Following on from my last post (part 2), I have completed the minor design changes and made and fitted the new parts. I’ve also received the 80 Watt heater cartridge that I ordered so I’ve fitted that too. I finally managed to track down a 60 Watt heater which I have only just ordered so not yet received. I’ve also received the higher flow rate fans and have made and implemented some design changes aimed at improving air flow distribution over all six heat sinks. I’ve carried out a lot of thermal testing in various configurations and have ended up with a configuration which has exceeded my expectations.
2. Design changes from last post.
In my last post, I mentioned two things. Firstly that the cap-head screws holding the top plate onto the combining block made it difficult to get a tool in to tighten the heat sinks properly. Secondly, when I modified the inlet block with deeper holes for the Bowden clips, this meant that the heat break tubes stood up above the countersink which made it difficult to load filament. I’ve now made new side plates which are 4mm taller. This effectively raises the inlet block by 4mm so the tops of the heat break tubes now sit below the countersink. Here is a picture with the clip inserts fitted but without the actual plastic Bowden clips.
You can see that the heat break tube is below the countersink at the bottom of the larger hole as it was originally designed to be. So now when I load filament, as it exits the Bowden tube it is guided to towards the centre of the heat break tube rather than catching on the edge.
Here is a close-up picture of the countersunk screws which replace the cap head screws in the inlet plate.
Here is a picture with the heat sinks fitted.
Now that the cap-head screws aren’t in the way, it’s much easier to get a tool in and tighten the heat sinks. I tighten the two inner ones first, then the four outer ones. The plate is just under 4 mm thick so that the bottom of the copper nuts seal against the combining block below. I have to be careful not to over tighten them because it’s easy to bend this plate. A couple more screws at the left and right hand edges would help but I don’t have room to fit them unless I make the combining block bigger. I’ll see what happens when I start testing and if there are any leaks, then I may need to make that design change.
3. Heater tuning
Now that I have an 80 Watt heater, I ran the PID heater tuning again, which failed when I only had a 40 Watt heater. This time it worked really well. I was expecting the 80 Watt heater to be far too powerful but that isn’t the case.
I will digress a bit at this point because a few people have said, “What does it matter if the heater is too powerful – the PWM control will take care of it.” This is true but my take on it is that if a Mosfet fails, it can fail in such a way that the heater remains on at full power. When I tried the 80 Watt heater cartridge on the Diamond 5 colour, the predicted full power temperature was around 660 deg C. That’s the temperature at which aluminium melts! “So what…..” some people have said “…..just fit a thermal fuse”. Well for sure I could do that. But it’s an added component, which needs additional wires. So it adds complexity and cost and if the thermal fuse itself fails to trip, or it falls off, then the machine could still be in a dangerous state. So choosing a heater of the correct wattage, means the machine is inherently safer, at a lower cost and with less complexity than choosing to use an over powered heater. At least, that’s how I see it……
Getting back on topic, after tuning this 80 Watt heater, I had no warning message from the firmware about potential over-temperature. The gain was 403.5 which I’m lead to believe is a fair indication of the temperature that it might reach if left on at full power. That’s not outrageously high given that Polycarbonate and some other “exotic” filaments need in the region of 300 deg C plus. So I’m reasonably comfortable with that and of course, that is without any filament. Once I get 6 lots of filament loaded, the maximum temperature that could be reached might come down. On the other hand, I do have plans to eventually make and fit a silicone sock which will have the opposite effect. In which case, I may need to drop down to a 60 Watt heater. These are all decisions for the future. But the PID implementation using Duet is really good. From cold, the temperature rises quickly to 200 deg C and levels out with only a 0.3 deg overshoot which soon settles.
4. Temperature Gradient Test.
This was the same procedure as my previous post. That is to say, I used a thermocouple to measure the temperature inside a filament tube from the nozzle tip, up to the top of the “hot block” in 5mm increments. I wanted to run this test again for two reasons. Firstly, the previous test was done using the 40 Watt heater which could only attain a temperature of about 160 deg C. Secondly, this 80 watt heater is only 20mm long but the tube into which it fits is about 32 mm long. So the heater sits in the bottom of the tube and I needed to make sure that sufficient heat would conduct to the top of the block. Chris from Slice Engineering assured me that it would but I wanted to double check.
