Mindset Update - August 2018
August has come and gone, and with it comes another update. This one will focus on tests we’ve ran on our EV2s.
Jacob Flood, September 17, 2018
This month we’ve been pushing hard to test out EV2, so we can plan the next steps for our DV run. We have a lot of technical content below about the good and the bad with our latest prototype, as well as a deep dive into the current audio specs.
The EV2 units are here and assembled! Having tested them thoroughly, there are three minor issues (which we’ve now solved) and one major issue (which we’re currently solving). We’re aiming to have the tooling cut in October, and the DV batch by the end of the year.
There are plenty of great pictures below, so check it out!
Last month we talked about some of the different parts of EV2: the armbands, ear cups, clips, and wiring. Shortly after the post, we assembled our first full EV2, and spent the last month running tests and redesigning pieces of the now-assembled units.
While it’s not perfect, assembling and holding the EV2 prototype has been an incredible experience. This update will focus on some of the subtler features that still need work, and the engineering steps we’ve taken over the past month.
Overall, the prototype feels great. The few issues we found were:
Electrode casing defects, which caused improper spring action of the electrodes
Improper ear cushion fabric and thickness
Lower band foam was too stiff
Difficult cabling assembly, which is costly and causes sliding issues for the arms
Right now, the top three issues have been resolved – these were merely kinks from the assembly. The part that still needs to be resolved is the cable assembly – having the equivalent of two products in one means double the number of wires, which is problematic for the sliding arm adjustments. This is the current challenge we are actively working on.
Now, as explained in previous updates, one of the important tests of the headphones was to put it on our specially designed head jigs, as well as test on real subjects. The result of this test has showed that the current band is now able to fit on all head jigs, and meets our requirements. You can see an example of the headphones when put on the jigs below.
The photos above show two different examples (a small and a large head). You can notice two of the issues previously mentioned:
The electrodes were not compressing into the band due to a defect of their casing. This has been resolved.
The grey lower band was too stiff and did not compress enough when put on the jig. This did not have a significant impact, but will result in a better fit.
Below are a few additional pictures, when worn by real subjects.
Although the cushions were too thick during the current test, the headphones were shown to properly fit on tested subjects. Decreasing the thickness of the ear cup foam will further improve the fit. In addition to the decrease the foam thickness, we may change slightly the shape of the arm to better match the angle of the head on the sides – a small change that won’t affect any other part of the design.
Since the beginning of the project, there has been considerable work done on the ear cushions - multiple competing factors such as comfort, appearance, and sound quality, have made it a difficult optimization. Below are various iterations of the ear cushions.
The original industrial design had fabric which was intended to be like micro-fabrics. It turns out that this fabric isn’t very comfortable on the skin – it’s rather rough, and can get quite hot. The fabric itself also reacted poorly to frequent use. You can see what the original fabric currently looks like on the same device as the above photo:
You can see in the picture that the fabric is very worn out, the initial vibrant colors have faded, and the texture has changed. Developing the Engineering Prototype, we went through several iterations.
This version used fabric that was significantly more comfortable, while still agreeing with the industrial design. However, the breathability of the fabric caused two issues: the sound quality was very poor due to air leaks, and it was in breach of a patent on active noise-cancelling technology. As a result, we could not use this fabric directly.
We then tried with a different type of fabric. Our goal was to test acoustics using two layers of materials: a bottom layer that wasn’t breathable, and a top layer of regular fabric. We wanted to confirm that we can achieve a cushion with comfortable fabric like the industrial design, while keeping the physical properties of a non-breathable fabric. The conclusion of that test was that while it worked well, it increased the cost of the device.
For this version, our manufacturer suggested a fabric that looked a lot like the industrial model and had good acoustic properties – on par with the pleather alternatives. The downside is that these acoustic fabrics are normally used on surfaces of speakers, and are not intended to directly touch the skin. It ended up not being very comfortable.
We ran acoustic tests on those samples, which turned out quite well. This option remained a potential backup, however we didn’t think the comfort was high enough to make it into the final design.
After significant setbacks in the comfort of the earpads, we considered reverting to regular pleather to ensure proper sound quality and comfort. We made samples of those to test with our headphones, comparing to the fabric options above.
