Watercooling A Canon DSLR Leads To Serious Engineering Upgrades

The Canon EOS R5 is a highly capable, and correspondingly very expensive camera. Capable of recording video in 8K in a compact frame size, it unfortunately suffers from frustrating overheating issues. Always one to try an unconventional solution to a common problem, [Matt] decided to whip up a watercooling solution. What ensues is pure, top-notch engineering.

The watercooling setup is amusing, but the real star of the show is the custom copper heatsink that transforms the camera’s performance without spoiling its practicality.

Upon its original release, Canon had the R5 camera simply shut off on a 20 minute timer when recording 8K video. When the userbase complained, an updated firmware was released that used an onboard sensor and would only shutdown when excessive temperatures were reached. Under these conditions, the camera could record for around 25 minutes at 20 °C. [Matt] set about disassembling the camera to investigate, figuring out that the main processor was the primary source of heat. With a poor connection to its heatsink and buried under a power supply PCB, there simply wasn’t anywhere for heat to go, leaving the camera to regularly overheat and take hours to cool down.

After whipping up an amusing but impractical watercooling solution and verifying it allowed the camera to record indefinitely, [Matt] set about some proper thermal engineering. A custom copper heatsink was produced for inside the camera, bonded directly to the processor and DRAM with thermal paste instead of poor-quality thermal tape. This then directs heat out through the plastic back of the camera. In cool environments, this is enough to allow the camera to record continuously. In warmer environments, simply adding a small fan to the back of the camera was enough to keep things operational indefinitely.

[Matt] finishes the video by pointing out that Canon could have made the camera far more useful for videographers by simply investing a little more time into the camera’s cooling design, while also generating more profits by selling a cooling accessory for extended recording. We’ve seen some of [Matt’s] work before too, such as this DIY 4K projector build. Video after the break.

55 thoughts on “Watercooling A Canon DSLR Leads To Serious Engineering Upgrades

    1. They don’t even have a cinema line 8k camera yet. The c500ii only does 6k and c700 does 6k. The Venice does 6k. A Red helium vv does 8k and its 50k+++. Ppl wanting 8k unlimited for $3800 is just not gonna happen from Canon. It was always meant to shoot for 20mins, but even after that the codecs are basically unusable unless you got a 10k cup, and even then it bogs down. The technology on processor and GPS side for codec not quite their in canon raw bc and canons h265. If they would have used their pro codec xavc intra h264. Then you would be able to use it. That tells you all you need to know of how Canon protects their cine line. Sony does same thing. With advent of A7siii and the Canon R5. The line b.t mirror less and cinema is basically gone. Look at Fx6 which uses mirror less a7siii sensor n processor to put in Fx6 body. Ofc u get xlrs. And built in ND which is why you pay more.

      But 8k is coming to mirrorless in early 2021 Sony is doing 8k and it will use all intra xavc i.class 300 codec. It will be a 50mp A9R …

      8k 24p 30p and 60p xavc I class 300
      4k 120p same as A7siii
      2k 240p

      Its gonna be basically the perfect hybrid and will retail for around $5k !!!

      Start saving now..

  1. The project is very impressive. But, and I don’t want to join the army of the anal retentive here, an EOS R5 is not a DSLR.

    This is a pretty basic terminology error and should be corrected. It’s a Mirrorless camera, also known as MILC, CSC, ILCC and a bunch of others. But it’s not a Digital Single Lens Reflex, as among other things – it lacks a Reflex mirror. Hence, mirrorless.

    1. I have been reading this site every day like clockwork for 15 years now… started back when it was actually 1 hack per day. I think the mistakes are what make this site great. It doesn’t feel to professional, just geeks reporting on geek stuff… its part of its charm. Even though this has been my daily routine for 15 years if they charged a subscription i wouldn’t pay it.

      I always find the comments about grammar or small details that don’t matter amusing, why complain, you could be spending that time and energy on creating something cool that gets its own article some day.

