We’ve made huge progress in 2019, and we’re excited to show you pictures of our first production batch!
Jacob Flood, December 23, 2019
This past month has been incredibly exciting – we’ve been recording data from our latest batch of headphones, and so far things are looking very optimistic. We’re really excited to share our progress with you before heading into the holidays.
We’ve got lots of great pictures of our latest batch, and the data coming through the devices. I hope you enjoy!
Our batch is finished, and it looks incredible! We’ve included some amazing images of the products hot off the production line, as well as our unbranded packaging.
This update we went behind the scenes into our fabric factory, where we show the process we used to assemble the ear cup and lower band foam. We also show some EEG data from a few experiments we’ve run with the latest units. We also update our timeline: we’ll be producing the next batch in January as expected, but likely shipping closer to February.
Finally, we’re looking for backers interested in receiving the first units, and helping us debug our product and app. If you’re interested, please read the end of the update, and email email@example.com!
At long last: our first production batch is complete!
Following all of the testing of each of the individual parts and subassemblies last month, this month we were able to put together our first full mass manufacturing batch of headphones!
Having scrapped a few parts, we were able to assemble 34 headphones in total. The most crucial aspect, the EEG, was functional in all units except 2, where defects in the cabling were found. Several glitches on audio, ANC, and mechanical fit were also diagnosed and corrected on a few units, but no major issues were discovered.
Having diagnosed the issues on the failed units, we can comfortable begin progressively scaling to larger production sizes for our next batches. As we mentioned last update, this batch will be approximately 100 units, and will include the first units shipped to backers.
We’re incredibly excited to finally have our units in-hand. Behold, in all their glory!
The latest prototypes, fully assembled!
A finished prototype, next to a PCB being tested.
The units look great, and feel even better. We’ve been able to get feedback across our office and partners, and the response has been overwhelmingly positive. It’s a great feeling to see the final production coming together so smoothly.
Each of these units were hand-tested across a range of mechanical, acoustic, and EEG-related functionality. Because EEG technology in this form is novel, we’ve had to design our own tests to validate the EEG quality in real time in the assembly line.
The upper band fabric, mid-assembly, using the new method.
When each device is fully assembled, it’s place on this jig. There, a script is run that sends a known signal through each of the electrodes, reads the output from the headphone, and confirms that the EEG signal chain accurately processes the signal at each stage. In conjunction with the PCB jigs that we described last update, this quality assurance process will confirm that each headset that goes out the door adheres to our quality standards.
While there are still some minor things to fix to improve assembly time and overall finish of the product, this is by far the best run we’ve done so far, and we’re excited about the results. At this point, we have begun ordering materials for the next few manufacturing runs – approximately 1500 units to start. This will enable us to scale up production as fast as our production quality can allow, and avoid delays due to long lead-time components. Batteries can take as long as 45 days to order, so preparing in advance is crucial.
In parallel, we’re working with suppliers to improve the finish on some units. Subtle changes in the assembly process can help us smooth our wrinkles in the fabric, kinks in the wires, and other small surface finish imperfections. With each subsequent batch, our factories will get better at ensuring a fully repeatable and seamless process, which will help avoid any future issues that could hinder batches from going out the door.
Of the 34 unit batch, the majority are currently being used for internal testing. In particular, there are two types of tests we are currently conducting:
Reliability tests and certifications. These include mechanical tests (drop-tests, clamping force tests, temperature and humidity cycling, etc), electronics tests (RF emissions, bluetooth signal strength, etc), and acoustic tests (audio curve, ANC curve, etc).
Internal EEG tests and app tuning (EEG signal quality, neuro-adaptive music system, brain training system, Bluetooth data transfer, etc.)
These tests will continue through the holidays and into early January. Our goal is to finish the testing in time to kick off the next batch in late January as we intended – we expect that we will be able to achieve this goal.
Finally, we have been working on the structure and aesthetic of the final packaging. While we aren’t going to share the artwork publicly, we wanted to share sneak peak of the box without the coloring & print!
The new silicone sleeve. Note that the final version will be black.
We’re really excited about these pictures – please share your thoughts about the images and excitement in the comments!
Naturally, getting to 34 full units required many steps! We shared a few of the subassemblies in our last update – in this post, we wanted to walk you through the remaining steps to getting to a full unit.
Our prototype electronics jigs, setup to test the PCBs.
