History

This project has been in the works for a while. It was originally conceived in 2008 as a CD player built around a mini-ITX PC. In 2009 I designed a set of PCBs and a custom chassis to hold everything. Some of the design principles were established then: galvanic isolation of all digital signals, 2-channel and 8-channel DACs and ADCs, 4 modules per chassis. I acquired a Digilent Nexys2 FPGA board for the digital interfaces. However, my FPGA skills lagged behind my ambitions and the PCBs I had designed weren’t very well thought out. I more-or-less abandoned the project and, for my own stereo system, continued using a CD player and an analog active crossover.

A few years later, I ended up doing research in digital circuits. This made the digital design for the FPGA seem much less intimidating. In 2013, I decided to finish the missing pieces and resurrected the original years-old hardware. I was able to demonstrate the concept of an FPGA-controlled USB DAC with the DSD1792A. However, I found many problems with the original PCB designs and my architecture was complex and unwieldy. Also, the Nexys2 FPGA platform was no longer supported by current tools; in 2015 I switched to a ZTEX board with a modern FPGA. In 2016, after finishing graduate school, I decided to revise the PCBs and construct a complete chassis. This became the “baseline” implementation of the project, with plans to add other interchangeable components in the future. In December 2017 I started using the device as a digital crossover and streaming server in my main stereo system. I rewrote some of the HDL code that was license-encumbered and published the project on GitHub in April 2018.

photos/DSC_0234.jpg

Testing of an earlier version of the platform in early 2017. On the left side are prototype Jung regulators with a 3-terminal form factor; they were later integrated onto the bottom side of the backplane. At right is a 2-channel DAC module based on the DSD1792 and a new discrete I/V stage. (This DAC worked quite well and I plan to port it to the current platform.) In the background is the ZTEX FPGA development board connected by a ribbon cable.

Progress has been slow due to work and family commitments, but I intend to continue soliciting community input, making improvements, and for those brave enough to build their own version, supporting the project indefinitely.

Fun facts

This list will be periodically updated with stories and lessons learned from building and testing this project.

  • You may be wondering, where does the name “Samoyed” come from? When I was working on many of the boards you see here, my wife would comment with enthusiasm about samoyeds. A samoyed is a large, playful, fluffy white dog. She talked about them so much that I decided to name the project after them.
  • Don’t try to drill out your own holes for XLR connectors unless you are comfortable with: nasty vibration, chuck falling out of the spindle, and irregular holes larger than you wanted. This happens even with the cloth trick. Send them to the CNC shop.
  • When I tested the current set of baseline modules, I wired the XLR connectors with pins 1 and 2 for differential signals and pin 3 for chassis ground. This is wrong; pin 1 should be chassis ground and pin 3 should be a signal pin. However, because I was doing loopback tests between the DACs and ADCs (both wired in the same incorrect way), everything seemed fine. It was only when I connected the DAC outputs to an amplifier and asked “why is there still ground loop hum?” and “why is there so little gain?” that I recognized the problem.