Mindset Update - June 2018
June has come and gone, and with it comes another update. This one will focus on our electrode stack!
Jacob Flood, July 6, 2018
We’re well into July, and so it’s once again update time. This time we’ll be focusing on the EEG stack – from the electrodes, to the electronics, and finally the software. This has been our focus for the past week, and we’re really excited about the results.
We’re finished our mechanical optimizations – The last step is choosing the acoustic fabric we’ll be using for the ear cups. The parts have been sent in for EV2, which we’ll receive in 2-3 weeks – at that point, we’ll share a better estimate of our shipping timeline.
The top electrodes have undergone a redesign, which allows the inner piece to be removed while the outer shell stays in the headphone. This will allow for future electrodes to be retro-compatible with the device. The ear cup electrode design has been frozen as well.
In our last update we described some thorough mechanical changes we’d made to the headphones for our EV2 prototype. At this point, we’re satisfied with the mechanical design (which includes the industrial design, the materials, the assembly process, and importantly the audio). We don’t intend to make any more changes in this area before cutting the tooling.
The last change in this area was related to the audio. Following our first few rounds of audio validation, we found that the air leakage through the earcups was a concern. On the industrial design shown in the Kickstarter campaign, the fabric selected by the designers was regular fabric. You cannot simply use any fabric on a pair of headphones - it has to have the right acoustic properties. As we are completing the final “golden samples”, we had to find proper acoustic fabric that matched the look, feel, and requirements of the design.
At first, we considered a blended approach – a traditional ear cup pleather, covered by the softer fabric that our industrial design preferred. This allowed for the comfort and aesthetics of the fabric, while maintaining the acoustic seal of faux-leather.
During this experimentation, however, we found that a family of technical materials called acoustic fabrics performed dramatically better. This material has special properties and coatings that block out acoustic leakage among other things, significantly impacting the sound quality and tuning. We quickly met with a few different suppliers and had initial custom samples made for our product.
In addition to the audio benefits, we improved the comfort vis-à-vis the previous material by choosing a fabric that is soft, and doesn’t feel too warm on the ear. The industrial design’s choice was a natural fabric, which tends to wear out quickly and risked getting caught in earrings. Our final choice is a more tightly-knit, similar-looking acoustic fabric that conserves the look, while improving the product’s function and ensuring comfort. We’re really happy with this choice.
At this point we’re good-to-go on the mechanical portion of EV2 – we’ll be sending in the order shortly, and should have units in hand in ~3 weeks.
Moving away from mechanical design, our focus for the past month has been on the electrodes. In the process of finishing the golden-samples, we decided to make a few changes to the assembly of the electrodes.
To improve assembly and removal of the electrodes – since this was one of our stretch goals – we changed the kinematic structure of the electrode. Before, the removable part of the electrode contained the majority of the assembly: the contact legs, the moving spring, and the housing structure. This wasn’t ideal – removing the electrode, while feasible, was a complex process of detaching wires, and ensuring the twist-contact connected.
In this redesign, the spring mechanism will remain within the headphones, while the metal sensor piece will be attached directly to a spring plate. If the electrode is to be removed or replaced, only the metal bit will be changed. This improves the stability of the whole mechanism – which is crucial for signal quality purposes, and longevity of the electronics. In addition, this more modular design allows for the possibility of manufacturing various types of electrodes in the future, using different materials – this will be useful for academic research, as well as sensor upgrades as the state of the art in conductive materials continues to improve.
As a comparison, below is the previous electrode iteration.
Top electrode, previous design. Note that the entire housing is removable.
Top electrode, new design. The internal piece comes out, separating from the housing.
In parallel, we’ve also been talking to several factories that specialize in medical-grade material plating in order to optimize this end. As the interface between the scalp and the electrode is the biggest predictor of signal quality, choosing the right material – one that has the right electrical properties, while not decaying over time – is crucial. Leveraging insights from experts in Shenzhen that have manufactured medical-grade biosensing devices before has been extremely fruitful.
