Can't get level first layer/Convex bed

[Jrjones]

I've been having problems printing items that are bigger than 100mm, because I can't get the first layer to go down evenly. Its too thin in some spots, and too thick in other spots.
I believe its mostly due to that the bed is convex, even with the glass clamped to it.
I loosened 3 of the six holding the PCB to the snowflake, which helped a little bit, but I'm not sure how to get the bed nicely flat (to a degree that it isn't noticeable).

To the best of my abilities to check with a digital level finder and a metal square, the towers are perpendicular to the bed. All belts are tight, nothing is loose.
I've run the least squares calibration method manually when I had a Rambo board installed, and I could never get the deviation below 0.200.
I've seen other people post that they have a deviation after running a calibration script that is less than 0.05, with typical values of 0.010 to 0.020 (correct me if I'm wrong).

I've recently installed a Duet and have tried all the calibration routines multiple times (6, 7, 8, and 9 factor) and the deviation is still 0.180 to 0.210.
I've added trigger height corrections to the calibration script as outlined here: https://duet3d.com/wiki/Calibrating_a_d ... bed.g_file
Which didn't really make much of a difference.
I've tried bed mesh compensation after calibration but I still can't get an even first layer.

Machine specs are:
Rostock Max V2 (with longer towers for a homed height of ~670mm)
Modifications include:
Duet Wifi
JohnSL FSR kit
New Ball-cup carriage arms and cheapskates
12V 350W PSU
Yellow Jacket Concepts Quick Change board

Current Config settings:
M665 R143.548 L290.8 B135 H667.946 X0.558 Y0.202 Z0
M666 X-0.40 Y0.75 Z-0.35

Here are some heightmaps, towers indicated on graph:
After S6 Calibration

After S7 Calibration

After S8 Calibration

After S9 Calibration

Also attached the data points as an excel document

I've run out of ideas of what to check or change to improve my first layers. I'm kind of thinking that I have to force the bed to be flat with a piece of aluminum, but I'm not sure if I can force the bed flat that way.

[Jrjones]

Here is a picture of the Onyx if I hold a straight edge on it. About 1mm gap on the other side. This is a little bit less with the glass clamped down to the bed, but its harder to take a picture of.

I've tried raising the Onyx so that it is only supported by 3 of the 6 screws, and its raised off the snowflake in case that was bulging it up, but it seems the convex shape is the relaxed shape of the onyx.

[Jrjones]

So it's actually worse when the heated bed is on.
This picture is with the straight edge balanced and fully touching the bed in the center, and you can see the gap at either end.

The bed will actually flex with power applied to it... I knew it did it a little bit, but not so much.
The center of the bed moves up when I heat the bed, and drops down as soon as the power to it is turned off.

[Doug68]

One method for getting the bed flat is to loosen off all the screws holding it down, then heat it up and whilst its still hot screw it down again.

[Jrjones]

I've tried that multiple times, it doesn't help. I actually can't get the bed to sit flat while it is heating up. The force needed to flatten the bed starts to bend the melamine underneath.
It also doesn't help that the bed is more convex when it is heated up than when it is room temperature.

[Dale Eason]

If the bed is Convex and symmetric about the middle then adjusting the Horizontal radius will compensate and make it be flat to the hot end.

Dale

[Doug68]

If the bed is Convex and symmetric about the middle then adjusting the Horizontal radius will compensate and make it be flat to the hot end.

Dale

That'd make it work in a practical sense, but the parts produced won't actually be flat, they'll be curved to the same degree the bed is and that curve would be replicated all the way through the height of the part.
If it were 0.1mm then that should be good enough but at 1.xmm I'd not like it.

After taking the thing off of the machine if it's just banana'd on the bench then attempt to persuade it to be flat, as at that point there's nothing to lose I'd have thought. I suspect the only real fix is to get a new bed.

[Jrjones]

Adjusting the horizontal radius doesn't help. Probably because its not symmetric.

Yea, I'm suspecting that the only solution is to get a new bed. I can't force it to stay flat no matter what I do.