DSL (Datasound Laboratories Ltd)

Switching contract manufacturers is rarely a simple decision. Not because of the process but because of the risk. You don’t move a production run of hundreds or thousands of boards without proof, and that proof usually comes at a cost. Small sample runs are expensive, and in many cases, they act as a barrier to making a change at all. So teams stay where they are, even when there are concerns. We removed that barrier. A free PCBA sample run, so you can validate quality before committing to volume, because changing supplier shouldn’t feel like a gamble.

Meet the team: Jack Thompson, Purchasing and Quotation Assistant. "The best preparation for tomorrow is doing your best today." In Jack's role, that is not a motivational quote. It is a job description. Purchasing and quotation in electronics manufacturing is one of those functions that is invisible when it works and very visible when it does not. Jack reviews requirements, prepares accurate quotes, and navigates a supply chain that moves fast and changes constantly. His job is to balance cost, availability, and quality, three things that rarely pull in the same direction, while building the supplier relationships that make reliable sourcing possible in the first place. One of the harder parts of the role is tracking down components that are difficult to source. As obsolescence accelerates and lead times shift unpredictably, the ability to find alternatives, manage supplier options, and deliver accurate information to production teams on time is what keeps programmes moving. For Jack, purchasing is about more than placing orders. It is about understanding the market, responding quickly to change, and helping turn a design into a working product. Outside work: online gaming, and time with his fiancée and baby daughter. Link in comments.

We need to talk about what "rugged" actually means, and how to tell if a supplier's hardware actually qualifies. "Rugged" is one of the most overused terms in industrial hardware specification. It appears on datasheets for products that have never seen a production floor, a vehicle installation, or a field deployment. Here is what rugged actually requires, and how to verify it: Operating temperature. A genuine industrial range is typically -40°C to +85°C for extended temperature variants. "Wide temperature" and "industrial grade" mean nothing without the numbers. Ingress protection. IP54 is dusty office. IP65 is splash-proof. IP67 is temporary submersion. The deployment environment should determine the IP rating, not the marketing copy. Shock and vibration. MIL-STD-810 testing is the reference standard. If a supplier cannot cite the test standard and the result, the claim is unverified. Lifecycle support. Industrial deployments run for 7 to 10 years minimum. A hardware platform without a committed supply and support window of at least that duration creates a mid-programme obsolescence risk. These are not difficult questions to ask. They are just rarely asked before the hardware is specified. Link in comments.

We have joined the Sizewell C Supply Chain network. Here is why. Nuclear electronics does not forgive shortcuts. Components that fail in a consumer product get replaced. Components that fail in a nuclear environment create problems of an entirely different order. The reliability bar is not higher. It is categorically different. That is why supply chain decisions for projects like Sizewell C carry so much weight. Every electronics partner in that chain needs to demonstrate not just capability, but process rigour, traceability, and a track record of building things that last. We have joined the Sizewell C Supply Chain network because this is exactly the kind of work DSL was built for. Thirty-five years of electronics design and manufacturing. A 5-year warranty on every production unit. Full component traceability from source to assembly. Made in Britain. If you are working on electronics requirements for Sizewell C or other high-reliability energy infrastructure projects, we would like to be part of that conversation. Link in comments.

If you are an engineering manager who has ever had a batch fail under thermal cycling, this is for you. A defence electronics client came to us after exactly this situation. Their previous manufacturer had introduced a defect that only became apparent under thermal stress in the field. Not in inspection. Not in functional test. In use. Three months of rework. A delayed certification. A supplier relationship that could not be salvaged. The root cause traced back to solder voids beneath QFN pads, undetectable by the 2D inspection system their previous manufacturer was running. Within six weeks of taking on the assembly, we had implemented a revised test regime, cleared the backlog, and produced documentation sufficient to support the certification review. The product shipped. The programme recovered. If you are currently dealing with a field failure or a quality issue that your manufacturer cannot explain, the problem is almost always findable and almost always fixable. What it requires is a manufacturer with the inspection capability and the process discipline to find it. Link in comments if you want to talk through a re-sourcing situation.

You don’t bring manufacturing in-house without a reason. For us, it was a pattern that built over time, costs increasing four or five times during the component crisis with no clear explanation, suppliers unable to justify what had changed, and in one case, a build where the design was altered and the boards failed catastrophically. At that point, the risk was no longer acceptable. So we took control. Design and manufacturing, under one roof, with full visibility over cost, quality, and process, because when you’re accountable to your client, “we don’t know” isn’t an acceptable answer.

