I am planning to use a Modushop Galaxy 1GX388 2U 330 mm by 280 mm by 80 mm chassis for a pair of Zenductor 2 amplifiers. To find out whether the 10 mm thick corrugated sidewall profiles are sufficient as heatsinks for 15 W to 25 W dissipated power per side, I did thermal measurements with a DC power supply, a 100 W 8 Ohm power resistor, and a thermal camera built into my cell phone.
The answer is that the chassis should work as is, i.e. without replacing the side walls with real heatsinks.
I attached the power resistor with a pair of spare M3 screws and nuts (supplied with the chassis) to the center region of the flat middle section of the 2-channel side of the Galaxy sidewall profile, and ran tests at about 18 W and 39 W, both with the wall free standing and as part of the assembled chassis (with the rear panel partially slid in, to leave an opening for the resistor power cables, and the remaining opening mostly blocked against air movement).
Temperatures stabilized after 20-30 minutes, and were lower than expected. From the measured temperature rises, and correcting for the heat loss directly from the resistor body to ambient, I compute a thermal resistance of about 1.4 K/W or °C/W.
Graph of four measurements (two power levels, freestanding and chassis-integrated heatsink):
Ambient temperature (the zero level of the graph above) was about 22°C. Interestingly, the equilibrium temperatures for the heatsink (corrugated 10 mm sidewall) were lower in the chassis than free standing. The increase in effective convection surface, from bolting on the other four panels of the chassis) must outweigh the shielding effect of the case.
Between the second and third measurement series, I installed (slid, with the screws and nuts already in place) the sidewall into the chassis.
Also interesting (to me) is that radiation losses are not negligible. A sidewall (448 cm^2, counting both surfaces, but not the corrugations) at 40°C radiates about 5 W more power into a 22°C room than it absorbs. I chose NOT to correct for radiation, since the effect will be of similar magnitude in the real application of the chassis.
As mentioned, I did correct for the direct convective losses from the power resistor, by measuring the resistor equilibrium temperature of 120 °C (!) at 9.68 W, with convection shielded from the resistor mounting surface (sitting on an insulating surface), giving a thermal resistance of 10.2 K/W.
The thermal resistance from the resistor case to the heatsink computed as about 0.5 K/W, which should be similar to achievable case-to-heatsink resistances for transistors. No thermal compound was used.
Example of a thermal image (equilibrium image for 18 W power, resistor mounted to side wall inside the chassis):
The resistor mounting position 'shines through' to the outside surface; I used the maximum, center temperature as heatsink temperature for my calculations. This choice should give correct results for the estimated case temperature of the device to be cooled. The average heatsink (side wall) temperature will be lower than computed from the effective thermal resistance of 1.4 K/W.
Resistor mounting position:
The answer is that the chassis should work as is, i.e. without replacing the side walls with real heatsinks.
I attached the power resistor with a pair of spare M3 screws and nuts (supplied with the chassis) to the center region of the flat middle section of the 2-channel side of the Galaxy sidewall profile, and ran tests at about 18 W and 39 W, both with the wall free standing and as part of the assembled chassis (with the rear panel partially slid in, to leave an opening for the resistor power cables, and the remaining opening mostly blocked against air movement).
Temperatures stabilized after 20-30 minutes, and were lower than expected. From the measured temperature rises, and correcting for the heat loss directly from the resistor body to ambient, I compute a thermal resistance of about 1.4 K/W or °C/W.
Graph of four measurements (two power levels, freestanding and chassis-integrated heatsink):
Ambient temperature (the zero level of the graph above) was about 22°C. Interestingly, the equilibrium temperatures for the heatsink (corrugated 10 mm sidewall) were lower in the chassis than free standing. The increase in effective convection surface, from bolting on the other four panels of the chassis) must outweigh the shielding effect of the case.
Between the second and third measurement series, I installed (slid, with the screws and nuts already in place) the sidewall into the chassis.
Also interesting (to me) is that radiation losses are not negligible. A sidewall (448 cm^2, counting both surfaces, but not the corrugations) at 40°C radiates about 5 W more power into a 22°C room than it absorbs. I chose NOT to correct for radiation, since the effect will be of similar magnitude in the real application of the chassis.
As mentioned, I did correct for the direct convective losses from the power resistor, by measuring the resistor equilibrium temperature of 120 °C (!) at 9.68 W, with convection shielded from the resistor mounting surface (sitting on an insulating surface), giving a thermal resistance of 10.2 K/W.
The thermal resistance from the resistor case to the heatsink computed as about 0.5 K/W, which should be similar to achievable case-to-heatsink resistances for transistors. No thermal compound was used.
Example of a thermal image (equilibrium image for 18 W power, resistor mounted to side wall inside the chassis):
The resistor mounting position 'shines through' to the outside surface; I used the maximum, center temperature as heatsink temperature for my calculations. This choice should give correct results for the estimated case temperature of the device to be cooled. The average heatsink (side wall) temperature will be lower than computed from the effective thermal resistance of 1.4 K/W.
Resistor mounting position:
Last edited:
