San Francisco is an unlikely place for a manufacturing business, much less a business that makes printed circuit boards (PCBs). The city, a center of dot-com wealth in Silicon Valley, is better known for disruptive, software-based technology that has changed how the modern world functions, and less so for the production of PCBs.
But here we are. Tempo Automation can be seen from the U.S. Route 101 overpass, taking 43,000 square feet in a nondescript white warehouse building on a grungy, industrial pocket between SoMa (South of Market) and the Mission, a neighborhood known as Potrero. The gleam of the high-rises of the Financial District are close enough, as are the sources of venture capital (VC)—and thousands of startups with bright ideas—products that are smart and often electronic, yet without the means to produce them.
Highly automated PCB production is nothing new. The Chinese trading site, Alibaba, lists 29,000 such PCB assembly facilities, many of them offering instant quoting and customer friendly interfaces, and all of them eager to do business with the world’s electronics firms. Tempo’s advantage appears to be that it is built for speed, is committed to small production runs (less than a thousand pieces), and is next-door—not separated by oceans and time zones. With China under suspicion for spying on its citizens, inserting hidden code in software, and copying IP enough times that entrepreneurs in other countries think twice about sending them their ideas, U.S. firms are starting to think twice before doing additional business with Chinese companies. In addition, when there’s a rush to market starting with a test batch of a product, time is everything—cost be damned. When a startup loaded with VC cash needs a couple hundred devices to demonstrate appeal or salability for the successful round of financing for its fitness wearable, Internet of Things (IoT)-enabled baby monitor, or another next big thing, who has the time to wade through thousands of PCB manufacturers across the globe much less certify them? Besides, most of PCB manufacturers deal with large volume production—and you are not there. Not yet.
Tempo is in San Francisco to service Silicon Valley’s startups, at least the ones whose software needs hardware to run on, whose production runs are still not considered mass production, and whose habits and expectations have little patience—or means—to create and maintain a facility for electronics manufacturing. To think like a startup, you have to be a startup. It is a culture that orders their meals on DoorDash. They want to order their PCBs the same way. They want to order it online, have it delivered to their doorstep not long thereafter, and have no humans to interact with. That’s the modern method of ordering poke bowls and PCBs.
Push-Button Manufacturing for PCBs
The design engineer who finishes their PCB design would like nothing better than to push a button that turns the schematic on their screen into reality. Isn’t all the information already there? Isn’t everything specified? Now just make it and send it to me. Spare me the drama, the sourcing, the bidding, the supply line that is no longer intact (thanks, COVID-19), the parts unavailable because some factory in China is closed, the capacitors that are in a container stuck in some harbor.
Tempo will take care of it all, says Tempo’s CMO Brady Bruce. Well, almost all of it. You still need to be the electrical engineer with the design. A mechanical engineer with an idea and a back-of-an envelope sketch of the next big IoT-ready gadget will not suffice (I tried).
Thou Shalt Not Wait
The investors and young entrepreneurs of Silicon Valley ache to disrupt each and every industry with innovation. In nearby Fremont, Elon Musk has succeeded in disrupting the automotive industry with fast and beautiful electric cars. A tractor pulls a trailer that uses AI to assess the condition of a specific plant in the field and delivers a precise dose of nutrients—its aim to disrupt farming. We have the companies that disrupted the taxi industry. Restaurants, disrupted by COVIDS-19, may be saved with apps that deliver their food to customers. Why not take any mature industry and disrupt it with an app, with immediate gratification and customized products?
How about a startup dedicated to push-button manufacturing right in our neighborhood? How did that escape our attention? We meet Bruce so he can fill in the gaps in our knowledge.
Next door, we have two startups, says Bruce, who is Zooming with us from his home office. Tempo will optimize the PCB design for manufacture and for electrical performance. We can do autorouting, too, he says.
Tempo was founded in 2013 by Jeff McAlvay, Jesse Koenig, Katherine Scott, and Shashank Samala, all of whom were frustrated by the inability to develop new electronic products quickly. The founders looked into the market, paying attention to the way that PCBs were assembled for electronics, says Bruce. With a little bit of seed money, they rented the space near downtown San Francisco and began to build the robots.
Tempo has reached $74.6 million with a series of funding rounds, most of it from venture capital plus at least two manufacturers: Lockheed Martin and Hitachi Metals.
What follows is the rest or our conversation with Bruce.
So, why San Francisco?
People do ask, “San Francisco, are you nuts?” There are a couple of reasons to be here. A), you want to be where the really great engineering talent is, especially with those regarding software and B), also the hardware. Basically, the people we want to hire were here.
