Study on the Characterization of Filters for a Direct-to-Chip Liquid Cooling System (40th SEMI-THERM Symposium, 2024)

Overview

This paper is an experimental study presented at the 40th SEMI-THERM Symposium in 2024. Its authors include Pardeep Shahi, Ali Heydari, Dereje Agonafer and Mohammad Tradat. The study starts from the premise that filtration in the coolant distribution unit (CDU) is key to protecting cold-plate microchannels from particulates.

The experiments were carried out on strainer filters in the secondary loop of a direct-to-chip system. Using a CDU and rack, filters of 25, 50 and 100 micron mesh were compared with the no-filter case; CDU pump speed was varied between 20% and 100% by a variable frequency drive, with the coolant held at 32 C. The authors measured pressure drop across the filters, the resulting loop flow, and filter condition over time.

According to the results, the pressure-drop difference between no filter and the 25 micron filter grew from about 0.11 psi at 20% pump speed to about 0.64 psi at 100%. At maximum pump speed loop flow was about 390 L/min, and flow differences between filter cases were around 1% or less. With 100 and 50 micron filters at 80% speed, flow was 308-309 L/min, and the 25 micron filter gave the highest pressure loss. These findings show that finer mesh adds a measurable but small hydraulic penalty.

As a limit, the results are confined to the flows tested, a single CDU and rack arrangement; behaviour may change with a different coolant, contamination load or long-term clogging. Even so, the study offers a measured reference for how filter micron rating can enter the hydraulic calculation, and it should be read alongside the OCP Cold Plate Cooling Loop Requirements.

Key points

1. Filters of 25, 50 and 100 micron mesh compared with no filter at CDU pump speeds of 20-100% (VFD), coolant held at 32 C.
2. Pressure-drop difference between no filter and the 25 micron filter grew from about 0.11 psi at 20% pump speed to about 0.64 psi at 100%.
3. At maximum pump speed loop flow was about 390 L/min; flow differences between filter cases were around 1% or less.
4. With 100 and 50 micron filters at 80% speed, flow was 308-309 L/min; the 25 micron filter gave the highest pressure loss.
5. Filtration in the CDU is presented as key to protecting cold-plate microchannels from particulates.
6. Finer mesh adds a measurable but small pressure drop and flow penalty.

Using it in design

• Include the filter micron rating and its pressure-drop curve in the hydraulic calculation for CDU and loop objects.
• Default to finer filtration (25 micron) for microchannel cold plates, since its hydraulic penalty is small.
• Add monitoring of filter condition over time to the maintenance plan.
• Account for the pressure-drop difference growing with pump speed when selecting pumps.

Who should read it

Engineers designing CDUs, filters and secondary-loop hydraulics for direct-to-chip liquid-cooled systems.

Related standards

• OCP Cold Plate Cooling Loop Requirements

Publication details

Authors

Pardeep Shahi; Ali Heydari; Himanshu Modi; Lochan Sai Reddy Chinthaparthy; Anto Barigala; Qusai Soud; Uschas Chowdhury; Dereje Agonafer; Mohammad Tradat; Jeremy Rodriguez

Publisher

SEMI-THERM

Year

2024

Document number

40th SEMI-THERM Symposium; NSF PAR 10537801

Event

40th SEMI-THERM Symposium (2024)

Document language

English

Source type

Industry body

Access

Free to read

Licence

Copyright retained; citation only — the publisher’s terms apply to the original document

Citation