From a Toaster Oven to Production: Tips and Tricks for Scaling Hardware

The first 200 SwitchBlox boards were assembled by hand. They were among the first boards BotBlox ever made, and we had no proper lab space or production equipment, let alone a reflow oven. But our founder, Josh, had the designs and the parts, so he found a way. He modified a spare toaster oven to reflow the solder, then tested every board himself.

A lot of hardware companies start this way. With limited money and equipment, you make do with what you have. At that stage, the priority is getting a working product into customers’ hands.

A 24 year old Josh soldering a prototype his Dad’s garage

Then customers started ordering ten at a time. The process that had made SwitchBlox possible could no longer keep up with demand.

SwitchBlox was the first board we scaled, and it taught us how much changes when you move beyond a handful of units. At hundreds or thousands, component availability becomes critical, small assembly problems become costly, and testing each board by hand takes more time than you can spare.

First ever hand-soldered version of SwitchBlox

Hand-packed by Josh

Here’s what we learned about designing for production.

First, get the product into people’s hands

Before making a thousand units, find out whether people want to buy the product.

Start with a handful. These early units will be expensive and labour-intensive to make, and you may even lose money on them. But they help you work out whether the product is worth scaling and catch problems before you commit to a larger run.

So how many is “a handful”? That depends on your confidence in the design and how sure you are that people will buy it.

For our new products, we’ve found that 20 to 50 units usually gives us enough for internal testing and evaluation by selected customers. For a more intricate design, we lean towards 20 because there’s more that could go wrong. If the board is simpler and we’re confident in the design, we’ll make 50.

Losing money on those first units can be okay, provided you have a credible estimate that the product will be profitable at your intended production volume. That first batch helps you establish demand and understand what needs fixing before you invest in a larger run.

Planning the initial design

Your first few boards may look different from the ones you eventually send to a manufacturer. Even so, a few early design decisions can make scaling up much easier.

Start with component selection. A part that’s easy to source for ten boards may be much harder to find when you need a thousand, especially if it comes from a small supplier or regularly goes out of stock. Choose suppliers with dependable availability from the outset. Switching components later can mean redesigning and testing parts of the board you thought were finished.

Sometimes a specialised part or material is essential. Our radar antennas, for instance, require specialised high-frequency fiberglass because standard fiberglass attenuates radar signals unacceptably. It’s expensive and difficult to source, but the application justifies that complexity. Where a readily available material or component meets the requirements, we’d choose that instead.

Gaps between glass fibers in substrates with different weave densities.

Image: Fiberglass weave styles 106, 1080, 2113, and 2116. “What Is the Fiber Weave Effect in a PCB Substrate?”, ZM Peterson, Northwest Engineering Solutions, 2020.

Then think about how your product will actually be assembled. At BotBlox, the small size of our Ethernet boards is central to their design, so we’re always pushing to make them smaller. But packing components tightly together can leave too little room for assembly equipment to work.

Take transformers, for example. Electrically, two transformers might work perfectly well placed close together. During assembly, though, the machine’s placement tool could knock one out of position while placing the other. The layout needs to accommodate the equipment that will build it.

Talk to your assembler early. Can their equipment place the components where you’ve put them? Is there enough room to inspect the solder joints and make repairs? It’s much easier to move a transformer in the layout now than to fix its placement after production has started.

Use customer feedback to your advantage

Some problems only become obvious when customers start using the hardware.

Set aside a few units from your first batch for selected customers to evaluate in their applications. Alongside evidence of demand, you’ll get feedback on how the hardware performs outside your own test setup.

We do this with all our new products. Customers put our hardware through conditions we can’t always recreate internally, and sometimes wouldn’t think to test.

In one case, customers told us their cables were disconnecting under vibration. They’d resorted to gluing them in place. The electrical connection worked, but the connectors weren’t holding the cables securely enough for their application.

We changed the connector, which solved the problem. Fortunately, we’d only made about 20 units at that point, so we could make the change before committing to a larger batch.

Different setups, unexpected environments and long periods of continuous use can expose problems your internal testing misses. Getting that feedback early lets you address those problems while the number of affected units is still small.

Know which compromises are worth making

Customers choose our products because they are unbeatably small. Naturally, we push each new design to be as tiny as possible. But making a board smaller can increase costs elsewhere, so we need to know which reductions are worth pursuing.

Shrinking a board introduces complications. There is less space to dissipate heat. Routing becomes tighter, making signal integrity harder to manage. Components become more difficult to place, inspect and repair. These problems are solvable, but addressing them takes time and money. Some design choices can also reduce manufacturing yield.

Say you save a couple of millimetres by placing two components very close together, but the change sharply increases the number of boards that fail during manufacturing. The product is now smaller, but each usable unit costs more to produce. Are those few millimetres worth it? Sometimes, yes. Sometimes, no.

It comes back to your customers. What do they need, and what are they willing to pay for it? If space is tight in their application, a slightly smaller board may justify a higher price. If those few millimetres make little difference, they may prefer the cheaper option. Make the tradeoff based on what your customers actually value.

Build testing into the production process

Design validation and production testing answer different questions. Validation checks whether the design meets its requirements. Production testing checks individual manufactured units for faults.

Even when the design has been validated, a manufactured board can have a bad solder joint, a damaged component or an incorrectly fitted connector. As volumes grow, you need an efficient way to catch those faults.

At BotBlox, we use a bed-of-nails fixture: a set of spring-loaded pins that make contact with test points on the board. This lets us power and electrically test each board without manually attaching a collection of leads every time.

An electrical test doesn’t catch everything. A board may have visible damage or solder blocking a mounting hole. A connector could be fitted incorrectly in a way the test does not detect. These defects can make the board unusable even if its circuitry works, so we pair functional testing with visual inspection.

Don’t ignore testing in your cost estimate. A few extra minutes per board may seem negligible in a small batch. Across a thousand units, they add up to days of work.

Plan how you’ll test and inspect the board while you’re designing it. That includes leaving accessible test points for the fixture and enough room to inspect the areas most likely to need attention.

Scaling up

Once the first batch has given you confidence in the product and evidence of demand, increase production gradually. That might mean moving from 20 boards to 50, then 100, 500 and eventually 1,000. There is no fixed recipe. The right pace depends on the complexity of the design, the cash available and what you’ve learned so far.

Before committing to a larger order, take stock of the previous batch. Are customers buying the product? What faults appeared during manufacturing or use? Does anything need to change? If you revise the design, validate those changes before making hundreds more.

Larger batches generally bring the unit cost down, but they also tie up more money in stock you still need to sell. Let customer demand guide your batch size. Saving money per board is only useful if you can sell enough boards to justify the larger order.

For BotBlox, that process started with a modified toaster oven and the first SwitchBlox boards. The setup was basic, but it let Josh build a working product, get it into customers’ hands and establish that there was demand. That gave him a reason to invest in the next stage.

Each larger batch should be a decision you can justify with what you’ve learned from the last. Keep resolving faults and refining the process as demand grows, until you can reliably build the quantities your customers need.