Some updates, and some finished units – read on to learn more!
Jacob Flood, November 13, 2019
We’ve been hard at work with the latest prototype batch, getting our tests done so we can send units in for certification and start shipping. We’ve made some great progress this month that we’re excited to share!
The T0 units have brought to light some tweaks, that we’ve diagnosed and implemented to prepare for mass production. We’ll be assembling more units incorporating these changes over the next weeks, to send in for certification in December. We’re still on track to ship our first batch in January.
Read on to learn more about the upper band fabric, the color matching, the electronics jigs, the ANC tuning, the Bluetooth data transfer, and more!
Without further ado, the part we’re all excited about: product pictures!
These pictures are straight from our factory, from the T0 batch we’ve been working on. At the time of the photo some of the parts were not perfectly color-matched, but these pictures are very representative of the final design. We’re really excited about how well they turned out.
The latest prototypes, fully assembled!
We mentioned in our October update that the lowerband cushion in our latest prototypes were too thick and too firm, and were therefore interfering with the electrode contact. Taking a trip to the fabric-wrapping factory, we were able to do some hands-on prototyping with the team there to resolve the issue. We’ve since gone through 4 iterations of the foam, and have found a variant that works properly.
Namely, the changes we made were as follows:
Changing the foam in favor of a memory-foam with lower spring force
Reducing the thickness of the foam
Reducing the thickness of the fabric
Improving the fabric-wrapping method for a tighter, more reliable fit
The samples we’ve received since have passed our inspection, and are confirmed to be replicable in mass production. We think the changes from this process are a net positive for the product.
The upper band has been the most mechanically complex part of the product since the start. One piece of this has been the fabric wrapping: a labor-intensive job that has a low margin for error, and was often difficult to reliably complete from a quality assurance perspective.
When we met the fabric-wrapping supplier, we took the opportunity to discuss the upper band assembly method. We tested a few variants, and concluded that with a higher-elasticity fabric, we could avoid the assembly process entirely by machine fabricating the fabric and simply sliding it onto the upper band like a sock.
A few tests later, we’re now very happy with this new assembly method. The new technique will greatly improve the efficiency of the manufacturing, and reduce the assembly cost of this process. Since higher reliability leads to lower scrap rate, this should help us scale the assembly much faster moving forward.
The upper band fabric, mid-assembly, using the new method.
The silicone sleeve that protects the electronics cabling was found to have some tolerancing issues, which was causing abnormal bending in certain areas once-assembled.
Note that the sleeve is pictured grey below, but is usually black on the product.
The new silicone sleeve. Note that the final version will be black.
We discussed the issue with our factory, and found that the issue resulted from a tolerancing error in the tooling. The molds edited, and the issue has since been resolved – we’re now satisfied with this part.
Currently, the 50 units we are making are being assembled by hand, to help us develop a firm understanding of the best way to test each aspect. The end product of this will be a Quality Assurance (QA) and Quality Control (QC) process to ensure that each product that goes out the door hits our rigorous standards.
A core part of QA/QC are electronics jigs. These are mechanical assemblies that probe the electronics boards at key locations, and run a series of automated tests to validate that the electronics are preforming their function correctly.
In the picture below, you see an example of very early prototypes of those jigs. Incorporated into the design of each of the electronics board are specific pads (test points) used to test the circuit. When installed on the jigs, a number of probes lower onto and touch these test points on the PCB. Inside the clear plastic casing, we install the test electronics containing software that analyze the boards through the test points to validate the function of the circuit. In the case of the image below, the jigs are used to send fake EEG signals into the device, and ensure the electronics properly process the biosignals.
Our prototype electronics jigs, setup to test the PCBs.
A closeup of the electronics jigs testing the PCB.
Once the jigs are validated, several copies of each jigs will be installed on the assembly lines. During the assembly, the team will place each board on the jigs, press a button, and within 10 seconds the system will have evaluated whether the boards meet our specifications. In the case of a failure, an error code will help diagnose the reason for the test’s failure, helping us understand the most common failure points and improve with each manufacturing batches.
Over the past month we’ve been testing out the jigs on our electronics, and preparing them for use in the assembly line. So far everything has been working smoothly – the next step is to assemble the units and get our results!
One of the subtler parts in the headphone is the anti-dust fabric. This small piece of cloth covers the speaker, in order to protect the internal mechanisms from dust particles. Per our factory’s recommendations, we followed standard practice with this piece in order to ensure proper audio quality.
We wanted to share this short anecdote about this part. This month we had to make the last decision on this: how to arrange the lettering to indicate the left and right orientation. Our instinct was to make the letter bold and “filled,” such that it’s as easy as possible to see. Little did we know that this small change may cause secondary effects: blocking the sound, and therefore changing the audio profile of the headphones.
We found this anecdote to be a strangely appropriate metaphor for our experience with hardware design: innocuous changes can have immeasurably complex effects. It really takes a deep and deliberate consideration to all facets of the experience to create a quality product, as this anti-dust fabric shows.
The final lettering on the anti-dust fabric.
Naturally, we decided to go for the low-key “unfilled” lettering. Keep it simple.
One of the later-stage design changes we’ve been experimenting with has been the cushion electrode location. Our goal is to allow for a constant, repeatable, and comfortable contact with the skin that doesn’t interfere with the normal headphone experience. The final validation of this had to wait until we had final units, since the fit is so closely a function of the final tolerancing.
Following many tests, we decided this month to move the cushion electrode from the back of the ear to the front. This change avoids effects from the jawline, which is very differently expressed across different head shapes. We found that we got a more reliable contact on the front of the ear than in the back, across different head shapes, hair styles, and facial hair styles. This region of skin also tends to be less sensitive, and therefore should improve comfort across the board.
Alongside testing the product function, a large part of this month was spent preparing for mass production. A key part of this is ensuring the final CMF – colors, materials, and finish – are ready for our shipped batch. This month we spend a lot of time working out the color matching of the parts.
The challenge for our product is to properly match the color between the plastic and the aluminum parts, to keep the quality feel consistent. Since color-matching is an iterative process, we started by producing reference parts for alignment.
In our case, we started with the plastic ear-cups parts. There are two processes used in the painting: applying the matte paint, followed by oil painting. This two-step process gives the parts a look more similar to metal. You can see the difference between the two steps below:
Step 1: matte paint.
Step 2: oil paint.
You can see below a few samples that we inspected to confirm that the color met our requirements. Next to the samples, you can also see the injection molding machine where the parts were made.
Several ear cups being color-match tested.
The resulting finish is quite strong, and will not scratch easily. The finished part looks like this:
The final painting on the ear cup.
Once the first samples are approved, the real challenge is to ensure the metal part matches the finish of the plastic part. This process requires iteratively tuning the color over several batches, to align them more closely at each step. Suffice to say, this isn’t the first time we repeat the color matching on these parts.
We had to go to the oil painting supplier with a few samples of the unpainted arm, and iterate over the painting method to get proper end result. The primary difficulty with matching metal and plastic parts is that the two materials absorb the paint in a different way – even if you use the same color on both, they will look different next to each other. It requires a bit of a trial and error, combined with knowledgeable engineers, to get it just right. The good news is that once you find the right recipe, you simply copy it for the rest of the production.
For us, that looks like the following!
The result of our color-matching!
In parallel to the hardware development, we worked to finalize the Bluetooth firmware for the product.
In our prior test, we had tested various platforms to transfer EEG data using Bluetooth Low Energy (BLE). However, in collaboration with our factory we recently found additional chip-level constraints that limit the bandwidth, packet size, and RAM allocation for BLE running in parallel to audio. Rather than compromise the EEG’s sampling rate, we decided to move from BLE’s notification-based streaming to a more standard Rfcomm approach using Bluetooth classic.
Baring any further issues, we don’t forsee any resulting changes in user experience, as we’ll still be able to benefit from the simultaneous high-bandwidth audio and EEG streaming.
Finally, with the off-tool headsets now built, we’ve received the most recent ANC tuning results from the factory!
We are really happy with the latest results, as it confirms that the cushion electrode does not cause any noticeable disruption in ANC function. The total isolation is consistently below -20dB across the entire [100, 500]Hz range, with a peak attenuation of -33dB. These results place our headphones alongside the major players in the industry, something we’ve worked very hard to achieve. Great ANC and great audio were a must for our product – with these results, we’re as excited as ever to get your feedback about how working with ANC changes the way your office feels.
That’s all! As always, feel free to share your thoughts and excitement in the comments, and we’ll answer each one over the next few days.
Our team’s hard at work to keep up with our January shipping timeline – we still expect that we’ll be able to get our first batch out before Chinese New Year. Until next time, lots of love!
– The Mindset Team