The Invisible Dust Cost: Why Your Filters Lie

Adam Wiggins runs the North American data center segment for Camfil, and at Mining Disrupt Dallas he talked about a problem miners cannot see. The particles floating in a sunbeam are about as small as the human eye can register, around ten microns. The dust that reaches a mining operation is mostly far smaller. As Wiggins put it, "99% of all particles are below one micron in size." Those particles get into miners through the filters and around the cracks, settle on the heat sink and keep air from pulling the heat off.

Miners throttle or need repair and hashrate falls. The number most buyers use to choose a filter is its MERV rating, and Wiggins argued that number is not enough. MERV stands for minimum efficiency reporting value, and it comes from a controlled lab test. He said, "The test is in a lab environment and lab environments aren't like West Texas." A lab does not see variable air flow, variable particle counts or wildfires.

There is a second problem. Many filters carry an electrostatic charge, and that charge helps them hit their rating on day one. As dust, dirt, pollen and sand build up, the charge dissipates and the efficiency goes with it. An appendix to the standard test, called Appendix J, discharges the filter with isopropyl alcohol and runs the test again.

Wiggins showed two filters that both tested as MERV nine the first time. After discharge, one only tested as a MERV seven-A. Camfil's own filter still tested as a nine. His advice was to ask suppliers for MERV-A, which he called cradle-to-grave performance, and he said Camfil guarantees its MERV-A filters will test to that rating from day one to day end.

Buyers see a four-dollar filter next to a six-dollar one and buy the cheaper one. He said, "that's assuming they all perform the same way. And absolutely, they do not perform the same way." The filter price is only part of the total cost; the bigger driver is the energy spent pushing air through the filter, then labor, then how often the filter is bought, shipped in and hauled away. In his three-year comparison the longer-lasting filter was changed three times instead of nine and used roughly a million fewer kilowatt hours.

In a thirty-megawatt data center the air filter is the smallest line on the budget, under one percent of capex and opex, so he told the room to treat it as a profit lever, not a cost center. The full story is on Mining Disrupt at miningdisrupt dot com, and many more like it play around the clock on Bitcoin Radio.