Sanding (or planing or scraping?) magnetic sheet for flatter bed?

A quick search didn’t turn anything up. Both of my corexy beds aren’t as flat as I’d like, though the new one is worse than the first. For making machine tools flat, there is a technique called “scraping” where the high spots are marked and removed in a process that takes some hours but can end up with a remarkably flat surface.

It occurred to me that I could use sandpaper, a very fine plane, or even make a scraping tool, and shave off the high spots of the magnetic sheet holding on the steel top sheet, until the bed is flat enough that it appears flat to a probe.

Has anyone here heard of anyone having done something like that?

When I want some thing flat, like the bottom of a plane. I have a granite flat stone. Put some water and wet and dry sandpaper down then the piece I’m flattening face down. Press on the piece and sand in rotating motion.

Don’t know if it’s applicable to your application.


My surface plate is only 8" x 12" — so smaller than my 3d printer beds :relaxed:

I’m curious whether anyone has done this specifically with magnetic sheet to flatten a printer bed. I couldn’t find references to that in a google search, and so far on twitter when I asked, I got only an old friend telling me I’m nuts and that I should send it back to the vendor. (All I can find when I search is people sanding the aluminum bed to make it flatter.)

SecKit are taking the problem with my SK-Tank seriously and agree that the .5mm variance is not acceptable, but it got me thinking about the case where I am the vendor. Having the SK-Tank up and running (currently relying on bed mesh compensation while trying to solve its out-of-flat bed problem) frees me up to do maintenance and redesign on my old corexy.

The bed I made in my first corexy that I designed and built in the skeleton of a tronxy x5a is not terrible — I can print across the whole bed successfully with PETG — but (without measuring) it’s clear that it’s not perfect. I could mount a dial indicator to the carriage and get a precise measurement everywhere and make it more flat. That bed is 330mm x 330mm, so I would need a much larger surface plate, and some pretty massive sheets of sandpaper, to make that strategy work for the bed. And as I think about sanding the magnetic sheet, it occurs to me that it might be a pain to get magnetic dust fully cleaned off of a magnet.

That’s why I thought about scraping, where you identify the high spots and selective remove material to make them flat. Shavings seem likely easier to get under control. This led me to thinking about a plane for better shavings. But a small one for good control. I don’t own a finger plane, so clearly I should do something about that!

I’ll see whether it’s any good. From reading reviews for that and the Stanley 12-101 that it is modeled on, I’ll probably want to use the “sandpaper on the surface block” approach to make sure the bottom is very flat. (I know, a real machinist would never do that — but it’s only a grade B surface block anyway!) Then I could imagine using shim stock under it to use my diamond cards and arkansas stone to hone the blade to shave a precise thickness off.

I still don’t know whether I’ll try it, but I incline toward the attempt.

1 Like

Am I reading this right you plan to flatten a metal plate with a woodworking hand plane?

Putting emery paper or wet and dry paper on the bottom of a hand plane might work after insuring the plane’s bottom is flat?

From my experience, a $10 plane will not be made with quality steel and therefore not hold an edge for shaving wood much less aluminum.

I thought he was looking into flattening the magnetic sheet material since that is a sandwiched between the bed plane and the removable bed surface material.

1 Like

@dougl exactly!

I know something about how to make metal flat. My mill doesn’t have enough Y travel for me to face mill the aluminum flat with my face mill, and tramming the head perfectly enough to use a huge shop-made fly cutter to flatten the aluminum would be a long exercise.

But I’ve already used very tenacious adhesive to attach a magnetic sheet several mm thick to that aluminum bed. Ultimately that’s the reference plane for the removable print surface. And it’s relatively soft, so probably easy to shave with a plane, I would expect.

I could find references to people flattening the aluminum, even doing it successfully by sanding it, using a straightedge for a flatness reference. But I didn’t find any references to trying to make the top of the magnetic sheet a reference surface.

Perhaps the solution is to sandwich a thin layer of mild steel between the magnet and the removable plate.

The top surface of this MS needs to be flat , but the lower surface just needs to be held by the magneic surface. Your thin flexible removable plate therefore has a flat surface on which to rest.

