This isn’t a paper but it feels a little like one, so I’ve included a bunch of material about how I did all this, and used that format. The TLDR: you can improve bed temperature evenness as well as reducing heat times if you insulate the underside of your heated print bed – but not if you do it the way you’re shown in youtube videos.
Adding insulation to the underside of an FDM printer heated bed is a not-uncommon modification to reduce bed heating time and conserve electricity by reducing heat waste. However, the effect of this modification on print bed temperature evenness does not appear to be often considered by hobbyists, and thermal imaging provided by youtubers showing this modification also show visible plate temperature unevenness both before and after this kind of modification.
On my Ender 3V2, I performed two versions of this mod, taking extensive temperature measurements beforehand, and again in each modified configuration. I found that while adding insulation across the entire plate did not worsen plate temperature irregularity, adding insulation across the underside of the plate except for the central temperature sensor resulted in decreased bed temperature irregularity, creating a more evenly-heated bed than stock.
Measurements also showed that a glass plate performs better on this metric than the metal heating plate itself, almost certainly due to its added thermal mass and inertia.
Background: I print a lot of objects with large areas of contact with the bed plate on my Ender 3V2 – storage system drawers, for example, where the entire base of the drawer is on the plate.
My most-often recurring problem is “corner lift.” This is a failure of adhesion, where the print doesn’t stick to the bed well enough to overcome the stress of PLA cooling, and the corners of the print curl away from the bed, creating a warped print.
This can happen to any model at any size, but it hasn’t really been a problem with small objects. Prints which reach into the corners of the print bed, however, are much more likely to have recurring issues.
As part of trying to solve this issue, I’d begun thinking about evenness of bed temperature. Bed temperature in general is critical to adhesion, and bed temperature variance – both cross location and time – is known to be a source of adhesion problems. I’d even bought some general-purpose thermal paste to play with, to see whether I might use that between the glass bed and the metal heating plate to improve heat transfer, hopefully making the bed more thermally stable.
(It was too goopy, -10 do not recommend.)
Recently, I’ve seen a few people talking about adding insulation to the underside of their heated beds to reduce bed heating time, and in the case of print farms, electricity use. Both are fair points and observationally supported. But they aren’t talking about bed temperature evenness at all.
I think they’re missing something by not doing so. I’ve watched an assortment of videos on how to do under-bed insulation. Some of them have had infrared cameras showing bed plate temperatures. On every one of those that I’ve seen, there’s been a clearly visible range of temperatures on the bed plate, always with a large hot spot in the centre and lower temperatures towards the edges – particularly in the corners.
While the people making the videos clearly demonstrated time-to-heat and electricity consumption reductions, the print beds shown did not appear to become any more evenly heated.
Of course, no one commented on this, as they weren’t even looking for it. But I think it might matter.
Methods: Sadly, I don’t have an infrared camera. But I do have a laser thermometer and I know how to use it.
I placed one sheet of plain white thin printer paper on each plate, cut to size and taped down at edges. This is to create a uniform surface for temperature measurement, as surface differences can exaggerate completely invalidate temperature readings.
Spot-checking for good paper contact as I went, I measured temperatures in an asterisk-shaped grid across the print bed, taking seven samples running along lines from corner to corner, as well as mid-plate edge to opposite mid-plate edge. This resulted in a total of 25 samples per plate. (Usually. I had to throw a couple out as I figured out a little late what paper distortion looked like.)
Using these readings, I calculated mean average, standard deviation, and margin of error at an artificial 99% confidence (vs. mean as reference) as a measure of actual temperature deviation range. All temperature measurements were taken at a bed set to 60°C, waiting two minutes after reaching stable temperature according to the printer’s control panel before measuring.
Measurements were taken of the glass bed as stock, the metal heating plate as stock, the metal plate with added insulation across the entire underside, the metal plate with insulation across the underside except the area containing the temperature sensor, and the glass bed with insulation across the underside except the area containing the temperature sensor.
In all cases, the bed levelling screws were installed as per factory. The bed was not levelled, as that was not considered relevant to the question at hand.