Here are the results.
|Nozzle – 5||199.0|
|Nozzle – 10||200.6|
|Nozzle – 15||200.6|
|Nozzle – 20||198.3|
|Nozzle – 25||196.8|
|Nozzle – 30||195.0|
That looks fine to me (sorry for doubting you Chris). Ultimately, I tend to print at whatever temperature gives me the best results and I largely ignore the filament manufacturers’ recommendations. With the Diamond, I found the ideal temperatures to be at or even slightly below the recommended minimum. So for example I tend to print PLA at 180 to 185, PET-G at 210 to 220 etc. Maybe the long melt chamber has something to do with it? In any case, if the core temperature is a couple of degrees hotter than the nozzle tip and the temperature at the top of the block is a couple of degrees cooler, I don’t think that will have any adverse effects whatsoever.
5. Heat Break Assembly Cooling Re-visited.
While I was waiting for the higher flow rate fans to arrive, I gave some thought to the potential air flow pattern or rather the distribution of air flow across the heat sinks. One thought I had was that I designed the hot end with slots which would allow a heater or thermistor cartridge to be removed without any disassembly of the hot end itself. But it occurred to me that these slots would allow air to escape from the top and bottom, rather than flow over the heat sinks. Here is a picture by way of explanation.
You can see the two upper slots, just in front of the fans. Of course, the same applies to the bottom, and the sides are open too (of necessity when both fans are pushing air in). It occurred to me that the air flow leaving the fan blades might have a radial component as well as an axial component and if so, much of the air that should be flowing over the heat sinks is being wasted. A couple of internet searches revealed that I might be on to something and one solution was to simply fit a tube onto the outlet of the fan. Also fitting a bell shaped inlet apparently helps but I don’t have room for that without making the carriage mount bigger and potentially losing some axis travel. Other people on various forums have also mentioned turbulence and again, internet searches resulted in a lot of technical papers about the design of vanes to make the air flow more laminar. Much of this was beyond my comprehension as I’m not an Aerodynamicist. However, I decided to print a couple of adaptors. One is just a simple tube, the other has cross shaped vanes in it. The tubes themselves are only about 13mm long.
Here they are.
Here is a close up of the bottom of the hot end with tubes fitted.
Finally, the carriage mount itself partially obscures the side outlets. So to be fully representative of what I might expect when the hot end is installed on the printer, I thought it best to fit it to the carriage mount. For readers who may not be familiar with my machine, I use two parallel rails for the X carriage, and the hot end “hangs down” between the two.
So this was the “test rig” that I used. Which is all as before, except that the hot end is now fitted to the carriage mount which is clamped to the printer frame, rather than using the two threaded rods.
5.1. Heat Break Assembly Cooling Tests
To avoid any confusion, I’d like to quickly explain my interpretation of some terminology. To me, a hot end (which is itself a confusing term) consists of three parts. A hot part and a cold part separated by a thermal barrier called a heat break. The heat break itself is usually a short section of low conducting, thin walled tube. But of necessity it has to have some means of fixing to the hot side and usually also has some form of heat sink on the cold side. So I call this tube with it’s fixing and heat sink a heat break assembly. In the following text, when I refer to a heat sink, I mean that part of the heat break assembly which is attached to the thin walled tube, just above the heat break zone.
As I stated in the introduction, I ran a number of tests with various configurations. As before, I marked the thermocouple by putting a piece of tape around it so that when the tape was level with the top of the Bowden clip, the thermocouple tip was at a point between the lower two fins of the heat sink, just above the heat break (+- about one mm). I ran the heater at 200 degrees C, waited for everything to stabilise, recorded the temperature, then moved the thermocouple to the next heat break assembly, waited for it to stabilise and repeated for all others. Rather than present all the results as a single spread sheet which would need to be referred back to, I’ll present each set of results and discuss them individually.
Unless otherwise stated, the tests were done using “push – push” fan configuration. That is to say, both fans blowing inwards towards the heat sinks and exhausting out of the sides.