      1. Indeed, I’ve also been here since the beginning and read the articles here religiously, but I would not pay a subscription fee. This site has always had a few spelling mistakes ect., it’s part of the charm and character.

      2. I would pay to have AI auto filtering grammar OCDs that add absolutely nothing to the subject at hand.

        I’m surprised how simple (and rushed) the R5 is, tape? Seriously? For 4k$?

        Looks almost like a volountary crippling to get you to jump to cinema cams after getting a taske of raw 8k’ing

  2. I watched this vid last week and was utterly amazed by it… But also utterly shocked how a 3-5 grand camera could be engineered so poorly by a multi billion dollar/pound company, and yet a maker online who runs a youtube channel is able to design a heat spreader that vastly improves the performance of the camera.
    There really is no excuse on canons part for this gaff. And frankly they owe it to the customers to engineer a clone solution and install it free of charge and offer an extended warranty.

    1. It’s done purposefully to differentiate models and performance levels, while maintaining the same basic internals.

      It is of course not the case that the Canon engineers *couldn’t* sort it, just that they didn’t want to.

      It’s worth noting though that even assuming they were not trying to differentiate models, it’s a design tradeoff with camera size, waterproofing, touchable external temperatures etc. Make
      Prosumer camera record endless 8k, which is not that common a request, or make it small, lighter and easier to build….

      1. Well in that case Canon are even worse in my light, As the price they ask for this camera is well within the range for film makers and content creators. Which is what this camera is mostly marketed towards.

        The trouble also is whether 8k is a fancy unused feature, 4k isnt. And this camera does 4k by downscaling 8k footage…

        Which also produces that same heat.

        If they wanted to keep this a controlled limited feature they would have kept that limit hardcoded in the firmware.

        This “hack” only works because they unlocked that limit.

        For the cost of a small sheet of copper, they could have made the world best prosumer film camera.

        Instead they crippled what is a VERY expensive camera out of choice. Rather than because of design limitations.

        1. Businesses so this all the time. All the time. And it’s part of a logical process, you don’t go all out straight away, you don’t show your hand until you need to. You beat the competition by a bit, then when you’re beaten in return you pop a bit of extra effort in and beat them again. That way you get a constant upgrade route for buyers and can market your advances. Most automotive manufacturers sit on huge technological improvements and only release them in a staged way to maximise profit. They’re not here to make everyone as happy as they can be, just as happy as they need to be to buy a new X.

        2. I’m pretty sure this is mostly a matter of Canon’s designers for prosumer cameras not being used to dealing with an SOC with anywhere near this kind of heat dissipation. Before someone had them come up with 8K video as a bullet point it just wasn’t a concern. As far as I can tell even their full-frame EOS Cinema models are only barely above 4k.

        3. Sometimes it is hard to have an accurate estimate on real life power estimation until you finally have working firmware/app running on new parts. By then the mechanical are already done. I try to be a bit more proactive in getting power estimates and talking to my mechanical/thermal/SI people *before* doing a layout. The usual work flow leaves that as an after thought check mark items in a design review when it is too late. It is a lot more work, but beats a churn.

          It”ll take more than a sheet of copper. There is only so much that can do when the case is plastic instead of aluminium. No doubt that there are some other forces at work on that decision.

        4. I mean it’s priced like a small second hand car but that doesn’t mean I’d buy the camera to drive to work. There’s nothing bad about this. if you’re a film creator and you bought this camera then you’re an idiot who should actually buy a camera more suited to the purpose. If you’re looking for something small and light which takes excellent quality PHOTOS then this is the camera that was designed for your purposes.

      2. “It is of course not the case that the Canon engineers *couldn’t* sort it, just that they didn’t want to.”

        Not so sure about that. Engineering teams come and go like water these days and are often stretched thin across multiple projects and must move from project to project very quickly. Then we also have companies increasingly relying on outsourcing.

        Projects get more complex, use faster, more powerful processers – all fine until a critical-mass of heat is produced at which iterative design improvements are not enough. But,sorry, not enough time to redesign, re-engineer and stay within release schedule – so take what we churned out or leave it.