Findings: In the stock configuration with glass plate, surface temperature measurements ranged between 50°C and 57°C, with the more central area around the temperature sensor maintaining a tight 57°C. Only one location showed 50°C; the more typical lows were 53°-54°C in the corners. Mean temperature was 55.5°C, standard deviation 1.50, margin of error at 99% pseudoconfidence vs. mean of +/- 0.683 or 1.23%.
Removing the glass and measuring directly against the heating plate produced more highly variable numbers overall, though without the 50°C low. Readings ranges between 52°C and 58°C, with surprisingly sharp dropoffs even in the middle, with readings of 53°C within 10cm of 58°C readings. Corners and edges were again mostly the coolest points on average. Mean temperature was 55.72°C, standard deviation 1.99, margin of error at 99% pseudoconfidence vs. mean of +/- 1.025 or 1.84% – noticeably worse than without the glass.
Disassembling the plate, adding the insulation to the entire underside (as per manufacturer intent), and then reassembling the plate produced a noticeably faster heating time. (As that was not the point of this experiment, I didn’t specifically measure the difference.) Temperatures were somewhat less uneven in this configuration than previous measurements against the bare metal, though were still rather more uneven than stock with glass, ranging from 53° to 58°C. Mean temperature was 56.2%C, standard deviation 1.37, margin of error at 99% pseudoconfidence vs. mean of +/- 0.74 or 1.31%.
I then disassembled the plate again and removed the insulation from over the temperature sensor. It is worth noting that the temperature sensor as shipped from factory already includes a layer of insulation, though thinner than the particular typical insulation I ordered.
This resulted in a metal heating plate with sharply more temperature evenness than the stock glass plate, with temperatures ranging from 56° to 59°C. Surface temperatures also improved further relative to machine reading, taken from the underside. Mean temperature was 57.4°C, standard deviation 0.69, margin of error at 99% pseudoconfidence vs. mean of +/- 0.36 or 0.62%.
The final step was to add back the glass plate, after cleaning both contact surfaces. This resulted in a very tight range of temperatures between 56°C (all of which were rounding down to whole degrees) and a high of 58°C (rounded up to that), for a probable real range of somewhere under 1.5°C. Mean temperature was 57.0°C, std. deviation 0.45, margin of error at 99% pseudoconfidence vs. mean of +/- 0.23 or 0.40%.
This is dramatically better, and a temperature variance reduction of roughly two-thirds.
Results and discussion: It’s not reasonable to produce a widespread conclusion from a single printer. But in this particular case on this particular printer, bed temperature evenness was substantially improved by adding under-bed insulation, but that difference was only meaningful when the sensor was not included in the insulation area. The maximum range of temperature variance across the fully-assembled glass bed was reduced from 7°C (or 4°C discarding the worst case) to around 1.5°C, which is approximately the margin of error of my laser thermometer.
The modification also produced temperature results marginally closer to that shown by the actual temperature sensor, moving from a mean of 55.5°C to 57°C, vs. a sensor reading and setting of 60°. This is flirting with margin of error for my thermometer so may not actually be meaningful, but it’s close enough to meaningful that I’m pretty sure it actually is.
I was not surprised at all to see that the bed surface temperature in shipping configuration was uneven. This had been visible in multiple videos of printers shot using IR photography showing clearly uneven temperature, and there was no reason for my printer to be any different in that respect. However, both the range of surface temperatures and the degree – HA! DEGREE! – of improvement surprised me.
I further expect that the stock bed’s tendency towards a hotter centre would be exaggerated by large-contact-area prints, given that the plastic itself will form another layer of top-of-plate insulation, often directly over the temperature sensor. This would all but certainly make the situation even worse.
Hopefully, this new configuration will result in fewer issues with curling on prints with large bed contact areas. As I don’t have two printers (much less two identical printers) I’m not going to do side-by-side comparisons, but it would be interesting and possibly useful were someone else to make similar modifications, compare the two via additional tests, and report back with results.
Maybe somebody even will. If you try it, let me know. ^_^
Posted via Solarbird{y|z|yz}, Collected.
![[personal profile]](https://www.dreamwidth.org/img/silk/identity/user.png){kind=small}
