to a falsely misleading small number). But the human ear/brain is very adept at noticing and being bothered by brief transient sonic errors, especially when their waveform is so different from the audio signal, and especially because the worst error occurs during the brief period when the audio signal is at its momentary quietest, and thus the distortion error stands out above the audio signal during that brief time period.
Thanks to the sonic perniciousness of this class B distortion, virtually no high fidelity amplifiers are pure class B. Instead, most power amplifiers use a hybrid class of operation for their output stage, called class AB, which blends class A and class B, thus giving you some of the benefits of each class. Class A can be blended to varying degree with class B, in each power amplifier design, thereby giving you a varying degree of the properties of class A vs. class B in each such amplifier design. Creating a class AB design with a rich amount of class A in the blend gives more of the better sound of class A, but also generates more heat, as is class A's wont. Conversely, creating a class AB design with only a lean amount of class A in the blend generates less heat, but doesn't sound as good, since the distortion endemic to class B is obviously more prevalent when class B is more predominant in the blend.
Why does a class AB blend sound better with a rich proportion of class A? When there is a rich proportion of class A, the two halves of the output stage overlap significantly in their 'on' time, and thus also overlap significantly in amplitude. This means that, at the moment when one half of the output stage is turning off or turning on, and thus is committing its crossover distortion error, the other half of the output stage is playing pure, clean, undistorted signal, and is playing it quite loudly. Therefore, the brief burst of nasty crossover distortion from one half is masked and hidden by the loud presence of good clean signal from the other half. The richer the proportion of class A in the class AB blend, the louder the good, clean signal half will be playing when the distorting half starts up or turns off, thus the better the class B crossover distortion problem will be masked, and (all other things being equal, which they are not quite) the better the sound will be. But, again, the richer the proportion of class A, the more heat will be generated (in conventional class AB designs), so the bigger and more expensive the heat dissipation chassis package needs to be.
High end monoblock power amplifiers, and some two channel amplifiers, can afford the large chassis space to dissipate a lot of heat, so they can utilize a rich proportion of class A in their class AB blend. But multichannel power amplifiers, and certainly multichannel receivers, cannot dissipate much heat per channel, so they need to restrict themselves to utilizing a much leaner proportion of class A in their class AB blend. The result? Conventional multichannel power amplifiers, and certainly multichannel receivers, simply don't sound that good - not nearly as good as dedicated monoblock separates can.
Enter the Arcam AVR600. Its power amplifier section stands uniquely alone, among multichannel products, in utilizing a very rich proportion of class A operation. In fact, each of the AVR600's 7 channels stays in class A all the way up to an astoundingly high 20 watts or so (into an 8 ohm load, and proportionately even higher into higher impedance loads). To put this into proper perspective, note that, on most program, if you play your system so hugely loud that you're on the verge of clipping all 7 of the AVR600's 120 watt rated channels on peaks, then the average power level you're using will almost surely still be less (usually much less) than 20 watts per channel, so you'll be hearing the AVR600 playing in gloriously clean and pure class A for all of this program that's at or below average level. That's one key scientific reason (and just the first of many scientific reasons) why the Arcam AVR 600 sounds so superior to the competition, and sounds so much like a $210,000 assemblage of 7 dedicated 300 watt high end monoblocks.
Now, conventional power amplifiers operating with such a rich, 20 watt per channel class A portion, would generate a lot of heat, too much heat to be adequately dissipated by a moderate size chassis such as the AVR600's. So how then does the AVR600 manage to pull off this trick? Some clever engineering. The AVR600 employs a dynamically adaptive class A, instead of conventional class A. Conventional class A generates a lot of heat because it is always running at full tilt, even when the audio signal is quiet or zero, so it is always producing a lot of heat all the time. In contrast, the AVR600's adaptive class A runs at a very low (hence cool) quiescent current when the audio signal is quiet (or at zero), and then it dynamically increases its current as the instantaneous signal amplitude increases (thereby tracking the signal and always staying on, up to about 20 watts, thereby staying class A up to about 20 watts). Because it runs cool for a lot of moments in time, and then runs hotter only at those moments when the instantaneous signal level demands it, the AVR600's adaptive class A produces much less average heat over time - and, since the size (and expense) of a chassis heat dissipation package only has to be sufficient to dissipate the long term heat average over time, the AVR600 can get away with a far smaller, less costly chassis heat dissipation package (i.e. 7 channels in a conventional surround receiver), while still delivering to you the wonderful sonic benefits of class A sound for most of your program.
Sonic Benefits from Rich Class A
This amazingly high level of class A operation, in the AVR600, provides you with dual technical benefits. First, you get to hear the clean purity of class A operation for the majority of your program. Second, the crossover distortions of class B turn-on and turn-off occur more rarely, and when they do occur, they are very well masked and hidden by the very loud 20 watts of class A coming from the other half of the output stage at that moment.
