Sourcing DC-DC Converter Modules From China: Read the Curve, Not the Headline Number

DC-DC converters are the quiet workhorses of any electrified product. They sit on a board doing an unglamorous job — taking 24V down to 5V, or isolating a noisy rail from a sensitive one — and when they fail, everything downstream fails with them. Sourcing them from China is easy in the sense that there are thousands of suppliers, and hard in the sense that the datasheets are written to sell, not to inform. Here is how to read them.

Isolated or non-isolated — decide this before you compare prices

The first fork is isolation. A non-isolated converter shares a ground between input and output. It is cheaper, smaller, and fine for stepping voltage within a system that already shares a common ground. An isolated converter breaks the ground path, which you need when you are crossing a safety barrier, feeding an analog sensor from a switching rail, or meeting a regulation that requires galvanic separation.

Suppliers will quote a suspiciously low price and then mention the isolation only when you ask. So ask first. An isolated module typically costs several times what a non-isolated one does at the same power, and if the manufacturer's isolation voltage is not stated, treat it as zero.

Efficiency curves, not efficiency numbers

Every datasheet has a headline efficiency figure like "up to 92%". That number is the best point on the curve, usually at exactly the load and temperature the marketing team liked. What you actually care about is the curve shape across your real operating range. A converter that hits 92% at half load and collapses to 70% at 10% load is bad news if your product idles most of the time.

Ask for a plotted efficiency curve against load. Reputable Chinese suppliers like MORNSUN and others in that tier publish these. If a supplier can only give you a single percentage, they either do not have the data or do not want you to see the shape.

Ripple, thermal derating, and EMC

Three specs that decide whether the part works in the field:

• Output ripple and noise. Given in mV peak-to-peak, and usually measured with a specific capacitor on the output. If your load is noise-sensitive, this number matters more than efficiency. Ask how it was measured.
• Thermal derating. Almost every module delivers full power only within a narrow ambient window. Above that, output drops. A 30W module rated to 60°C might only give you 20W at 85°C. If your enclosure runs hot, this is the spec that bites.
• EMC. Does the module meet EN 55032 or CISPR 32 on its own, or only with external filtering you must add? Modules that need a pi-filter plus a common-mode choke to pass are common, and the filter components add cost and board area. Get it in writing.

The overcurrent and protection story

Check what happens on a short circuit or overload. Good modules do hiccup mode or current limiting and recover. Cheap ones die, and sometimes take the load with them. Ask specifically about short-circuit protection, overvoltage protection, and thermal shutdown, and whether they are self-recovering. "Protected" on a datasheet is too vague to rely on.

Certificates: who actually holds them

This is where buyers get burned. A supplier shows you a CE or UL file, but the certificate belongs to a different company name, or to a component they resold, not to the module you are buying. If your product needs UL recognition, the UL file must cover the exact model you purchase. Ask for the file number and check it against the model number. Do the same for CE and for any IEC/EN safety report.

For medical or automotive applications, the compliance burden is heavier and the module should carry clear evidence of the relevant standard. Do not let "we meet the standard" stand in for a report.

MOQ, lead time, and stock

Standard off-the-shelf modules from large manufacturers are stocked and ship fast, often with MOQs measured in single units or a small tray. Custom output voltages, special pinouts, or wide-temperature versions need a production run and push MOQ into the hundreds or thousands. Lead time for custom is typically four to eight weeks, longer if a transformer is wound to order.

One practical tip: ask for the golden sample and the production test report for the lot you receive, not just the approved sample. Some suppliers ship a pristine sample and then deliver modules that are a different revision. A test report tied to the shipment catches this.

Bottom line

Buy the curve, not the headline. Decide isolation early, insist on derived data for ripple and derating, verify that certificates belong to your exact model, and confirm the protection behavior before you commit. A converter is cheap relative to the product around it, so there is no reason to gamble on the spec that keeps everything else alive.