Certainly larger heatsinks will help. A 250mm X 210mm heatsink has a dissipation of 0.19 deg C/watt per Modushop ratings.
As such, with each device dissipating 66 watts (132 watts/ch), then 66 watts over a 0.19 degC/watt would be ~ 13 deg C rise over ambient. I don’t know what your ambient is, but let’s assume 25 deg C.
25 + 13 is close to 40 deg C on each heatsink piece. That’s much cooler.
Fans and cpu coolers are certainly options for any heatsink size, and that’s up to you and your situation. I aim for max of 50 degrees C overall (between the front chassis, heatsinks and rear chassis), but I am a scaredy cat compared to most. I also avoid fans personally.
Have you checked the temp directly on your MOSFETS? Where are your MOSFETS mounted? At the bottom 1/3rd of the heatsink or somewhere else?
Best,
Anand.
As such, with each device dissipating 66 watts (132 watts/ch), then 66 watts over a 0.19 degC/watt would be ~ 13 deg C rise over ambient. I don’t know what your ambient is, but let’s assume 25 deg C.
25 + 13 is close to 40 deg C on each heatsink piece. That’s much cooler.
Fans and cpu coolers are certainly options for any heatsink size, and that’s up to you and your situation. I aim for max of 50 degrees C overall (between the front chassis, heatsinks and rear chassis), but I am a scaredy cat compared to most. I also avoid fans personally.
Have you checked the temp directly on your MOSFETS? Where are your MOSFETS mounted? At the bottom 1/3rd of the heatsink or somewhere else?
Best,
Anand.
I know you've solved your problem but just to say my forced cooling solution works well and is silent at the listening position. It's all about displacement. 80mm fans are never going to displace much air quietly. A 200mm fan will flow much more air at much lower revs with much less noise. Go big!
@andyr I went for a different solution, see here: link
The problem with CPU coolers is they are designed to have a large surface area within a small volume: so they need high velocity air to shift the heat. The higher the velocity: the greater the noise.
I found a used big pin heatsink to force cool with a 200mm Noctua at 9V ish. A conventional finned heatsink would work just as well. The difficulty is in getting good data to predict what you will have dissipation wise, but I reckon if you have a 200mm Noctua you can go with 25% of the heatsink area in free air, and still have a good margin.
The problem with CPU coolers is they are designed to have a large surface area within a small volume: so they need high velocity air to shift the heat. The higher the velocity: the greater the noise.
I found a used big pin heatsink to force cool with a 200mm Noctua at 9V ish. A conventional finned heatsink would work just as well. The difficulty is in getting good data to predict what you will have dissipation wise, but I reckon if you have a 200mm Noctua you can go with 25% of the heatsink area in free air, and still have a good margin.
Hi Anand,
Understood!
However, I really would prefer not to have to go to a 500mm deep case.
Given your above figures ... as these are my "winter amps" (where ambient is 22 deg C - courtesy of the central heating), with each device dissipating 66 watts into its own heatsink, then 66 watts over a 0.19 degC/watt heatsink would be a 13 deg C rise over ambient = a total of 35 deg C. Which is low.
So ... I would like your opinion of whether the 400mm deep case will be acceptable. This has a 200mm long heatsink for each output device. With a dissipation of 0.28 deg C/watt (Modushop's figures) ... this means 66w x 0.28 = 19 deg rise over ambient. Which = 41 deg C.
41 deg C seems to me to be entirely reasonable?
Thanks,
Andy
Understood!
However, I really would prefer not to have to go to a 500mm deep case.Given your above figures ... as these are my "winter amps" (where ambient is 22 deg C - courtesy of the central heating), with each device dissipating 66 watts into its own heatsink, then 66 watts over a 0.19 degC/watt heatsink would be a 13 deg C rise over ambient = a total of 35 deg C. Which is low.
So ... I would like your opinion of whether the 400mm deep case will be acceptable. This has a 200mm long heatsink for each output device. With a dissipation of 0.28 deg C/watt (Modushop's figures) ... this means 66w x 0.28 = 19 deg rise over ambient. Which = 41 deg C.
41 deg C seems to me to be entirely reasonable?
Thanks,
Andy
Andy,
Going with the real world number data as Vunce has provided can be more useful than calculations.
In the 5U/400 chassis he has, if you build it with the AN 4 ohm version that you have (66watts dissipated per sink or 132 watts overall/ch), your overall temp on the heatsinks may rise to 60 deg C.
So that says that there is certainly a fudge factor here and obviously heeding Vunce’s experience would be advised. One of the reasons might be that the heatsinks in the 400mm and 500mm depth version of the Modushop Dissipante chassis are not linked to one another. If you design in a heat spreader plate to adhere them as one (which is more work and more expensive), then you might get data closer to an ideal. Moreover, I would recommend use of Aluminum oxide pads + thermal paste instead of the ubiquitous Keratherm. Another consideration is accurate measurement of the actual ambient temperature where the MOSFETS exist. Finally, we are asking all of this from just 2 MOSFETS. They are incredible devices with 280W power ratings, but placement of the MOSFET on the heatsink for best thermal dissipation results is key.