Furthermore, this high level of class A operation already explains why the AVR600 can be sonically so dramatically superior and in so many aspects, as we heard in our listening evaluations and reported above, thereby corroborating and justifying our enthusiasm, both for the power amplifier alone and for the receiver as a whole. As noted, the rich class A means that class B crossover distortion energy (aka byproducts) occurs less often, and not at all during virtually all average program loudness levels - and at the only times it does occur, during brief program peaks, it is hidden and masked much better by the very high, clean and pure program signal energy of these same brief program peaks. This explains why, firstly, the AVR600 sounds much cleaner and purer, since you don't hear the garbage sound of distortion energy.
Secondly, this also explains why the AVR600 sounds more transparent, revealing more of the subtle detail in recordings, than competing products. You see, in competing products, frequent and pervasive bursts of this garbage noise energy (from class B crossover distortion) compete with and obscure the subtle detail in recordings - but in the AVR600 there is no such garbage distortion energy, at most program levels and for most of the time, so the subtle details of recordings can emerge into the sunlight and be heard clearly, against a background of awesomely black intertransient silence, instead of a background of competing noisy garbage as found in competing units.
Thirdly, this also explains why the AVR600 has shockingly superior spatial imaging, for surround sound and two channel stereo. The sonic cues that define (and enable our ear/brain to decipher and hear) spatial imaging (e.g. hall wall reflections and reverb) are very subtle details, and the AVR600 reveals these subtle imaging cue details very clearly, against its black silent background free of spurious garbage distortion noise energy - whereas competing units obscure these same imaging cue details, burying and hiding them amidst a background that is noisy with garbage distortion energy, so their spatial imaging is not nearly as superb as that of the AVR600.
Fourthly, this might also explain the AVR600's amazing bass quality and its authoritative control of woofers, even at and below their resonance frequency, and even at loud levels (amazingly loud for a receiver). Class A has some technical advantages in woofer control, compared to class B or conventional class AB amplifiers having a lean portion of class A in their AB blend. A class A output stage can have more current quickly available to drive a woofer, which is especially important in the low bass, where a woofer's current demands are severe, both because of the larger excursions and also because of the load impedance curve's wild gyrations. A class A output stage has twice the number of output devices doing the driving, which not only provides greater current reserves, but also doubles the transconductance and halves the source impedance, thus making control of the loudspeaker more accurate and intimate, and also helping the amplifier to better fulfill its crucial role of sinking those nasty back emf reactive currents that a violently moving woofer sends back to the amplifier.
As a bonus, the AVR600 stays in class A beyond 20 watts when the load impedance is higher than 8 ohms, and a dynamic woofer's load impedance skyrockets around its system resonance frequency, so the AVR600 stays in class A far above 20 watts in the neighborhood of the woofer's resonance frequency, thereby being better able to tightly control the woofer through its resonance region, precisely where the woofer wants to behave loosely, therefore precisely where it most needs extra control (control that the AVR600 so superbly sonically furnishes, sounding far better than other multichannel receivers).
Incidentally, it's worth noting that Arcam's own previous power amplifiers, including the multichannel P7 and the AVR300 and AVR350 receivers, all employed conventional class AB output stages, with a pretty lean conventional proportion of class A in the blend, and all shared a sonic thumbprint of solid state bipolar sound (tamable by using an appropriate power cord design like the Wan Lung). On the other hand, the new Arcam AVR600, with its much richer class A diet, does not have this bipolar solid state coloration, instead sounding eminently natural and neutral (and it does not need a taming power cord to change or ameliorate its sound).
Class A is only part of the story in the AVR600 power amplifier. The AVR600 also employs class G operation. What is class G? It's an operating mode that employs tiered devices in the output stage, operating at different voltage levels, and is thereby much more efficient, producing much less heat on typical dynamic program. This saves on heat generation, for a given power output capability, and thus gives the design engineer an extra heat budget that he can choose to spend in various other ways, for a given chassis package having a given heat dissipation capability.
How does this class G work, and why is it more efficient? In a conventional power amplifier, say the ubiquitous class AB type, there is only one voltage rails for the power output stage. This rails voltage represents the absolute maximum that the power amplifier can output (and, incidentally, some internal circuit losses reduce this a bit). The output devices in the power output stage act merely as gates or valves, letting out some portion of this rails voltage to your loudspeakers, this portion depending on the varying instantaneous signal level (which in turn depends on the varying program and of course on your volume control setting). So some portion of the rails voltage is let out by the output stage gate/valve to your loudspeakers - but what happens to the rest of the rails voltage that is not let out? Simply speaking, that remainder must be dissipated internally by the amplifier's output stage, dissipated as heat.
Now, with typical dynamic program (music, film soundtrack, etc.), the instantaneous program signal level is far below its maximum peak output for most of the time, and thus is also far below the power amplifier's maximum peak voltage output for most of the time. This means that, for the majority of the program's varying amplitudes, and for the majority of the time, only a tiny fraction of the rails voltage is let through the output stage gate to your loudspeakers, and the remaining large fraction of the high rails voltage causes pressure on the output stage gate, forcing this gate to expend energy (and dissipate the resulting heat) holding back the high voltage/pressure from these high voltage rails in this output stage.
