonics, it may make the bass sound worse!

Crucial to the argument so far is the assumption the NFB (Negative Feed Back) keeps impedance down at all times. But it certainly won't do on a fast, feedthrough pulse. The feedback signal can't get around the circuit fast enough to make the correction: even an XR4 x 4 has a finite transit time, and the result is TID, (or if you prefer another mouthful, transient intermodulation distortion).

In common with a number of other designers, Dr. Anton Rauch feels that this malady has been overstressed by certain amplifier manufacturers, doubtless out of a desire to 'prove' better sound, flash up the advertising (TID = 0.00012%' looks great in a brochure), or in seeking a scapegoat for their bad sound, despite some impressive, conventional specifications. The reasoning is that fast feedthrough pulses don't occur in real music programme, except when there's some electronic switching or patching going on at the console (say). So whereas TID is a probable phenomena in odd bursts, every now and again, it certainly can't be viewed as a regular sonic detraction. It's certainly not worth wrecking an otherwise good design, in going overboard for the 'zero TID' medal. That's because granted a healthy Slew Rate, coupled with well defined RF filtration up to 20MHz, Zero TID is guaranteed by design.

Returning to NFB, in any system or amplifier, this inviolably falls off with frequency (it's a physical fact of the universe), and the only discretion the designer can exercise is to decide how far out up the audio band the flat portion of the curve can be pushed, and how steep the inevitable HF roll-off should be (Figure 5). The Rauch DVTs have been designed with this in mind; notice how the feedback ratio/damping factor doesn't fall off to a miserable x 20 or x 30, at 20kHz!

Figure 5. Damping Factors compared

This leaves us with the bass-end controversy, where bi-polar amps are frankly alleged to possess more guts. To cut a long story short, there have certainly been some bad sounding MOSFET amps, but the difference is entirely down to driver topology, power supply current sourcing, and the heat exchanger. Let's focus on this latter point. Often, impressive heatsinks are let down by an outrageous thermal bottleneck, between the inside of the power-device (where the heat is coming from), and the bulk of the heatsink (where it's going to). It's rather like the M25: great stuff, having 4 lanes near the A127, but what about the Dartford tunnel? Dr. Rauch's radical thermal design knocks off just 0.4°C per watt which sounds insignificant, but translates as a 40°C reduction in temperature, per 100 watts of dissipation. This goes a long way to explaining how the compact Rauch heat exchanger manages to shift up to 3kW of heat, without the power devices turning into plasma! More subtly, a little of the unprecedented weight and size to thermal capacity ratio is traded off for lower operating temperatures. This in turn helps to keep the open-loop output impedance low. (In case you hadn't guessed the lateral thinkers behind the DVT series have already modelled the heat-exchanger as if it were an electrical feedback network, built around the amplifier circuitry!)

Leaving aside the sci-fi systems modelling, the heat-exchanger ends up around a third of the size of a conventional 'heatsink' (itself a 1950s invention), so if we keep thinking laterally, it won't take long to realise that this leaves plenty of scope to lose the odd rack unit. What's more, we now have space for a chunkier power supply, which contributes far more to bass quality, than any amount of damping factor. Put another way, there's really no such thing as bad MOSFET sound, just a chronic disease called 'fettered thinking', that results in bad sounding amplifiers and some ripped ears all round. Here lies the sting in the tail: those who have made rude comments about MOSFET bass in the past, have shut up, since the DVTs have made their debut.

Next month we examine power supply design, crosstalk, the stereo image, and output housekeeping.

Series - "Analogue Equipment Design"

Read the next part in this series:

All parts in this series:

Part 1 | Part 2 | Part 3 (Viewing) | Part 4 | Part 5

More from these topics

Browse by Topic:

Design, Development & Manufacture

Electronics / Build

Previous Article in this issue

Falling Into Line

Next article in this issue

Digital Sampler/Delay

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

The current copyright owner/s of this content may differ from the originally published copyright notice.
More details on copyright ownership...


Home & Studio Recording - Mar 1986

Donated & scanned by: Mike Gorman


Design, Development & Manufacture

Electronics / Build


Analogue Equipment Design

Part 1 | Part 2 | Part 3 (Viewing) | Part 4 | Part 5

Feature by Ben Duncan

Previous article in this issue:

> Falling Into Line

Next article in this issue:

> Digital Sampler/Delay

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 2024
Issues donated this month: 0

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

Funds donated this month: £0.00

All donations and support are gratefully appreciated - thank you.

Magazines Needed - Can You Help?

Do you have any of these magazine issues?

> See all issues we need

If so, and you can donate, lend or scan them to help complete our archive, please get in touch via the Contribute page - thanks!

Please Contribute to mu:zines by supplying magazines, scanning or donating funds. Thanks!

Monetary donations go towards site running costs, and the occasional coffee for me if there's anything left over!

Small Print

Terms of usePrivacy