Non-Backlash Explanation of Z-Banding

I will put in some experience on this: First thing to know is you have an SD3 with Sang board. I have no idea if they were only shipping aluminum Z table bracket at this time or if they still shipped some ABS Z tables. It would be nice if you could tell us which you have, makes all the difference.

Now, from my personal experience with my self-build, you are absolutely correct that heat expansion during bed warm up changes the Z height to some degree. I have measured this myself. My self-build I had use C channel to support the bedplate. It was not terribly stable so I added a polycarbonate piece to make it more planar and stable. By doing so, my Z height calibration was quite different when leveling cold vs leveling with the bed warmed up. My bed heater is well insulated, but it did not matter, still too much heat.

I will say this though, once the bed had been soaked (by this I mean the heater reaches setpoint and then sits there for 5 minutes more since the glass on top is not directly measured for heating and needs additional time to equalize) then the Z height of this setup was extremely repeatable. As in, if I warmed up my bed for 6 minutes each print, the first layer was spot on with the Z tab setting not being changed. The warping in the bed was repeatable.

I also measured my all-metal Solidoodle 3 bed for changes in Z height during 20 minutes of heating from cold. It changed a total of 0.005" during the beginning of warm up and thats it. Now my bed is not very flat (stock Solidoodle) and I have a 2.5mm mirror clamped onto it in two places. I attribute this change to the mirror. Regardless, I have no noticeable Z wobble. I do have moire which is predictable, and I also have random shifting that follows a print job that I attribute to the nozzle hitting curling on overhangs and shifting the bed on those screws, but every calibration column and large prints without overhangs have no shifting, no banding, and no wobble. On stock Solidooodle 3.

On my self build, I have some Z issues related to X-Y carriage torsion common with H-BOT setups, so I cannot say the same performance just yet. I have changed the Z bed to use a ceramic sheet instead, which stabilizes the Z frame, but I was using that as previously as insulation with the PCB heater and to stabilize bed warp. So now I need to get a thick piece of aluminum for the heater, or a thicker piece of glass that will not bend to the shape of the PCB (cheap PCB heaters are warped if the copper on the back side is not etched).

So if you are having large changes of Z height with bed heating, there are two things you need to consider: Are you waiting long enough for the bed to soak? And if you are, then I would say the heat is slowly getting to something else farther away (like plastic pieces not directly connected to the bed frame, but connected to the Z rods and heat is transferring slowly over aluminum bed frame. But I can tell you from experience that this change in height vs temperature is very predictable. So the solution is get rid of the plastic parts of your bed architecture, or simply wait long enough for the temperature to stabilize.

Your test should provide some decent empirical data. I also hope you understaood what I meant about 'soaking', since the temperature probe for the bed is under the aluminum and it takes a lot longer for the glass to come to temperature after the graph shows the probe to be at temperature. A rule of thumb I have always gone by is to wait at least 15 minutes after the bed reports getting to temp before printing.

I am also hoping that you are not saying that the +/- 2.5 degrees of bang-bang control is actually making the Z height change significantly. But your test will provide data on that.

This is from about a week ago showing that the banding is constant up the part and that soaking is probably not a factor past the first 10 minutes or so.

Hold that next to your z thread and tell me it's not the same pitch...

Well, the shared pitch would lead one to think that the rotation of the thread and the banding might in some way be connected, right?

It's likely the thread being slightly off-true pushing the bed around: this lines up with the rotational angle of the thread, and thus in height this corresponds to the thread pitch. First step would be to address potential wobble with some kind of slop-nut arrangement - that's my recommendation at least.

Note that tiny movements will show up as banding, it's quite a sensitive indicator of both wobble and layer height inconsistency. Getting rid of it is a bit of an achievement, and I look forward to seeing your results! Happy hunting.