Olimpico

Reviewer: Srajan Ebaen
Financial interests: click here
Sources: Retina 5K 27" iMac (4GHz quad-core with Turbo boost, 32GB RAM, 3TB FusionDrive, OSX Yosemite. iTunes 14.4), PureMusic 3.02, Audirvana 3, Qobuz Hifi, Tidal Hifi, COS Engineering D1, Denafrips Terminator, Soundaware D300Ref as USB bridge/SD transport
Preamplifier: Vinnie Rossi L2 Signature with WE 300B/VT52 or Elrog ER50/300B; Nagra Classic, Wyred4Sound STP-SE Stage II, COS Engineering D1, Vinnie Rossi LIO (AVT module)
Power & integrated amplifiers: Pass Labs XA30.8; FirstWatt SIT1 monos, SIT3. F5, F6, F7; Goldmund/Job 225; Wyred4Sound mINT; Nord Acoustics NC500 monos; LinnenberG Audio Liszt monos, Bakoon AMP-13R, SAG AIO
Loudspeakers: Audio Physic Codex; Cube Audio Nenuphar; Kroma Audio Mimí; EnigmAcoustics M1; Albedo Audio Aptica; EnigmAcoustics Mythology 1; Boenicke Audio W5se; Zu Audio Druid V, VI & Submission; German Physiks HRS-120; Eversound Essence; Fram Midi 150 & 120; Acelec Model One; sound|kaos Vox 3f [on loan] Cables: Complete loom of Allnic Audio ZL 3000; Zu Event; KingRex uArt double-header USB cables; Tombo Trøn S/PDIF; van den Hul AES/EBU; AudioQuest Diamond glass-fibre Toslink; Black Cat Cable redlevel Lupo; Ocellia OCC Silver
Power delivery: Vibex Granada/Alhambra on all source components, Vibex One 11R on amps/sub
Equipment rack: Artesania Audio Exoteryc double-wide 3-tier with optional glass shelves, Exoteryc Krion and glass amp stands
Sundry accessories: Acoustic System resonators
Room: 4 x 6m with high gabled beam ceiling opening into 4 x 8m kitchen and 5 x 8m living room, hence no wall behind the listening chairs
Review component retail: €13'350 ex. VAT

This is a companion piece to a prior CanEver review which covered all of the company's background and design philosophy already.

Voltage versus current amplification. n- channel versus p-channel Mosfets.

Variable impedance which loudspeakers present to our amplifiers makes ubiquitous voltage drive not exactly ideal. Yet so-called transconductance drive that would be impervious to their behavior is very rare even if once again proposed in Esa Meriläinen's 2010 book. Mario CanEver's amps from Venice are inspired by Esa's research but, given the impedance requirements of modern speakers, not text-book transconductance amps. They mate a 6N6P twin-triode voltage gain stage to current outputs which, like Constellation Audio's, combine two single-ended amplifiers set up counter phase to operate push/pull with just n-channel transistors for superior symmetry. Mario's two stages couple via bifilar interstage transformer doubling as phase splitter. Its secondaries present perfectly mirror-imaged signals to the output-stage halves which couple to your speakers via 600-watt output transformers. Sophisticated micro-processor bias control insures that the output transistors can't drift from perfect symmetry. The third tube works as shunt regulator for the voltage stage.

In one criminally simplified paragraph, this condensed what Mario admits tends to "can't possibly work" reactions. For €22'500, he calls his working proof to the contrary the two-chassis LaScala stereo amplifier with already a few most favorable reviews. It delivers 90wpc/8Ω in class A with a beefy external power supply. For €45'000, you'd call it two LaScala bridged to 180 watts/8Ω/ea. For €13'550, it's today's Olimpico. It transforms the sibling's big linear PSU into a constant-current smaller SMPS. That was originally planned to contain within the main chassis. When it grew, it split off into a separate unit. "50Wrms in pure class A was the original target for the single enclosure. When I moved to an external power supply, I raised the available power. There's a trimmer in the supply and another in the amp's bias circuit. If I accepted just below 60°C on the heat sinks, Olimpico would make even more but for decent headroom and to run things cooler, I set it to 70wpc/8Ω (140W/4Ω) for full-spectrum real continuous power into a resistive load. Your loaner was tested for 4 hours at 180Wrms into 4Ω. Olimpico can easily bridge for twice the power by simply connecting the + of the left channel to the – of the right channel or vice versa and the free terminals to the loudspeaker. Bridging two output terminals is possible because the output transformers isolate the load."

The external PSU has a double pi filter, toroidal inductors, teflon capacitors and electronic power-factor correction for 99% claimed power transfer. It also means zero feedback, zero signal-path capacitors, a minimalist two-stage architecture and just one pair of lateral Exicon Mosfets per channel. Where Olimpico differs from LaScala is thus only in the power rating, switching vs linear supply and fixed not configurable bias points.

"I originally designed this switch-mode power supply for another far more demanding application than audio. Just two weeks ago in fact that passed all of its official EMI/RFI compliance tests. In medical applications, it's not a matter of whether you can or cannot hear the effects of ultrasonic noise on critical circuity. As a matter of life or death, you cannot interfere. So you can't imagine what becomes possible when you must beat such hyper-stringent EMI/RF regulations which enforce zero emissions and zero sensitivity to external noise."

