Microchannel cold plates — the hard part is keeping them open

By hydraulic diameter. Below about 1.00 mm is generally called microchannel, roughly 1.00 to 3.00 mm is mini-channel, and above 3.00 mm is a conventional milled channel. The labels matter less than what they cost you. Going finer packs more wetted area into the same volume and raises heat transfer, but pressure drop climbs far faster than the channel shrinks, so the pump has to work much harder for each step down. Most instrument plates land in the mini-channel band because that is where the thermal gain is still worth the pumping power, and true microchannel is reserved for genuinely concentrated flux — laser diodes, power semiconductors, that class of problem.

Because thermal performance is the part everyone gets right and blockage is the part that comes back. In a closed loop, every particle stays in the loop: shredded PTFE tape from tapered threads, braze flux residue, machining swarf that was never flushed out, corrosion product from a mixed-metal circuit. All of it migrates to the narrowest passage, and in your system that is the inside of the cold plate. The symptom is not a leak. It is one device running a few degrees hotter each month until it trips, months after commissioning, which makes it one of the hardest faults to attribute. Designing the plate is straightforward; designing the loop so the plate stays open is the actual engineering.

Face-seal, never tapered-thread-with-tape. A tapered thread such as NPT seals on the thread flanks, so it needs PTFE tape or liquid sealant, and both shed debris on make-up and again on every service disturbance. In a closed loop that debris circulates until it finds the narrowest passage. A parallel BSPP thread with a bonded washer, or an O-ring in a machined seat, seals on a face instead: nothing consumable sits in the flow path, and the joint can be broken and remade for service without adding anything to the circuit. We machine the ports that way as standard and will say so on the drawing.

Three options with an honest ranking. A bolted lid sealing on an O-ring in a machined groove keeps the plate serviceable, which on an instrument that may need flushing in the field is often worth more than the last few percent of performance. A soldered or brazed lid removes the gasket from the maintenance list and takes higher pressure, at the cost of never opening it again. A welded lid goes highest on pressure and is the least forgiving of a mistake in the channel layout. For quantity one to ten on an instrument, bolted is usually the right answer and we will say so even though the sealed versions bill more.

It matters more than the plate material itself. Aluminium 6061 is the default: light, machines cleanly, conducts about 167 W/m·K. Copper roughly doubles that and earns its cost where flux is concentrated under a small device. 316L stainless is for aggressive or contaminated coolants. But the deciding question is the circuit: an aluminium plate in a loop containing copper or brass is a galvanic pair, and once the glycol inhibitor ages, the aluminium dissolves to protect the copper — corroding from the inside, which also generates exactly the particles that block the channels. Keep the wet circuit to one metal, or nickel-plate the aluminium, or run properly inhibited coolant and change it on schedule. 7075-T6 at about 130 W/m·K buys nothing thermally, and 304 at about 16 W/m·K is a structural choice rather than a thermal one.

Cleanliness is a deliverable here, not housekeeping, because the passages that catch tape debris in service catch swarf in manufacture just as easily. Cross-drilled intersections are deburred, bores are flushed, and ports are capped before packing. Pressure or leak testing is available on request, with the method and hold time agreed at quotation rather than assumed. If your loop has a filtration specification, tell us and we will match the flush to it — there is no point flushing to a coarser standard than the filter you intend to run.

No, and it is the same answer we give on the cold plate page. Volume programmes with CFD sign-off, friction stir welding lines and vacuum brazing furnaces are what the large thermal houses sell, and they will beat us on both price and paperwork at that scale. Our lane is ones and tens to your geometry: the plate your instrument actually needs, the plate that replaces one nobody makes any more, or the prototype that has to work before the volume decision is taken. When your programme reaches the size where a thermal house makes sense, we will tell you.