This review page is supported in part by the sponsors whose ad banners are displayed below

I asked Carlos Candeias to ship the CS2 directly to my office. I still did not expect such a heavy load and was clearly impressed with his wooden box of nearly 60kg. The last integrated to have been a real pain in the back was the Emillé Labs Cha’am. This time BMC's loaner was even more massive and sans any facilitating handles. Heavyweight for a relatively light sticker. Might that be the introduction to the feature set? The stabilized 2KVA toroidal transformer is guilty for this brute mass. Such a generous power supply is most rarely seen in integrated amplifiers except perhaps for big Musical Fidelity class A statements. An innovative stabilization circuit in the power supply filters remaining ripple and other disturbances from the supply voltage and keeps it constant even under changing load conditions. The circuit is designed for low impedance across the audible range to not restrict dynamics or interfere tonally. The numerous small but fast power-supply capacitors draw a lot of current on demand to justify this XXL power transformer.


The large illuminated display confers a strong visual identity. More visual thrills for me came from the wraparound heat sinks and the sweet satin finish of the entire enclosure. If the CS2 does not exude the svelte aesthetic refinement of the Devialet Premier, its external appearance achieves impressive heights for a device in this price range. I asked Carlos Candeias more about this impressive casing and he confirmed that very unusual attention was given to the enclosure whilst keeping costs controlled. This tour de force is based on aluminum parts produced through an extrusion process of liquid aluminum. Then the profiles are cut into matching pieces and merely a small amount of CNC machining remains as the shape is already formed. The complex heat sink thus requires no more work than a front panel. This process allows for a further improvement in controlling their aluminum’s quality and density and it avoids heat stress from extensive CNC machining. The satin surface of the CS2 amplifier is not easy to make. Few specialists are able to deliver constant quality. First the aluminum is sanded then sand-blasted with a defined filtered grain size under high pressure. Finally a special anodizing process applies to give the treated surface a little sheen to achieve the final silky look and texture.


The central power meter is flanked by two large knobs, the left for power, the right for volume. A small button between  meter and power knob controls illumination level, a matching one between meter and volume toggles sequentially through the inputs which can also be selected from the multi-component remote wand of machined aluminum. The central dial displays selected input and indicates volume setting in 1dB increments from 00 to 66. Remote operations are very friendly and the settings fast and precise. The big BMC aluminum remote controls the amplifier together with the matching DAC and CD player. The AMP C2’s attenuation occurs electronically via an array of transistors and resistors. The knob itself merely triggers selection impulses. Input switching activates an auto fade to attenuate the previous input, then restore volume with the new one. These fast precise non-intrusive operations demonstrate the innate quality of the new kid in town. When the CS2 comes on, the display begins a 10-second countdown during which all supply voltages ramp up. When the amp is ready to go, volume is automatically adjusted to a conservative ‘10’ which avoids unpleasant surprises with any kind of speaker. The peak-reading meter displays power in watts (for a 4-ohm speaker) and dB relative to 1W.


This fully balanced amplifier is claimed to deliver 200wpc into 8 ohms and an impressive 350 into 4. It has two pairs of balanced inputs, three pairs of single-ended inputs, their sockets placed to either side of the central IEC power inlet. The rear panel does not add anything else except for a simple pair of biding posts at either end. The most salient features of the CS2 are inside the casing. Carlos Candeias defines the CS2 as a load-effect-free amplifier operating neither in class A nor B. Candeias’ LEF terminology implies that his product doesn't use energy-wasting class A operation to eliminate crossover distortion and have transistors operate within a more or less linear range because such designs usually still require some global feedback to reduce distortion. If distortion never occurs in the first place due to the choice of transistor operating characteristics, it needn’t be corrected.


Candeias explained that LEF technology is a new kind of current amplification neutralizing transistor nonlinearities without overall negative feedback but instead a local current servo. It is not complicated to design a feedback-free voltage amplifier with low distortion when the main distortion characteristically caused by transistors happens in the current buffer due to strong variations of VCE and/or IC. These variations are usually part of normal playback and have been considered non-avoidable. Hence feedback compensation became de rigueur. LEF separates signal reproduction from power variations to work the gain transistors with virtually constant VCE and IC. That’s the overall concept.