You may be better off with 3 strong , temperature tollerant (Alnico?) ) magnets that accurately position a MS plate at defined points. The flexible build plate adheres to the flat MS plate. Flattening the MS is much easier than working with ceramics.

This is the normal flexible magnetic sheet. Like a high-temperature (130° curie) fridge magnet. It’s not ceramic. As I said, “it’s relatively soft”. I don’t see the value of adding mild steel to the sandwich. One more layer between the heater and the print.

Machining the bottom of the bed to insert high-curie magnets is a thing that plenty of folks have done, but is also something that would have made more sense before I installed the heater, if I were going to do that. In a more general sense that doesn’t apply to the printer I designed but could otherwise be relevant, using a few magnets like that messes with inductive probing.

When I first made that bed, I was expecting to adhere permanently a thick sheet of PEI to the surface. If I had known that I was going to use a magnetic bed, I probably would have embedded high-curie magnets in the bed in the first place. But I’m not peeling off the heater to do that now! Even if I wanted to, that sounds like a lot more work than making the surface flat…

At this point it’s pretty clear that no one here has actually done the thing I’m asking about. I’m probably going to try it after I’m 100% certain that the new printer is ready to be my workhorse, as part of other work to improve the older printer. If I do, I’ll share what I learned.

Points taken.
I guess the trick then is to hold the flexible sheet in such a way that it is flat and ‘rigid’ enough to be flattened without deforming.
Ideally this would be the surface it normally rests on so that irregularities in it are accounted for…

You might get some tips on getting a flat surface from this video

Perhaps a sheet of glass could be used as the starting reference surface.

The flexible magnetic sheet is permanently bonded to the aluminum bed. There’s no trick to holding it. It’s not going anywhere, and removing it is possible but will destroy the sheet. I’ve done this once before with this type of adhesive, and the half hour I spent with acetone cleaning the adhesive off the aluminum was no fun.

The underlying bed should be flat.

I don’t know how precisely the thickness of the magnetic sheet and the adhesive layer used is controlled in manufacturing.

What matters most is whether it is tram to the movement of the head. Therefore, measuring its flatness relative to the head path is the key thing.

The head moves only on linear rails, so all the movements should be straight. This means that there is really only one kind of distortion in bed height feasible, which looks like “twisting”. Therefore, if the apparent shape of the lack of flatness had been twisted — that is, if opposite corners were offset from flat in the same direction, that would be a likely sign that the frame itself was not square, and the head was not moving in a plane. So if, for example, rear left and front right were too high, and rear right and front left were too low, that would most likely indicate an assembly problem with the frame itself twisted.

But the lack of flatness doesn’t follow that pattern. Therefore, the surface itself must not be flat.

In any case, I have the finger plane now, and I have a chisel I can try using as a scraper, and if those don’t work well I can buy a gooseneck cabinetmaker’s scraper.

1 Like

A fun problem and exercise :wink:
As you have linear rails above the catastrophe, perhaps they could bee used to guage the plate and get a map of the high and low spots.

Even something as simple as a felt tip pen mounted on the carriage and swept across the surface will pick up the high spots which could be addressed first.
I was originally thinking of a cam follower type ball or probe that could vary the capacitance to ground and give an accurate profile. Knowing the shape might assist.
I was not thinking of twist and skewness in the bed, rather of surface irregularities. What sort of dimensions are you talking of?

If it is gross deformations in the substrate, perhaps some light panel beating could remove the twist or other gross maladies?

Perhaps you can go the other way and build the surface with a very thin layer of (epoxy? polyester? silicone?) liquid that will fill the low spots.

Just thinking aloud.

I’m really not looking for suggestions how to flatten the cast aluminum.

There’s no catastrophe with my old corexy bed. I haven’t probed exactly, I expect it is within 0.15mm or so based on the quality of prints. This is not gross deformation.

I keep saying that gauging is not a problem. I have a dial indicator. I can mount it to the carriage. Similarly, I could mount a probe (either bl-touch or inductive) to the head and use software to make a map, as I have in my newer printer where it has an inductive probe.

It’s even possible to do manual bed mesh visualization using paper to measure nozzle height above the surface. I did that years ago when I discovered that my old troxy x5s had a horribly warped bed. That was what set me on my path to tear it apart and build an otherwise completely different printer within its frame. But it sure took a long time to manually measure 25 different locations! And that’s not really enough for this exercise; I think 7x7 49-location grid is as coarse as makes sense. But here’s what I did long ago with gnuplot entirely manually:

Today, OctoPrint’s Bed Mesh Visualizer and bed probing makes this a lot easier.