What was Tempo’s initial reason for existence?
The initial goal was to make robots that would make PCBs, but the company pivoted. We found there were some really fantastic robots associated with PCB assembly already from a number of companies around the world. Our contribution was to find a way for all of the robots in a modern PCB assembly shop to communicate with each other. And to automate not just a machine but to automate all the machines in the factory in sort of a unified system that allowed them to be auto programmed, that allowed all of them to communicate with each other about their status and that the quality of the boards that they were producing at each stage, and then for the system to learn over time and improve.
They made the pivot about a year in to focus on software and automation—software in particular. We’re a hybrid. We are as much a software company as we are a PCB assembly factory. And we develop the software that would allow third-party robots to function better in harmony and provide better results faster for engineers who were working through prototypes and new product introduction. Then we needed to build a factory that could actually do it.
A physical PCB would be the crucial last step. How do you do it?
At Tempo, it is that all of our engineers are in downtown San Francisco. They’re in the same building as the factory. The engineering team also sees that factory as its lab because they’re constantly monitoring the data that is spun off by the normal operations of the factory and looking for ways to improve its operation. We are able to reintroduce improvements on a daily basis into the operation of the factory.
As a company, we’re focused on prototyping and low volume manufacturing. Our customers may be in the prototyping phase. They have a new design for a PCB for a satellite, autonomous vehicle, a medical device, a launch vehicle, a jet aircraft—whatever it might be. Our job is to build them a set of prototypes—as many as they want—and deliver it to them quickly, as perfect a representation of the design they gave us.
A company still needs electrical engineers to come up with the PCB design created with their own EDA software, correct?
Yes. Our customers are by and large electrical engineers working with Mentor, Cadence, Altium … others—the usual suspects. But they do the design work and present it us.
Do you optimize the design for manufacture [DFM]?
We do provide plug-ins and DFM guidance to make sure the products are manufacturable. Because of the accumulation of knowledge we’ve gathered over all the previous builds we’ve done, we’re able to make recommendations, where warranted, back to the design engineer and say, “This will be difficult to manufacture because of this choice you’ve made for trace widths,” or “This part you’ve selected is no longer available anywhere in the world, but here are two alternative components that would work there.”
DFM is not 100 percent automated, but it is enhanced by software. We have people also looking at the designs, aided and abetted by our software.
What distinguishes Tempo from other printed circuit board assemblies (PCBAs)?
We diverge from others who also build prototype and low volume PCB assemblies. Most take in Gerber files. That’s a 40- or 50-year-old representation of a PCB. We work with the native digital output from their design application and the IPC-2581 file format. Once we receive their digital file format of their design, our software goes to work.
Our software assists in the preparation of a quote. It looks at the design and the bill of materials (BOM). Its estimations are based on other designs we’ve done. We know what it will take to build that board. We’ve got data from databases of the trustiest North American component suppliers. Our software finds out who has those components in stock. Based on our experience, we make probabilistic assessments about who’s most likely to get us those parts by 8:00 a.m. the next morning. There will be some parts that will take longer, but our goal is typically that if [we] get a design in by midafternoon, we work to have the parts in by 8:00 a.m. the next morning.
Tempo handles the sourcing of all the components on the board?
Yes. We’re a turnkey manufacturer.
Tempo does not store or warehouse any components. You depend on just-in-time delivery?
Yes.
What else happens after a customer provides you with a board design?
We develop a digital twin of that board. It’s a photo-realistic digital twin of that board, a complete 3D representation of the board. We call it the Fabsim, for fabrication simulation. It is indistinguishable from the finished board. We’ve challenged people. 90 percent of the time they can’t guess which is the Fabsim and which is the actual board that came off the assembly line.
When we provide feedback to the engineer based on the design for manufacturability study and include the Fabsim, it’s fascinating how much an engineer will pick up, even after having lived with the design and been in deep with the electronics, how much new information they get from seeing this photo-realistic image of a finished board. They see things they hadn’t seen when they were designing it, and they can make design changes before manufacturing begins.
How is the board presented to the electrical designer?
We transmit all of it. The legacy companies will be transferring this information back and forth. It’s a conversation between the manufacturer and the design engineer with emails, phone calls. Instead, we provide a unified interface on our site. The information we share with the customer comes through this unified interface. All the files are there, all the communication takes place there. It’s all done online in one place. Because when people are making phone calls, sending emails, etc., it’s easy to lose the thread. We’re aiming to improve not just the speed and the quality of the output but also the customer experience.
