tions over the long term).
Incidentally, class G can be made even more efficient, with even less heat generation, by simply making the voltage lower on the low voltage rails, i.e. a smaller fraction (say one fourth rather than half) of the voltage on the high voltage rails. This works so long as the lower rails voltage is above the average program level - which is easy to accomplish, since the average level of most program material is merely 1/10 the peak level or even less.
Thanks to class G's heat efficiency, the designer suddenly has a substantial heat budget surplus to play with (which was previously needed to support a given conventional class AB power amplifier having single rails voltage, but now with class G no longer needs to be accommodated as a constraint). The designer can make use of this gift, this new heat budget bonus, in many ways. He could make his product package smaller and less expensive. He could increase its rated maximum output power capability. He could accomplish some other sonically beneficial design moves. Or he could engineer a customized mixture or blend of these various benefits.
Wisely Spending the Heat Budget Bonus
When we saw the first announcement of the forthcoming AVR600, we immediately recognized that something unusual was afoot. Although the new AVR600 was to employ class G, its rated output power per channel was to be merely 120 watts, scarcely more than the 100 watts per channel of Arcam's previous class AB surround receivers, and certainly well shy of the perhaps double power rating that theoretically might be available. Indeed, the AVR600 was shown to be having a somewhat larger chassis than Arcam's previous surround receivers, and the increased heat dissipation capability of this larger chassis package could itself provide for the increase from 100 watts to 120 watts per channel, without bothering to even introduce the change to class G. Clearly, then, Arcam, in changing to class G and thereby receiving a large heat budget bonus, had not chosen to expend much if any of this heat budget bonus on the showoff spec of a much higher rated output power capability. But what then did Arcam choose to expend their newfound heat budget bonus on, if not greater power?
Better sound quality.
As a first step, consider the AVR600's rich class A, and all its sonic benefits, as discussed above. The richness of the AVR600's class A, in its class AB blend, was in part made possible by spending some of that heat budget bonus realized by the change to class G. As discussed above, richer class A, even while it provides many important sonic benefits, also extracts a penalty by generating more heat. Some of this extra heat generation was pared away by Arcam's use of a dynamically adaptive class A. But Arcam's employment of class G gave them a further bonus in their heat budget, and they thus could make the class A even richer in the AVR600. In consequence, you hear in the AVR600 the sonic benefits of class A for even more of your program, for even more of the time, and over an even bigger range of signal levels. Better sound quality.
Second, consider output current capability. The spec that gets all the attention is rated maximum output power capability, which depends largely on the rails voltage available for output. But having a high maximum output current capability can be just as sonically important. Just ask the proud owner of a Krell or Electron Kinetics power amplifier, which boast very high current capability, and have the heretofore requisite large chassis (and large price tag) needed to obtain this high current capability.
High current capability gives the sonic benefits of authority and control, even at sustained loud volume levels, which we found to be superb in the AVR600, indeed shockingly so for a mere multichannel receiver. High current capability also provides better sonic accuracy in the voltage waveform from a power amplifier, yielding better perfectionist reproduction of all program, including subtle details, as we found to also be superb in the AVR600.
How does this happen? High current capability usually implies low source impedance, both in the power supply and in the output stage, and low source impedance improves sound, by improving both control and accuracy. Furthermore, the primary job of every power amplifier is to instantaneously source to your loudspeaker (and sink back from your loudspeaker), whatever instantaneous current is required, in order to maintain in its voltage output waveform a perfect replica of the voltage waveform input to this power amplifier. It turns out that these instantaneous current requirements can be shockingly high, especially with reactive loudspeaker loads.
Most power amplifiers cannot supply (or sink) instantaneous currents this high. Not only does this failure make them sound weak, wimpy, and poor at controlling the loudspeaker, but this failure also corrupts the voltage waveform, actually causing distortion that you can hear as such (and this waveform distortion does not show up in the usual sine wave distortion measurements driving easy resistive test loads). If and when the power amplifier cannot quite supply the very large instantaneous burst of current needed to maintain the output voltage waveform as a perfect replica, then the output voltage waveform obviously becomes a not perfect, i.e. distorted, replica during that burst, which you hear as bursts of ugly (inharmonic) distortion, probably brief enough in most instances so they sound like background dirty crud, which also veils and obscures the program, degrading transparency.
Since high current capability makes a power amplifier sound better, more accurate and transparent, plus more authoritative and in control, why do so few power amplifiers provide high current capability? Heat. If there's a demand from the loudspeaker load for the power amplifier to source (or sink) a higher current, and the amplifier correctly complies, then that higher current generates more heat in the output stage that needs to be dissipated, even if the signal voltage itself has not changed to itself cause more heat.
Thus, higher current capability requires a designer to set aside more of his chassis package's heat dissipation capability budget, to providing for high currents. But this takes away from the heat budget he can use for a higher rails voltage. And it's higher rails voltage that sells amplifiers, since a higher rails voltage translates into a higher rated output power spec, and that's what sells power amplifiers. Therefore, most power amplifiers are current starved, usually deliberately (in the power supply, and/or the output stage, and/or by compressive limiting in the name of thermal and current overload protection), to save money and to effectively protect the amplifier against generating the excess additional heat that would be generated if high currents were allowed.
