Product Update - May 2018
May has come and gone, and with it comes another update. We've got some great pictures of EV2, so be sure to check it out!
Jacob Flood, May 28, 2018
This month we focused on the Engineering Validation 2 (EV2) build, as well as the NASDAQ protocols we were testing. Lots of news to report on both those ends.
Below, we’ll be going through a thorough analysis of the differences between EV2 and our original industrial design. In addition, we’ll give a progress report on when EV2 will be finished, and touch on some of the software tests we’ve been running. Finally, we’ve gotten several questions in our comments and facebook community related to timeline and finances, which we’ll respond to in detail below.
Let’s dive in!
The EV2 design is finished – we’ve sent in half the parts to be manufactured and will be sending in the rest this week. By next update we will have made, assembled, received, and started testing EV2.
Unfortunately, some electronics mishaps related to our data acquisition prototype cost the software team 2 weeks to fix. We’re now back on track with the new prototype, and running our user-experience NASDAQ.
Naturally, timelines are very difficult to estimate – that said, we’re currently still on track to deliver in September. We’ll be able to give a better estimate next month once we receive EV2.
In previous updates, we discussed the recent changes to the design: the goal was to improve comfort and ensure functionality, while making sure that the product remains cheap enough to be viable. At this point, we’ve finished the design and implementation of all these new features, which will be present in our EV2 prototype. These units should turn out identical to the units we’ll be shipping you all soon.
Below, you’ll see a comparison between the product at different stages:
First, the industrial design, as displayed throughout the pre-sale campaign. This was the intended design, prior to any of the changes described in the months following the campaign. The prototypes we built prior to the campaign launching were based on this design.
Second, our EV1 prototype. This version you saw a physical build of in previous updates. This design still used 3 electrodes (despite having 2 in the pictures), and included the complex, flexible mechanism we described in our march update. We built this design to validate the mechanical assembly of the whole headphone, and use it as a bouncing plate for the latest round of design changes.
Notably in this version, when worn the ear cup foam was noticeably too thin - your ear would impact the frame of the cup. The change in curvature of the lower headband necessary for the electrodes to touch the head was also jarring, and left too little space in the middle for foam. These changes were reflected in our post-EV1 version.
Third, our post-EV1 iteration. This design was proposed by the manufacturer, as an attempt at satisfying both the feel and function requirements for the product.
You can see that already, the ear-cup size had grown quite a bit. This was following an acoustic analysis of the original design: in order to satisfy sound quality, active-noise cancelling and ensure comfort for the ears, a size increase was necessary. The downside of this, is that the result looked quite bulky and square.
We also noticed while wearing the headphones that the overall curve of the band was very wide. While they looked great in their closed form, the curvature made them appear quite large when worn, and prevented proper contact of the electrodes. You can see this in an image below - Notice the large gap between the side of the head and the slider of the headphones when worn.
Fourth, is our EV2 prototype. Using our learnings from EV1 and our thorough understanding of mechanical design constraints (refer to blog post of Ergonomics and Anthropometry), we reconstructed the design, making sure at each step that the design and mechanical constraints agreed. This process resulted in the new engineering prototype, which is now being built and assembled. By next update, we should have the prototype in hands, and some results from our acoustic tests.
Each iteration included hundreds of small, subtle, structural changes necessary to move the design closer to manufacturability. Aside from these, the main structural differences from original design to EV2 are as follows:
1) Total number of electrodes was reduced from 5 to 4, removing the center electrode.
2) The thickness of the lower headband was increased. This is made of a compressible material, providing adjustment and comfort to adapt to different head curvatures
3) The band curvature was reduced to improve electrode contact and minimize gap with the side of the head when worn. This results in a more anthropometrically accurate design overall.
4) The pivot was centered on the ear cup to balance the clamping force and ensure proper comfort and achieve better active-noise cancelling.
5) The cup size was increased to ensure the full ear can fit inside without being compressed. This also improves overall sound quality.
6) The controls were moved to the right ear cup, with the except of the AUX port on the left side. This makes more sense from a usability perspective and is easier to assemble.
Our simulations have demonstrated that all requirements for this design are met. In the next few weeks, we will run additional tests on our physical head jigs to validate head fit. Once that is done, we will be ready to cut the molds, and run an initial test batch (DV1).
In parallel, this month we’ve been doing quite a lot of work on the electronics.
There are two main sections in electronic design. First, we have our analog signal processing chain - this includes all the EEG electronics, which ensure that the EEG signal is read, amplified, filtered, and generally uncorrupted until it’s sent to the computer. The second is the digital portion, which controls all communications, audio, microphone, and active-noise cancelling.
