Luca Chiomenti on the previous page: "You absolutely nailed the character of my amp. That was the design goal. About the bass: yes, to stay a little bit in what you call the 'elasticity zone' was a deliberate choice. A damping factor of 16 gives this result. The general question requires a long discussion. Briefly, in my opinion lower Zout gains something only with some speakers which are innately underdamped but loses something else: the richness, articulation, harmonics and speed in the bass. Obviously like any choice, it is a compromise. The Zout in the 0.3-0.6Ω range is mine. And I appreciate how just a mention of any advanced concept could seem like big empty words without providing more detail. My real challenge is mostly having to take time away from production. There is something very concrete about my approach and I'm putting something together for you. As to tech specs, I could provide you with all manner of charts but the real difficulty are the power specs; not because they're secret but because defining and measuring power is one of the key aspects of my design which makes comparisons with traditional circuits difficult. I use 10wpc/8Ω class A as a common reference but it is really more. Just so, it becomes very hard to define power limits when distortion increases monotonically and clipping is so soft as to be difficult to find. And how to define static vs dynamic power in the first place? How short is a transient? How much distortion and what shape of distortion are acceptable during short-term peaks? There are no standardized answers, hence I have difficulty presenting fixed power figures that correlate with the 'norm'.

"The PSU was carefully designed not just for raw power but sound. It's about 200VA; sufficient for a 10wpc amp I think. Five power supplies use two transformers, one for pure audio circuits, the other for control and operational services. The tube branch is pi-filtered and stabilized. The power branch uses another pi filter and a technique I've used for 25 years; distributed capacitance*. Instead of two big slow capacitors, I prefer many small fast caps, with the last one extremely close to each power transistor. And there are no protection circuits, only fuses on the power rails. Key for the signal path is no global feedback anywhere and just minimal local. The single-ended ECC82 voltage amplifier uses none whatsoever and that stage wholly determines the amp's sonic character. The complementary outputs mix BJT and Mosfet and require no phase splitter or transformer. The motorized Alps sits at the very input before the first stage."

* Here we might cite Steve McCormack's so-called DNA or distributed node amplifiers from some time ago which implemented the same; or very recently, the Denafrips Terminator DAC with its multitude of small capacitors.

The thinking reader has already done the math. If Luca's psychoacoustic model were true—designers from Nelson Pass to Jean Hiraga have said the same things for years—any gain circuit that accounts for correct distortion behaviour should, over neutral speakers, sound congruent, easily accessible and harmonically weighty. Nelson Pass too shares the belief that remaining distortion products should be simple not complex. Still, Pass track their sales. Those show a near 50:50 split between amplifiers voiced predominantly 2nd or 3rd-order. Whether by biological imprint or later overwrites from culture, exposure and musical preferences, half their audience prefers the sweeter lusher sound of low amounts of even-order harmonics, the other half the crisper leaner more separated version of low odd order. Are there two expressions of the basic distortion model? Riviera's is clearly the first. Just as clearly, it must contain some 3rd order spice to avoid the cloying effect of 'deep triode' which Dennis Had of Cary espoused back when. By listening to well below 10 watts, I also sat below 0.2% total harmonic distortion in the first place. This wasn't the noxious perfume-doused stranger sitting down unbidden one table over only to drown out my dinner in a heavy cloud of sickly sweet scent. This was rather more subtle. Now Luca explains a bit more background.

"About 25 years ago I presented my first design in formal production. In its intro I wrote that 'we believe that over the last 20 years, high fidelity ended up in a blind alley. We see a pointless race to technical perfection. For too long the goal has seemed to reach the limits of our test gear. This obsession forgot the real function of an audio amplifier: to reproduce music through an electroacoustic transducer.' 25 years later the situation seems still more confused. Today there is a general consensus. Common tests for audio amplifiers do not correlate with their actual sound. As an indicator of subjective sound quality, audiophiles largely reject technical specs and bench tests. Nevertheless, perfect measurements seem essential in the market. Personally I cannot synthesize 25 years of study and research on the relation between subjective experience and laboratory measurements. I base my studies on direct experiments and hundreds of pages of bibliography covering over 80 years. I hope to soon complete a white paper about this. For now I will try to present certain key points on the foundation of Riviera Audio Labs. Here we go back to the beginning. An audio amplifier must reproduce a signal with the highest fidelity; for the human ear, not test gear. Now it becomes critical to understand certain aspects of how the human hearing system works and consequently, how to define the characteristics of the reproduced signal for the human ear, not for an electronic measuring system. Let’s start at the ear. When we hear a pure tone, many studies verify the creation of harmonics inside the ear, specifically the cochlea. This is no new discovery. First reports on this came from Fletcher (yes, the famous man behind the Fletcher-Munson curve from the 1920s). More precise reports came from H.F. Olson (Acoustics, 1947) and many others later. It is interesting how the ear generates really high levels of second harmonics; about 10% for 90dB SPL (not 120dB or more!). Higher-order harmonics decrease with the order of harmonics. Now we can define a spectrum of the harmonic distortion of the ear. The shape of harmonic distribution is very important. There is a high predominance of lower-order harmonics in a decreasing spectrum which changes with sound pressure but is too complex to examine here. Its key points are: 1/ the high level of distortion the ear generates itself; 2/ that the ear/brain system cancels out those harmonics and the resultant perception is one of an absolutely pure tone.