        Sometimes it is anti-cannibalization, sometimes it is just-in-time industry improvements, sometimes it is incompentence, and sometimes it’s combinations of these or other factors. Bottom-line is that I refuse to carry water for these companies – especially at these prices.

        1. You make valid points, but in this situation (Canon) I’m going to agree with the people that say this is done on purpose for some reason. Canon releases cameras every year. They charge different prices for those cameras depending on what they think the market in any location will pay. I live in Japan and bought a Canon camera– manufactured in Japan and exported to the US– FROM the US for half the price I could buy it in Japan. Shipping it around the world twice was cheaper than buying locally for the exact same product. This is not a logical company from a customer’s perspective. The logic lies elsewhere. I bought Canon because they have the best tech, but they cripple it artificially so they’ll have room for upgrades on the next year’s model. If they didn’t do that, development costs would probably be prohibitive.

          1. There are many infuriating aspects to dealing with Canon products, such as their lack of integral intervalometer or GPS features. The only reason to not include them is so they can sell overpriced accessories. Especially the intervalometer. You can get hacked firmware for canon point and shoot that allow for intervalometer and scripting features, including things like lightning detection. Unfortunately there isn’t as much available to the DSLR and Mirrorless ranges. The processor is plenty capable, they just lock you out.

        2. It’s not the engineer, it’s marketing segmentation.
          Like in the good old days of canon 10D/D300, the later one was crippled in subtle ways to push more people to buy the 10D.
          It’s a constant at Canon, their trademark with backward incompatibilities (some will say induced obsolescence).

      3. Sony’s the worst with consumer vs pro video. None of their consumer cameras will record internally in 10-bit, for no other reason than to push you to their pro line.

        I was going to pull the trigger on a ZV-1 until I saw what the iPhone 12 Pro Max looked like in FilmicPro’s high-bitrate 10-bit mode. The phone blows the ZV-1 away in anything but low light. If/when I can actually make money, then it’ll be something like a BMPCC 6K.

    2. > also utterly shocked how a 3-5 grand camera could be engineered so poorly

      The thing is, shooting continuous video on a digital STILL camera is a bolt-on gimmick that these types of cameras were never supposed to be doing in the first place, because of the rolling shutter effect of the sensors they’re using. Older cameras with CCD sensors were much better suited for video production, but bazinga, they don’t make them anymore because it’s a more expensive sensor type, so everybody went for the grainier and less sensitive CMOS with software noise reduction and other clever tricks like multiple exposure stacking, but that’s na.

      It makes for a terrible video camera anyways, and people simply ignore the fact because as a “prosumer” product it’s still cheaper than a proper purpose-built DV camera, in a sense that you don’t need two devices where one will just about do. People are simply “abusing” these things for shooting video, which is OK as long as you don’t pan the camera, you don’t zoom while you’re filming, and you’re mostly filming static scenes – like the typical one-man youtube production here. It’s not the right tool for the job, but it is a tool that gets the job done, like hammering a nail down with the other end of a big screwdriver.

      Taking continuous video is merely possible, so they had to put it in as a selling point. Rather than re-engineer the cooling solution that was sized for taking still photos, they simply made a software timer that shuts the camera off after 20 minutes, and that’s how they made the marketing happy: now it is possible to shoot 8K video on it, great, put it on the box and shove it out the door.

      1. RE CCD sensors: This was true at the start of CMOS sensors but no more. They’re now far superior to CCD in every way. BSI CMOS sensors like Exmor R have a true global shutter and more dynamic range than cinema negative film ever had. BSI CMOS has also enabled such fast readout times for global shutter operation that for all practical purposes it might as well be a global shutter.

        CCD still has a foothold in machine vision for fast moving stuff, but even that’s quickly fading.