Enter the Arcam AVR600. Thanks to its use of class G, and Arcam's wise decision to not spend all of class G's heat budget bonus on a much higher rails voltage and the flash of a much higher power rating spec, a lot of class G's heat budget bonus in the AVR600 is left over, to spend on other factors that make for better sound. Factors such as higher current capability. Armed with the foreknowledge that the AVR600 had heat budget to spare, the Arcam engineers could then bravely employ premium output transistors that have high current capability and a large SOA (safe operating area, the ability to handle both high current and high voltage at the same time, to do more work controlling your loudspeakers accurately and with authority). This in turn meant that these transistors not only could deliver high current, but also would not have to be protected by prissy, premature current limiting protection circuits like most other amplifiers and certainly all other receivers have (the AVR600 does of course have current protection, but it is set at a significantly higher current level than competing units, with substantially wider margins). And the Arcam engineers could arm the AVR600 with a beefy, high current power supply, knowing that their product could literally take the heat when high instantaneous currents are demanded, for achieving perfectionist waveform fidelity. As a further benefit, power supplies and output devices with high current capability also tend to have low source impedance, which in itself further promotes more perfect sonic and waveform fidelity (hence better transparency, spatial imaging, etc.).
There's an even further benefit, arising from Arcam's decisions on how to spend the heat budget bonus from class G. Some of this bonus was spent on giving the AVR600 a richer class A, reaching to higher (louder) signal amplitudes. In fact, the AVR600's class A reaches so high that it nearly or virtually meets the class G switchover point, from the low voltage rails to the high voltage rails (how close it comes depends on the load impedance, and with load impedances higher than 8 ohms the two can actually meet). Now, during class A operation there are two output transistors sharing the current output duties, so the output stage can put out twice as much current as a single high current transistor is rated for. And, when class G is working off the high voltage rails, there are likewise two output transistors sharing the current output duties (the upper that has just turned on, plus the lower that has always been on), so again the output stage can put out twice as much current as a single high current transistor is rated for, and can work harder to authoritatively control your loudspeakers, since two power devices share the power dissipation.
Because these two amplitude regions virtually meet each other in the AVR600, this means that this benefit obtains for virtually all signal levels, with exactly two transistors sharing the current output duties, thus giving the AVR600 much higher output current capability than its single high current transistors are rated for. Additionally, this also gives the AVR600 higher local thermal capability, and greater SOA (safe operating area), than each high current transistor is rated for. The power devices share the work and share the dissipation, so they can be driven harder and do more work, authoritatively and accurately controlling your loudspeaker. And this transistor doubling might also further reduce output stage source impedance, with attendant improvements in sonic fidelity.
All these factors above combine together, to give the AVR600 its superior sonics. The doubling up of output transistors, combined with a high current design, combined with a heat budget surplus to spare, all together mean that the AVR600 can output large currents and can do harder work with ease, and that's why it sounds like an array of 7 dedicated 300 watt monoblocks directly at each loudspeaker, with the power, authority, and control you would hear from such an array. Even though each AVR600 channel's voltage clipping level is obviously lower than that of a 300 watt monoblock, it sounds just as authoritative as a 300 watt monoblock, because Arcam have gone for a high current design and true sonic authority, rather than for the superficial 'marketing watts' spec that comes from high voltage clipping level but without high current to deliver authority. This high current design also means that the AVR600 can continue to output surprisingly large currents, including with difficult program and/or difficult loudspeaker loads, without running into excess heat dissipation problems, and without having to include or implement premature current limiting protections, as competing units must.
If the AVR600 does perchance generate too much heat for normal silent heat dissipation to fully vent, a fan kicks in to cool thing down. But we drove an array of 7 B&W 802d loudspeakers, known for being a difficult load (with low impedance segments, plus wildly varying impedance, plus very reactive phase angles), and we drove them continuously (for hours on end) to very loud levels (filling a huge 15,000 cubic foot castle ballroom) - and the AVR600's fan never came on even once. Clearly, this AVR600 package has a lot of moxie, and a lot of moxie in reserve.
There's yet another area where Arcam's decision on heat budget spending provides you with better sound. As noted above, one could make class G even more efficient, and generate even less heat, by lowering the voltage of the low voltage rails, say to one quarter of the high rails voltage instead of one half. This would make the switchover, from the low voltage rails to the high voltage rails (from the farmer's shallow pond dam to Hoover dam), occur at a lower signal level. But Arcam elected to spend part of the AVR600's heat budget by not choosing this more efficient, lower heat option. Why spend some the AVR600 package's heat dissipation budget on this less efficient version of class G?