"In short, our hearing system suppresses its own self-generated distortion. Even more interesting, it suppresses that range of harmonics even if they are of external origin, under the condition that their shape and pattern be the same. Obviously then our biological system is programmed to cancel that shape of distortion. It cannot distinguish whether the origin of said distortion is internal or external (some interesting musical phenomena are related to this, i.e. the missing fundamental note). If the harmonics differ from this shape, the ear/brain system detects them as different tones. From this we believe that an amplifier which generates distortion similar to the human ear will sound extremely transparent and clean even if its THD (measured in isolation) is relatively high.

  • This mechanism depends on SPL. The higher the sound pressure, the higher the harmonics distortion which the ear generates and accepts. This leads to a preference of amplifiers whose THD increases monotonically with output power.
  • The higher the sound pressure, the higher the order of harmonics which the ear generates. At higher levels, it processes an increased quantity of higher-order harmonics.
  • Detectable distortion depends, among other things, on 1/ the ratio between peak and average signal value; 2/ the duration of each signal peak. Studies proved that distortion can reach very high values yet remain inaudible if peak lengths are very short.
  • Masking is a well-known phenomenon whereby a low-level tone in close proximity to a louder tone is inaudible. This can limit the influence of IM distortion in an amplifier with predominantly low-order harmonics. This effect seems to be even higher if the distortion spectrum's shape of the amplifier approaches that of the ear.
  • Feedback is the classical method to reduce THD and improve measurable circuit behaviour (THD, IMD, bandwidth, noise and more). Unfortunately feedback reduces lower-order harmonics (less harmful and more benign to the ear) far more than it does higher-order harmonics. What's more, high feedback acts as a generator and multiplier of higher-order harmonics which the ear detects as dissonant and noisy. Even if these high-order harmonics remain below the audibility threshold, the mechanism itself generates a noise floor which seems to be strictly undesirable to the human ear. This suggests that it is better to avoid feedback (especially global feedback) or reduce its use to a minimum.

Starting with this theoretical base, we defined the ideal characteristics for amplifiers dedicated to humans, not test gear. Our amplifiers are optimized for the bench test in effective relation to this concept, not to pure technical virtuosity. We focused on these points:

  • Optimization of distortion for amplitude and frequency. THD needn't be extremely low but it must track the shape of the ear's own. This means a predominance of low-order harmonics with their regular distribution across frequency (the higher the order, the lower the harmonic level, with a relation between these harmonics which is similar to the human ear). Also, the distortion level must increase monotonically with power. The amp should have soft clipping and, if possible, the distortion spectrum must be similar to the ear spectrum even in this area (or to the highest level possible before losing the right shape).
  • The use of zero global feedback and minimum local feedback to minimize its negative effects. This is the best way to reach the desired distortion behaviour.
  • Good open loop bandwidth.
  • Reasonable damping factor like the best tube amplifiers without chasing after meaningless records: between 15 to 20. This contributes to articulation and harmonics richness in the bass.
  • Total stability into any load.
  • Absence of protection to avoid their sonically negative effects and dynamic choking.
  • Extreme care in the design of the power supply.

"These theoretical points were then implemented in a real amplifier design. We used zero global feedback and minimum local where it was absolutely indispensable. This became the first step toward reaching the desired distortion behaviour. The use of class A in all stages became the next logical consequence. We wanted maximum linearity without corrective tricks. The hybrid solution became the next logical consequence. A triode is the best voltage amplifier and, mostly in a single-ended configuration, offers a natural distortion shape that's very close to the required one. Silicon devices and especially Mosfets are the best choice for power and low impedance. If properly used and driven, they too can offer a good distortion shape. In the topology adopted, they also give the desired output impedance. Protection circuits damage the sound so we only use fuses on the PSU rails. This led to a realistically large power supply without coke-can capacitors. A secondary consequence was the need for adequate mechanical dimensions. Also, we wanted a laboratory-level finish and obviously Italian cosmetics. Finally a few hard specs. Sensitivity is 400mV relative to 10W/8Ω. Input impedance is 50kΩ. Bandwidth at 1 and 10 watts is 11Hz - 55kHz minus 1dB. At minus 3dB, it is 7Hz to 105kHz. Residual noise is 95µV A-weighted. SN/R is 99.7dB A-weighted. Power consumption is 110VA."
To be continued...

Riviera Audio Labs website