For the EEG electronics, all the blocks are designed, built, and functional. There remains still some tweaking to be done to the filters, cabling and PCB shape – these are minor changes that can be done over the next few iteration batches. We will be validating this latest batch over the next couple weeks, following positive simulation results we recently got. In total, we are currently at our 5thfull revision of the electronics.
In theory, the digital electronics should be handled fully by our manufacturer. In reality, we’ve had to spend quite a lot of time supporting them in ensuring the data is handled properly. The main challenge is ensuring both EEG and audio data streams operate in parallel, with no data loss, on all devices – this is a common problem with all modern day IoT devices. In our case, the bandwidth requirements end up pushing our factory’s chipset close to its physical limits. Our manufacturer has now implemented a solution, which in our first tests works fluidly. The next step is to evaluate Bluetooth in various settings to ensure the data is communicated properly in all environments. The last thing we want is choppy audio or EEG when using the device, the bane of all wearables.
All these parts, both mechanical and electronics are either assembled and ready or being manufactured and tested as we speak. We expect two fully functional units, integrating all the changes described above, to be completed by mid-June, before next update. If all major tests are passed, then we will be ready to move to Design Validation (DVT) phase, and design molds for mechanical parts.
Last update, we talked about two major tests we were working on for the software: the electrode evaluation test (EET), and the neuro-adaptive system data acquisition (NASDAQ). This month, we intended to test out each of these two systems – unfortunately, as you’ll see below, problems got in the way.
The first couple weeks focused on testing the EET with our newest electronics. The goal of this process is to create a standardized test for evaluating the signal quality of the entire electronics chain, now that we’ve sufficiently tested each component separately. In theory, this is a simple process that we didn’t expect to take long.
In practice, this process ended up costing us more time than we anticipated. To begin, repeatability in testing ends up being very difficult – since the electronics prototype isn’t in a headphone form factor, we struggled to get the electrodes to make a similar contact each time. Also, when comparing our electrodes to a medical-grade equivalent, setting up the EEG alongside our own rig often took over an hour. Finally, since the tests were running through our yet-unfinished app, small bugs often corrupted the data acquisition mid-stream, resulting in us having to restart a few hours later.
On their own, none of these slowdowns would have been problematic. More significantly however, were the electronics issues we had. The EET was the first time we were running a full data acquisition on the new electronics our product team had designed for EV1. The chipset we were using was ~50% complete – it didn’t include several of the front-end components that are critical to the quality of the signal. Our goal was to test the system as-is, to then compare the results when we finished the full stack. While running the tests however, it became clear to us that we couldn’t achieve any valid results with this incomplete prototype.
The tradeoff between design and prototyping is a constant issue in the engineering process. Product team’s priority is designing the final hardware - spending time defining requirements, simulating designs, evaluating the results, and formalizing the parts that work. In parallel however, the software team developing the app and algorithms are constantly eager to work with the latest prototype, given that the signal quality increases slightly from one iteration to the next. As a result, the product team has to take time periodically to build prototypes, both to test their own design and to provide software team with a new model.
For the EET and the NASDAQ, it was critical that the data we collect be as close to the final version as possible. As a result, in April product team committed to making the 50% prototype described above. As we found out this month, we jumped the gun too soon, and ended up wasting time running tests on a chipset that wasn’t sufficient for this purpose.
Once we realized this, we put the EET on hold. We spent the remainder of the month prepping the NASDAQ, which we’re running on an older, fully-functional prototype. In fact, I’m running the NASDAQ as I write this update, collecting data that we’ll use to design our neuro-adaptive system.
Ultimately, this time was not entirely wasted – we made several strong moves forward this month setting up our cloud infrastructure, synchronizing the GUI with our data streams, and setting up UX elements that will ultimately carry over to the final product. In practice, I’d be lying if I didn’t say that these issues were immensely frustrating. In total, we lost approximately 2 weeks of development time from this.
The silver lining is that as of next week, we’re back on track analyzing the data from the NASDAQ. We’ll be analyzing these results for the next two weeks, following which we’ll make some structural changes to the user experience and run NASDAQ 2. This will last until beginning of august, at which point we’ll run NASDAQ 3: our fully assembled, ready to ship version of the app, in all its glory.
Each update we typically give an estimate of our timeline to production. Given the slowdown on software, and the difficulty with the Bluetooth streaming described above, we’re currently slightly shifted from the schedule we announced in March.