Given that the electrode circuits are inherently a “stack” – the signal flows through each part sequentially - these kinds of optimizations make a huge difference. Any one component intrinsically effects each other part of the circuit, meaning that each change has to be validated for compatibility with rest of the pipeline. Getting this right is part of what separates a high quality device from a cheap one.
The image below shows our most recent sample of the electrode, which are mounted inside the headphones. We’re slightly tweaking the leg size, diameter of the tip, spring force, and degrees of freedom to improve the electrode comfort while keeping good signal quality. These changes are minor, and can take place in parallel to the DV and PV rounds over the next few months. What’s important is that we freeze the overall design, and we’re very confident that we can do that now.
This modularity improvement will make the subsequent production iterations faster, and should mitigate any future electrode issues. Overall, this is a big step forward for the EEG stack.
Note that gold-plating is only one of the finishes currently under test for signal quality. Medical-grade electrodes can use Gold, Silver or Silver-Silver-Chloride for their electrodes, depending on the application. For our purpose, we are testing all of them to provide long-lasting results and avoid requiring early replacement of electrodes.
EAR CUP ELECTRODES
In addition, we have started running test for the assembly process for ear cup electrodes.
After several tests, going back and forth between different ways of contacting the ears to optimize for comfort, consistency, and signal quality, we ended up deciding on the pad approach: using a conductive fabric pad to keep contact against the ear lobe inside the ear cup.
There are many reasons why this is better – without going into too many details, we have found that this alternative provided the best comfort and signal quality. Key to the design is that the fabric electrodes are made to be movable within the ear cups – because ear shapes vary widely across users, we decided to allow you to adjust it very easily so that it touches your ear comfortably.
We have found that touching on the outside surface of the ear anywhere other than the lobe became hot or uncomfortable over time – a reality we weren’t comfortable shipping with. In addition, having a sensor touch behind the ear where glasses typically sit – our previous solution – often resulted in both comfort and signal quality that was inferior.
As mentioned, we are currently assembling test fabric electrodes with our manufacturer to nail down the assembly. Having run several experiments on the fabric electrodes over each prototype iteration, as well as working alongside several conductive fabric suppliers, we’ve found that our high-end fabrics perform even better than the top electrodes. This month, we added our secret sauce – which unfortunately we can’t talk about for IP reasons – to the sensor to improve their conductivity by a full order of magnitude. We are quite happy with the signal quality that we have been able to reach with those electrodes, a crucial part of the EEG stack.
On the electronics side, we spent a lot of last month focused on the electrode pre-amp boards.
These components are responsible for buffering, amplifying, and filtering the analog signal. Simply put, this component is a crucial part of the EEG stack that allows us to maintain a very high signal quality despite the uncertain environment in which the headphones will be worn (an office, rather than a lab).
The pre-amp boards are one of the only components of the headphone that we are very secretive about – this piece is a crucial part of what makes Mindset so unique. EEG is an inherently noise-sensitive signal, which until now has struggled to exit a lab environment. We’ve spent countless months optimizing the electronics stack to allow for a high-quality, low-cost, high-fidelity signal that can read attentional metrics from a headphone formfactor. The pre-amp board represents a large part of our IP, and therefore we can’t share any development pictures of it publicly – sorry!
In parallel to the pre-amps, however, we’ve been working on the latest batch of Bluetooth boards (seen above). This is the full digital stack that handles all the rest of the headphone logic. Having now frozen details related to data transfer rates, battery specifications, and Bluetooth protocols, the Bluetooth boards for EV2 are currently in pre-manufacturing, and should be ready to assemble in the next couple weeks.
With the electrode evaluation test fully finished, we’ve been focusing all of our efforts on the app UX, and the neuro-adaptive system.
The electrode evaluation test (EET) was a side-priority over the last couple months – developing a series of experiments we can run to validate our electrode’s ability to measure changes in brain-state. This is crucial to compare electrodes empirically, once the simulation analysis and first batch are produced.
We ended up using two tests: occipital alpha during eye closing, and steady state visually evoked potentials.
The first test is a measure of increase in alpha activity (8-12hz brainwaves) that arises in the occipital lobe when your eyes are closed. This is a quick, repeatable test that can confirm the ability for our electrodes to measure a distinct brain signal. As alpha spindles are also a common metric used to quantify attentional shifts, it doubles as a sanity check for the types of measurements we’ll be doing in the final product.
