Our development plan is taking shape. Last week, I started to outline the schematics for the major components of our design, like the AD7606 analog-to-digital converter used for input CV signals. I will complete these over the week-end (6 more to go), then use some PCB development software to get complete schematics and gerber layouts for some of our boards (Keypad, CV, and Backplane).
In parallel, I will order evaluation boards for all the chips that we are planning to use. This will allow us to validate the selection of our major components, and to start developing some software for them. My goal is to have all components validated by the end of May. From there, we will start prototyping a few boards.
We will start with the Keypad board, because it is functionally simple, focusing exclusively on controls and indicators. Most importantly, it will give us a first experience with the ATmega328 microcontroller, which has become a cornerstone of our architecture from a user interface standpoint. The only challenging part should be the integration of the TUSB2077A USB hub, especially with regards to power supply. Fortunately, the diagram shown on page 13 of this datasheet gives us a great starting point.
To get a working Keypad board, we will have to go through the process of selecting a PCB development application, learning it, designing schematics and layouts, selecting a PCB manufacturing contractor, ordering parts in small quantities, getting the board manufactured, testing it, fixing hardware bugs, and possibly going through a few more iterations before we get something that we like. This should keep us busy throughout the Summer.
After that, we will focus our attention on the backplane. Its most complex component is the AD8191 HDMI switch, and we might decide to remove it from our design if it proves too complex to integrate. After all, we could easily salvage boards from existing HDMI switches, like we will do for the AVB switch. The backplane will also include two ATmega328 (one for each group of 4 knobs), and one TUSB2077A hubs, but we’ll have become familiar with these components by then. Otherwise, we will just have to route a bunch of tracks for implementing our mesh network, but these do not require any complex components. With a bit of luck, we should have a prototype board ready by the Fall.
At that point, we will be able to start playing with the Parallella boards in a more realistic context, using 4 of them connected through their mesh network and the Gigabit Ethernet network. With such a setup, we will start validating that our mesh network can actually be used for propagating audio and CV signals internally, connecting the 4868-9102C, 4868-9018EK-8, EVAL-AD7606EDZ and EVAL-AD5360EBZ evaluation boards to the backplane. If we manage to get some audio signal into our system, modulate it with some external CV signal, then produce some sound and derivated CV output, we will know that we are on the right track.
From there, we will move to the CV board, which will give us our first introduction to mixed signals. There, the main challenge will be one of PCB density, and we might realize that we need two boards instead of one, but we will play it by ear.
Once we have a CV board, we will move on to its companion portplate. This will give us a first experience working with metal plates and screen printing, starting with a relatively small project. We will go through as many iterations as we need to become comfortable with the process. This should prepare us for the development of the faceplate, which is massively more complex.
Winter should be upon us by then, and we’ll get ready to embark on the most challenging parts of our system: the audio ADC and DAC boards. If we decide to stick with the ESS converters, we will need some help, in the form of training and consulting. I originally thought that we had found it, but our lead did not pan out, so we’re back to square one. Fortunately, ESS Technology, Inc. is local (Milpitas, CA), and several of my LinkedIn contacts are directly connected to many of its executives. With a bit of luck, we’ll get the help we need from the source. Otherwise, we’ll read a few more books, and burn a few more chips…
If we get that far, we will know that we can do it. The backplate, amplifier board, and PSU board should be fairly straightforward at that stage, especially if we start from an existing design for the power supply unit. At this point, all that will be left will be purely mechanical or software based. On the mechanical engineering front, we will work on the faceplate, display panel, and frame at the same time, to ensure that all the pieces fit together nicely. For that part of the work, we will get help from an industrial designer, and we will work closely with a local metal shop. We should consider ourselves lucky if we have a first prototype sometime in the Spring of 2016.
After that, it will be all software, and I doubt that we’ll be able to get anything working in less than 6 months, which takes us to the end of 2016 for a first product release. It’s a full year later than what we had anticipated, but it sounds a lot more realistic. And by release, I mean something that would be shipped to a handful of early adopters, with incomplete software and a ton of bugs. I would expect that it will take another year to get something complete and stable. So we’re looking at 3 years worth of work to get a first product that could be brought to market. That’s an awfully long time to wait, but that’s what it should take, at the very least…