Anyway, that’s literally using the linear rails to gauge the plate.

330mm x 330mm

My point here is that it’s not clear that magnetic sheet is manufactured to tighter tolerances than cast aluminum plate. It’s quite reasonable to think that the bed surface might not be perfectly flat even if the aluminum bed itself is perfectly flat.

My thought is that maybe my idea could be a more widespread fix that anyone with a 3D printer with a flexible magnetic sheet, with no specialized machine tools, might be able to perform to improve their bed flatness.

  • Until satisfied with bed flatness:
    • Do bed mesh visualization
    • Lightly scrape the high spots
    • Remove all scrapings thoroughly so they don’t interfere

I was just curious whether anyone else had seen this idea.

It seems pretty clear that it’s a new idea here since everyone is telling me to tear my printer apart and do something else instead.

Hi Michael,
I guess you have answered your own question

  • doesn’t look like anyone else has done this, or will talk about it
  • decide on the toolset you will use
  • decide on what to use as the reference ‘plane’
  • test gingerly and make adjustments where necessary

Is it possible that you have air bubbles or dust between the aluminium and mag plate?
A bit like a cell phone glass protector screen. You may need to find a way to remove the air.

All things are possible. :relaxed:

I cleaned the aluminum plate with acetone and IPA, then wiped with a lint-free cloth. I slowly rolled it on from one edge with a large, clean, heavy aluminum roller, pulling the backing off the adhesive incrementally with a hemostat to avoid touching the plate. So at least I did what I understood to be best practice to avoid air bubbles and dust. Whether I succeeded? Who knows? I don’t see evidence of bubbles from visual distortion of the upper surface.

Sorry, I didn’t mean to sound annoyed.

And if I succeed I’ll post about it. Or, well, if I fail I’ll post a warning! That’s always a possibility! :grin:

So here’s an idea based on the thread dialog up to this point …
First let me state, on a CNC router there is often a spoil board that needs to be machined flat every now and then.
Usually the operator has already made a gcode for the surfacing operation and there will be a large shell cutter to zig zag over the whole spoil board.

This is the idea …
Unscrew the filament nozzle and replace it with a short bolt that has a large flathead.
Apply adhesive backed, fine grit sandpaper to the bolt head.
Create a zig zag gcode to traverse the entire bed.
Run the “Z” down to barely touch the bed and then run the gcode.
Since the operation is a bit of an unknown it’s probably best to go conservative on the speed and “Z” load.


I hadn’t thought of using an M6 bolt in the hot end to hold sandpaper. That’s not obviously crazy.

The trick would be a large enough bolt head to not clog quickly. I could actually make a tool like this on the lathe, something the same length as the nozzle, but I could cut it from 25mm aluminum stock and have a nice wide surface for a piece of sandpaper, and use double-stick tape to hold it in place. I’ll ponder that! :relaxed:

I wrote a surfacing gcode generator and posted it here not too long ago…

1 Like

SecKit wanted me to pay shipping and part of the material cost for replacing the bed, which I wasn’t satisfied with.

I did finally receive the replacement magnetic sheets (with the additional flex beds, so I can have one super clean for PLA, another with hairspray as a release agent for PETG, and one in reserve for whatever else I come up with). So now I feel OK trying some things out; worst case, I can recover by replacing the magnetic sheet. And if I need to do that, I can use a dial indicator on the carriage to indicate the flatness (or lack thereof) of the raw aluminum bed before I put the next sheet on.

So now I decided to experiment on the SK-Tank, since I have the ability to recover and it’s really badly out of flat. I figured the worse that happens is that I contribute to the journal of negative results, and honestly I’m getting kind of curious here.

I attached a dial indicator to the hot end and moved it around, and it’s pretty clear that what the inductive probe is reading is actually true. This is not just, say, variation in magnetic field causing the probe to read different in different parts of the bed. And what I see looks about the same whether I’m probing with the PEI spring steel sheet installed or not, so it’s not varying thickness of that sheet. It really is the substrate.