      2. The DSLR and Mirrorless cameras can be used as video cameras just fine technology wise – the sensors are superb now, focus and zoom controls are usually very capable for video as well (and if the stock firmware doesn’t bound to be something like magic lantern that vastly improves things)

        The real problem with them is ergonomics the DSLR style cameras are designed for point and shoot not sustained tracking . For my money this camera is superb for what its for, and can do 8K video really damn well by all accounts, just not all day without extra cooling. If you wanted to shoot video in that quality and all day you’d be buying something with the ergonomics and balance to be a video camera (and it will also have huge batteries to keep it running that long, and help it balance neatly)…

        The big selling point for mirrorless is the SLR high quality stills with LIGHTER WEIGHT – so they don’t want to toss in some extra few hundred gram of copper to let use a video mode infinitely, as you probably didn’t buy the camera for video, and you picked mirror less presumably because its lighter than DSLR, so it better be properly lighter.

        Personally I’d want proper cooling to use all the mode freely, but I can see some good arguments against making it heavier to use the one mode you probably didn’t want (as you’d have spent similar money on a more purpose built video camera if that was what you really really needed)

      3. They are *not* “still” cameras. They are in a form-factor *similar* to the film SLR’s of old, but they are full hybrid cameras. The major differences between them and cine cameras are entirely ergonomical.

        Rolling shutter is common across all digital cameras, whether they’re for “cinema” or not.

    3. There is no gaff. You’re applying an engineering requirement to a piece of equipment for a different purpose. the R5 wasn’t designed to operate with the sensor on continuously and not for thermal noise performance on long exposures. It was designed to be light and small for high performance photography. And it achieves that goal quite well.

      If you bought this camera for continuous recording then you’re the one who made a gaff. They don’t owe customers anything.

  3. Corporations and normal people often find hardware hackers quaint.

    This is a damn cannonball fired point blank at Canon. Nothing quaint about it.

    The people often featured on this site prove to me, at least, that yes, there often is a single intelligent human brighter than entire corporations. I never take someone with a brain and technical chops for granted.

    This is the real power of an ethos- open it, own it- and make it better.

    I’ve never seen anyone watercool a camera, CCD sensor yes, but not a full camera. Well done!

    1. Clearly not that intelligent if you don’t understand that this isn’t a movie camera, and never will be, regardless of how long it shoots 8k video.
      Yes, you can shoot great video on a Canon DSLR (can’t speak for this non-DSLR camera, but it’s probably pretty good), but it’s not a movie camera. There are significant and much more fundamental differences between these and the C100 etc.

  4. I was wondering a bit why anyone would spend USD4000 on a camera that overheats so easily, so I had a peek at the manual.
    Maybe the setting for “overheat control” was wrong (Page 372)

    ( Quote )
    Overheat Control
    Set to [On] to conserve battery power and help prevent the camera from overheating while waiting to shoot.As a result, it may enable you to record movies over a longer period.
    ( /Quote )

    Source:
    https://cam.start.canon/en/C003/manual/c003.pdf

    1. That “while waiting to shoot” part tells me this function probably turns off the processor while it’s live viewing but not capturing and processing video. It’s like a video pass through of some sort I’d wager. That’s not the problem shown in the video where capturing for 20 mins causes the camera to overheat.

  5. First Cannon is a group of people working together to sell stuff. That’s what a business is.
    Second the complicated part is Cannon isn’t JUST engineering, to quote someone “their are no unsolvable technical challenges.”
    That said it is VERY unlikely the “engineers” are just to blame.
    Politely put A team effort means a team failure. To fix the blame is to never fix the problem. It appears Cannon has some management issues between product development technical design and production. As for “they didn’t want to”, I doubt anyone but the people involved know that for a certainty. The more likely scenario is someone in management said “no you can’t fix the problem, get it out the door YESTERDAY”. Remember they have to answer to someone (the engineers).
    Then their is the “almighty” schedule some people worship.

  6. I noticed that almost everyone assumes that Canon got it wrong – no one has asked the question of what was the metal plate that was removed?

    While the main focus has been on the thermal transfer it achieves, it also separates a power supply board from the processor and memory. Both aspects tend to be a bit of a nightmare from an EMC point of view – almost any metal plate will help reduce the psu board affecting the cpu+memory from an electric field point of view but what of the magnetic field?
    If the metal plate was “Mu-Metal” then this replacement to copper sheet may well have *impaired* performance rather than improve it.