A closeup of the electronics jigs testing the PCB.
In our last update, we focused on the individual parts, and the color matching of the metal and plastic. With this batch, the plastic and metal parts look incredible – just try and guess which parts are metal and which aren’t!
The pre-assembled band is assembled by our core factory, before being shipped to a factory that specializes in fabric-wrapping. There, the ear cups, upper band, and foam pieces are assembled, wrapped, and integrated into the product.
The various fabric pieces required are laser cut before being sewed and assembled by hand. To this day, fabric work remains one of the most difficult materials to automate production, and often requires a very dexterous combined with a set of specialized molds to assemble repeatedly.
The final lettering on the anti-dust fabric.
In the picture above you can see some of the pre-cut fabric pieces used for the batch we assembled. These were later sewn together and assembled on to a piece of memory foam to become the ear cushion.
Step 1: matte paint.
While we were at the fabric wrapping factory, we also made our golden samples of the ear cushions to ensure the quality was acceptable. The goal, ultimately, is to minimize the fabric wrinkles on the surface of the ear cup. The golden sample serves as a reference product to which future production batches will be compared during the quality assurance process.
Several ear cups being color-match tested.
Once the ear cup was assembled, the team tackled the lower band. The lower band’s fabric is molded into its shape before adding the electrodes and structural pieces. Pre-shaping the fabric to the size and shape of the foam in the headphone’s resting position attenuates any creases that may arise during use.
The final painting on the ear cup.
In our recent trip to the fabric wrapping factory, we were able to produce the golden samples for the upper and lower bands, agreeing on quality standards for each of these parts.
Finally, once the lower band is assembled, the upper band is assembled using the tube-sock method described last update. This subassembly is then sent back to the main factory, where they integrate the electronics, the ear cups, and perform the final quality assurance validations. The pictures below show the subassembly ready to be shipped back to our main factory.
The armband off the mold, after post-processing, before painting.
The ear cushions are assembled onto the ear cups at our main factory, where the audio is also integrated. We’ve been very impressed with the audio quality of the units that have come off the assembly line thus far.
The result of our color-matching!
And with that, the product is ready to test and use!
With the latest batch of headphones, we’ve been able to sample the EEG data coming through the final electronics and form factor. While similar to our previous experiments, these tests help give us a sense of the signal quality we’ll be able to reliably achieve while using this product in the wild.
We wanted to share some preliminary results from this data with you!
One of the quick-and-easy ways to validate the signal from the electrode is a blink test. In this experiment, we wear the headphones and watch to see whether the electrodes can detect the electrical activity produced by your eye muscles when you blink. As muscle activity is higher-magnitude than brain activity, these responses are often very clear in the data from merely visual interpretation, and don’t require sophisticated machine learning to parse.
Below you can see data from our upperband sensors (C3 & C4), filtered, and referenced to the right ear cup (A2). With virtually no processing, you can see very clear electrical spikes at the beginning and end of the 4-second window due to blinks.
This experiment demonstrates that the signal acquired from the electrodes match in phase and amplitude in response to common signals – a requirement for EEG devices. This also means that we will likely be able to use blink-rate as a reliable feature for measuring focus levels, as blink-rate is well known to correlate strongly with cognitive engagement.
Continuing on, the next graph shows another example of muscle activity: jaw clenching. The first half of the segment represents a continuously-held jaw clench, followed by raw brain activity.
As the size of the muscle increases, the electrical signal generated increases as well. In the case of your jaw muscle, the signal produced is significantly stronger than the underlying brain activity. Over time, we will be working to develop algorithms that filter these effects out from the signal in order to continue monitoring the focus levels during these segments. This signal may also be used as a feature in our analysis, as jaw clenching and facial tension are correlates of stress levels, which themselves are related to focus.
Eyes open & closed
Finally, the last experiment we will share demonstrates a neurological effect that comes when you close your eyes. When you close your eyes, your brain produces a spike in alpha wave (8-13Hz) activity in the occipital cortex (the back of your brain). This quick experiment serves to show that the electrodes are measuring brainwave response, and not simply muscle activity.
The two graphs above show a frequency graph (amplitude of signal at each frequency) for the top two electrodes. You can see a sizeable increase in amplitude in the ~8-10Hz range during the eyes-closed condition, demonstrating the expected alpha response. This experiment has been very repeatable using our headphones, which is a great validation of our signal quality. We can also use this effect as a reliable measure of fatigue, since a similar alpha response often occurs when you become tired, even if you keep your eyes open.
These small experiments are fun ways to show the effects of environmental conditions on brain activity, and demonstrate our product’s ability to measure these neurological phenomena. Over the next months we’ll continue to perform mini-experiments (most notably: SSVEPs, P300s, and other go-to EEG experiments) and will eventually share the results in a white paper showing our headset’s utility as a low-cost EEG research tool.
If any of you are interested in the product for these purposes, or know anyone who would like to be involved in this process, please reach out to firstname.lastname@example.org!
In previous updates, we have already talked about the Active Noise Cancellation (ANC) tuning for this product – the results so far have been very positive. With this first mass production batch finished, the ANC tuning curves have been completed and validated. We don’t expect any major changes from now until you get your unit.
As a result, we wanted to share the results of our ANC tuning with you before the holidays.
The graph above shows the combined ANC + Passive Noise Isolation attenuation response. Passive isolation works better at higher frequencies, which are naturally absorbed by the ear cup fabric. The quality of the isolation depends primarily on how well the headphones create a seal around your ear. To contrast, active noise cancelling is tuned to prioritize frequencies between 100Hz and 1000Hz, which is the range where passive isolation is weakest.
When we combine both types of isolation, we get the following above. You can see that in the low frequency spectrum, the isolation hovers around -30dB. Measured in terms of perceived sounds, this means you can expect a reduction of volume of roughly 90% at low frequencies (10dB is equivalent to cutting the volume in half). This level of attenuation places our product in the same category as the other high-end ANC products on the market, something we’re extremely proud of.
Last month we also made a big addition to the product, adding an audio pass through mode accessible via the function button on the headphones. In this mode, external sounds will be amplified and fed through the headphone speakers, so that you can hear your surroundings despite the passive isolation. We think that this will enable a wider range of experiences, including making calls, and running EEG coaching sessions, and conducting research experiments.
When the headphones are turned on, they will default to ANC mode – this is optimal for deep work sessions. By double clicking the main function button, you’ll be able to cycle between three audio modes:
We made this decision after much feedback from you, our backers, about the ways in which you intend to use the device. So thank you – your support, suggestions, and feedback continue to improve the product experience, and allow us to create the perfect work headphones.
For the past 3 months, our timeline has been to produce the DV batch by November (which we achieved), and then to produce the subsequent batch in January to start shipping. We’re currently still on track with the timelines associated with this goal, which we’re quite proud of.
This year, however, Chinese New Year is extremely early. Chinese New Year holidays officially start January 25th, however we’ve recently learned that our factory will be ramping down production speed as early as January 10th, and re-opening only at the beginning of February. We’re on track to start the next batch of ~100 units in January, but we think that it’s more likely that we’ll be shipping the first units in February, after the holiday rather than before.
We’ll be spending this time during the CNY holidays testing the first few units coming off this batch, to ensure that each of the aesthetic tweaks we applied from our recent batch yield fruit. We will aim to share these units with a few backers in January if possible, to get external feedback during this time as well.
This brings us to the final point: we are looking for beta testers interested in receiving the first few early units. It’s very likely that the first few units that go out will experience issues that will take time and effort to debug – it’s extremely important to us that the first few backers that receive their units are willing to help us with this debugging process. We understand that this kind of support isn’t for everyone, so we’d like to extend a call to any backers who are particularly interested in helping us in this way.
You are comfortable receiving a product that contains small imperfections
You are willing and eager to use the headphones and the desktop app every day, despite the occasional glitches.
You are willing to chat with our team periodically to share your feedback about the product, app, and experience.
You are interested in helping us debug issues with the headphones (software knowledge is a plus!)
You are interested in using the external API to stream data from the device outside of the app, and willing to help us develop support for this functionality.
We can’t wait to ship our first units, and we’re really excited to get your feedback to continue improving the experience. If you think you fit this description, and you would like to receive a unit from our first batch, please send an email to email@example.com, and we’ll reach out shortly!
Thank you so much to everyone for your continued support this year!
Here’s wishing all our incredible backers a great holiday season, and a happy new year. Best wishes, and looking forward to an incredible 2020!
Lots of love,
– The Mindset Team