I have applied Wakefield paste between the front panel and heatsinks as well as in between the heatsinks but to be honest, unless the heatsinks + 10 mm aluminum front panel are one contiguous unit, it seems to make only a minor difference. Moreover, you cannot over tighten the heatsink brackets to the front panel because if you do, the rear of the enclosure does not line up with the top/bottom panels.
Paraphrasing from Ti Kan of AMB Audio who wrote a very good summary:
That’s P = V*I which we have already calculated in Andy’s 4 ohm build to be about 132 watts.
Next, we determine if a heatsink (given its thermal resistance rating in °C/W) is sufficient. To do that, we calculate what the MOSFET internal junction temperature would be and check if it would approach the maximum rating of the device with some headroom.
The formula is:
Tj = Ta + (P * (Rjc + Rcs + Rsa))
Where,
I realize the above will obfuscate some of you reading it and that certainly isn’t my goal. But as Hugh said very early in this thread, the biggest negative to these single ended Class A designs is the enormous thermal needs they have. This is the reason I will be rebuilding my next version as mono-blocks.
Best,
Anand.
Going with the real world number data as Vunce has provided can be more useful than calculations.
In the 5U/400 chassis he has, if you build it with the AN 4 ohm version that you have (66watts dissipated per sink or 132 watts overall/ch), your overall temp on the heatsinks may rise to 60 deg C.
So that says that there is certainly a fudge factor here and obviously heeding Vunce’s experience would be advised. One of the reasons might be that the heatsinks in the 400mm and 500mm depth version of the Modushop Dissipante chassis are not linked to one another. If you design in a heat spreader plate to adhere them as one (which is more work and more expensive), then you might get data closer to an ideal. Moreover, I would recommend use of Aluminum oxide pads + thermal paste instead of the ubiquitous Keratherm. Another consideration is accurate measurement of the actual ambient temperature where the MOSFETS exist. Finally, we are asking all of this from just 2 MOSFETS. They are incredible devices with 280W power ratings, but placement of the MOSFET on the heatsink for best thermal dissipation results is key.
I have applied Wakefield paste between the front panel and heatsinks as well as in between the heatsinks but to be honest, unless the heatsinks + 10 mm aluminum front panel are one contiguous unit, it seems to make only a minor difference. Moreover, you cannot over tighten the heatsink brackets to the front panel because if you do, the rear of the enclosure does not line up with the top/bottom panels.
Paraphrasing from Ti Kan of AMB Audio who wrote a very good summary:
Heatsink sizing considerations
The heatsink should have low enough thermal resistance to support the amount of heat to be dissipated. Bigger and more efficient heatsinks have lower thermal resistance specifications. To determine whether a heatsink is adequate, the power disspation must first be calculated.That’s P = V*I which we have already calculated in Andy’s 4 ohm build to be about 132 watts.
Next, we determine if a heatsink (given its thermal resistance rating in °C/W) is sufficient. To do that, we calculate what the MOSFET internal junction temperature would be and check if it would approach the maximum rating of the device with some headroom.
The formula is:
Tj = Ta + (P * (Rjc + Rcs + Rsa))
Where,
- Tj = junction temperature
- Ta = ambient temperature
- P = power to be dissipated
- Rjc = thermal resistance of junction to device case
- Rcs = thermal resistance of device case to heatsink
- Rsa = thermal resistance of heatsink to ambient
- Ta is the highest ambient temperature you expect the circuit will operate under (note that if the circuit is going to be in an enclosure, then Ta will be higher than room temperature!). P is the power dissipation we calculated. Rcs is the sum of the thermal resistance of any mounting isolation pads, heatsink compound, etc., and you can look that up from their datasheets. And Rsa is the thermal resistance of the heatsink itself.
- Calculating for Tj, and comparing it to the MOSFETs' maximum junction temperature of 175°C, you can determine if the heatsink is good enough. Even though these MOSFETs are very rugged, you should avoid letting the MOSFET junction temperature exceed 100°C to prevent reduced device lifespan. Under extreme conditions, forced air cooling (e.g., fan) may become necessary.
I realize the above will obfuscate some of you reading it and that certainly isn’t my goal. But as Hugh said very early in this thread, the biggest negative to these single ended Class A designs is the enormous thermal needs they have. This is the reason I will be rebuilding my next version as mono-blocks.
Best,
Anand.
One other consideration is how much aluminum metal is in the rest of the chassis and how well thermally bonded it is to the heatsink panels.
On my prototype A40 amp with 300mm deep fins, running 29.5v and 1.7A, the fins were 58C and the main case was 42C. Thus, a significant portion of the heat was also distributed onto the main chassis. I am expanding the chassis to 375mm deep so I am hoping it will drop to about 50C max.
The A40 chassis has 4.7mm thick walls besides the heatsinks.
On my prototype A40 amp with 300mm deep fins, running 29.5v and 1.7A, the fins were 58C and the main case was 42C. Thus, a significant portion of the heat was also distributed onto the main chassis. I am expanding the chassis to 375mm deep so I am hoping it will drop to about 50C max.