This change, however, implied a drastic change in design of the headphones as compared to the original design. The pictures below illustrate the comparison – we were very reluctant to compromise the industrial design in this way.
Additionally, it is important to note that the upper band where the electrodes are mounted cannot use pleather – it would wear out too quickly in contact with the hair, and start peeling in under a year. An added challenge with the pleather is to solve the issue of stretch marks, as seen in the picture below:
In parallel to the fourth version, our manufacturer prepared an additional sample made from a comfortable fabric. The sample above is not the proper color, but it does reflect the texture. We combined this fabric with a non-breathable layer underneath to obtain the sample. This is very promising as it is both comfortable and it meets our acoustics requirements (to be discussed later in this post). In the end, it looks like we’re close to the finish line when it comes to the ear cushions.
After internal tests, we’ve also determined that for comfort reasons, it is not necessary for the cushion to be as thick. We are preparing additional samples with Version 5 cushion fabric in a smaller size to test comfort. Having smaller cushion pads would greatly improve the look of the headphones when worn, and may improve the fit on the head, without causing discomfort from impact of the ear on the ear cup casing. The images below show the impact of reducing the ear cushion thickness - let us know your thoughts about the cushion and fabrics in the comments!
Other than the slight appearance issue of bad tolerances (which is an easy fix) we are happy with the control buttons. We wanted to avoid a crisp clicking noise that would resonate loudly in the ear cup when you click the buttons - for this reason, we were extra careful in the design of the controls. We’re happy with the result: the clicking sound of the button is firm, yet quiet and warm.
We’re currently debating internally whether we modify the buttons to add a location feature. It turns out that while using the headphones, it’s difficult to know which button you are currently touching, which leads to accidental miss clicks. While it’s a small detail, we expect to fix it in the next batch of prototypes.
As previously mentioned, the cabling assembly is currently our biggest challenge. Because EEG signal are so small and sensitive to noise, the design is quite intricate, are requires many cables. Our challenge is to fit all those cable without compromising the function or appearance. In the previous updates, we showed a few pictures of the cables passing through the arms. Since then, we have done quite a lot of work to figure out the best way to proceed.
The major issues related to the number of cables are as follows:
Assembly complexity (more assembly means higher cost)
Wire stiffness limits lower band compression (the foam under the electrodes)
Wire stiffness limits ear-cup pivot rotation
Wire stiffness prevents sliding of the arms
As we work to solve these problems, we wanted to get your thoughts on the following option: our manufacturers are suggesting passing the cables inside the headband through the bottom of the clip, rather than inside the arm band. This change would allow for smooth slider and ear cup adjustment, and simplify the final assembly. The downside, is that this would make the cable visible outside of the headphones, between the ear cups and the band.
What’s your opinion: would having the cables pass outside the arm band be a good idea? Do you have any alternative suggestions? Leave your thoughts in the comments!
AUDIO, ANC, AND AIR FLOW
After assembling the last prototype (EV2), we were informed that the current design may lead to ANC instability due to air-leaks. According to our manufacturer, the way the control panel (where the buttons are) is designed could lead to ANC leaks. This needs to be tested further, but we’re considering our options as a precaution.
Below you’ll find the suggested change from our manufacturer: make the control panel wrap all the way around the ear cup, instead of only on the bottom. Naturally, we prefer to mitigate changes to the industrial design – sticking to the design aesthetic we promised in our crowdfunding – but we’ll be testing this option against our current design.
Let us know what you think about the change!
ACOUSTICS AND TUNING
For many months now, we’ve had requests asking to show some sound curves from the acoustic tuning. Nearing the final design, we now have some interesting data we want to share.
Before we dive in, it’s worth noting that sound curves are not evidently indicative of quality – we’ll probably be writing a longer blog post about audio design later. Here are a few points to keep in mind while looking over the details:
There are no perfect sound curves. Each pair of headphones have their own sound curve, and there’s no obvious “best”. While there are some useful metrics and goals to aim for in the design, there are many instances of headphones with “poor metrics” that sound great. It turns out that acoustic tuning is as much an art as a science.
The consequence from (1) is that in the end, what matters is how the headphones actually sound and feel, rather than what they looks like on paper. For this reason, one of our validation steps is to have someone at ONKYO referred to as the “Golden Ears” review the prototype and provide suggestions for improvement.