Fig. 1 shows a traditional single-ended current amplifier. Just one transistor transforms the high-impedance music signal from the voltage amplifier to a stable low impedance output. This amplifier must work in class A to avoid current clipping. Therefore a current source must supply a constant maximum peak current. For 300Wrms at 4Ω this would mean a 12A bias current. The transistor has very high variations in VCE and IC and a bias current of 12A is utterly unreasonable. Fig. 2 shows same amplifier with an added floating cascode. Now the signal transistor undergoes no more variations in VCE. However IC variation and the high bias for required class A operation remain unchanged.


In Fig. 3 and to support true LEF operation, the signal transistor from the circuit in Fig. 2 must eliminate high current variation and very high bias current. For this purpose an IC servo is added. It includes a sensor for the signal transistor’s current, power current sources for either polarity and a high speed regulator. The bias current for the signal transistor now can be 50mA. This is a very small class A amplifier with excellent dynamic response. As soon as the signal transistor current becomes other than 50mA due to music signal and speaker load, the servo adds just the right current in the right polarity.


A hifi power amplifier is a voltage source. It has a very low-impedance output and the output voltage should be independent of drawn current. The IC servo measures and provides current alone. It ‘knows’ nothing about signal voltage and is not involved in it. The IC servo’s power-current source impedance is very high and incapable of influencing the signal transistor’s output voltage. The time relation between voltage and current (technically phase) is ‘unknown’ to the servo as well to render its action phase independent. This now is a true LEF amplifier. The signal transistor is free of any variation in Vce and Ic and thus free of distortion. All load effects are neutralized and the stability of the voltage amplifier is improved remarkably. Fig. 4 shows how for best audiophile performance a balanced LEF amplifier is required. This doubles signal processing speed and reduces distortion further. Truly balanced means differential operation. Summing up, LEF allows single-ended operation with high power output, making global negative feedback obsolete (even if strict minimum negative feedback should be employed to diminish excessive gain). LEF also makes possible the following technologies.
Another significant difference with common amplification devices is called discrete intelligent gain management (DGIM). DGIM recalibrates amplifier gain without attenuating input signal. Carlos Candeias does not  use a standard attenuator prior to his voltage gain stage. Candeias’ circuit changes its gain factor according to the desired volume whereby the input delivers constant uncut source signal without attenuation. This avoids the common paradox of high fixed gain that results in attenuating the input signal only to amplify it again over multiple gain stages with concomitant distortion and losses involved. Here the gain is set directly at the speaker level’s output voltage to prevent signal attenuation and redundant amplification.


For DIGM bipolar transistors switch precise resistors in different combinations to match the desired gain factor. The input current turns into the kind of voltage which corresponds to the required speaker output voltage and adjusts in 1dB steps. Integrated CMOS digital potentiometers cannot be used because of high requirements on current and voltage strength. Hence a specially tailored circuit meets these demands with SMD-fitted modules. This scheme also prevents amplification of noise. Traditional amplifiers reduce noise from the source when they lower the level but the noise of the output stage remains constant. DIGM reduces gain at lower levels to decrease both source and amplification noise. DIGM also eliminates the need for a preamp so a source may connect directly to an amplifier.


As a few current amplification specialists like Krell have done, BMC has equipped all components with their patented current injection design. The main difference of BMC’s patent is based on the fact that the dedicated CI input doesn’t require proprietary socketry but uses common XLR interconnects. Krell’s CAST patent is also based on a feedback loop with inverted input to lower impedance. In a BMC machine the original current which the signal source injects into the low-impedance CI input flows through the amplifier circuit and the resulting signal voltage is defined only by the termination at its end. The CI input uses the original signal until it reaches the output voltage whereas voltage inputs just create a copy of the signal as do Krell’s amplifiers. In that sense Candeias’ work is closer to JMF Audio. The termination can be a simple resistor or a resistor network as with DIGM. Of course it is necessary to use other BMC devices appropriately designed to optimally drive the CI input and unleash its ultimate capabilities. Other combinations may in fact produce inferior sound from a CI than standard XLR input due to insufficient current.  I thus could not assess this specific feature by having no matching BMC loaners on hand.