I decided not to start with the replacement nozzle small sandpaper block. It’s an interesting idea, but if I overdo it, there’s a lot of flex in the head compared to a machine tool, and if it catches an edge it could dig in instead. So I’m putting that at the back of the queue of ideas to try.

I honed the finger plane on an arkansas stone to make sure the blade had a good edge. However, the aggressive angle appropriate for wood is too “grabby” in the magnetic material. So the finger plane may be useful for a wood project but won’t be useful here.

I honed a typical ½" chisel sharp and tried a few things. The only thing that worked at all was to use it basically like a shaper, holding it straight up and down and pulling it flat-side-first rapidly across the bed while putting very consistent pressure on it. I tried clamping it to a square wood block to make sure that it was aligned, but that was an abject failure; it did nothing.

Like scraping, I identified high spots and worked them in a cross-hatch pattern to bring them down. Every once in a while I’d clean the bed, put the flex sheet on top, probe 121 points (11x11) twice each, and visualize the mesh again.

That was really slow, and I found that I was not working out all the slight gouge marks from not keeping the chisel perfectly upright. Additionally, it was really bad for the chisel; it didn’t stay sharp long under this kind of abuse for which it was not designed. So I stopped using the chisel.

A gooseneck cabinet-maker’s scraper plane might do the trick, but I don’t have one.

I attached 60-grit sandpaper to a flat block and started sanding the high spots. The resulting magnetic dust is harder to clean up than scrapings. But it is working faster than anything else I’ve tried so far. I’ve spent a few hours working on this project, and I am not done yet. I started out worrying about removing too much too fast, but that hasn’t been a real problem. The sanding block clogs with gummy lumps of magnetic surface bits within 10-15 seconds and then doesn’t remove so much. I then took several minutes to pick the lumps off the sandpaper with the tip of a nail. Sanding a 90° corner of a block of wood cleans them out in less than a minute, but it’s still one third sanding two thirds cleaning at best. And it’s still really slow going.

I know that when cutting some soft metals (like brass) you don’t want any top rake, so I tried clamping a HSS lathe tool blank with a 5° pre-ground edge to an aluminum block as a kind of plane with a 90° cutting angle and 5° relief, but that didn’t work much better, and it was back to sanding.

I’m to the point where I’m starting to worry about having enough of the magnetic sheet left to hold on, and I finally realized that I have some shim steel stock in the shop in one, two, three, and five thousands of an inch thickness. I also have empty soda cans, which are about three thousandths. The soda cans aren’t magnetic because they are aluminum, but for a layer underneath some steel I might get a sandwich that makes the bed more level.

@shorai you were on to something!

Overall trying to flatten the bed by removing some of this gummy magnetic sheet has been an exercise in frustration.

1 Like

Wild idea! You might try a cabinet scraper. I am not sure how it will cut nor how long an edge will hold but it may be worth a try. Woodworkers use this tool and technique to get very fine surfaces that are pretty flat after plane-ing. There are lots of YouTubes on how to use and sharpen.

At least you would not have to worry about it “loading” up :slight_smile: !

Not many woodworkers can use and sharpen this tool and it may not work in your application but your description of using a chisel as a scraper jogged my memory.

There is a range of costs based on the quality of steel used. The cheaper ones, just like any cheaper tool steel need more sharpening than the more expensive ones.

You will need a good single-cut file and a burnishing tool to sharpen and then roll the edge (burr). I made a sharpening jig to hold things at the correct angle to create and fold the edge. If you decide to try this I can provide more info on the jig. It only takes me a few minutes to re-sharpen during use. I use an old lathe alignment tool as a burnisher. The shaft of most cutting tools is hard enough to burnish the edge of a scraper.


yuppers! I had been thinking about a gooseneck shape to avoid the corners:

I read up on the burnishing process, and wondered about getting the angle right. I recognized that the right burr might be different between wood and the magnetic sheet. The uncertainty that kept me from going that route. I saw how differently my lathe tool blank worked on wood vs the magnetic sheet; it actually put a nice surface on the a piece of baltic birch but gouged the magnetic sheet.

I now realize that the technique means that straight would be fine, I wouldn’t need the gooseneck. I didn’t know that until I started reading though!

1 Like