The following results are preliminary, based on the tests performed by our manufacturer. They have not yet been validated or tuned by our audio partner ONKYO. Final audio tuning should be similar, however.
The curves are shown using our latest ear cushion samples (version 4 & version 5). This is important to note as the type of materials for the cushion will have influence on the sound.
We’ve simplified some of the explanations for brevity – let’s dive in!
Total Harmonic Distortion
The THD curve measures how different the sound is between input (what it should play) and output (what it actually plays). Generally, lower is better. However, in the real world, there is no perfect system – there’s always distortion, even if very small, that slightly colors the sound. The three curves above represent different tests and fabrics conducted with EV2.
Most headphones will attempt to reduce distortion below 1% around 1kHz. Premium headphones (including ours) tend to reach distortion levels below 0.2%, which indicates faithful reproduction of the sound.
You also want to avoid spikes and variations across the range. These are caused by subtle resonances in the earcup, both due to the design of the system (the materials used and how well it’s tuned) and simply the physics of headphones (like the resonance frequencies of the drivers and cups). Designing and attenuating these spikes helps avoid unnatural changes to music.
Finally, you may ask why the distortions are higher at the low frequencies (below 100 Hz)? This is a problem that all headphones must face: lower frequencies attenuate less quickly and resonate strongly, and therefore it’s harder to control distortion in that range. A good rule of thumb is to keep low-frequency distortion below 10%.
Above, you can see several frequency response curves. The pink curve is the reference chosen as a baseline to tune our unit. The Green and Red curves are the actual tuning of the latest engineering prototype. Blue and black are other tests we were running.
The frequency response curve illustrates the amplitude you will hear a given sound, across the spectrum of frequencies perceptible by the human ear. An important part of tuning a headphone is the iterative process of making mechanical, electrical, and acoustic changes to the ear cups in order to achieve a desired frequency response.
In our case, we spent some time at the beginning of the design process testing out several headphones, in order to see what fit best – the right middle between high fidelity and natural sounding profiles. Having tested the audio (and tried it out ourselves!), we’re really happy with how it turned out – we’ll be diving into more detail about this in an upcoming post.
We will be ready to send one unit for sound validation by ONKYO in Japan shortly. After this is done, there will be some slight tuning adjustments, at which point the acoustics will be ready.
In our last updates, we talked about some of the software projects we were working on – the electrode evaluation test and the neuro-adaptive system in particular. This month, we’ve been focusing more strongly on the app itself: the desktop interface where you’ll interact with Mindset, and with your data.
In particular, we spend August working on:
User login– being able to create profiles, and log in to save your profile info
Authentication– making sure your login is secure
Database– setting up the cloud system for storing EEG data
Dashboard– allowing for creation of widgets that calculates and displays insights on the data collected
Hardware integration– making sure the streaming from the headset is smooth, and reconnects in case of issues
We’ll be spending the next month building out more of this app-centric functionality. In parallel, we’re working with a UI design firm to help us create the experience around the app – taking the framework and UX that we’ve built and tested, and designing a visual interface to best display the information to users. This will be followed by several test iterations, which will span for the next years as we optimize the experience.
We’ll be sharing mock-up app interfaces the moment we have some. Until then, we’re seeking feedback on the kinds of widgets we should build: what kind of information would you want to see in the app? What kinds of tools would you want made available through the interface?
Please post your suggestions in the comments!
As a final note: thank you to everyone who reached out to be part of our beta testers. We had way more demand than we expected (the hundreds of emails was almost overwhelming!), so I apologize that we haven’t been able to answer all of you personally. I want to express that we haven’t forgotten about you - when we get our beta units, we’ll make sure to reach out to each of you to schedule some beta testing.
Thank you so much for the support – I can’t wait to get the products into your hands!
That’s all folks! Please share your thoughts and feedback in the comments – we really appreciate hearing your thoughts on different parts of the app, how you intend to use it, what other tools you’re already using, all of it. The whole team reads those comments, so keep them coming!
As always, send any personal questions to firstname.lastname@example.org, and we’ll do our best to give you a prompt reply.
Lots of love,
- The Mindset Team