It occurred to me that what I didn’t do the first time was establish a baseline with no cooling at all. So here is the first test with the fans not running.
|Heat Break Assembly Temperature tests at 200 deg C hot end temperature|
|FL=Front Left, FC=Front Centre, FR=Front Right, RL=Rear Left, RC=Rear Centre, RR=Rear Right|
This was quite a surprise to me and illustrates just how important having some air flow over the heat sinks really is. What is really significant is that the temperature is way above the glass transition temperature for just about any filament. For PLA, the glass transition temperature is about 60 deg C, for PET-G it is about 75 deg C, and for ABS about 105 deg C.
So any static filament inside the heat break with no fans running could soften and deform and potentially cause a blockage (note that I have emphasised the word “static”). This problem will be worse with all metal hot ends that have large, often stainless steel, heat sinks with high thermal mass. Slice Engineering tell me that the mass of cold-sided metal (i.e. metal above the heat breaking zone) in thermal contact with the filament, with the Mosquito heat break assembly is about 2gms. If the filament is moving forward, very little room-temperature filament needs to pass through the tube to cool 2 grams of metal to a “clog-safe” cold-side temperature. So for “normal” printing the filament itself will absorb the small amount of heat that gets past the heat break providing that filament is moving forward. Slice engineering also tell me that (quote) “if you don’t pause the flow of filament for a significant period of time (5 seconds is pushing it for some PLAs) you can print any material indefinitely with an uncooled Mosquito, even a Mosquito heated somewhat via conduction through metal bracketry connected to a hot extrusion motor in a passively heated chamber”.
But mixing hot ends have a unique problem in that it is not always possible to keep all of the filaments moving forward, all of the time. I could probably get away with having one or two percent of most colours mixed in with everything else, but white filament is a particular problem. This will become apparent in my next post but be assured that even 1% white mixed with red will turn “Ferrari Red” into “Barbie Doll Pink”. So having a low heat break temperature at all times, when filament might be static, is important to this mixing hot end.
It is sound advice and well known that one should never turn off a printer until the hot end has cooled sufficiently, because if the filament is static inside the heat break, with no cooling being applied to the heat sinks, that filament could soften and deform as it cools. So the next time the printer is used, one could start with a blocked heat break. Obviously that depends on the filament type and in this respect PLA is one of the worst culprits and heat break/heat sink assemblies with high thermal mass will be worse than those with low thermal mass. One thing I always do is a retraction at the end of a print as it helps prevent oozing when starting another print from cold and the hot end is heating up. It has just occurred to me that this retraction would also relieve any pressure in the Bowden tube which might help alleviate any filament deformation in the event of the cooling fans being turned off prematurely (but this is just conjecture on my part).
Moving on, the next test was to establish another baseline using the original quiet but low flow fans. Here are the results.
|42.5||60.6||50.2||49.0||51.6||38.9||Low flow fans – no adaptors|
This is the difference that just that small amount of air makes
|91.5||79.8||85.4||93.1||99.0||102.1||Difference compared to no fans|
This result is similar to the test I ran previously with the hot end at 160 deg C. That is to say, there is a spread of about 20 degrees C between the hottest and coolest heat sinks. The centre ones were the hottest, then the heat sinks on the left of the fan (so front left and rear right), and the heat sinks on the right of the fans were coolest (front right and rear left).
The next test was done again with the low flow fans but this time, I added the straight tube adaptors (without vanes). These are the results of that test.
|36.9||50.2||41.7||36.6||47.6||36.7||Low flow fans – plain adaptor|
That’s fairly significant. If we look at the difference compared to no adaptor we get this.
|5.6||10.4||8.5||12.4||4.0||2.2||Difference compared to no adaptor|
So the plain adaptor reduced the temperatures by between 2.2 deg C to 12.4 deg C.
That hottest temperature of 50.2 deg C is almost 10 deg below the glass transition temperature for PLA, so I’d be happy with that but cooler still would be even better. Also, there was still a noticeable difference between each heat sink. If we compare them with the front left, we get this.
|0.0||13.3||4.8||-0.3||10.7||-0.2||Difference compared to FL|
For this next test, I decided to reverse one of the fans so they were in push-pull configuration, from the back of the hot end towards the front. Still using the plain tube adaptor. Here is the result.
|52.2||40.9||51.7||44.7||37.5||42.8||Low flow fans – plain adaptor|
Things did change around with the coolest heat sinks becoming the hottest and vice versa. But the front three heat sinks were all hotter than the rear three, probably because they were in the “lee” (to use a nautical term) of the rear heat sinks. Overall there seemed to be no net gain so I decided to abandon “push-pull” configuration and revert back to “push-push”.
