firmware surface kwality bug 14.07 vs 14.03

[Daid 306]

I still find it odd that this can effect quality like this.

Do you know if your bed temperature is stable without the PID enabled?

 

[gr5 2,172]

I don't have the latest firmware but I tested the 24V supply *at* the nozzle wiring on the PCB with the nozzle powered fully and made two measurements one with the bed off and one with the bed on. With bed off I got 24.0V on I got 23.8V. That's a 0.83% difference in voltage or a 1.7% difference in power. I believe this is true: average_nozzle_watts * constant = nozzle_temp - room_temp. So if printing at 220C or delta_C of 200, this change of 1.7% should change the temp no more than 3.3C before the PID controller adjusts. So at least for my power brick, this isn't the problem.

However maybe dimensioneer's power brick is slightly weaker. Also I didn't have servos moving which draw additional power. Also the power supply never had a chance to warm up (was only on for about 30 seconds). But my theory is looking weaker than before.

I still think the problem is with the nozzle varying temperature significantly (e.g. 10C) but I'm not sure how having the bed in bang-bang mode could cause this. I definitely don't think it is caused by the bed changing temperatures as that shouldn't affect anything more than 5mm above the bed. Plus I built my own heated bed for UMO 1.5 years ago and bang-bang mode kept it within 1C so I suspect the same is true for the UM2 bed in bang-bang mode.

 

[johank 7]

Has someone tried the P(ID)- approach Daid suggested with a really high P coefficient and the I and D values ignored?

 

[Dim3nsioneer 557]

Yes. See my post on 11/03 above.

[johank 7]

Dim3nsioneer: Thanks, I missed your post

That is really interesting. I guess it's too late to fix this problem hardware-wise for the UM2. The question is if it can be tweaked in software so that the EMC stays within limits.

 

[synvox 0]

I don't know what PWM-frequency is actually used, but it should be possible to use a very low one for the PID of a heating system, like a few Hz (5 - 10), because the time constant of the heatbed system should be quite large. And a low PWM-frequency should have a lot less impact on the EMF.

For example many soldering stations etc. use that kind of very low frequency PWM for their PID temperature regulation.

Best regards

Neni

 

[gr5 2,172]

Another interesting not - more firmware changes - possibly related to this issue. This is from Simon aka illuminarti:

 

It looks like after 14.09 was released, the PID sampling rate was quadrupled to try to give a more stable head temperature. I'm not sure if this was ever released at 14.09.1 - but if it was, it would have required a factory reset for the new defaults to get picked up. Now there is 14.11RC7 build which is presumably included in the corresponding beta version of Cura. That version includes the new pid settings, and forces the settings to get picked up at first run, without requiring a manual reset.

 

So I'm thinking the PID update rate for the nozzle was quadrupled so that the "bang-bang" changes to the heat bed no longer mess up the temperature at the nozzle which I assume caused the problems in the pictures above.

Note that even if you got 14.09.1 you might not have gotten the changes because Cura only does a "factory reset" if the version changes a significant amount. So 14.09 to 14.09.1 does *not* load the new settings without a manual "factory reset" but 14.09 to 14.11 *does*.

 

[Dim3nsioneer 557]

In my case I can rather exclude this possibility. I compiled Ultimaker2Marlin from the Github source and had the new behavior with a version on the level of 14.09.1 (the change was obvious on the display of the temperature in the 'heat up nozzle' option). And I had the banding.

When I changed from bang-bang mode to PID mode, I hardcoded even the numbers I found with the autotune feature into the Marlin code. I always upload my self-compiled versions with the Arduino software not with Cura.

 

[Nicolinux 288]

Just a side note - I now have a 3DR Delta printer and I see the very same issues with bang-bang vs. PID mode. The surface quality is worse with bang-bang mode. I am using http://www.euclidframeworks.com/products/hbp/ heated bed, a Rumba board and a 12V/30A power supply.

 

[Dim3nsioneer 557]

The 3DR is also running on Marlin, isn't it?

 

[Nicolinux 288]

Yes it is.

 

[geert_2 557]

I still think the problem is with the nozzle varying temperature significantly (e.g. 10C) but I'm not sure how having the bed in bang-bang mode could cause this.  I definitely don't think it is caused by the bed changing temperatures as that shouldn't affect anything more than 5mm above the bed.

 

Could this sort of problem be caused by long common ground wires, used for both power and sensing? Or something similar? So that some reference shifts up or down when a high power device (e.g. heat bed) is switched on and causes a voltage drop in these lines? I don't know the board schematics, so I am just guessing.

Further, what part of the printer exactly causes the EMC-overshoot? Heated bed, wires, controller board,...? Could that be handled in a simple way by adding an (bigger) earth-line somewhere, or by shielding?

In that case, might it be possible to officially say that: "On the condition that you modify your printer so and so, the bed-PID may be enabled"? Or would that violate the laws too?