DuetWIFI migration

mhackney

David, I have finished making a suitable PEI bed to use with the IR probe following the instructions on your blog (painted back side, bake in oven). I then recalibrate the probe and that seemed to go well. But…

My calibration is bad (typical deviation of 0.086) but very good when I switch over to my FSRs (I have them both installed so it is very easy to change from FSR to IR Probing) it is pretty good (deviation 0.45) although not as good as I'd like.

Some questions...
I scuff the surface of my PEI with very fine sandpaper (600 grit) to leave a matte finish on my parts. I installed this PEI with the original glossy side up (painted the matte side as per your instructions) and then scuffed the top. Do you think this would have any effect on the sensor? I wouldn't think so if PEI is IR transparent.

dc42

I don't think it would make much difference which way up you put the PEI. I think the next thing for you to do is to measure the trigger height at a few widely-separated probe points. If you find significant differences, then use feeler gauges to see whether they are due to the edge of the sensor board being at different heights when the nozzle is touching the bed - which would indicate that effector tilt is a problem..

mhackney

Thanks David, I posted a rather long post on the delta google group. I installed a bulls eye level and my effector is dead level everywhere it moves. I also see very different calibration results between 6 pt and 3 point probing - 3 point is VERY low deviation (0.007). I need to digest this and think about next steps. But I think measuring trigger height makes sense.

dc42

I suggest you measure trigger height differences using the FSRS too. You may find that the FSR setup is more sensitive when you probe close to one of the FSRs, and less sensitive when you probe at the centre.

mhackney

Good suggestion thanks. Then I can compare results across both probes.

This printer has always been a bit persnickety. It is 3+ years old and was one of the first laser cut Rostock Max kits. I've replaced everything except the frame and towers. My other 4 deltas have FSRs and Duets and calibrate to <.0.03 deviation repeatedly.

mhackney

David, I have now run a set of trigger height and calibration tests. I've attached a screen capture of the spreadsheet I used to collect the data. Some comments before the data:

This data was collected on PEI with a black back.
Hot end = 190°C
Bed = 55°C

Data Collection Process

Home
G0 to the Probe position (with Z5.0)
Do the paper test and run G92 Z0 to set Z=0
Remove paper and run G0 Z5 then G30 S-1
Record trigger offset run 1
Raise 5mm and run G30 S-1 again
record trigger offset run 2

Repeat 1-6 for each point on spreadsheet.
Once data collection was complete, I ran the auto calibration twice (6 factor) and recorded the deviation.

I ran the IR Probe first then reconfigured and then ran the FSRs.

Note that my FSRs are not aligned with the columns or midway between columns, they are actually 30° offset from the column. The FSR positions are shown in the spreadsheet. This was done for installation simplification (the Rostock Max has a snowflake insulator that mounts in such a way that this is the natural alignment)

I calculated average, standard deviation, minimum, maximum and delta (max-min) for each run.

And here is the data:

Some comments:
Firstly, I was surprised at the poor FSR calibration deviations. I am not sure why this is. I have checked everything about the machine and it was calibrating much better the other day. But this is the data I collected. (Update as I write this: I realize that I have been probing on a 100mm radius in the past, so the results described below are more in line with this printer's past performance.)

I do have a level installed on the effector and I looked at the level at each of the probe positions and there was absolutely no deviation of the bubble.

Interestingly, the FSR results at the 240° probe position (alpha/X tower) is the worse and this is the location where I saw (and see) the nozzle "skate" slightly on contact with the bed. I thought this might be a belt issue and tightening the belts did improve things a bit. I plan to eliminate this probe location and rerun auto calibration and see what happens.

Note that if I decrease the probing radius:

(FSR results):
120: 0.137 (as I collected in the table)
110 decreases to: 0.071
105 decreases to: 0.062
100 decreases to: 0.053

(IR Probe results)
120: 0.135 (as collected in the table)
110 decreases to: 0.115
105 decreases to: 0.080
100 decreases to: 0.094

I verified these results with multiple runs. FSR probing IMPROVES with decreasing the probing radius and IR probing GETS WORSE with decreasing the probing radius.
I have a hypothesis about this but I'll wait to hear comments and feedback! David, I would greatly appreciate your feedback on this - including if this was a valid test and experimental design!

