The Raspberry Pi 4 presents some challenges for enclosure design. It has a different physical layout and has the capacity to produce more heat. I wonder if now is the time to consider a change of official enclosure design? If so maybe a radical change may be considered alongside a subtle change.
It may be advantageous to add significant passive or active cooling. (My next enclosure is likely to have heatsink fins exposed.)
Maybe the enclosure iterates more subtly with changed connector location and a small fan.
Maybe the enclosure becomes more modular allowing greater flexibility of integration and customisation.
Maybe the community provides inspiration, e.g. suggestions via forum, collaboration on design, competition, etc.
We may each design our own wonderful and eclectic enclosures but I suggest now may be the time to consider what the official Zynthian looks like.
I agree, there needs to be a re-think of the case.
Modular design would allow for different interface/control elements depending on usage. For example audio I/O via XLR or 1/4 inch jack, CV/Gate for euro-rack, larger displays (I chose a Waveshare 7" USB for my self build)… even have a case that could be used on the floor with footswitches and pedals.
Modularity is going to mean some compromise but the increased flexibility is worth it. Cooling should be no problem, several fan and/or heatsinks are available for Pi-4 already.
The power over ethernet hat included a fan which strikes me as indicative. I’ve found metal cases on rpi3’s to be more thermally well behaved than plastic boxes but certainly it is something to be considered. The idea of modularity is nice and I’ve tried most config stacking audio cards et Al seems a pretty modular response but is limited by the need to mount the pi in the corner of an enclosure to allow access to the connectors so I think the compromise of just using a larger case is sensible. Quite what will constitute a standard build is starting to emerge but we are beginning to introduce complexities with the a/d d/a and digi i/O. Quite how this gets implemented in a standard build is certainly open to much discussion.
We need to consider fabrication complexity and cost. It needs to remain accessible and viable.
I quite like the idea of exposing the required interfaces via bus cables, e.g. I2C bus that can connect to each relevant device. That might give more freedom for positioning hardware within the enclosure, placement of heat sinks and flexibility of which hardware devices were supported, maybe requiring device specific interface cables which could be described in the support pages.
Regarding the next Official Zynthian Alu-Case for the Pi4, Ramon and me are thinking about it quite a lot. Our goals for the new design are:
Maintain size and price
Solve heat problems. If possible, avoid fans
Allow to connect “expanders” (i.e. ), probably using a ribbon cable
These are some of my thoughts about the heating problem with the RBPi4:
RBPi boards use Heat-Spreading Technology. That means that the heat generated on the CPU and other ICs are “pushed” into a kind of “thermal layer” inside the PCB sandwich. You can read more about this subject here:
For that reason, normally, when operating without a case, the RBPi4 doesn’t get overheated. At less, this is my experience. So, in an open space (no enclosure), the “natural convection” is enough to maintain the board under 80ºC, even without a simple heatsink!!
So, the real problem is evacuating this heat from the enclosure, not evacuating the heat from the RBPi. (OK, the true real problem is maintain the CPU under 80ÂşC , but you know what i mean.)
The more simple way is to use a fan in the case to force convection around the RBPi4. Of course, a heatsink would help too …
The not so simple way but a lot more elegant, reliable, silent, etc. is to NOT USE A FAN. We have 2 main lines here:
Use the Alu-Case as a heatsink, “conducting” the heat from the RBPi’s PCB (CPU?) to the alucase’s chasis. And i say “conducting” in the more physical sense of the word. First, we should replace the current “nylon” separators by “brass” separators. “Brass” is a good conductor and will help to evacuate heat from the RBPi’s PCB to the case. Also, we could try to conduct the heat directly from the CPU’s metal pad, but it’s more complex and probably we will end putting a heatsink here.
Also, iw we achieve to conduct enough heat to the case surface, we could “glue” a heatsink to the case bottom, or try to integrate the fins in the design.
Add extra ventilation slots on the front-side and probably make bigger the back-side and bottom slots. We have to take all the advantage we can from the “natural convection” process. A Finite Elements simulation would be nice, but i don’t know if we can afford to do that. It’s not a 2-days work … (BTW, somebody wanting to do that? ;-))
Note that the “ventilation” slot in the front side could be used for passing the ribbon cable that allows connecting an “expander”.
Only ideas. Nothing has been tested yet … so … please, i’m very very interested in your ideas and suggestions. Don’t hesitate to post in this thread!! Let’s brain-storm, guys!
We should aspire to passive cooling being sufficient for “full load” at “normal” ambient temperature. I think active cooling may be desirable and acceptable to cope with higher ambient temperature. I would expect that studio use would be at “normal” ambient temperature and we would expect a silent device. Live performance may be in a more hostile environment but may also have higher ambient noise level, e.g. audience, other musicians, etc. What constitutes “normal” is open to debate but in the fairly clement climate of England I would suggest aiming for core temperature of 70C at full processing with ambient temperature of 22C not triggering the fan to start.
Peltier is interesting technology but I would be concerned about the extra complexity and power required to run it. If it sat between the processor and the heatsink it would have to be permanently active. It could be used attached to a case exhaust fan to help cool the air within the enclosure and hence be enabled with the fan or as part of a staged cooling process:
Passive cooling from heatsink attached to processor, venting to enclosure vents / heatsink.
Peltier drawing heat from enclosure interior to exterior via enclosure mounted heatsink.
Fan assisting Peltier.
Fan assisting CPU heatsink.
Maybe overkill though. We should model solutions and build prototypes to identify best solutions.
It’s quite inefficient and it would consume too much power …
1.) Let’s try to conduct as much heat as we can from the Pi to the alucase chasis.
2.) Let’s optimize the natural convection by smart placing the ventilation slots.
If this is not enough, we would consider more complex solutions …
I have a RPI4 in a 3D printed enclosure with the peel and stick heat sinks on the system, wifi, power and ram. There is a little 30mm 5V brushless fan bolted to the top of the enclosure which is flush with the top of the USB connectors, leaving them exposed… When the system is doing a video compression using VLC running at 2GHz with the video subsystem running at 600MHz, the maximum temperature reached is 59 C. Watching the processor utilization graph and the temperature, when the utilization is about 50% the temperature is 55, when at about 80% the temp is 59. I also observed that dropping the video to 500MHz drops the temperature by 5 C. When the system first started up the temp read 34 C.
.