Better sound. You see, class G has a glitch distortion problem, much like class B's crossover distortion problem. Class G is able to turn on the higher voltage rails, when needed for a momentary peak, quite smoothly and gradually, without glitches, and then gradually fade it out when no longer needed for this momentary peak, again without glitches. However, the low voltage rails needs to be switched off when the high voltage rails is gradually turned on, and then the low voltage rails needs to be switched on again when the high voltage rails is gradually turned off. With the present state of the design art, this on and off switching, of just the low voltage rails, cannot, alas, be accomplished gradually and without switching glitches. Consequently, with class G, a small glitch appears in the output signal waveform, only on high peaks that trigger the need to use the high voltage rails, and there are two glitches, one on the leading edge of the peak and one on the trailing edge.
In spending some of the AVR600 package's heat budget by choosing a less heat efficient version of class G, i.e. by choosing a relatively high voltage for the lower voltage rails, Arcam ameliorated the detrimental sonic effects of these class G glitches. Indeed, Arcam killed two birds with one stone here, to give the AVR600 better sound. Choosing a relatively high voltage for the lower voltage rails, fully half of the voltage of the higher voltage rails, means that the switchover point is set at a much higher signal amplitude level. This in turn has two benefits. First, the switchover needs to occur much less often, so class G's distortion glitches occur much less often. Second, when a switchover and consequent glitch does occur, the amplifier is playing pure, clean signal at a much louder level, and this naturally hides and masks the glitch very effectively, so that it's almost impossible to hear anything amiss.
The proof is in the pudding. We have been very critical of the sound of previous class G amplifiers, especially in their dirty distortion at high frequencies, a sure symptom of this class G glitch distortion being highly audible. And indeed, based on this listening experience with other previous class G amplifiers, we approached the AVR600 with trepidation and negative prejudice, expecting to hear the worst. Instead, we were pleasantly shocked to hear that the AVR600 has some of the cleanest, purest, fastest, most natural, most transparent, most detailed and articulate high frequencies we have ever heard from any power amplifier, with no hint of the class G glitch gremlin. This is proof that Arcam have indeed conquered class G, and that Arcam have indeed spent their heat budget bonus wisely, to give you truly better sound.
Perfectionist Design Details
The AVR600's dramatic sonic superiority we heard, both over its competition, and even over state of the art high end separates, is further explained and justified by many perfectionist technical design details Arcam has engineered into this product. These perfectionist engineering details you would be lucky to find in the most expensive high end power amplifiers, but here they are, designed into a mere surround receiver. These perfectionist technical details take an already great sounding product over the top, extracting the last little bit of which this great product is capable, in sonic aspects such as transparency, subtle inner detail, speed, articulation, clean purity, effortless clarity and authority even when the program gets loud and complex, spatial imaging, etc. These perfectionist engineering details take the AVR600 beyond the cutting edge of the state of the art, so that its sonic performance crates and defines a new standard for the state of the art, just as we found in our critical listening evaluations.
Here are but a few examples of Arcam's thoughtful engineering details that they managed to come up with, over the AVR600's three years of hard development effort. Again, you'd be lucky to find these perfectionist details in expensive high end 300 watt monoblock amplifiers - and that's why the AVR600 sounds like a $210,000 array of 7 such amplifiers.
Arcam spent extra money on the output transistors, not only to get high current ones with a large SOA, but also spending extra to get a model that has its own temperature sensing diode built right into the silicon. This gives a faster, more intimate, and more accurate sensing of the output transistor's instantaneous, constantly varying temperature than the usual cheaper technique of mounting a sensor on the heat sink outside the transistor. This more accurate information in turn allows Arcam to more accurately stabilize each output transistor's operating points, transfer characteristic, quiescent current, etc., even as the program signal dynamically changes and tries to monkey with these parameters, via its changing thermal impact on the output transistors. And this stabilization in turn makes the AVR600 more linear and accurate in reproducing the input signal.
This superior, perfectionist linearity and accuracy puts the perfectionist icing on the AVR600's cake, in many sonic aspects. Clean purity gets better, because better linearity equals less distortion. Transparency gets better, because less distortion means less background garbage, veiling and obscuring the program's subtle details. Revelation of information gets more accurate, because the amplifier's transfer characteristic is maintained more accurately, instant by instant, even as the dynamically changing program tries to monkey with it.
Articulation and individuation get better, as do black background and intertransient silence. That's because nonlinearities smear and disperse program signal energy over time, especially when they are thermally induced or thermally correlated (since thermal time constants are long); this smearing makes the original program transients fuzzy and inarticulate, and then smears this energy to a later point in time, where it gets dumped as garbage that fills in what should be the black of intertransient silence. In contrast, the AVR600's superior linearity, instant by instant, plays the full transient accurately and articulately when and as it occurred, and there's no energy smeared to and dumped at a later point in time, so intertransient silence is better and the background is blacker (which in turn further helps transparent revelation of subtle details).
Dynamics get better, because the full impact of each transient attack is played exactly when and how it occurred, instead of having some of its energy being shorn away and smeared to a later point in time, as happens with nonlinearities. Spatial imaging gets better, because spatial imaging cues are very subtly encoded in a recording, so it takes absolutely maximum transparency to fully reveal them, and thereby re-create a much richer, more believable spatial image (both for surround and also for simple stereo).
(Continued on page 166)