In certain ways, Mario's Olimpico stereo amplifier is a triple hybrid. It combines

• tubes + transistors
• transistors + input + interstage + output transformers
• tubes + switching power supplies + class A bias

For small-signal tubes, the latter isn't a first. Among others, Nagra do it. For power tubes, it's still rarer but Manley do it for their Absolute headphone amp, Ancient Audio for their new 300B SET. McIntosh have used transistors and autoformers for decades. Hybrids which marry tubes and transistors in either sequence have been around for ages. It's once we look for precedents that cross off all three of our bullet points that Mario's could just be first. Ditto for replacing the usual macho heat sinks with just four visible small chimney types for the output Mosfets. That's most unusual given the butch 50-watt rating in pure class A. This clearly is no cookie-cutter circuit that paints it in by the numbers. We'll wait for Mario to finish its final tuning—"neither easy nor fast with a switch mode supply pushed to beat my linear supplies"—before listening begins. Until then we do a Gertrude Stein:

A hybrid is a hybrid is a hybrid.

To return to transconductance drive, "by definition, mine isn't a transconductance amp where the output current is proportional to the input voltage and both input and output impedance are ideally infinite and in practice still very high. That wouldn't work with loudspeakers whose impedance with typical peaks spans 50-100Ω.

"But Esa Meriläinen proposed a more realistic implementation which became the starting point for my interpretation. In a schematic, my first stage is the voltage gain block, the second stage the current gain block whose input converts voltage to current then only amplifies current. With a sub 1Ω output impedance, it's simply not a classic transconductance affair.

"My circuit is a minimalist implementation because experiments convinced me that less is more relative to the most natural sound. From the onset I decided on a 100Wrms class A circuit using the lowest possible number of parts. That meant just six by way of an input transformer, a twin triode, an interstage transformer, a pair of Mosfets and an output transformer. Such circuit simplification and parts reduction minimizes the fingerprint which electronic circuits have on the signal. So my first stage connects its triode halves in push/pull which easily generates all the necessary voltage gain. The second stage connects two Mosfets in push/pull which easily manages the required current.

"For push/pull to work optimally, the devices for either phase must be the same. Hence I use twin triodes. It puts two triodes inside the same glass bulb to work at the same temperature, to be made from the same parts with the same tolerances on the same machine. Then I very tightly select my tubes to ideally match their two halves. The input transformer driving them is a precision phase splitter with identical windings.

"Likewise for the Mosfets. They are both n-channels so matching them is possible and simple. Again the interstage transformer driving them is a precision phase splitter with identical windings. It's also designed to account for the transistors' Miller capacitance which usually limits bandwidth. This transformer isn't a standard catalogue item but my own design which—sincere thanks to Per Lundahl for his patience—is then wound by Lundahl. [At left, a stock LL2746 three-section dual coil C-core tube amplifier stepup/interstage transformer.]

"The interstage transformer allows me to only use n-channel Mosfets, to avoid coupling capacitors and to avoid a third Mosfet to compensate for bias drift since I'm using only lateral Mosfets to begin with. To close this path, the output transformer becomes the impedance match of push/pull Mosfets to loudspeaker.

"That allows the Mosfets to work at their best linearity in the ideal area of their load line. The output transformer neatly avoids the need for the typically enormous output capacitor and its unavoidably huge impact on the sound. It also automatically protects the speaker via galvanic isolation. What about push/pull's cancellation of even harmonics to retain only the odd? Being mindful of this behavior means using identical bi-phase devices which don't introduce distortion harmonics in the first place of which then only even-order content gets cancelled. That relies on stringent pre-selection of my triodes and carefully matched n-channel Mosfets from the same wafer to be within 0.1%. What about bandwidth limits with transformer coupling? That is a very real challenge. All transformers impose bandwidth limitations and phase shift when compared to circuits without them. Once phase shift occurs beyond the audio band, the effects of limited bandwidth aren't audible. Here we need a minimum bandwidth of 40kHz. My three signal-path transformers have a bandwidth of, respectively, better than 150kHz, 75kHz and 100kHz for the input, interstage and output magnetics. Phase shift thus begins at 37.5kHz for the custom interstage transformer well beyond our hearing threshold."

Like an old-school 'heavy iron' valve amplifier, CanEver's Olimpico prefers transformer to capacitor coupling. Unlike an old-school 'heavy iron' valve amp, it uses an in-house designed medical-grade SMPS and transistor outputs. Like Nelson Pass FirstWatt transistor designs, it celebrates circuit minimalism in the pursuit of a more natural sound. That insists on non-switching transistors (class A), zero negative feedback and just two stages. Unlike FirstWatt designs, it delivers 50 watts into 8Ω. It's really very much its own effort. "Olimpico evolved step by step. I closely studied many amplifiers like the Phase Linear 700B, Harmon Kardon Citation 12 and FirstWatt F6 and finally arrived at my own take which obviously couldn't overlook transformers because I find their behavior to imprint on solid state a sound close to tubes. But I wouldn't say that I invented anything. I simply did the job of an engineer." Spoken like an actual engineer with a Master of Science in digital electronic engineering and a second-level Master in radar system maintenance.