Toddimus

Another data point…

I also have a Rostock Max (V2) with FSRs and the JohnSL board. I installed some Hall Effect sensors in lieu of the mechanical ends stops at the top, which may give slightly more repeatable results for end stops (as opposed to the mechanical stops). I have the JohnSL board sensitivity set to whatever was default. I didn't have any extra jumpers lying around.

One thing I did differently than your notes above show is that I only did an initial home/delta auto-calibration at the beginning of the whole test run. It sounds like you homed after each point. I'm assuming this probably doesn't matter. I also took 5-6 probe values at each point. Interestingly, the values measured at the same probe point varied by around 0.04 to 0.06mm across the multiple probes done one right after another.

Your offset numbers look similar to mine, except that mine were slightly lower (average around -0.24mm, min -0.19mm, max -0.33mm). With a 140mm probe radius (6 points at periphery, 6 halfway and one center), I'm getting right about 0.100mm standard deviation after an auto calibrate. I also saw this number much lower (around 0.06-0.08mm) when I used a 120mm probe radius. I found that I needed to adjust the Z offset for the probe (i.e. center point) from the original values I found during calibration, in order to get a correct first layer height.

As an aside, I notice that one of my FSRs takes more force to trigger. I can tell this when I'm probing between the towers and only one of the LEDs lights up, when two theoretically should light up. I designed some little FSR mounts that use the snowflake mounting holes and keep the sensor aligned with the axes, as opposed to being offset as you said yours were. In other words, the mounts/sensors are concentric with the snowflake mounting holes and frame base mounting holes adjacent to the towers. The backside of the sensor itself is centered over the top of a carefully prepared countersunk hole that places the top of a flat head screw co-planar with the floor of the FSR mount. I'm still not sure I like my FSR mounting setup because I'm concerned it might be a bit over-constrained in the X-Y directions. This may be the cause of the one sensor taking more force to trigger.

Which brings me to my next question... does everyone like the little orange silicone pads that come with the FSR kit? I'm wondering if they have too much "springiness" or softness. I put one on the base of each plunger that comes into contact with the sensor itself.

mhackney

@Toddimus (thanks) Yes, I did home after each point primarily because if I did not, I ran into situations where Z=0 was set at an earlier probe point but is slightly above the bed for a later probe point (if you know what I mean!) I could not probe below Z=1 to do the paper test.

I actually used he snowflake mounting holes but on the V1 RMax these are not aligned with the towers (there were 6 mounting positions on the snowflake on V1 - I think this is the same on all of them). You've probably seen my setup on the SeeMeCNC forum. Mine is not as over constrained as most I've seen but I have a design that is constrained properly but haven't make/tested it yet.

I actually have tested those pads against cork, neoprene washers (with a hole in them) and craft store foam. The foam was horrible but cork and the washers and silicone were all very repeatable. This printer has the washers - 1 on the plunger that contacts the top of the FSR. The bottom side of the FSR is stuck to the printed mount with it's adhesive backing.

mhackney

I've taken this to the Google delta robot group: https://groups.google.com/forum/#!topic/deltabot/zwfpSvCNjR8

Meanwhile, my Duet WiFi has been working well. I do have the occasional disconnects (one a day) but I also get occasional disconnects on my other printers. They rarely occur simultaneously. I don't actively monitor network traffic but maybe I should do that for a week or 2.

mhackney

Well, after much testing (and some time off to do a few other projects) I've come to the conclusion that the probing issues I've had with the IR Probe are due to the fact that I use CA to fix scratches and divots in my PEI. PEI has different IR characteristics. I developed this "fix" for PEI several years ago to allow me to have a perfect first layer as that is the most visible on the fly fishing reels I manufacture. I love the idea of the IR probe but its nearly impossible to visually tell where a CA fix has been made in order to change the probing location. I've gone back to my FSR probing and everything has been working great.