The XMOS Processor
The XMOS processor primarily handles the interfaces, including USB connections, SPDIF/AES-EBU connections and I2S input from the WiFi/Network audio module. All inputs are decoded, passed through a memory buffer, then DSD/DOP decoding is applied and then MQA decoding. Also, the entire user interface, display control etc. are handled by the XMOS processor – the 16-Core / 2,000 MIPS.
From the XMOS the I2S PCM signal or DSD signal is passed to the FPGA for any filtering/conversion etc. and from there through a galvanic isolation barrier to the final re-clocking and DACs. This makes sure that all the noisy ‘digital’ processing and its noise is contained on the digital module and cannot affect the DACs and analogue side.
The audio clock is placed next to the re-clocker and DACs and on the ‘Audio’ side of the isolation barrier. This clock is programmed extremely precisely with a < 0.04ppm precision (hence, beyond the ‘femto clock’ level) and is either fixed (USB input) or is adjusted to the long-term average of the incoming data (SPDIF/AES-EBU/I2S from Wifi) which together with the memory buffer makes sure any source jitter is ‘locked out’ completely. This system is the same as found originally in the AMR DP-777 where it was implemented in an FPGA. For the Pro iDSD we have ported it to the XMOS, which is more suited to this kind of task.
After all that digital complexity (which is there for very good reason), the rest of the Pro iDSD is delightfully simple.
The DACs are used in voltage output mode. The signal from the DACs is direct-coupled to a RCLC passive low-pass filter at around 80kHz and from there either to the volume control and the input of line drive circuit or directly to the input of the line drive circuit.
The line drive circuit is in essence a modern take on classic 1950’s tube studio circuitry. It is conceptually based on late generation TAB/Mahaik Studio Console line-driver modules, which had two low-gain triode stages and a choke-loaded cathode follower driving an output transformer with a modest amount of looped feedback to lower distortion to what was considered ‘adequate’ by the stringent requirements of the German Public Radio/TV system.
Our circuit is not capacitor / transformer coupled but instead Direct- Coupled and able to drive 200 Ohm (or lower) loads without the need for an output transformer. It is fully discrete and a single ‘non-op-amp’ structure modelled on the way tube circuity works, eliminating for example distortion between stages caused by the non-linear input impedance of solid-state devices compared to tubes.
We use either triode (NOS GE 5670) or J-Fet input, with a special second stage in solid-state that offers extremely high input impedance and high linearity, similar to a triode circuit. The current drive comes not from a cathode follower, instead we use a Mosfet Follower loaded by a CCS with a ‘current booster’ using bipolar transistors as output buffer. Again, a modest amount of loop feedback is used to achieve ‘studio standard’ measured performance.
This looped feedback can be lowered to almost zero in the Tube+ setting, meaning the input tube operates without feedback and the solid-state second stage and output are enclosed in a shorter local feedback loop. This causes much higher measured distortion, subjective sound as always is a different story.