Reviewer:
Srajan Ebaen
Financial interests: click here
Sources: 27" iMac with 5K Retina display, 4GHz quad-core engine with 4.4GHz turbo boost, 3TB Fusion Drive, 16GB SDRAM, OSX Yosemite, PureMusic 3.01, Tidal & Qobuz lossless streaming, COS Engineering D1 & H1, AURALiC Vega, Aqua Hifi Formula, Fore Audio DAISy 1, Apple iPod Classic 160GB (AIFF), Astell& Kern AK100 modified by Red Wine Audio, Cambridge Audio iD100, Pro-Ject Dock Box S Digital, Pure i20, Questyle QP1R
Preamplifier: Nagra Jazz, Esoteric C-03, Vinnie Rossi LIO (AVC module), COS Engineering D1, Wyred4Sound STP-SE Stage 2
Power & integrated amplifiers: Pass Labs XA30.8; FirstWatt SIT1, F5, F6, F7; Crayon Audio CFA-1.2; Goldmund Job 225; Aura Note Premier; Wyred4Sound mINT; Linnenberg Allegro monos; Nord Acoustics One SE UP NC500MB
Loudspeakers: Albedo Audio Aptica; EnigmAcoustics Mythology 1; soundkaos Wave 40; Boenicke Audio W5se; Zu Audio Druid V & Submission; German Physiks HRS-120; Eversound Essence; Albedo Audio Amira [on review]; Ecobox Daydream [on review]
Cables: Complete loom of Zu Event; KingRex uArt, Zu and LightHarmonic LightSpeed double-header USB cables; Tombo Trøn S/PDIF; van den Hul AES/EBU; AudioQuest Diamond glass-fibre Toslink; Black Cat Cable redlevel Lupo
Power delivery: Vibex Granada/Alhambra on all components, 5m cords to amp/s + sub
Equipment rack: Artesania Audio Exoteryc double-wide 3-tier with optional glass shelves, Exoteryc Krion and glass amp stands [on loan]
Sundry accessories: Acoustic System resonators
Room: Rectangular 5.5 x 15m open floor plan with two-storey gabled ceiling, wood-sleeved steel trusses and stone-over-concrete flooring
Review component retail: Older cosmetics €6'980; new cosmetics as reviewed €7'890 (both factory-direct, customer adds shipping)


Neither did I know the emailer nor his company. But in true picador fashion, my curiosity was lanced within seconds. "My name is Mircea Naiu. With my firm Naiu Laboratory GmbH, I manufacture the Ella amplifier. This has been compared to many established brand and the consensus is outstanding. It's based on Naiu Digital digital/analog technology. The circuit is fully differential and the concept combines the advantages of analog and digital with specs like small-signal bandwidth of DC-1.1MHz, 40ns propagation delay, 450/860wpc into 4/2Ω power, short-circuit protection and a 4KW digital power supply. I would like to discuss a review." His website added that "the aim was to unify the advantages of class A and D without their disadvantages. This sounds trivial and logical but took years of fundamental research to accomplish. Our patented circuitry doesn’t fit any established matrix. With our amplifier we set new standards for power bandwidth, phase precision and sonic stability. Ella reaches the tonal advantages of class A with an efficiency of ~80% which resides in the area of pure class D amplifiers."

Old-style chassis.

For their circuit, Devialet make a similar A+D claim but execute it very different. And who ever heard of class D with 1MHz bandwidth? Further specs are a >800 damping factor, <0.02% distortion, 2.25Vpeak input sensitivity for full output, input impedance of 20kΩ at DC, phase shift of <0.0002° at 20Hz and <0.03° at 20kHz, with peak power consumption of 2'500VA. All of it is encased in a 44x32.5x11cm WxDxH chassis weighing just 20kg. Doing online recon, I came across a 2008 post by a gent who'd worked at, then acquired the post-bankruptcy assets of Erlangen's AN+DI Electronic to take on service of their Revolution amps. Those were Mircea's earlier sliding-bias class A designs.