The second test is a steady state visually evoked potential (SSVEP). This test measures a reflexive neurological response in the occipital lobe caused by perception of a controlled-rate flashing light in your visual field. A resonance in the brain under this stimulus causes a spike at the frequency of interest, which mirrors the frequency of the stimulus. Measuring this gives us a clear, quantifiable indication of the quality of the signal measured.
We tested both of these using two data acquisition platforms, and several electrodes – both ours and medical grade – in order to see where we lie in the mix.
Both of these tests have passed the software team’s check, and were used last week by the product team to validate the latest batch of electrodes.
With that our of the way, we’re testing the rest of the app on ourselves nearly-nonstop. Our goal is to hit 105 hours of collected data by the end of this week, and we’re on track to accomplish that. Analyzing this data will give us a final sign-off on our machine learning stack, so we can focus on the neuro-adaptive interface, and the app design over the next months.
Unfortunately, we aren’t yet comfortable sharing images of the app interface – we’ll be working on some key features over the next weeks that we don’t want to present until they’re ready. We’ll be sure to post some more details as soon as we can.
Our goal this past month was to sprint through the final stages of the design of EV2, to get the parts manufactured prior to this update. Unfortunately, as is often the case in complex projects, the complexity rose and we were unable to fully complete this milestone.
We are currently in progress freezing the design, a crucial phase in any development process. As such, we aren’t willing to cut corners at this point in order to get the units quicker – doing another EV iteration is to be avoided at all costs.
Two crucial events must occur in order to validate the EV units and produce the molds, which will become the foundation for the production runs:
The first is the electrode validation – from the mechanical design described above, to the EET we’ve been working on recently, this process has been a crucial part of the EV2 run. This was also a source of delays – since the EEG stack is not something we can risk getting wrong, we needed to ensure the design of the parts was 100% final before moving forward, which ultimately cost us time.
The second is the engineering validation process, which will follow the EV2 batch. Our factory informed us that the barrage of tests they’ll be doing post-EV2 in order to cut the tooling will take 2-3 weeks, and require two full prototypes – far beyond what we’d been initially quoted. This kind of change, while alone isn’t sufficient to cause delay, stacks on the electrode testing and the prototype delays we discussed last month. All in all, we’re behind schedule right now on EV2.
We’re still incredibly optimistic – while the 10,000ft view is sometimes cloudy, the progress we’re making on the factory floor is incredibly tangible. We’re very, very excited to ship you your mindset headphones.
The unfortunate reality is that hardware design is a complex, multifaceted, non-linear process. Unlike software development, where bugs can be resolved in real time, any improvement or experiment frequently takes 3 weeks to produce. What we call a delay is actually an acknowledgement that inventing, designing, optimizing, and mass producing a new technology takes longer than anyone – from us, to our experienced mentors, to the best headphone factory in the world – had anticipated.
That being said, we understand that delays are frustrating. From the inside looking out, it’s hard for us to convey why the non-linearity of design has resulted in not yet shipping our product, despite the crazy hours we’re working. Having been on both sides of delayed pre-order campaigns, we know there’s only one course we can take: tons of hard work, met with an equal amount of transparency and honesty.
Right now, the honest assessment is that until we get the EV2, it’s unfair for us to give a realistic update on our timeline. This isn’t for lack of knowledge – we have a full understanding of the steps ahead. Rather, it’s a lack of information: simply put, we need to know where we stand with the EV2 batch before we can paint a tangible roadmap to shipping that we’re confident we can follow. We don’t want to give a fake assessment that doesn’t reflect a proper understanding of the steps we need to take, and we need EV2 to do so.
We’re working day and night to run, not walk, to the finish line with Mindset. The moment we feel confident to share more detailed information, we’ll absolutely do so. Until then, rest assured that you’ll definitely get your headphones as soon as we can. I promise, it’ll be worth the wait.
As always, feel free to share your questions in the comments below and we’ll be sure to share any additional details!
Lots of love,
- The Mindset Team