There's something grounding about a colleague disappearing for a couple of weeks and coming back with photos like these. Oliver Simpson, our Junior Account Manager, recently trekked to Machu Picchu.  It's a bucket list challenge for good reason. Thin air, long climbs, and one of the most remarkable sights in the world waiting at the end of it. We're proud of you, Oliver. Brilliant effort.

Edge AI is no longer a pilot technology. In 2026 it is being deployed at the point of production. The hardware decisions being made now will define the next decade. For years, AI inference happened in the cloud. The data left the machine, went to a server, came back with an answer. Acceptable latency for many applications. Unacceptable for anything safety-critical, real-time, or operating in an environment without reliable connectivity. The shift to edge AI changes the hardware equation fundamentally. The industrial PC is no longer a terminal. It is an inference platform, running computer vision, anomaly detection, and predictive maintenance locally, on the machine, in the environment it operates in. What this means for hardware specification: AI inference performance is now a baseline purchasing criterion, not a premium feature. Security has moved to hardware level: firmware protection, self-encrypting storage, authenticated boot. Long lifecycle support is non-negotiable. A platform that cannot be supported for 7 to 10 years is not suitable for industrial deployment. The hardware specified today will be running production lines, medical systems, and industrial automation environments in 2035. That is not a purchasing decision. It is a programme decision. Link in comments.

5 myths about electronics manufacturing that are costing OEMs serious money. Myth 1: "All PCB assemblers work to the same standard." They do not. The inspection regime, solder joint criteria, and documentation requirements vary enormously between assemblers. Two manufacturers can both hold ISO 9001 and produce product of fundamentally different quality. The certificate does not tell you what the process catches. Only the process does that. Myth 2: "ISO 9001 certification means consistent quality." Certification means a quality management system exists. It does not mean that system uses 3D AOI, runs three inspection stages per board, or maintains full component traceability from goods-in to finished assembly. The badge on the footer and the process behind it are two separate things. Ask about the process. Myth 3: "Offshore manufacturing is always cheaper." Per unit, often yes. Per programme, frequently no. Quality escapes at volume, lead time failures, re-sourcing under pressure, and IP exposure do not appear on the original quote. The businesses that make the UK versus offshore decision purely on unit cost are often the same businesses rebuilding their supply chain eighteen months later, at a cost that dwarfs the original saving. Myth 4: "A 5-year warranty is a marketing claim." Only if it is not backed by a process. A warranty is a financial commitment. No manufacturer offers five years on every production unit without confidence in their inspection regime, their component sourcing, and their build quality. If your current assembler does not offer it, that is worth asking about. Myth 5: "Design and manufacture are separate problems." The most expensive mistakes in electronics product development happen in the gap between them. A design that cannot be manufactured as specified is not a manufacturing failure. It is a communication failure between two teams that should have been talking from day one. Keeping both under one roof changes what is possible and what gets caught before it costs money.

You have just sent your design to manufacture. The quote comes back higher than expected. The lead time is longer than planned. One component is on a 52-week lead time and was not flagged until now. None of this happened at manufacture. It was locked in at schematic stage. This is the part of electronics design that most teams underestimate.  The decisions that feel small at schematic, a part chosen for convenience, an architecture selected without modelling power consumption, a tolerance set without considering what it constrains downstream, are the ones that define what the finished product costs, how long it takes to get to market, and how much it costs to change when something goes wrong. Here is where it happens: Part selection. A component chosen without checking long-term availability is a hidden programme risk. When that part hits end-of-life mid-production, the choice is a costly redesign or a last-time-buy that solves this generation but not the next. Availability and alternate sourcing are schematic-stage decisions. Architecture. The choice between an FPGA and a microcontroller, or how many power rails a design carries, shapes every downstream decision. Revisiting architecture at layout stage is expensive. Revisiting it in production is a different project entirely. Tolerance stack-up. The clearances defined at schematic determine what is possible later. Impedance control, differential pair routing, return path planning: these cannot be retrofitted. They need to be in the design intent from the start. A free PCB Design Health Check reviews your design at the stage where changes are still cheap. Component availability, architecture decisions, design for manufacture: caught before layout, these cost a conversation. Caught after, they cost a programme. Link in comments.