Charly [above in an Ella demo] reminisced how "the Revolution tech rested on four pillars. The first was modified class A. It replaced classic high constant bias with a real-time analog comparator circuit which also controlled feedback. It ran AD 8000 series opamps with 240Mhz bandwidth. The complementary Mosfet outputs were driven by two rails, one for the signal, the other for a constant 100mA bias current. Three precision resistors fed back the combinant signal to the calculator circuit. The second pillar was the effect of the 100mA bias on each output transistor pair. Since the transistors were always slightly open, some delta loading not the full gate charge sufficed. This led to 3MHz speed. In the Revolution 2 this was deliberately limited to 1.8MHz to remain compatible with capacitive loads. The smaller Revolution 4 and 5 used 1.5MHz. The third pillar was a discrete control loop for each Mosfet to avoid reliance on matched pairs. The Revolution thus could get by on cheap IR 200V types. The fourth pillar was the 'current copier' which could scale up the circuit in parallel for ultra-low Ω loads. A safety feature shut off any defective power Mosfets, then scaled back output power to suit the remaining transistors. Unlike conventional class AB where only one transistor of a complementary pair is on (either the n or p channel), in Mircea's pairs both were. Now the constantly monitoring calculator circuit could instantly correct for anomalies in one phase without time delay. The price was idle power consumption of 600 watts; certainly not much for a 250wpc class A amp but still an impact on the utility bill."


The next descriptive sketch I found, now on today's switching Ella amp, was in hifitest.de's 2013 review. Then it was still branded Newtronics and had a €5'000 sticker. According to reviewer Christian Rechenbach, the amp's first stage creates a pulse-width modulated signal followed by a low-pass filter to generate the amplitude-modulated signal. The location of the output filter inside the feedback loop and modulator decouples it from the outputs to achieve load-invariant performance with a low constant output impedance. On a 2013 open-end-music forum thread, I finally hit upon a post by speaker designer Rainer Weber of Kaiser Acoustics fame. He felt that Ella was on par with amps like the Thrax Spartacus or Engstrøm TheLars. He also reported on a customer demo of his big Kawero speaker for Dutch clients. They drive their own Kaweros with the biggest Mark Levinson amp. They also felt that Ella performed in a much higher league. For those keeping tabs, by Munch HighEnd 2017, Kaiser showed with valve gear by Kondo who'd become their Japanese importers. Herr Weber has high standards and expensive tastes. His endorsement of Ella carried weight in my little black book. Mircea the picador had a few more lances in my hide.
 

If we distill the above, we net a fast amplifier, a concept which Ella shares with Goldmund, Linnenberg and other—admittedly non D-class—amps. Steep rise times and minimal phase shift rely on ultra-wide bandwidth. Why a new circuit? "It was impossible to implement our requirements with established gain technologies. They destroyed the power transistors with ultrasonics. They suffered unavoidable parts differences, thermally shifted load lines and small offsets between transistor switching times. Furthermore, we had to consider today's requirement of energy efficiency." It's why the final circuit is called class ND. That's short for Naiu Digital. Attentive readers might recall class T aka Tripath. So... Ella is an analog switching amp with a very powerful switch-mode power supply, 2MHz controller and a filter isolated from its outputs. From the above we also glean that the Naiu Laboratory Ella is the continuation of the Newtronics Ella whose very different precursor was the now long discontinued an+di Revolution range. Would Mircea's third attempt in business be the charm?


Another Charly post filled further gaps. "Mircea Naiu is of Romanian descent and studied IBM computer tech in Bucharest. He eventually was drafted by Siemens Medical to work for a Chinese joint venture in computer tomography. In his spare time he studied the theory and practice of gain circuits. After getting frustrated with traditional approaches, he finally developed what became the Revolution 1. He did so without regard for cost or compromise. This led to a proposed DM95'000 sticker for a stereo amp. I own the only two samples in existence. Their tech is still protected by a 20-point patent. Former Siemens colleague Walter Keller helped Mircea turn his circuit into a final product. Obviously the Revolution 1 prototypes were far too over the top. This led to the half-sized Revolution 2 with a 250wpc power limit. It launched in 1996 and exhibited at the Vienna show. Fully loaded, it demanded DM45'500, still too much for a newcomer. Only five sold. Unfortunately Mrs. Naiu and Keller had by now left Siemens because a distributor had promised turnover numbers which turned out to be over inflated by 99%. It led to the formation of an+di which I soon joined. I immediately recalibrated the 3-year business plan and insisted on more cost-effective models without all the features. I aimed at a DM15'500 price point. an+di already had similar ideas with the Revolution 3 from which we scaled down further the models 4 and 5. About 20 of the latter sold. That was insufficient to cover costs. Investor funding dried up. By late 1997, an+di declared insolvency. I acquired all assets except for the patent. This meant I could only service but never build any Revolution amps."


From sliding bias class A to a novel switch-mode amp, we've now resolutely landed on Ella's turf, with the Romanian revolution long behind us. For completion's sake, there are also the Cruz Audio class D amps based on a similar Mircea circuit from 1999 [above left]. Prior to Cruz, there were Arnold Heres' €15'000/pr StraDa amps for Audio Physic for which Naiu contributed circuitry as well. A good man with a solid idea is hard to keep down? With the branding now unmistakably Mircea's own Naiu Laboratory, it certainly did seem so.