This review page is supported in part by the sponsors whose ad banners are displayed below
First some words from the designer. "As you already know, there are just two transistors and one step-up transformer used to create this device. The idea dates back to the early 90s. During this period we used step-up transformers for our electrostatic loudspeakers. Their step-up ratio was 130. A prototype power amplifier used a modified version with a step-up ratio of 25. However we had no intentions to making a commercial version as this transformer was created to handle 200 watts. Designing a dedicated transformer was very expensive and due to very low gain compared to transistors, a tube could do a very good job instead. This was the moment when we launched our hybrid amps using the old Toshiba power Fets driven by 12AT7 double triodes.
So why use a step-up in the Aurix now?
10dB gain module size relative to standard CD cover for scale.
"There's just very little amplification needed to drive high-impedance headphones. 4-6 volts are enough to blow out your ear drums. The low-impedance versions need no amplification at all. They produce the same damage at 2 volts or less. In our case the design is based for use in conjunction with dacs whose output is based on the Redbook standard, i.e. 2Vrms max. Therefore we created two gain settings which you can manipulate with a switch on the back of the unit. 0dB is the low-impedance setting for no voltage gain at all. The signal goes from the input directly to the potentiometer. From its tap the signal then flows directly to the final power Fet voltage follower. For passive gain the signal is routed to another Fet also connected as voltage follower which sees a 60KΩ load driving our step-up transformer.
"Now to some potential issues and how we address them. Potentiometer problems can occur when loads on the center tap are too heavy to induce bandwidth alterations depending on pot position. We connect the center tap to the Fet's gate which means no load except for one 10MΩ resistor and the Fet's internal 50pF capacity. Our Fets are pure voltage followers with a bandwidth of several MHz. As already mentioned I don't like transistors for low gain since the only way to force them into such behavior is to use local or global feedback. We needed a maximum amplification factor of 3 (10dB) which would have meant copious amounts of feedback to reduce gain. We prefer no feedback.
The module's flip side.
"An associated issue are the slewing problems when fast signal rise times aren't tracked to cause instability. Of course there are more criteria about selecting the appropriate semiconductor such as noise figures, current handling, distortion etc. Tubes would be an obvious choice as their voltage gain is lower. Our issue was that their power requirements meant they'd not fit into the housing which we wanted to be the same as our Octave MkII. This rekindled our old notions of a step-up transformer whose performance is comparable to tubes for clarity and imaging.
With gain module installed.
"A question could be whether the chosen Fets might negatively influence the sound but that's not our concern since our Fets aren't configured for gain. They simply track the input signal but supply more current at their outputs which drives both the step-up and the headphones. Other issues can be transformer limitations caused by core saturation and internal capacity from too many windings. Ours has typical THD of 0.007% when connected to a 2Vrms source. This means now distortion could only increase with load complexity (16-600Ω). However our step-up doesn't see the load at all. It's fully isolated from the outside world. But it does mean that in the 10dB gain setting our effective bandwidth is limited by the transformer's 5Hz - 60kHz specification. The 0dB bandwidth would be 2MHz but we intentionally limit it to 500kHz with a 1st-order filter to avoid ultrasonic interference.
Heat sink alley with gain module removed to better show the 1 x LM337T P+ voltage reg, 1 x IRF510 n-channel Mosfet and 2 x LM317AHV voltage regs per channel.