The tests were done on a vcore4 400 (mazas2) and a vcore4 500 (tg73).
To evaluate the various mitigations, we use a static test, where a dial indicator / dial test indicator (DTI) is placed on the bed, measuring the change of distance between the gantry and the bed:
We also use the Beacon to log the distance it sees. For that to work the bed cannot be too far, and we did the tests with the head parked at 2mm above the bed.
The bed is unheated, the top panel is on and the front door is removed.
To make sure the tests have a similar starting point, the system was cooled down before each test for what we hope to be a long enough period. We either measured the temperature or the gantry to return to within 0.2C of ambient, or we saw that the distance difference had settled far enough so that no more than 10um difference was to be expected, based on the graphs. This is a bit of an inaccurate way to define the starting point, but unfortunately we did not have access to a constant-temperature lab, and even the variations of normal room temperature caused deflection to change by several microns. But that is something to consider: even though the measurements might be accurate, the ambient environment and the start-up conditions could cause an error that might be around 10..20 microns for the steady-state deflection.
The DTI affects the airflow and causes an error in the measurements. A better place for the base would be behind the gantry, where there is less air movement. The measurements starting from RUN-66 use that arrangement. The error depends on air flow direction and strength, and a guesstimate is that it could change the results with stock airflow by 10..20um.
The hotend temperature used in most of the tests was 250C, even though some tests were done with 100/150/200C. The tests not done with 250C are marked so.
To make sure the bed does not move, we also measured the distance from the bottom of the printer to the underside of the bed:
No significant bed deflection was observed.
We measured the temperature near the left motor:
We used a few cheap Dial Test Indicators (or Dial Indicators) with a 10um resolution and unspecified accuracy for non-automated measurements. The DTI used for main deflection automatic collection is Shahe GLS-5301-10-7 with a 1um resolution and a manufacturer-specified accuracy of +-3um. It not calibrated.
It turned out that the DTI exhibits some hysteresis or friction, which makes the readings jump over several resolution steps if the change is slow enough. To overcome this, and to make it possible to get sub-micron readings, we added mechanical dithering with a small unbalanced DC motor coupled to the printer frame. The speed of the motor was adjusted to give noise of about 1 LSB to the readings. After that, we averaged the readings for the final result.