All in plain view, like ordering at a restaurant with an open kitchen. You can see what you ordered being made?
Or like Domino’s. Our order tracker is like their pizza tracker. A very similar experience. They can actually see in what phase of the process their order is. The design has been ingested. It’s completed DFM. You’ve agreed to these things. These parts are in. We’re waiting on these. In the next phase, it’s now in the solder-paste machine, it’s now in the pick-and-place machine. They can actually sort of watch their order move through the smart factory.
So, we’ve ordered the parts, we’ve performed the DFM, we’ve done the back-and-forth through the interface. They’ve seen the digital twin. That Fabsim becomes very important at the last step of the process, because once we’ve all agreed on the design and the final version of the Fabsim has been created, it is that ideal representation of the design and the intent of the board against which all the boards that come off the factory are going to be judged.
The Fabsim is the gold master. All boards are compared to the gold master all the way to the last board in production. All the boards that match it are considered acceptable. We apply software in an environment where fatigue and routine can dull the senses.
Can you describe the automation from the beginning?
When parts are ordered and they arrive, they come into the factory and are scanned in with our software. We call it TempoCom.
TempoCom is watching because it knows the BOM for that particular order. It’s been waiting for the last part to come in. TempoCom will find the next production line that’s got an open spot and it program it. It takes a couple of minutes to get the production line started.
That’s in contrast to the way other companies do it. All of their machines, from the solder paste to the pick and place, they put all of the components on the liquid solder on the boards, do the optical inspection and the rest of it. They typically require a couple of hours—sometimes three or four hours with somebody sitting in front of them and typing in the instructions for the next build.
We’ve gone to the manufacturers of all the systems that we use in the factory and got the best, most modern machines we could find and asked for access to their low-level APIs. There are two benefits of that. One is that we can extract information from the machine at its operational level and we can also program the machine. TempoCom goes about the process of programming all the machines in the production line, telling them what they’re supposed to do. The entire instruction set and the parts are then loaded into carts, which plug into the various machines in the factory. And at this point, the production line runs itself.
Like a lights-out factory?
We still need people to manage the robots. But as one of our leads in the factory said, “At other companies, I was the robot. At Tempo, I manage the robots.”
It is the job of the people in the factory to manage the processes, looking for alarms. Their activity is actually managed by the same software that has programmed the production line. They carry mobile devices which tell them where they’re supposed to be in the factory, at which point of the day, what they’re to be monitoring and so on. And if an alarm goes off, it immediately appears on their mobile device.
This orchestration of people and robots, it comes courtesy of the digital approach we’ve taken to manufacturing from the outset and our own software. During production, the machines are chattering back to our private cloud—highly secure inside the factory.
How can you automate quality control and inspection?
TempoCom is monitoring each job as it moves down the different production lines inside the factory. It’s watching. If a machine makes an error, TempoCom can intervene. It will understand the error and make sure it’s not repeated.
As a board moves from the solder paste machine through the pick and place machine. It’s a wonderful machine. The same machine manufacturers may have except without our programming it with software, so there are no people sitting with keyboards introducing errors. After the solder hardens, the components are locked in place and then move on to the automated optical inspection. Again, it’s a machine found in many places, but in most places, without the automated optical inspection which will compare each board that comes off the production line to the gold master, the Fabsim model, the agreed-upon model in software, visually accurate, a predictor of exactly what the board’s supposed to look like. Boards may have imperfections introduced at various points in the manufacturing process and they will be rejected.
Our automated optical inspection machine has a camera moving around on an XY gantry at very high speed, looking at all the critical elements on the board—the orientation of components, the solder paste, everything. Then it moves into the x-ray machine where it’s directed by the same software to look for other critical areas in the board.
All of that, start to finish, has been directed by software that began with the ingestion of a digital design file rather than by people poring over a design, working to understand it and then working to program the production line. The result is we’re able to produce boards faster and get them back to customers more quickly than other manufacturers because our factory is so efficient, and the efficiency’s built in to the up-front, to the DFM study and the parts ordering, which is automated.
We have a similar initiative. We call it push-button manufacturing. We’re on a mission to try to make that happen.
Our ultimate goal, too. We’re not there yet. But we have big ambitions. We think of it as handing the keys to the factory to the engineer. The engineer is their design program, and just like their word processing program, they hit print and it sends the job to the printer and the printer takes care of it. There ought to be a manufacturer button.
We part but we have united, Tempo and engineering.com, both committed to making manufacturing as frictionless as possible, as easy as pushing a button at the end of the design—and waiting for the part to be delivered.