I can happily report thew new rails are quiet!
@edf @brent113 i think you questions are all spot on
My core observation is the height measured is very very consistent in any given location by the head sensor (as repeated in snapmakers own test). This implies to me the issue is possibly one or more of:
- issue with the sensor and how it detects surface height, maybe by being affected by the density of metal / differences in metal at different locations - aka it thinks it is at +0.1 across the board but in reality it is at -0.15 to + 0.15 - as such the mesh calculation is bogus
- there is a math problem with their bilinear mesh and catmull mesh calculations
- there is some level of repeatable imprecision on the Z axis that means the application of the mesh is physically incorrect
whatās most interesting to me is the test with the straight edge down the same Y axis as the pressure test shows that there is a physical deformation that matches the pressure difference (and that cause my large prints to be great at center and too close at edges causing filament to be welded to the surface).
As such i donāt believe this to have a root cause mechanical issue as snapmaker assert.
it also may explain why folks who move to glass get better results (the flatness of the glass plus not using the sensor naturally masks the issue) and possibly why those with the IR mod see less issues. I can do one more test, i havent done this test on glass - that can rule out the sensor and confirm or refute my hypothesis, will try and do it this week.
1 Like
hmm i did manual test on stock bed, so i guess it will be surprising if the glass results are different (or rather flatness of glass will likely mask the issue) we will see.
If density is interpreted by the sensor as distance (and I donāt know, I never read through all the threads on the sensor), then the sensor is never going to work well. I donāt know enough about this to comment but it is easy to check, if there is a way to directly read from the sensor - put a piece of steel and a piece of aluminum, machined to precisely the same dimensions, under the sensor and check the reading.
I doubt density is it, though. More likely itās a matter of the coarse resolution of the mesh measurements. If the sensor was sampling, say, a 1mm square area, then consistently taking the mean or median (or even one end of the range) of those points as the measurement, then a virtual mesh would be created that is probably more representative of the surface.
Iām starting to lean towards creating the mesh in software without the A350ās help. Have to think that over a bit. I remember @Tone did a bit of work on this at the beginning of the year.
That actually shouldnāt be too bad to debug, if we (being, the users of the forum) have an example of mesh coordinates that produce incorrect results. I havenāt actually looked at the source code with an eye towards building a test scaffolding, though.
That is probably the most frightening prospect.
I can conceive of a workflow where the printer warms up and then pauses at the center of the build plate, while measurements and such are taken, adjustments made, etc, before the job starts. Thatās not too unusual for high-precision work. Of course, any adjustments made on the touchsreen will be perma-saved (why isnāt that a configurable option anyways), so maybe this is better handled through the console. If it is only the Z offset we have to worry about (due to the imprecision of the Z axis), then a simple shim or gage block can be used to determine if the height needs to be changed, and this would only need to be done in one location. Basically an extra minute of setup. Not too bad.
ā¦ except for Z hops. Ugh.
I have a slighty-too-small glass plate (it was cheap and I had planned on machining a ferrous frame to hold it to the uild plate) as well as the PEI sheet you got. Never took either out of the box. Did I mention yet I donāt have the free time for this nonconsensual hobby? 
2 Likes
lol, my wife has been very very patient with me talking about this to her and the time i have spent
2 Likes
Very good info, Ive parked my machine a while ago for 3d printing because of this. Its really making me tempeted to rip it apart and just drive it all from a Duet WiFi board like all my other printers. Its just I do use the laser engraving and CNC from time to time for paid work.
Also I work in military aviation in UK for Eurofighter Typhoonā¦ always nice to meet some in the industry 
2 Likes
Snapmaker has reached out to me to let me know they are continuing to work on this issue.
This is a little more encouraging.
2 Likes
Man you had some really good calibrations there for a while. what happened?
new rails made the problem worse, plus i settled for too close at the edges to get it right in the center - thats ok until you have a PLA that likes to weld itself into the default surface
sigh, just installed new firmware
put glass on top of heated mat (firmly held)
did full manual calibration, saved (mesh looked excellent except for back row - need to figure out if that is bed tilt or procedural error on my part).
then homed
then i returned the nozzle (by hand navigating) to dead center of bed (i am super careful doing this) and found that at z = 5.5 it is hard pressed against the glass.
only conclusion is is the z offset is not saving correctly - i have seen this once before, after another firmware upgrade - i think a powercycle + recalibrate might fix itā¦ be careful out there people and donāt trust that it knows where z = 0 actually isā¦
1 Like
This fixes it.
