How VoltaStream ZERO was born

Hi. I’m Philip. The guy behind VoltaStream. The development of the ZERO board was a very long and painful journey. This isn’t just another boring Linux tinkering board with bad software support that will get thrown into your tinkering stockroom after a couple of days. Instead it was created with a lot of passion and love for detail by a single person.

Everything started in 2013. I had a vision. One day my own music system built from scratch. Fully open, free of planned software obsolescence and affordable.  At this time I was a software developer with very limited experience in designing embedded systems. If I had known that it will take four years for the first fully integrated audio solution to roll of my own production line I’m not sure if I would go that way again today.

 

The first steps

So in 2013 PolyVection was founded. The company behind VoltaStream. In the beginning it was a software focused company only but with the time more and more hardware solutions were developed either for customers or for our own online shop. Most of them had something to do with audio. Behind the curtain not a single day passed without thinking about a Linux audio board from scratch.

The first concepts were based around the Raspberry Pi. Very soon I noticed that it’s a really good board for teaching children but for serious applications it’s a no-go (my personal opinion).  So something better had to be found. Something that can be customized and with good availability, even better documentation and best audio connectivity.

Raspberry with a DAC and a flexible connection for easy integration.

 

Going the individual way

After searching a lot on the web without any result the only way was to design something from scratch. Of course the first idea was to look at one of the open source boards like BeagleBone. That looked very promising. Everything is open, the TI Sitara SoC is easily available and in general this should allow a short time-to-market. After copying the complete design into the PCB CAD program I used at that time and ordering a few prototypes it started to feel a little more complex than initially expected. The PCB prototypes were really expensive as is was a 6 layer design, the SoC and the PMIC were expensive, DDR3 memory was expensive + hard to get in low quantities and of course the boards didn’t boot. Additionally I noticed that DesignSpark PCB wasn’t good for routing high speed signals and placing all the SMD components by hand really wouldn’t be an option for mass production. So without a better PCB CAD software and a pick-n-place machine this was a dead end road. All commercial solutions were just too expensive at this time.  Furthermore it gave me headache when thinking of the final cost per unit. My target was to have something not far from the cost of a Raspberry. Not three or four times as much.

The first PCB layout using a TI Sitara SoC wouldn’t boot.

 

Choosing the right SoC

After some frustration there was a break for a while. Randomly I looked into the market of ready-to-use system-on-modules. There were several good ones but they were either too expensive or were leaking audio features and documentation. At this time the cheap Allwinner SoC’s were common but didn’t want to use them because of their bad availability, the not-so-good Linux support and their high energy consumption.

Some time passed and someday a friend told me that the CERN contributed some nice features for high-speed routing to KiCAD. I knew KiCAD from their pre-OpenGL time and never liked it much. I decided to look into it again and was surprised how much it was improved.

The first prototype using the i.MX6UL arrived.

At the same time Freescale (now NXP) introduced their i.MX6UL SoC which was full of useful features, not too expensive and the evaluation kit was based on a 4-layer board. Great, several lines from my list of problems could be deleted. So I started to design an audio board based on the i.MX6UL in KiCAD.

Several weeks later the first prototypes arrived and they finally worked. What a great feeling that was! Of course there were a lot of ideas and the following prototypes got bigger with more features and more expensive. There were several iterations until concentrating on bringing the price and the complexity down.

A prototype with integrated amplifier, ADC, TOSLINK and wireless radio. Too expensive.

Reducing cost

Okay, so building a working Linux board with a lot of audio features like amplifiers, ADC’s and DSP’s wasn’t a problem anymore. But what about the cost? The plan was that the first product should be as affordable as possible. So some hard decisions had to be made in order to get the price down. No PMIC, no DSP and of course no WiFi or Ethernet. The PCB was limited to 65 x 30 mm so that our expensive but excellent European PCB fabrication house could still be used. For network connectivity an USB-A connector was added and all components were chosen to be SMD for easy manufacturing.
On the audio side a Texas Instruments DAC (PCM5121) with a great price/performance ratio is feeding a 3.5 mm TRS line-out jack that also embeds a TOSLINK transmitter for optical audio output.

VoltaStream ZERO KiCAD layout available on our GitHub.

 

Setting up a production line

One problem that wasn’t solved until this time was the manufacturing of the boards. I didn’t want to give that task over to an external company as the quantities weren’t clear and they are charging a lot of money for their services. While browsing the web for a solution I found a Chinese company that just introduced a very tempting machine. The NeoDen 4. Two cameras, conveyor, four nozzles, 48 feeders and actually affordable. At this time they had no distributor in Europe but an excellent sales service. Their offer looked so good that buying was the only option. A couple of weeks later it arrived and after some hard time in setting it up everything worked perfectly. What a great machine! That in combination with a nice stencil printer and the modified oven solved all of the problems on the production side.

VoltaStream ZERO gets populated with SMD components by NeoDen4.

During that time Freescale was acquired by NXP and they introduced the i.MX6ULL. It is pin compatible with the UL version and offers in the fastest configuration nearly twice the speed of the first prototype using the initial i.MX6UL SoC. So switching over from the 528 MHz i.MX6UL to the 1 GHz i.MX6ULL (actually only 996 MHz) was an easy decision.

Now everything was prepared for starting a production run. What I was a little mad about was the absence of a case. Okay, it’s a tinkering PCB sold as an evaluation board / kit to developers. Nevertheless some basic protection would be nice. It needed to be cheap as it should be included free of charge. 3D printing wasn’t an option as it just doesn’t look nice enough and manufacturing is slow.

 

The case

What about laser cutters? They are really cheap nowadays and should be sufficient for a simple case. So I bought the cheapest one available from china and a few days later it arrived. Setup was easy and the first cuts were good but not perfect. After an hour of aligning the internal mirrors and burning several holes into my fingers the quality was really good. The machine can easily handle a panel of 24 plates. Cutting is straight forward. They are looking nice and aren’t expensive.

The last step was to create some guides and documentation as well as this website. I don’t want to bore you with details about that.

Laser cutter engraving the front of the case.

 

 

I really hope that you enjoyed reading. See all technical details and the full documentation of VoltaStream ZERO in our shop HERE. In the next blog posting you can read more about PolyOS – the operating system created from scratch to match VoltaStream ZERO perfectly.

Subscribe to our NEWSLETTER below or follow us on TWITTER @PolyVection !