RoadTest: 3D Gesture-Controlled MGC3130 Hillstar Dev Kit
Author: jgmarcelino
Creation date:
Evaluation Type: Evaluation Boards
Did you receive all parts the manufacturer stated would be included in the package?: True
What other parts do you consider comparable to this product?: Hover board (also MGC3130-based), Cypress CapSense MBR3 Evaluation Kit, CMUcam5 (the Pixy)
What were the biggest problems encountered?: Windows-only configuration software and not very intuitive, "Extended interface" on MGC3130 board was useful but hard to use without rearranging the whole system, board connections are a little fragile, 2mm pitch connections instead of the more standard 2.54mm.
Detailed Review:
I had great expectations for the MGC3130 Hillstar Development Kit, but somewhat tempered by past experiences with systems such as Cypress CapSense (problems in longer range sensing) or the CMUcam5 system (terrible low light performances). I also had tried a Hillstar that was on demo at a conference but it didn't work properly there, which I now understand was probably due to misconfiguration.
Most 3D gesture recognition systems use cameras and very complex image recognition, using for example using OpenCV or in devices like Leap Motion. However these require a great amounts processing power and hardware so they tend to be expensive, need a lot of energy to run, can't be hidden away and lighting conditions are critical. All huge obstacles for systems that would ideally go anywhere. The MGC3130 promises to change that.
Well let cut to the chase, I'm happy to report my initial expectation were well and truly blown away. The MGC3130 is a fantastic little chip and the Hillstar kit does it good justice. In particular I love the module nature of the kit, you can plug just what you need and get the signals you want all via largely independent modules. You can even easily create your own and add them to the signal path. The modularity extends to the electrode as well, you can remove easily remove the provided protective cover, a nice touch.
Essentially the kit comes in three parts: the electrode board, the MGC3130 board itself that generates the e-field signal and interprets the readings from the electrodes and communicates to a host via I2C and finally a PIC-based I2C to USB interface that connect to the MGC3130 board for configuration and readings.
There are cheaper kits out there for 3D gesture sensing also using the MGC3130, such as the Hover, but you can't replace the electrodes on that one and you don't get a USB interface for configuration via your PC. That said if you don't need the full configurability of the Hillstar the Hover is a pretty good piece of kit too. I guess the MGC3130 is just a fantastic chip wherever it's used 
Gerrit already posted full unboxing photos so I'll skip those.
While in operation I measured the TX output of the MGC3130 board with my oscilloscope. This is the signal that generates the e-field:
Sure enough there was the ~100Khz field (the bottom square waveform, top is showing the I2C control signals flying through). So this is really how it works, no alien technology here Which is even more impressive considering the MGC3130 is interpreting minute differences in capacitance measured in picoFarads in real time across 5 electrodes.
I didn't fully test the GestIC APIs since I mostly used the kit with microcontrollers over the I2C bus. I only used the supplied Aurea software for configuration.
That leads us to the negative parts, nothing major, just small annoyances. First the software is Windows only. As a Mac user it's always disappointing when vendors don't see you as a worthwhile target group, but such is the reality of the EE world - fortunately that is changing, albeit slowly. I did run Aurea under VMWare and it worked fine, except for the cube demo which didn't work at all - but that's just a demo, nothing major.
The Aurea software itself was a little confusing, but again nothing new in the EE world. Last week I had to use a Chinese application in some weird windows codepage that didn't render it's own text properly to configure a NAND controller, compared to that Aurea is a walk in the park
While the modularity is great, I'm not sure why they picked 2mm pitch connectors. Board size perhaps? Lower noise? If possible I'd prefer 2.54mm connectors so I wouldn't need to use special cables. This would be particularly convenient when connecting to rapid prototyping stuff like the Arduino. The pins, especially on the electrode board were also very fragile.
Maybe it's just me but I'd also appreciate current measurement test points on dev boards and it would be nice to have one on the MGC3130. Battery power and portable applications are taking over the world these days and it's good to measure how well our power management is doing without wiring up a lot of cables. Same goes for easy access to the Extended Interface port that we need to test the self wake up mode, not sure how we're supposed to use them while connected to the electrode board.
A extra electrode board would also be nice, such as the 1.65 inch circular one. I ended up producing a copy of that one at OSHPark, but surely a small PCB (even 4 layer) with one simple connector can't add much cost to the BOM of the kit.
But again these are minor things, nothing a little tweaking, some soldering or a little PCB production doesn't fix. All considered this is an great kit for an amazing chip and I can't wait to do much more with it. You can read my blog posts about what I've done so far and my future plans for it:
Connecting 3D gesture sensing to devices using Bluetooth Smart
Testing the MGC3130 Hillstar for gesture sensing in shop windows
Building electrodes for the Microchip MGC3130 Hillstar Development Kit
Thank you for reading, please leave any question in the comments.
