2018 · Nest (Google)

One of the major projects I worked on at Nest was the Nest Cam IQ Outdoor – a premium wifi enabled security camera for homes and businesses. We built the camera optimizing for the following features:
During the early development stages on this camera I was working closely with the Audio team – we wanted to make sure to stake out enough real estate for a best in class 2 way talk system. This required both sufficient back volume for a loud enough speaker with good bandwidth to compete with street noise, and enough microphones with good placement to pick up / isolate the conversation from someone standing close to the camera.
We started by building up several speaker box back volumes and drivers with acrylic enclosures, a tried and true method for quickly prototyping speaker boxes. Our first stop was the ID team. Industrial design often favors many small holes over fewer large ones, but small holes act as a high frequency filter and can make a speaker sound muffled. After a few iterations, we settled on a radial pattern of 1.5mm holes in a 1.5mm thick housing, a good balance of appearance and acoustics.
Next we aligned with Product Marketing on how loud the speaker needed to be. Research on street noise levels suggested designing for a ~70dBA environment. That is quite loud, but if users need to converse clearly with a package delivery person outside a city apartment, it is a reasonable threshold. For the demo, we chose the roof deck of one of our offices that was within earshot of the highway and was measuring about 68dBA. With this setup, Product Marketing agreed with our speaker performance targets, and the final design used a 38mm driver with full duplex capability.
Working with the other requirements on the product, we settled on the design shown below with an enclosure made up of a die cast metal top (for heat spreading), a 20% glass filled PC bottom (to match the Aluminum’s CTE), and a donut form factor to allow the hinge wires to pass inside. This box was sealed to the enclosure by a series of O rings and gaskets during assembly.



When discussing requirements for microphones, one important element for our algorithms was the ability to recognize not just how loud a sound was but where the sound was coming from. This would allow us to focus on audio of interest that was being delivered from near the camera, without having to rely on exactly how loud it was. We used multiple microphones and some DSP algorithms to accomplish this. Two microphones allows you to understand which side of the device sound is coming from, three microphones allow you to locate a sound source in 3 dimensions, and 4 microphones gives you a huge improvement in the quality of the signal, with some redundancy on the 3 microphone system. After much discussion with marketing and ID, we were able to agree to a 3 microphone system with 2 on the camera face and one on the bottom side. For 2 way talk, it is important to isolate the microphones from the speakers as much as possible, so the performance of the 3rd microphone is slightly inhibited by it mounting to the main housing, where the speaker is also mounted. Microphone port diameters and lengths were based on waterproofing and simulation, and finally validated to ensure that any filtering from the port geometry would not inhibit the algorithms too severely. In the end, this feature came out great and 2 way talk works seamlessly as long as your wifi bandwidth can support it.

After surveying existing security cameras, we realized that they were a pain to install, difficult to set up, and usually very invasive to the home. As a result, we wanted a product that required minimal holes drilled into your house, could be mounted quickly and easily, and pointed in any direction the user desired. This combination of goals led us to a single USB-C connector in the base to plug the camera in, a mounting mechanism that allowed mounting of the device to the wallplate with a single push on, and a hinge that could articulate freely and without limit once installed.
The hinge for the camera, as described above, needed to be able to be rotated in a nearly unlimited range to make installation as simple as possible. After many rounds of brainstorming and different limiting mechanism designs, our team decided that we would allow unlimited rotation through the use of a slip ring. This would also make the hinge unlikely to break; if the camera was struck by something like a football, the hinge would simply slip and sustain no damage, and the user could reset the direction of the hinge once again.

Since this hinge both needed to be buttery smooth but also high friction force, we used a combination of low friction materials and a 250N spring. The friction pairs for the 3 degrees of freedom are anodized aluminum against acetal, anodized aluminum against Teflon fiber impregnated Acetal, and die cast aluminum against a die cut PTFE ring. The resulting hinge is a pleasure to interact with – well damped, smooth, and also high enough force to stand up to vibration, wind, and any birds that might land on the camera.
These stringent reliability requirements fit within a bold industrial design required a lot of innovation from our team. The IK07 test is pretty intense.
To survive this sort of impact, we did two main things. First, we created a die cast aluminum part that was glued into the main housing. This had the added benefit of being a great heat spreader to cool our SOC – the aluminum insert part was a vital part of our thermal pathway that tried to spread heat across the largest surface area of the outer housing in order to maximize cooling.
Second, to make sure the camera could not be knocked off its mount, we worked with ID to grow the base by 1mm and used that space for a 1mm thick, full hard stainless steel plate to anchor the locking mechanism. To build confidence in the design, our team went at the mounted camera with baseball bats and a hammer, and couldn’t knock it off the wall.
If you haven’t had the privilege of using an IPX6 tester, you might think that this test is easier than IPX8. After all, it’s a lower number. But dig in a bit and you’ll see that the test is performed with a fire hose – the power in this stream will actually snap pieces of plastic and water has a nasty habit of wriggling its way into even the tightest enclosures.
This camera had 15 ingress paths and all of them had to be sealed, across all devices. Working with our reliability team at Nest, I built several custom fixtures to allow them to test one seal at a time. There were several strategies that had to be employed:

While this realm was left to the industrial design team, I am really proud of the collaboration that occurred between our mechanical team, the industrial designers, the manufacturing and supply engineers, and our factory team. This product comes out of the box looking and feeling friendly, rugged, and high quality.





