Asking Mircea for further details, "I started to develop this technology in 1999 but since then, the parts industry has made major progress. To take advantage of the very latest components, I utterly revised my original schematic which of course also meant a new layout. Our boards are gold-plated due to the active components' ultrasonic activity (VHF). We use four different types of feedback: differential, common mode, automatic offset and modulator feedback. Our amp will withstand a 30A/2Ω peak value at the zero crossing so there are no issues with 90° phase shift. I don't know of any other amplifier like it.

"For speaker safety, I limit overload to 28A/1Ω. Even under short-circuit conditions, the amp will deliver 10A almost indefinitely. The inputs are protected against ±10V over voltage. At 10V there is more than 4 x overdrive but even then a 60A shut-down in the power supply protects the speakers. The outputs are in absolute phase so each delivers one half cycle. Those are added in the 'recombinant zone' to create the amplified replica of the input signal. The differential outputs are driven by just one transistor so we avoid all common issues of running multiple transistors in parallel."

With Ella's 860 watts into 2Ω rating, those clearly were no ordinary transistors but industrial high-power specimens. Creative marketeers on occasion refer to a solitary output device per phase as single-ended push/pull because for such a circuit, four transistors (two per channel, one per half cycle) are the bare minimum. Class A/B or A solid-state amps of the type exist from Constellation, Gamut, Gato, Pass and Reimyo. For class D however and on that point, Ella probably has even fewer or no competitors.

"The amplifier is built across three sections as you can see in the block schematic. The modulator works in PWM and on absolute value on the first floor. After the filter and second floor we have the amplified/rectified input signal. The third floor is a power bridge which switches according to the polarity of the incoming signal. To do all of this, we use a type of analog computer. There are also some parts in SiC tech (silicon carbide). The modulator works at 2MHz and the inductors use special HF Litz windings. There are a lot of other state-of-the-art features. The amp is very robust. None have ever been damaged. The amp from the 2013 German review was from the first production run and thus different from our current units. Newtronics ceased their activities after our first run sold out. I also have a preamplifier with a linear 25MHz bandwidth and a volume control without any phase shift regardless of amplitude. For finish we have black or silver. For simplicity's sake, I drew the transistors as bipolars but they are Mosfets. They work under a constant Uds of 6V so power dissipation is low. More I cannot disclose. As to Cruz Audio, I merely contributed some schematics but they had no connection with Ella. The StraDa for Audio Physic did represent the concept at a very early stage some 17 years ago. The parts available then were far from what we have today, however. Again, Ella is a completely new development."

Mircea with the original Ella.

We see how exploded bandwidth to minimize or banish phase shift—which amounts to an optimized time domain—is one of Mircea's core concerns. Knowing what a hifi designer stands for or believes in, exactly, creates instant identity and relevance. A potential buyer isn't left wondering. One understands the design goal. Now one can either agree or disagree with its validity and relative importance. S.P.E.C. of Japan for example, another proponent of ground-up class D but there viewed as a modernized evolution of tube technology, focus on very different aspects. It comes down to what you want from your hifi. Which qualities are paramount to your enjoyment? If you don't yet know what type listener you are, you have some auditioning to do. Only that will tell you whether you prioritize rich tone, fast transients, heightened dynamics, maximal density, fluid breathiness, precise capacious soundstaging, relaxed mellifluousness or hart-hitting excitement. From what we've learnt so far, where do you imagine Ella is positioned on the map of many possibilities?

WIth shipping confirmation, I received this email: "After about 3 hours of standby, the casing will be about 12° C higher than the ambient temperature. If you use the RCA inputs, also use the included XLR shorting plugs. Otherwise you'll get a bigger offset at the outputs due to the input bias current and 20kΩ input resistor on the minus input. Your sample was burnt in for at least 72 hours. That is the industrial standard. I won't burn in amplifiers for 400-500 hours. To me that makes no sense and one could also say that it would make a new amp already partially used." Mircea also included a drawing on how to remove the top cover. Then he added the following: "Ella is the name of my wife. The basic circuit came about between 1985-1993 when I was transferred to China to build up a joint venture for CT scanner production. Conditions at the time were limited. I didn't want to waste my free time. To use it productively, my wife suggested that I start the research which became the Ella amplifier. Building it from independent functional block as shown has the advantage that the power supply controller, modulator and output stage can each be tested and adjusted separately outside the chassis.

"My switching power supply is very special. Generally using such a type for audio is problematic. During periods of no signal, a typical SMPS goes into burst mode. That is audible. It happens because a modulated pulse cannot be created from 0-100%. Usually it occurs from 5-90%. For those 5% you need a minimum load to keep the SMPS stable. This load produces appreciable heat and reduces total efficiency. Our SMPS is stable without any burst behaviour even during zero-load conditions. Just a small resistor on the secondary discharges the filter caps. Again, the transformer windings and filter inductors use high-frequency Litz wire. To reduce copper losses, their cross sections are quite large. The inductor's Litz is doubled up for a solid 10mm². The transformer can transfer more than 4'000 watts. To achieve that with a linear 50Hz-type power supply would require tremendous size and weight.

The three different modules - photos by Mircea.

"The primary current is monitored and immediately shut down if it exceeds the safety margin. The secondary current too is measured by a feedback loop. If it exceeds 60A, the controller forces the power supply to switch in constant-current mode. This creates an output voltage that stabilizes the output current at 60A. So in effect, our SMPS has a PWM range from 0-90%. The available power covers the maximum peak power which the two channels can deliver. This means that the output of the SMPS won't sag regardless of load. This is very important for very low bass. The diodes are SiC types. The controller is another building block with its own power supplies for the auxiliary voltages. Here I don't use a standard IC because their steering relies on flip flops. This steering is far too important to be exposed to irregular flip flops so I designed my own controller which can work in hard switching or phase-shifting modes. For the timing sequences, I use delay lines so my bridge is very safely driven at a switching frequency of ~80kHz, a good compromise between switching losses and size. Feedback compensation is as minimal as possible to keep the response time during load changes ultra short.

New-style chassis in black, photos by Mircea.

"The fully differential DC-coupled amplifier has each output drive one half cycle of the signal. If you then subtract the negative from the positive output, you arrive at the correct non-inverted signal. If you do the opposite, you get the proper inverted signal, with the input ground as reference and the output at ~13V above input ground. There is a certain protocol whereby we add the two half cycles in the so-called recombinant zone of ±2V. The incoming signal is rectified, parametrized and used as reference for the common-mode feedback all in real time due to our DC coupling. The operating points of the power transistor bridge are created by three different feedback loops. The power resistors are all zero-inductance film types. By working our modulator at 2MHz, the filter values can be very small. This increases overall bandwidth. For sufficient precision in our real-time analog processes, we use resistors of 0.1% precision. The modulator is driven directly from the outputs because its transfer function is very complex, including differentiation, integration and synchronization of the outputs. I personally wind the modulator storage choke, the primary current transformer, the output filter inductors and the power transformer. The silk Litz of the modulator bobbin uses a 3x2mm profile for the smallest possible size. There are no output relays. The outputs connect directly to the power transistors. To control the transient behaviour at on/off, the power supply starts up softly. There's a similar procedure for the start-up of the bias. By the way, this technology would be very interesting for car amplifiers and could also be integated in a hybrid chip for televisions; or fully integated."