I realized after last post the xyz coords was totally broken.
- z reported 5.5 touchscreen when at 0 and was pressed hard against glass
- x reported it was 150 on touchscreen when it actually was at 160
- y reported it was at 67 on touchscreen when it actually was at 170
I know this because i lined the nozzle up on the intersection of the marked lines 170 x 160 both times - before powercycle touch screen reported incorrectly, after reboot it reports correctly.
i recommend everyone hard boot after either application of firmware or first calibration - not sure which, i only had to hard reboot for it to sort itās self out, i didnāt need to recalibrate.
2 Likes
This is just an observation for folks playing along at home with this thread.
I use an imperial square because the gradings are closer than on a mm.
This explains the issue i was having where a fixed XY coord at the rear of the bed was one time no pressure on the card and one time has pressure. Does this matter? I have no idea. I can say the last set of v1 rails did NOT have this issue. The new v2 rails do.
This may be what caused my weird calibration mesh
Still working on why the concave nature of my bed+pei plate is not corrected for, have confirmed that glass does help (aka avoids the issue) - but still means fundamentally calibration IMHO is not accounting for actual shape of a surface.
I have more tests planned over the holiday (as requested by snapmaker) to identify the issue with calibration overall.
2 Likes
Based on what Iāve read here, Iām going to take a level and measure the bed from corners and centers to map the shape of the bed. At the moment, Iāve got the laser bed on it. After that Iāll measure the heated print bed. I also have a leveling app on my phone which might be useful.
Then Iāll take the bed off and measure the flatness of the frame between the holes with a straight edge.
Just to see what I can see.
A level will only tell you which direction the high points are canted in.
With good straight edge you can use a flashlight to spot the gaps where the low points of the bed lie. A feeler gauge can be used to determine the size of the gaps.
The last calibration I did was on a glass bed, so I used manual calibration and a gauge the set the height. The same process can be used to map your bed, if you have a feeler gauge.
Iāve been meaning to take the frame off and put it on a surface plate to determine how high/low each mount point is ā¦ but Iām afraid of what I might find 
Thing is for me the platform (what you are calling the frame i think) doesnāt exhibit the same issue when placed on flat granite - its way more likely to be shitty manufacturing of the heated bed IMHO - its thin and flexible compared to the platformā¦ and i have two platforms (because snapmaker insisted one was an issue) it wasnāt and the new frame was no different - both flat. to be clear i am not discounting others having warped platforms either.
1 Like
My original bed was getting old and marked and giving me flatness and skirt and adhesion problems so I got a nice bit of thick smoked glass (6mm) put that on the heater bed and then tried a 4x4 calibration but the nozzle kept crashing into the glass on position 1 before giving up and going to the centre. Is the z sensor not seeing the ātransparentā surface. I tried putting the opaque calibration at position 1 but it still tried to go through the surface? Do I need to manually set the machine z coordinate to account for the extra thick print bed and how do I do that?
FYI, Iām not as bright as you guys so will never go into the code to make adjustments!
Help!!
Also, how do I centre the damn nozzle anyway, the x and y āfactory settingsā are nearly1cm off.
@heterodynebob you can only do manual calibration with glass (its worth it).
you set this in the settings > 3d printing menu on the touchscreen
Yeah, platform is what I meant.
Good to know. I have been suspicious of that bed for awhile. One of my glass-bed mounting experiments (I got one smaller than the full SM bed, 300mm sq) was to bold two strips of 1/8" A36 steel to the platform, with the heated bed sandwiched between them. When I did this, I noticed that the gaps under my straightedge seem to have disappeared. Been pretty busy doing other things, but will sit down and measure that to make sure. It may just be that longer bolts and some through-holes drilled in sixty-cents-a-pound steel cures the problem (they could be under the heated bed as well as on top of it).
Yea, my 3 cents on the matter are detailed here: Carriage Tolerances - Unusable Over Distance >75mm From Center
My platform was 1mm out, and also you can watch the video of the dial indicator going around one of the āflatsā showing 0.003" of rise across the diameter - nuts.
The replacement platform I got was also 1mm out. And I had to shim the flats with masking tape to get the laser platforms level.
The solution from snapmaker support was loosen all but a couple screws on the heated bed and let it āfloatā since itās flatter than the platform 
1 Like
scyto - Brill. Many thanks. I did wonder about optical signals in glass. Will now have a go.
Just to be sureā¦ I manually z adjust to gently pinching test card for each x/y position then save?