For the next test, I used the tube adaptor with cross shaped vanes (still using the low flow fans). Here are the results.
|37.9||37.4||42.0||40.1||37.7||39.2||Low flow fans – vane adaptor|
This result actually stunned me. It was totally unexpected so I double checked everything and it repeated again really well. Comparing this with the same fans, in the same push-push configuration but with no adaptor we get this.
|4.6||23.2||8.2||8.9||13.9||-0.3||Vane adaptor vs no adaptor|
And comparing each heat sink with the front left we get this.
|0.0||-0.5||4.1||2.2||-0.2||1.3||Difference compared to FL|
So the low flow fans, with “vane” adaptor and a hot end temperature of 200 deg C, gives me a maximum heat break temperature of 42.0 deg C with 4.1 degrees maximum temperature deviation across all 6 heat sinks.
That was pretty amazing and exceeds all my expectations. But as I had already purchased some high flow fans, I thought it was time to test these. My expectation was that these would drop the temperature even further. The quiet but “low flow” fans were rated as being 3.5 cfm and the noisier “high flow” fans were rated as being 7.7 cfm. Given that I use two fans, that would increase the theoretical flow from 7 cfm to 15.4 cfm.
Here are the results using the high flow fans with the “vane type” tube adaptor.
|41.4||39.6||43.1||40.8||36.0||39.0||High flow fans – vane adaptor|
So this was a surprise. If anything one could say that overall, the temperatures were slightly higher, but certainly there was nothing to be gained. Out of curiosity, I decided to run these fans at lower speed just to see if that made any difference. So I changed the “S” parameter to reduce the PWM value from it’s maximum of 255 which represent fully on. I just measured the temperature of a single heat sink Here are the results of doing that.
|Temp (FL)||PWM “S” value|
So there was no improvement in running the fans slower and the temperature actually starts to increase at about 80% of full speed.
I wondered why these higher flow fans gave no further drop in temperature when compared with the lower flow fans. The fact that running them slower made matters worse seems to disprove a theory I had that all I was doing was increasing air pressure but not flow rate. I suspect it can only be that the shape of the fan blades (which are at a more acute angle) create more turbulence and so the tube adaptor and/or vanes inside that tube might need to be optimised for those particular fans.
Also, I did some “rough and ready” simplistic calculations which go like this. I cubic foot is 28,316.8 cubic centimetres. So 3.5 cfm is 99,106 cubic centimetres per minute. I use two fans so that gives me a potential air flow of 198,212 cm^3/min. Where the heat beak assemblies sit can be considered as an open sided cuboid shaped chamber of dimensions 5 x 3.8 x 4 cms. So the volume of that “chamber” is about 76 cm^3. If I divide the air flow in cm^3/min by the volume I get 2,608 air changes per minute or about 43 air changes per second. If only 10% of that air flow reaches the heat sinks, it still means that the air around them is being changed at 4 times per second. I would imagine that this is more than enough to carry away any heat that is being transferred from the copper heat sinks to the surrounding air. Hence the reason why increasing the air flow has no effect, and in fact these quiet, “low flow” fans are actually high flow for this application. That seems reasonable to me but I’m neither an Aerodynamicist nor an expert on thermal dynamics so I could be wrong.
Anyway, I carried out one final test which was to see if running the “low flow” fans at a slower speed had any effect. These are the results.
|Low flow fans|
|Temp (FL)||PWM “S” value|
There is not much to say except that I could reduce the speed to about 80% before I noticed any significant increase in temperature.
In conclusion, neither increasing nor decreasing the air flow gives any further gain in temperature reduction. It is possible that further refinement of the vanes inside the tube adaptor might give further gains. But the quiet “low flow” fans with the tube adaptors and simple cross shaped vanes give me a worse case heat break temperature of only 20 degrees above ambient (42 deg C actual) with the hot end running at 200 deg C which is more than good enough. So I’m not inclined to spend any more time refining these adaptors any further.
6. Air flow visualisation.
I had thought about using some sort of visualisation technique to evaluate the airflow over the heat sinks but it wasn’t a huge success. However, on the basis that it might help someone else, I thought I’d document my attempts here.
