One More Time

Now that we departed from the Eurorack form factor, we can be as creative as we want. It’s a bit of a dangerous game, but I like playing it anyway. Today, I asked myself the following question: if I were to reduce my audio equipment to as few individual pieces as possible (in perfect less.best fasion), which ones would they be? Of course, I would need a keyboard (or a grid) and a a pair of studio monitors. But could I combine my hybrid synthesizer, control surface, desired audio interface, and lusted-for audio accelerator into a single portable device? This is the answer I came up with:

Faceplate 0.0.9

Backplate 0.0.2

This beauty would have the same footprint as a MacBook Pro 13″ and would be 60 mm tall on the back, and 40 mm tall in front, with an inclined top surface. The 16 OLED displays would be replaced by two 5″ HDMI touchscreens with 800 × 480 resolution. Two knobs would be removed from the previous design, but ten SPST buttons would be added. And the 32 jack chassis sockets would be moved from the faceplate to the backplate, making room for 8 motorized faders with full 100 mm travel.

That’s for the outside. Inside, things go totally crazy. Instead of using the Enclustra Mars ZX3 for the core fabric and our submodules, we would use the Parallella-16. This board has the exact same footprint as a standard credit card and sports the same Xilinx Zynq-7020 as the Enclustra, but it adds an E16G301 16-core microprocessor with some incredible specifications.

  • 16 high performance RISC CPU cores
  • C/C++ and OpenCL programmable
  • 32-bit IEEE floating point support
  • 512KB on-chip distributed shared memory
  • 32 independent DMA channels
  • Up to 1GHz operating frequency
  • 32 GFLOPS peak performance
  • 512 GB/s local memory bandwidth
  • 64 GB/s Network-On-Chip bisection bandwidth
  • 8 GB/s off-chip bandwidth
  • 1.5ns network per-hop latency


And we would put 8 of these boards into our system, or one per channel. Internally, they would be connected in two ways: first, we would link them through a built-in Gigabit Ethernet switch, allowing the 8 instances of the Linux operating system to talk to each other; second, we would connect the 8 E16G301 microprocessors in a ring topology by using their mesh networking capabilities, allowing signals to flow between the 8 microprocessors and the 8 FPGA subsystems at 8 GB/s, with no more than 6 ns latency.

With such an architecture, we would not need any central module for coordinating signals across our 8 submodules. Instead, we would have a flat network of submodules, each capable of communicating with the others on their own. Each submodule would run its own io.js server on the Zynq, and two submodules would be used to drive the two HDMI displays. Finally, one submodule would be used to drive the buttons, faders, and knobs.

One benefit of this design is that it would give us enough room to accomodate all the circuitry that we need for top-of-the-line audio conversion, using 8 ES9102C ADC converters, and 2 ES9018S DAC converters, with proper power supplies and buffers. This would give us 16 audio inputs and 16 audio outputs with 32-bit sampling and a 384 kHz sampling rate. Sounds crazy? Yes, it does, but it looks like we might have found an expert in the field willing to give us a hand with that part of the project, therefore we’re a bit emboldened.

All we have to add to our design is a pair of XLR outputs that would let us drive a pair of studio monitors, and a solid pre-amplifier for mixing headphones. We’re still looking into these, but I’m pretty confident that we’ll find something suitable in the DIY market.

What I really like about this new design is that the Parallella board is totally open source and supported by a fairly active community. It also retails for $265, which makes it more affordable than any alternative we’ve considered so far. And it would give us a positively mind-blowing 256 GFLOPS, assuming that we could solve all the heat dissipation issues that we will likely encounter.

With all that goodness packed into such a small package, retail price is likely to shoot toward the $5,000 mark, but for that price you would get the following:

  • Control surface with 8 motorized faders, 10 buttons, and 14 knobs
  • 16 × 16 audio interface with 32-bit resolution at 384 kHz
  • 16-channel synthesizer with 16 CV inputs, 16 CV outputs
  • Audio accelerator with 144 CPU cores and 1,760 DSP slices

Now, I really, really want one…


2 thoughts on “One More Time

  1. Julien says:

    I was wondering when you would introduce a Parallella 🙂 . A thing to note, general platforms that can do many things at the same time had a hard time being successful, such as the Creamware SCOPE platform, which was fantastic from a technical and performance point of view. I believe people prefer dedicated hardware, that do one thing, but do it well. The Feeltune Rhizome is another example.


    • You know me all too well my friend… Yes, general platforms can be a curse: jack of all trades, master of none. I’m trying to do something slightly different though: an integrated system: interface + processor + surface. And I want it to be portable, which neither the SCOPE nor the Rhizome were. I want the audio quality of an Apogee Symphony I/O, the power of the fastest DSP card currently available, and the user experience of a Pro Tools | S3, but in a smaller package. Is there a market for that? Beyond myself, I don’t know. But if such a thing were to exist, I would buy it. And since it does not, I’m going to build it.


Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Google+ photo

You are commenting using your Google+ account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )


Connecting to %s