    1. I was ready to disagree with you, thinking that the increased thickness should make up for it not being a special alloy, but I looked up the permeability data first.

      Boy was I wrong. Copper is, it turns out, worse than literal air for magnetic field blocking.

  7. Not to diminish this slightly over hyped hack (yay bigger/better heatsinks make electronics run cooler what a surprise) but does anyone believe [Matt] started out this hack with an overbuild watercooler contraption?

    He used IRT to detect and isolate the heat source, then noticed the inadequate heatsink and terrible placement of thermal pads. How is the next logical step a copper angle bracket with attached watercoolers?
    Oh, I know – the sub-textual advertisement for the watercooler stuff (heat exchanger, pump, radiator and so on)…
    This is why I don’t particularly like his YT videos.
    The hacks themselves are good but all this completely unnecessary and misleading influencer/ad-platform stuff gets me on the wrong foot.

    More on topic:
    Why not just cut out a part of the plastic back cover behind the flip&turn display (left of the CPU) and place a slightly bend strip of the thinner copper plate there (similar to [Matt’s] “first” watercooler variant but with the copper to the left).
    It would dissipate the heat directly outside of the camera without going through the plastic and no need to 3D-print anything or build a new replacement for the existing aluminium heatspreader.
    Bridging the airgap between the rest of the backcover and the new copper cooler shouldn’T be much of a problem (hotglue, iso-tape, *-foam, etc.).

    1. Being him, with the other projects he has done the copper scraps and waterblock are likely out of the parts bin to prove cooling that spot is enough – its always possible that the worst offender for stopping it working isn’t actually the biggest source of heat, or that another spot needs improved cooling to run long term too… I know I’d throw water cooling at it as a test myself – as I have PC watercooling parts laying around that would be easy to use and certain to take as much heat as the copper could carry..

      I think if you want to radiate more out of the back you probably have to do major surgery to that back panel or accept a very odd bulge that makes the camera harder to use.

      My first thought was as yours seems to have been to cut the surface behind the flip screen down, giving a huge area outside to put flat sheet metal to act as heatsink (or strips out of it at least – so some bare copper gets direct access to the air) – but that would be really really tricky to do neatly, too much removed and you might end up warping the whole back part so it won’t fit back on (and you would still need to do a mod similar to his to get the heat out to the heatsink effectively enough – maybe just one of those extra flat heatpipes with light modification of stockplate)..

      In short his mod is very much simpler, safer (as I have no idea if you can even get a spare back case half let alone what it would cost…), and works so well I don’t think you would ever really need to do more – its a stills camera that can moonlight as a video device, not the video camera with massive batteries, better balance and ergonomics for use as a video camera so you are not likely to really need it to run forever…

    2. Oh, and I also thought of bringing the heat sink to the outside of the camera but it would seriously reduce the already minimal, but functional, IP rating of the camera when realistically, the overheating is only an issue when taking 8k or high quality 4k video, which to most of us is rarely.

    1. Very unlikely – there probably isn’t going to be any meaningful heat there at all when shooting stills, even using a live view on the LCD that probably will use that chip set its going to be at very low intensity compared to video shooting – so it can dissipate all the heat just fine.

      Canon definitely give their cameras a deliberate hierarchy, but they are not going to release a top end model that is bad at its primary function – taking stills, heck its really good at taking 8k video by all accounts with just one limitation, it can’t do it stock for longer than than approximately 20 mins.. Which is still a very very long time to shoot video for, particularly in 8K which will eat up all your bytes with astonishing rapidity. Still disappointing in such a pricey product, but at the same time everything about it screams designed for taking stills not movies, so understandable engineering choices, that they may just re-evaluate from now on…

Leave a Reply to DudeCancel reply

Please be kind and respectful to help make the comments section excellent. (Comment Policy)

This site uses Akismet to reduce spam. Learn how your comment data is processed.