Reading your comments every week, I’m aware that this is a huge priority for you all. As such, I want to give some insight into how we think of these timelines.
The reality is that these schedules are only ever estimates – the engineering process is often messy, unstable, and unpredictable. We’ve had some parts of the design that we anticipate taking one week actually take 4, and other pieces that we anticipate taking a half a month actually only require a single day. When we talk to other hardware companies, to factories, to investors, and to experts in the field, this is the purported “trade secret” that gets discussed most often – even for an experienced company like Apple, with 10x the staff and 100x the resources on each project, it’s nearly impossible to accurately predict how long a design process will take.
We’ve always been fully aware of this. We’re not exaggerating when we say we’re working with experts: our factory Grandsun is one of the top 3 headphone manufacturers in the world, our audio partners Onkyo ships millions of headphones each year, and our investor SOSV is the #1 hardware investor in the world. We consult with these partners every week to share our progress and get feedback on the process, and they’ve helped us design our timelines every step of the way.
I want to be clear: none of this is meant as an excuse. The delays we’ve had to announce have been difficult decisions each time – it’s unfortunate that we haven’t been able to ship the product on our predicted schedule. We’re truly sorry for that. What I hope to convey, is simply that bringing a product to life is difficult and uncertain, and that we’re all working incredibly hard to deliver a high-quality product that you’ll be excited to use every day. This never ceases to be our motivator, and number 1 priority.
All that being said, our best estimate of when we’ll start shipping is still September. The delays this month are minor, and can be recaptured at a later step – they don’t justify a readjustment of our entire schedule. Next month when we receive EV2, test out the whole stack, and finish NASDAQ 1, we’ll have a much better idea of the remaining tasks, and will share a more detailed breakdown of the final steps.
As always, feel free to share any questions in the comments, and we’ll be sure to provide more details any way we can!
One of our backers recently asked about our company finances. The implied concern was that we may run out of money, and be incapable of delivering the products as promised. We wanted to quickly address that concern: we are currently fully capable of delivering the products as planned.
As a general rule, we don’t share our company finances public. This is common – sharing detailed financial information would hinder our ability to raise investment, to seek out retail contracts, to sell products, and to find suppliers.
That said, by backing the campaign you earned the right to transparency. What follows is a rough breakdown of how Kickstarter hardware companies are made.
During a well-run Kickstarter, the cost breakdown typically looks something like this:
- One third of the money goes towards building the Kickstarter. This includes the 8% Kickstarter fee, the design of the page/video, the media necessary, and the marketing costs to let backers know about your product.
- One third of the money goes towards the cost of goods sold (COGS) for the first production run. This is essentially the cost-per-unit to make the product, at scale.
- One third remains to fund operations and development.
That last third, then, has to pay for all company expenses during the time that it takes to build the product. This includes employee salaries, prototype costs, travel to China, legal fees, molds for the units, and other company expenses. The honest truth is that in the majority of cases, this amount is not enough to sustain a company throughout production.
So what do you do? Two things: increase the money in and decrease the money out.
Firstly, we continue to sell units at thinkmindset.com beyond the campaign for this reason – the more people join our cause, the more resources we have to make it happen. And since we’re selling at a higher price now and aren’t paying Kickstarter a fee for purchases, a larger percentage of each sale can go towards operations. In addition, we can raise investment – we’re actually closing our seed round as we speak (side note: we still have a small amount of room left in this investment round, if any of you are interested! Send a shout to firstname.lastname@example.org).
Additionally, we strive to keep costs down. We pay ourselves only the minimum required for living - in Montreal, the cost of living is ~5x lower than San Francisco, which cuts expenses dramatically. We exclusively use budget airlines, we negotiate down every contract we get, and we’ve committed thousands of dollars of our own money for tools and supplies.
For this reason as well, we don’t offer refunds. In order to get the per-unit cost lower, we had to guarantee an order quantity to our factory. Any change to that quantity would end up coming out of the operational portion of our funds, in addition to the transactional fees we paid for processing the payments. This would ultimately hurt our ability to cover the remaining costs necessary to get to the mass production stage. The end result is that refunding a small number of people would hurt our ability to deliver the product to the remaining backers, a compromise we aren’t willing to make.
As it stands, we’re in a tight, but healthy financial standing – we’re fully capable of delivering all of the devices, and we intend to do so as quickly as possible. We’re 100% committed to delivering an incredible product, won’t let anything get in the way of that happening.
That’s all folks!
Comment below if you have any questions, and as always we’ll be sure to answer them.
Lots of love,
- The Mindset Team