I have seen smoke being used in wind tunnels. I also happen to use e-cigarettes and make my own “juice”. For anyone who isn’t familiar with these devices, the main ingredient which produces the visible vapour is vegetable glycerine, of which I have quite a lot laying around. I also have spare cartridge heaters and power supplies. So I made a simple “smoke machine” (actually vapour machine) by drilling two 6mm diameter holes in a piece of 8mm thick aluminium. I put a heater cartridge in one hole and filled the other with vegetable glycerine. I connected the heater to the power supply and switched on. It kind of worked (but with caveats). Firstly, there was no control over the heater temperature so although it started to produce quite a lot a vapour, eventually the vegetable glycerine started to boil rapidly and hot liquid started to spurt out if the top of the hole. If I made another, I would use a larger diameter but shallower hole and control the heater – somewhere about 220 deg C should work.
I managed to get a picture of sorts. Here it is.
The vapour shows up quite well against a black background but it needs careful lighting (and probably more vapour). Getting the lighting right in my garage is not something that I can easily do so I abandoned this little experiment. I can however testify that the vapour does not set off a smoke alarm. 🙂
7. Thermal images
I mentioned in my previous post that I would like to take some thermal images of the hot end. This desire has been largely negated by the temperature tests that I have done using thermocouples inside the filament “tubes”. However, my good friend Tony Lock from Duet has kindly sent me his “Seek” thermal camera that can be used with an Android ‘phone. Unfortunately it has a micro usb connector and my current ‘phone has a type C connector. I had an old android ‘phone laying around which has the micro usb connector but unfortunately it won’t detect the camera when I plug it in (I think I remember now why I changed my ‘phone). In the mean time, I have ordered a micro usb to type C adaptor so I’ll try this camera on my new ‘phone when this adaptor arrives. Chris from Slice Engineering has told me that to get accurate thermal images, I need to spray the hot end with matt black barbecue or stove paint. I’m not sure that I want to do that, given that I only have this one prototype hot end. We’ll see……
8. Noise test
I don’t have an accurate sound level meter but I downloaded an app for my android ‘phone which I thought would at least give some sort of comparison. According to that app, the background ambient was 28 dB. With fans running and my ‘phone placed about 50mm away from them, the noise level was shown to be 30 dB. To my (aged) ears they are barely audible. Then I heated the Diamond hot end and measured the noise level from the 50mm, 27 cfm fan that it needs to keep the heat sinks cool. Again, with my ‘phone about 50mm away from the fan, I recorded a noise level of 62 dB. So although the readings may not be accurate, I can say that this hot end is a lot quieter than the Diamond hot end.
9. Next steps
Well I guess the time has come when I should actually load some filament and see what happens. Keen eyed readers will also have spotted a tiny piece of swarf in one of the holes in one of the above pictures (there are no prizes for anyone who spots it). I didn’t notice that myself at first and although I have done what I thought was thorough job or cleaning, I don’t have compressed air or anything like that available. There is a reasonable chance that other small pieces of debris may be “lurking” inside the mixing chamber. So my plan is to drill out a nozzle to about 1.5mm diameter and use that to flush out filament and any debris, before fitting a 0.5mm nozzle.
But before I do that, I want to design some test parts which highlight the “stripey toothpaste” effect and print these with the Diamond hot end. Then I’ll be able to compare the two in terms of how they mix filaments together (or not). Then I have to fix the locating dowels in the carriage mount. Finally, I’ll have to re-configure my machine and position the six extruders so that the outlets line up more closely with the new hot end inlets, as well as make and fit new Bowden tubes. I did think about this possibility when I redesigned the extruder gantry, so there is no machining involved but it’s still quite a major job.
But I’m happy with progress so far. I do have a hot end that I can change the nozzle on. The Diamond comes with a lump of brass which holds the heat breaks, heater and temperature sensor and which is itself the nozzle. The only available size is 0.4mm and for any other size, the brass cone has to be drilled out. Not only can I change the nozzle but I can do so with one hand and don’t need to support the heat breaks. The hot end is also smaller, lighter and quieter than a 5 colour Diamond. Of course, none of that counts if it doesn’t work……
My next post might be in a couple of weeks time as there are other things than 3D printing which require my time and attention.
Until then and as ever, I hope readers will have found something of interest in all of the above.