To understand this better intuitively, consider as an analogy that exit tube at the base of Hoover dam, which can shoot a stream of water for hundreds of feet. Consider the gate valve that closes or opens this exit tube to varying degree. There's tremendous water pressure on this gate valve, precisely because the Hoover dam is so high, just like the rails voltage being high on a power output stage. Now imagine that you had to do the work of holding that gate valve partially open, to varying degrees, say by holding your 'very large' hand palm over part of the exit tube, thus holding back some portion of that tremendous water pressure.
If you held your large hand over most of that exit tube, letting out only a small stream of water (like letting out only a small signal to the loudspeaker), you'd still have to fight to hold back most of that tremendous pressure in the exit tube, coming from the high dam above (like the high voltage rails) - and fighting to hold back water against this pressure would be hard work, causing you to burn calories and sweat (to dissipate the excessive heat generated in your body by this hard work). Conversely, if you moved your hand palm off to the side, so as to let most of the water flow freely out of the exit tube (like letting a large signal out to your loudspeakers), you'd then scarcely feel any pressure on your palm at all, since all the pent up pressure from the dam above would be going into pushing huge amounts of water out the exit tube (driving your loudspeakers), and you'd scarcely have to do any work, so you wouldn't get hot nor have to break a sweat to dissipate heat.
This example is analogous to the typical power output stage operating with the typical single, high voltage rails. When the signal amplitude output to your loudspeakers is small, compared with the amplifier's maximum output amplitude capability, then the output stage gate valve has to do a lot of work, to 'hold back' that tremendous pressure/voltage from the single high voltage rails, to keep it from being output to your loudspeaker, and all this work generates heat that then must be dissipated. It probably seems counterintuitive to think that a power amplifier would have to work hard, in order to merely push out a small signal into your loudspeakers. But this dam analogy helps us to view the situation differently, as instead pertaining to holding back most of the high pressure/voltage from a high dam or a high rails voltage, and to the work that must be done, and consequent heat that is wastefully generated, even when only a small signal is output.
Everyone wants power amplifiers to have lots of rated power output capability (indeed, this numbers game is overvalued as a selling point). But, to get a higher maximum rated power output capability, one needs to raise the rails voltage, and that increases the pressure on the output stage gate valve by making the dam higher, which in turn makes the output stage work harder to hold back this higher voltage/pressure and keep it from getting out to your loudspeakers, and this harder work creates more wasteful heat to be dissipated, which then forces the chassis heat dissipation package to be made larger and thus more expensive.
In any given amplifier chassis package, there are limits on how much heat can be dissipated (long term), hence limits on how much heat can be generated by the circuit, therefore limits on how high a rails voltage can be allowed. When making an expensive monoblock power amplifier, the design engineer can always make the single channel chassis a little bigger and a little more expensive, in order to be able to raise the rails voltage and thus claim a higher spec for that vaunted maximum power output capability. But in a multichannel power amplifier, and certainly within the modest size constraints of a multichannel receiver, the chassis package's modest heat dissipation capability cannot be substantially increased, so there is a severe limit on the allowable maximum for rails voltage - at least with the conventional configuration, where the output stage works off a single high rails voltage.
Enter the class G output stage configuration. The basic concept of class G is quite simple. Class G simply has more than one voltage rail, and the plural rails are set at different voltages. The output stage uses only the lowest voltage rails when outputting small signals to your loudspeakers, and then changes to use higher voltage rails only when the instantaneous signal level rises enough to warrant this change. To return to our analogy, that's like having two dams as water sources. When you only need to output a small stream of water, you use a farmer's shallow pond with a small dam, so it's very little work for you to hold back and control the small water pressure (like small rails voltage) coming from a hole at this dam's bottom, and you don't work up much of a sweat doing this easy work (you don't generate a lot of heat that needs to be dissipated). Then, only once in a while, you need to output a big burst of water, but only for a brief time. So you quickly switch to using Hoover dam as your water source, and you can then output that large but brief peak burst of water.
Admittedly, during your use of Hoover dam, you are working very hard holding back and controlling the much higher pressure/voltage, so you are temporarily generating heat at a very high rate. But, and this is the crucial but, if the need for these large bursts only occurs a small fraction of overall time, and if each large burst is brief in duration, then your overall average work output over the moderate to long term will be low, and thus generation of heat that needs to be dissipated will be low. In point of fact, virtually all program we listen to, via audio power amplifiers, does have this blessed characteristic, of having an average level much lower than the peak level, and of having the peaks occur only occasionally, and of having peaks that are each brief in duration. Thus, on average, the work you do or a power output stage does, and the heat you or the power output stage generates and must dissipate, is not much more than it would be if you were using the small pond with the small dam all the time.
When the output stage outputs a signal whose level at that moment is low, then the output stage only has to act as a gate valve for the lowest voltage rails, so it only has to do the work of holding back this lower voltage/pressure, which is far easier work than having to hold back the high voltage/pressure from a high rails voltage, so far less heat is generated.
Again, this heat efficiency advantage for class G depends on the fact that virtually all the program we listen to does indeed have an output level that stays below the lower rails voltage most of the time, hence also has an average output level below this point, with only occasional peaks that are brief in duration (thus, any given single peak does not last long enough, and any series of peaks is not temporally dense enough, to severely impact the chassis' heat dissipation package, which func
(Continued on page 165)