The A40 chassis has 4.7mm thick walls besides the heatsinks.
Absolutely - I'm very grateful to Vunce for providing this real-world case.
Calculations using the temp coefficient numbers provided by the mfrs would appear to be suspect, by comparison.
For instance, the temp coefficient figures which Gianluca provided to me are these:
* 300 x 210 (my current heatsink) - 0.20 deg C/watt
* 250 x 210 - 0.19
* 200 x 210 - 0.28
Now, if the 200m long heatsink is specced at 0.28 and the 250mm long heatsink is specced at 0.19 ... logically, the 300mm long heatsink should be 0.14 to 0.15. Yet Gianluca told me it is 0.20 - this figure is provided to them by their heatsink supplier ... which doesn't make any sense.
Taking Vunce's real-life case:
* a 400mm Dissipante case has 2x heatsinks, each 200mm long (each side)
* so each heatsink takes the heat from one output device
* this is 28.5v x 1.7a --> 14 deg rise, per heatsink
* with 22 deg ambient temp ... this should make the heatsink temp 36 deg
* yet it's not - it's double the rise ... it's actually 50 deg!
Using Vunce's AN 8R numbers and substituting my own AN 4R DC voltages and bias current:
* 22v x 3a --> 19 deg rise per heatsink (theoretically!)
* guessing actuals ... suggests heatsink temp will reach 60 deg C - as you suggested
* which is still a bit hot!
So it seems like a 500mm deep case is required!
Or I see whether there's any spare space in a 400 deep case (compared to my current 300 case) to mount a couple of fans to blow at the heatsinks! (And run them at low voltage.)
I actually use mica washers (+ thermal paste) ... do you think those white Al oxide pads (+ thermal paste) do a better job of heat transfer?
Thanks,
Andy
Temperature coefficient numbers posted by Modushop were calculated at thermal saturation, which is a little less useful for our applications. Gianluca posted the study leading to the results in 2 picoDumbs toxic masculinity chassis thread, I believe. There is some wonderful data contained within the study. The Italian is fairly easily translated.
We're also generally concerned with the temperature at the tips of the heatsinks' fins for safety. The calculations were done using the maximum temperature of the heatsink, which is more relevant for thermal calculations. The study was also done in fairly ideal circumstances with a heatspreader across the bottom of the heatsink providing the heat vs. individual devices. In addition, one heatsink was loaded while the other was not. The chassis were fully built vs. testing heatsinks separately. Ambient also seemed a touch high. I had asked for confirmation of ambient, but it really doesn't matter too much. My guess is that the heatsinks were allowed to come to temp and the air temperature nearby (within inches perhaps) was used vs. the temperature of a controlled room prior to applying any heat to the heatsinks. That is speculation, however.
tl;dr - The coefficient must be 'de-rated' when the temps are anything lower than the saturation temperature, which for the Modushop study in some cases was taken to ~ 118C. I would not use those numbers. Numerous people have posted data and personal experiences re: recommended dissipation limits / temperatures for various chassis.
Nelson also posted a fantastic summary of his personal experiences with Modushop's various chassis. I don't remember exactly where it was posted. If I find it, I'll attach.
Edited to add - I need to confirm whether their calculations were done on the tips of the sinks or the max temp... thinking about it, the tips of the sinks makes a bit more sense (better thermal coefficient), but less relevant for thermal calculations to see how well the device(s) may fair. So, as always ... check the real data.
We're also generally concerned with the temperature at the tips of the heatsinks' fins for safety. The calculations were done using the maximum temperature of the heatsink, which is more relevant for thermal calculations. The study was also done in fairly ideal circumstances with a heatspreader across the bottom of the heatsink providing the heat vs. individual devices. In addition, one heatsink was loaded while the other was not. The chassis were fully built vs. testing heatsinks separately. Ambient also seemed a touch high. I had asked for confirmation of ambient, but it really doesn't matter too much. My guess is that the heatsinks were allowed to come to temp and the air temperature nearby (within inches perhaps) was used vs. the temperature of a controlled room prior to applying any heat to the heatsinks. That is speculation, however.
tl;dr - The coefficient must be 'de-rated' when the temps are anything lower than the saturation temperature, which for the Modushop study in some cases was taken to ~ 118C. I would not use those numbers. Numerous people have posted data and personal experiences re: recommended dissipation limits / temperatures for various chassis.
Nelson also posted a fantastic summary of his personal experiences with Modushop's various chassis. I don't remember exactly where it was posted. If I find it, I'll attach.
Edited to add - I need to confirm whether their calculations were done on the tips of the sinks or the max temp... thinking about it, the tips of the sinks makes a bit more sense (better thermal coefficient), but less relevant for thermal calculations to see how well the device(s) may fair. So, as always ... check the real data.
Last edited:
And here is the thread member @ItsAllInMyHead referred to that was started by Nelson Pass.
Member Eric, also posted a long while back an interesting study with regards which thermal interface is ‘best’ with MOSFETS.
Best,
Anand.
Last edited: