Determining heat dissipation of 3D printed box - final results!

[HendriXML]

I designed a 3D printed box for a battery (5x1.2V) to 5V regulation circuit. For that circuit I know it has more than enough dissipation power.

However I still like to know how much heat can be dissipated, just to get a sense of what happens when you put stuff in a box.

To determine this I will use the circuit which was posted here:
https://www.eevblog.com/forum/projects/very-stable-temperature-control/msg2514615/#msg2514615

I will use that circuit (Control schematic 1.3.37) and mount the offboard stuff (heating transistors, sensor etc) on a second perfboard (Schematic). The transistors will have the heatsink on them for which the box is designed for.

I posted 2 renders of the box, I've used the appearance of wood because it easy on the eye.

One of the features of the box is, is that the air slots in the side walls are zig zagged. So nothing straight can be poked trough them.

The questions I like to answer are:
How much power can be dissipated when the heatsink is at 40 deg. celcius above roomtemp (will be around 60 deg). in the following circumstances:
1) No box
2) Box without top
3) Box with top
4) Box with covered side air slots
5) Box with all air slots covered

[tunk]

I don't know the answers to your questions, but here is what I
would have asked myself: what is the current draw (i.e. power
needed to be bled off) and what size of heat sink. Also, what's
the size of the box?

[HendriXML]

The box can contain a 50 mm x 70 mm perfboard and a heatsink of 38 mm height.

So it is a small box, mainly for small projects.

Will post images of the real stuff as well.

[tggzzz]

How much power can be dissipated when the heatsink is at 40 deg. celcius above roomtemp (will be around 60 deg)

How does the box's material properties change when in contact with the heatsink at 60C?

[HendriXML]

The printed material PLA cannot withstand 60 deg very well. But the heatsink won't be in contact with the plastic. So I will also find out whether the inside of the box gets hot or not. In the covered conditions it will probably be close to that 60 deg.

[HendriXML]

The top part has finishend printing, so here's one image that shows the air slots on top. These are above the heatsink, but aren't zig zagged: small diameter stuff could be pushed trough it.

[Conrad Hoffman]

What are the external dimensions of the box, and the wall thickness. Which face is up? That's all you really need to calculate the temperature rise. And the type of plastic of course.

[HendriXML]

I don't know about whether calculating would end up with accurate K/W values. Parts of the box are not solid. The top takes up space inside the box. The air slots are of somewhat complicated design.

My guess is that having airflow or not will make a large difference. And even having good airflow vs lesser airflow. If you'd like to demonstrate calculating this problem, I could create a technical drawing with relevant parameters. For me such an exercise would stretch my math brain a bit too much.

[Mechatrommer]

since you provided air vent i guess this is similar to other ambient temperature analysis, since air can come and go to replace rised/expelled hot air. i think 60degC will be ok for PLA, its when 100degC and up it will start to deform..

[DaJMasta]

The airflow is going to be the main thing, as mentioned, as 3d printed boxes have infill beyond the perimeter layers, so they are quite good insulators.  I would not use PLA for this application, yes it shouldn't outright melt at those temperatures, but it does soften slightly, and if anything presses on the top above the heatsink, for example, it will slowly bend and eventually make contact.

You could try using PETG, it prints on most printers that will run PLA (not much higher temp) but is a bit higher temperature and a bit more rugged.  That said, my printer's extruder housing was PETG and after a few months of use in a 50C heated enclosure, the parts under mechanical pressure had bent enough to the point of being unusable.

I would design so that you have a clear airflow path out (looks reasonable for that) and so that you stay clear of the edges of the heatsink - something like a horizontal mount may be much safer for higher temperatures just because of the clearance to the nearest side.  From there, it's just up to testing.  You can get something that will read a few thermocouples or thermistors and go to town with running it in various configurations.  I'd estimate the box as it is will be good for less than 10W even when optimized unless you can guarantee that the airflow slits are unobstructed or you have some forced air.  If you, for example, made the lid out of metal and thermally coupled it to your heatsink, you'd easily get a couple Watts more dissipation, but you'd have to worry about the temperature of the interface of the lid to the part.

I think you could simulate it with free software (don't have specifics, but I know it exists), but I think the simplest method is going to be trial and measurement, and then building in a margin so you don't run into problems.


If you have a capable printer, you can look at higher temperature materials too.  Something like ABS, nylon, or polycarbonate holds up to the heat better.

[HendriXML]

Thanks for your response, the idea of using a different material like PETG sounds good.

88 deg of glass transition temp vs 60 is a lot.

The box used for this experiment is a throwaway one, has already a big hole in it for the wires. 

Like you said getting some airflow is what is under investigation. I hope the heatsink will create enough rising air / create enough pressure difference. I would be happy with 10 watts.

[Conrad Hoffman]

If it has airflow, it becomes more difficult, but IMO your chances of success are way better in terms of not getting too hot.

[DaJMasta]

I would consider the glass transition temperature as a guideline than a hard rule, though, and different blends from different manufacturers will behave somewhat differently.  My experience is that with moderate pressure at about 60-65C sustained, PETG will bend permanently given a while whereas if there isn't any load on it, I haven't yet seen it flex permanently.

I just had a PLA print start to warp and peel off the bed set to 65C while other PLAs (even from the same manufacturer, though in a different color) were fine with a 65C bed temp.  Lowering to 60C got me the adhesion I wanted, but it makes me think that the equivalent temperature for PLA to warp in a loaded situation is probably in the 45C ballpark, maybe even lower.

The plastic doesn't need to fully transition to be able to flow somewhat, and if you're looking for longevity, that's probably the temperature you want to design around.

[CatalinaWOW]

One of the beauties of 3D printing is the ease of prototyping.  It is far easier to measure the results than to predict them with any accuracy.  All that said I would not expect much dissipation capability from this configuration.  Have you considered making the heat sink one of the walls of the box.  Lets the heat get to outside world the easiest way possible and saves the printing time of that box side.  The big downside is that it isn't electrically insulated from the world which may be important to your application.

[HendriXML]

The reason for this little experiment is mostly to see what the maximal dissipation can be for this kind of box. It isn't optimized for airflow, it is also not fully closed. It is a box that can be used at the bench without any worries that a flying piece of clipped wire could short things out.

The question for me is, to what level of dissipation it still can be used. At this moment this is done in the name of science only. To answer the question how much difference the air slots make.

Maybe the pressure of the heated air is not good enough to push air through the top slots at a decent rate. In that case the temperatures in the box might become static and evenly spread, so even lesser amount of temp/weight difference of air can be created around the heatsink. In that scenario the vertical slots will be of not much use either, so it will be close to the performance of a completely closed box.

If it performs well then I'll have to find out what kind of effect covering top slots - a few at a time - have on the performance. Will is gradually perform less or will it collapse?

These thing might sharpen my intuition when designing other enclosures. However this kind of design will not be the best for high power stuff.

[HendriXML]

I think this experiment is delayed due to ordering fake LM335Z's from aliexpress. Off course the dispute period is over.. These weren't even that cheap. What a crap!!

[HendriXML]

Being short on temperature sensors I took one LM335Z from another project and used that.

So no controlling a temperature difference, only keeping the heatsink at 60 deg. The room temperature was about 25 deg.

The results give some nice insights.

I measured the voltage across the sense resistors (0.203 ohm combined) with a Fluke in Min/Max mode. The min/max values are in the table, but the avg is the one that matters.

So the best case is no enclosure in which 6.7W can be dissipated. In reality this could be higher if the heatsink temp was allowed to rise.

In the worst case 2.8 W was dissipated. In this case the ventholes where covered by duck tape. Because the top gets quite warm, this will give slightly better results than having no air slots at all.

The box in its intended state can dissipate 3.2W, only slightly better. This proves the idea that the slim slots are obstructive.

However the box with no top at all results also in a significant drop in dissipation capability relative to the no enclosure situation.

The conclusion that I draw from this experiment is that 60 deg vs 25 deg does not produce much air pressure. So it is easily obstructed. To be effective large slots are required.

Maybe I'll redesign the top to test whether that improves things. Or redo part of the experiment with a 80 deg heatsink temperature.

[HendriXML]

Will also change Resistor R11 in the PID control to something larger, that will make the (slow) oscillation a bit less. It used to control a large thermal mass, now its a lot less.

Another improvement in the test setup will be to isolate the LM335Z a bit, without isolation the non enclosed situation will probably have a stronger sideways cooling effect on this sensor, so the temp reading might be lower than the actual temp of the heatsink.

The isolation will have a small effect on the cooling capacity of the heatsink.

[HendriXML]

I updated the PID control, it is now much more stable (R11 -> 100K)

The sensor got isolation.

I did measure the room temperature with the LM335Z it was 27.8 deg.

The heatsink was regulated to 90 deg. A temperature difference of 62.2 K.

The free air wattage was 11.4 so the heatsink to air had a thermal resistance of 5.46 K/W that's a lot more than the 3.8 K/W (FA-T220-38E) of the datasheet.
https://nl.mouser.com/datasheet/2/303/sink_f_r-1265536.pdf
 
The difference between covered air slots and uncovered is minimal. I think the top slots are too small. This was one concern when I was designing those slots. Putting the circuit in a topless box, also makes a 27% difference, that is a lot.

With a 90 deg heatsink in the box, the top did become somewhat soft. It has to be said that driving the transistors at 90 deg at this power is most likely not a good idea. I don't know the thermal resistance of the silicone pads (probably rubbish) so I can't be certain about overheating.

[HendriXML]

If I look at the specifics of PLA
https://www.makeitfrom.com/material-properties/Polylactic-Acid-PLA-Polylactide
It show that it has a high Specific Heat Capacity (1800 J/kg-K) and a low Thermal Conductivity (0.13 W/m-K). So heating a 100 gr of PLA 62.2K would take 11 kJ.
That is 37 minutes of 5 watt all of the power was only used for heating up, no losses.

What if the most prominent cooling effect of the box came from heat capacity. That would explain the small difference between uncovered vs covered.
So to do this experiment well, I should leave it warming up for at least an hour. Maybe more.

My experiments had some duration, but not in that magnitude.

So I will redo the experiment of the uncovered/uncovered test cases at 80 deg. 90 deg is a bit too hot.

Also covered air-slots might be better than no air-slots, so there usefulness can only be proven not disproven.

[tggzzz]

If I look at the specifics of PLA
https://www.makeitfrom.com/material-properties/Polylactic-Acid-PLA-Polylactide
It show that it has a high Specific Heat Capacity (1800 J/kg-K) and a low Thermal Conductivity (0.13 W/m-K). So heating a 100 gr of PLA 62.2K would take 11 kJ.
That is 37 minutes of 5 watt all of the power was only used for heating up, no losses.

What if the most prominent cooling effect of the box came from heat capacity. That would explain the small difference between uncovered vs covered.
So to do this experiment well, I should leave it warming up for at least an hour. Maybe more.

My experiments had some duration, but not in that magnitude.

So I will redo the experiment of the uncovered/uncovered test cases at 80 deg. 90 deg is a bit too hot.

Also covered air-slots might be better than no air-slots, so there usefulness can only be proven not disproven.

The thermal capacity is only of interest when considering thermal transients. Naturally you should measure temperature-vs-time to ensure that the transients have "finished" and you are looking at the steady-state power and temperature.

Have you done the basic calculation to estimate the internal temperature rise of a box without vents? All the information you need is the power, the surface area, the thickness, and the thermal conductivity.

Have you compared that theoretical value with the measured value?

Then add in the effects of the vents. (Hint: forced air cooling is used for a reason, not for fun!)

[DaJMasta]

Also remember the way your slicer actually realizes the model.  If the boxes walls are thin enough, there should be just solid perimeters for walls and the thermal conductivity will be about what it specifies, but if they're thick enough for infill, they will insulate much better because of the built in air pockets.

If you take the same box and slice it for a solid wall and a second box with walls with a little infill (like 1mm of thickness), especially if your infill percentage is low, the box with the hollow wall will hold its heat better (what we don't want, in this case).

[HendriXML]

My thoughts when "designing" the box where that the chimney effect would do more than 80% of the cooling. So that a closed box would perform very badly.

The experiments until now seem to show something different.

They show a reduced cooling capacity of only +/- 50% without airflow at all, that's better than I would expect by intuition. Calculating the expected performance would be tricky. The box is partly solid, partly with low infill. The bottom is obstructed with a PCB. Calculating this kind of stuff needs a very good model.

Maybe "they" also calculated the thermal resistance of the heatsink. But in my instance it doesn't hold up. (Why?). That alone would mess up calculations.

With this experiment real values can be determined.

Looking at the thermal capacity however it seems to me it was still in thermal transient, that can take a long time at those low powers. (Something to remember!)
When things heat up from inwards to outwards more and more the effects of isolation will show.

During the first experiments I think the plastic that later become isolating where still heating up.

I'm thinking about monitoring the power with a DSO, not the best way, but the only one at hand. That way a graph can be plotted which shows the cooling capability over time. And will show stability when things are heated up to their end values.

I guess the cooling capability will drop for both uncovered and covered, but that the values will be further apart. I hope the uncovered will not be much worse. I think without active cooling the results then might be considered ok, if you take into account that just having side walls has already a dramatic effect.

[HendriXML]

For further testing I created a new model, with specific slicing properties.

The ventilation part doesn't have a bottom or top layer and has 1.2 mm infill line distance, thus creating a mesh.

Using the slicer, and not the 3D modelling tool for this gives better control over those detailed structures. I hope it prints fine.

I wouldn't like a more coarse mesh, so this is in a way an optimized (top) airflow.

At layer height 13, the solids are shown where a threaded insert can be placed. I've done that in the box that I actually use.

[HendriXML]

The mesh had some inner wall lines that could be eliminated, resulting in a bit more mesh...