multichannel package. It's an inevitable problem faced by all amplifiers and is not unique to Plinius. The trick in home theatre power amplifier design is to make the most intelligent compromise, the one that still gives you the best possible sound quality, with the most possible power per channel, in a package that is still of manageable weight, size, and cost.
      There's a further practical consideration. Even if you were willing to manage the enormous size, weight, and cost of a truly no compromise home theatre amplifier, all that extra heat radiating from the fins (three times as much as a two-channel stereo version of the same circuit) still has to go somewhere. The amplifier would need to breathe all its extra radiated heat out into the wide open space of your whole room. But the practical reality is that many users want to install their power amplifiers in shelving units that hem in the radiated heat somewhat, or worse yet hide them in closets, or worst of all stuff them into enclosed equipment cabinets behind closed doors. Even if a power amplifier is designed for perfection, without compromise for manageability or cost, to adequately radiate out all its heat, if you stuff it into an enclosure then it will soon bake in that enclosed oven, and all that expensive heat radiating capability will be for naught.
      The Plinius engineers also face a special design challenge regarding heat. That's because their premium power amplifiers operate in true class A mode. True class A is thermally the most inefficient type of power amplifier, so it generates the most excess heat that must be dissipated by heat fins. But in return true class A offers the sonic reward of the best sound, which is why Plinius employs it in their premium two-channel power amplifiers. Two principal factors give true class A the best sound.
      First, the amplifier, and the solid state devices therein, are always operating under pretty constant conditions, including optimum temperature. They are always processing a large amount of current and power, even when the overall music is quiet instead of loud, and even when the AC music signal, for each of its individual cycle swings through the zero crossing is small, at or near the zero crossing, instead of at its maximum (positive or negative) amplitude of the moment. This gives them the advantage of what we call stress imperturbability. True class A amplifiers are not significantly perturbed from their optimum amplifying state (and optimum sounding amplifying characteristic) when the music goes toward its maximum amplitudes on each cycle, nor when the overall music level gets loud instead of soft. Amplifier configurations other than true class A are significantly stressed as the audio signal changes from small to large during each cycle, and also changes from quiet to loud in overall average level over time. This stress in other amplifiers shows up as changed amplifying characteristics, which means distortion, often sounding like congestion, squashing, or outright grundgy distortion on music's dynamic peaks. But true class A amplifiers continue to sound their optimum best (clean, pure, transparent, dynamic), regardless of what the audio signal does. Incidentally, the technique of sliding bias class A in power amplifiers (popularized some years ago by Threshold, and now appearing in Krell amplifiers) does not offer this sonic advantage, since the circuit's operating points are constantly changing in response to the audio signal amplitude (indeed, such sliding bias class A designs might add their own peculiar distortions, since they have to change operating points so quickly at the beginning of every loud transient).
      The second sonic advantage of class A comes about because the solid state devices in a push pull circuit are biased so that they never turn off (cross the zero line of the AC audio signal), nor even enter the small signal region near the zero crossing. This eliminates several distortions common to push pull amplifiers that are not class A, including so-called crossover notch distortion. In push pull output stages that are not class A, the solid state output devices are turned off for about half of each cycle. The problem is that solid state devices suffer a turn on time delay when you try to turn them on again for the half cycle where they are supposed to be conducting and amplifying the audio signal. This time delay causes a missing gap in the audio signal every half cycle, seen as a cut out notch in the audio signal at every crossover of the zero crossing line (hence the moniker of crossover notch distortion). This missing gap or notch is of course distortion, which probably sounds edgy and harsh, just as many solid state amplifiers do in fact sound. Class A output stages do not have this distortion at all, since they never turn off their output devices.
      A further problem is that all amplifying devices (both solid state and tube) have a nonlinear region when handling very small signal amplitudes near the zero crossing. A push pull output stage that turns off its output device naturally also forces this output device to traverse this nonlinear (i.e. distorting) small signal region. This distortion sounds especially pernicious because, while the amplifier is distorting, the audio signal is very small, so that the distortion is actually a very high percentage of the audio signal that you are hearing at that instant. Note that conventional distortion measurements don't reveal this true high percentage of distortion, because they compare the distortion level to the maximum amplitude of the audio signal, not to the much smaller signal amplitude at this instant, and also because they temporally average the distortion contributed by the amplifier over full cycles, thereby minimizing recognition of the high burst of distortion momentarily occurring just in the small signal region near the zero crossings of each cycle. But the human ear/brain is very good at hearing and giving full recognition to fleeting transients, be they transients of the music we enjoy or transients of ugly distortion that annoy us. So we hear this ugly sounding distortion at its true high percentage, relative to the very small music signal level occurring at that transient instant. This distortion probably sounds like the artificial glare which plagues most solid state designs that do not employ class A. A well designed class A output stage will bias the output devices so that they stay above this nonlinear small signal region, and thereby do not generate this type of distortion. That's why a class A amplifier can sound more musically natural, without artificial glare (assuming the circuit is well designed in other ways as well).
      All of the above advantages help the premium Plinius two-channel amplifiers, with their true class A design to fulfill the promise of the Plinius sound as discussed previously, that beautiful natural musicality, without the typical sonic vices of many other solid state amplifiers. That's why the Plinius SA-250 weighs in at 156 pounds for merely two channels.
      Why aren't all solid state power amplifiers made in true class A format? Our old friend, heat. True class A, as the least thermally efficient format, generates the most heat. And heat imposes the penalties of size, weight, and cost, required to dissipate all that heat and keep the temperature under stable control. So most power amplifiers instead go for the next best type of design, called class AB, which is more thermally efficient, and thus doesn't require measures as heroic and costly to dissipate heat. Class AB doesn't offer the above sonic advantages of true class A in a pure, unadulterated form. But it does attempt to address the same problems to some degree, by going part of the way toward class A.
      Essentially, class AB still has the aforementioned flaws introduced by turning off output devices (crossover notch distortion caused by turn on delay, plus small signal distortion), but it tries to cover up these distorting flaws, committed during a given short time period by one side of the push pull output stage, via the simple tactic of keeping the other side of the push pull output stage fully turned on and playing signal purely (without distortion) during that same time period. Thus, the pure music being played by the side that is still turned on tends to cover up or mask the distortion errors being committed for the moment by the other side that is just turning on. The two halves of the push pull output stage overlap in their contributions to the overall output signal, during just those times when one half is distorting, so the pure contributing half overwhelms the distorting contributing half. In effect, class AB is a technique that reduces the percentage of distortion during those flawed time periods near the zero crossing, not by reducing the levels of distortion produced by the erring half, but rather by raising the overall amplitude of the signal (from the pure, nondistorting half), so that the percentage of distortion you hear is less. Class A eliminates these distortions, whereas class AB merely masks them.
      To what degree can class AB mask these distortions? Just how effective is class AB in approaching the desirable sonic qualities of class A? That depends on how richly the class AB stage is biased, which is a choice made by the design engineer. In effect, if the amplifier's design uses a high or rich bias, then the pure half of the output stage will stay on longer, and will be playing louder when the other half first turns on (and goes through its small signal region), i.e. when it commits its distortions. Thus the pure half, by being set to play louder during that critical moment for the distorting half, will be more effective at masking the distortions committed by that distorting half, when the bias is set higher in a class AB push pull output stage. Well, why then doesn't the design engineer deliberately set the bias very high, in order to give you the best sound? Our familiar villain, heat. The higher the bias is set, the more heat will be generated. Indeed, if the design engineer sets the bias higher and higher, then each half will stay on longer and longer, until we reach the point at which each half never turns off, which, presto, suddenly gives us a class A amplifier again, with its maximum heat problems.
      Let's briefly sum up the picture. Class A completely eliminates the distorting mistakes that devices make during turn on, because class A simply never turns on these devices while playing music, having never turned them off (or allowed them to go into the nonlinear small signal region). Class AB, on the other hand, does allow these devices to make these distorting mistakes (including crossover notch distortion and small signal nonlinearity), because it does turn the devices off and then on again during music playing. Raising the bias for class AB to a higher (richer) level does not actually reduce the amplitude of these distorting mistakes in any way, but it does effectively mask them, by overwhelming them with progressively greater amplitude of pure signal (from the push pull half that is still turned on), at the instant that the previously turned off half is again turned on and makes its distorting mistakes. But raising the bias for class AB also raises the excess heat generated, which gives rise to heat dissipation and high temperature concerns.
      Thus, the sound quality of a class AB amplifier is directly related to how high the bias is set, which in turn is limited by heat dissipation considerations. The design engineer of a class AB stage must choose the best compromise, which gives the best possible sound, within manageable constraints of heat dissipation, which in turn relates to constraints of manageable package size, weight, and cost, especially for a chassis that must dissipate the heat from six powerful channels, not just one or two channels. Class AB might not have the idealistic cachet that class A does, but in truth it can be an excellent pragmatic choice. There are some excellent sounding solid state amplifiers on the market which operate in class AB. They sound so good because the designer has biased them high, into rich class AB, and of course he has invested in the required heat dissipation package needed to adequately handle the added heat generated by this better sounding higher bias (a heat dissipation package which is still more reasonable [in size, weight, and cost to you] than that which would have been required had he biased the amplifier all the way up to class A). Which brings us back to the Odeon.
      With the above as background, you can now clearly understand the dilemma faced by Plinius engineers. They would dearly love to give you the exact same sonic purity in the six channel Odeon as they do in their premium two-channel class A amplifiers. But our villain heat won't allow it (at least if you want the muscle of anything like the 200-watts per channel that the Odeon gives you for each of its six channels). So the intelligent compromise that the Plinius engineers executed was the following. Start with their best circuit, the beautiful sounding circuitry of their most recent SA-102 100-watt-per-channel stereo class A amplifier. Increase the rails voltage so that this same beautiful sounding circuit could pump out 200 watts of muscle per channel instead of just 100 watts. Then, address the impossible heat problem that would result from running all six channels in class A at 200 watts on one chassis. This they did by reducing the bias, so that the output stage runs in class AB in the multichannel Odeon, instead of class A as in their premium two-channel amplifiers.
      Having made the decision to go with class AB for the Odeon, the Plinius engineers' next question was how high to set the bias. Remember, higher bias would bring the Odeon closer to the sound of the Plinius class A premium stereo amplifiers, but would also generate more of our problematic villain, heat. How high could the bias be set, for better sound, without running into serious heat problems? The chosen design solution was to set the bias higher on two of the six 200-watt modules, and designate these two as the primo channels with the best sound, to be used for the two front main channels and stereo music playback. The remaining four channels would receive lower bias, and would be intended to be used as secondary channels, the center and surrounds, for film soundtrack playback. Thus, this two primo and four secondary channel approach would generate less problematic heat than if all six channels were set to high bias, and would give better sound to the two front main channels than if all six channels were set to the lower bias that does not cause heat problems. This design solution even cleverly included more of the perfectionism we've seen so much of in the Odeon. The two bays chosen for the primo channels are the two center bays, a perfectionist choice because they have the shortest (best sounding) ground paths to the power supply, and also because they are the quickest to warm up to optimum sounding operating temperature.
      At first glance, this designation of primo and secondary channels seems to make sense. You'll want the highest sonic quality when using a home theatre amplifier for playing music recordings, most of which are stereo (two channel). And most film soundtracks do not intrinsically have the best sonic quality, so there's not much to lose by having the four extra channels at a slightly lower sonic quality level from their lower bias. Also, who cares whether those noise effects all around you from films are rendered with the ultimate musical purity?
      But we also had some misgivings about this split in sonic quality. The new high resolution audio formats offer surround sound capability, and they also offer even better sonic quality than stereo CDs, so a music lover wanting to use the Odeon for music playback would suffer by having the surround channels sound inferior. In addition, someone wanting to use the Odeon for triamping a stereo music system would certainly want all six channels to be of high (and equally high) sonic quality. Also, surround music of good sonic quality is offered by some concert videos, where you'd want the musically natural Plinius sound all around you, not just in front. Furthermore, expert Perry Sun says that the mix on many films includes music coming from the center and surround channels, not just the two front main channels, and so again you'd certainly want that musically natural Plinius sound from all channels, even for conventional home theatre film viewing.
      Still, pragmatic limits do sometimes have to be drawn somewhere. And a sonic quality split can yield good results, even in high-end perfectionist products, provided the sonic quality of the secondary channels is still high enough. For example, Audio Aero's new Prestige DVD player is the best sounding DVD player in the world that we have encountered so far, chiefly because it employs their expensive, resolution enhancing STARS upsampling processors. And it goes to the perfectionist extreme of using these premium processors for all six channels (which is why this player is so costly), thereby insuring high quality sound for all six channels. Yet even this perfectionist product draws a pragmatic line, and executes a quality split, by using tube output stages for only the two front main channels, and solid state output stages for the four secondary channels.
      Thus, we were on alert, during our probing evaluation of the many aspects of the Odeon's sonic performance, to always compare the sonic quality of the primo versus secondary channels, under the many varied operating conditions we subjected this amplifier to. The bottom line question here was simple: do the different bias levels on the primo vs. secondary channels make an audible difference in the Odeon's sonic quality, especially on music?
      It turned out that there was a major difference. The two primo channels (with higher bias) sound gorgeous, especially in balanced mode, and are everything you could pragmatically wish for in a home theatre amplifier. They stand tall as proud representatives of the musically natural Plinius sound. But the four secondary channels (with lower bias) sounded distinctly poorer, especially in unbalanced mode. In fact, in unbalanced mode, the four secondary channels with lower bias sounded much like many competing conventional solid state amplifiers, including a phony upper frequency emphasis and hard glare, which not only sound artificial in themselves, but also block, clog, and obscure music's natural subtle inner details. This glare was still audibly present in balanced mode, degrading the music, though it did sound less saliently obnoxious in balanced mode than it did in unbalanced mode.
      Clearly, the well known distortions of class AB near the zero crossing, which give rise to the artificial unpleasantries of solid state sound in conventional competing amplifiers, were not being adequately masked in the Odeon circuit by the lower bias setting (this lower bias setting has the pure half of the push pull circuit doing less work, to cover up the distortion sins of the other half just being started up, for each half cycle of the audio waveform). What we heard here was not the famous Plinius musically natural sound, and it was especially annoying during the first half hour of playtime after turning the amplifier on, before the secondary channels had fully warmed up. Incidentally, the Odeon does have a standby mode, but the amplifier's main circuitry is not engaged therein, so it doesn't warm up then. And we really don't recommend leaving a power amplifier fully on all the time to keep it warm, especially an amplifier with prodigious power capability like this one, since (for example) a glitch earlier in an unattended system chain, amplified by the all the Odeon's muscle, could blow your loudspeakers.
      Given the sonic disparity between the two primo high bias channels and the four secondary low bias channels, we asked Plinius engineering if all six channels could be set to the sonically superior high bias, and, if so, what would then happen to heat problems, vis a vis the heat dissipation constraints of the Odeon package. Raising the bias on the four secondary channels to the same as the two primo channels would give everyone the musical magic we were able to obtain from the Odeon at its best, for all surround applications. It would also allow the Odeon to be used to tri-amp a truly fine stereo music system. But what heat problems would it engender?
      The answer was great news. It turns out that the monster package Plinius gave the Odeon can actually dissipate the extra heat that would be generated by setting all six channels to the sonically

(Continued on page 74)