Irregular Surface Finish with Heated Bed?

dc42

I've knows that sort of layering to be present when using bang-bang control of the heated bed, and solved by using PID. When you use PID, how closely does the heated bed maintain temperature?

You've told us you are using 14 gauge wire for the bed heater. I assume you are powering the bed directly from the Duet, without an SSR. Are you also using 14 gauge wire from the PSU to the Duet?

Is there a noticeable drop in supply voltage when the bed heater is on? You can read the supply voltage by sending command M122 from either Pronterface or the GCode Console in DWC,.

alansrobotlab

Hey dc42, thanks for your prompt reply.

This generation RostockMax uses a pc power supply, and I have 6 20gauge wires from the psu going into the duet.
The wires are bundled but not soldered together, and go directly into the duet.
I'm not sure what's normal for this setup, but running all the heaters drops the +12 from 11.9 to 10.7, a 1.2v drop.

Here is what I'm seeing when running M122.

From reset, homed, bed + hotend off
Supply voltage: min 11.9, current 11.9, max 12.0, under voltage events: 0, over voltage events: 0

Enabled bed + hotend heater
Supply voltage: min 10.6, current 10.7, max 12.1, under voltage events: 0, over voltage events: 0

dc42

That's a rather large voltage drop, but not uncommon when using cheap PC ATX power supplies. Good named brands of PC power supplies are normally better, so are LED/CCTV power supplies.

If you were using bang-bang then I would say the problem is almost certainly caused by the hot end power dropping when the bed heater is on. Are you absolutely certain that it also happens when you use PID for the bed? If your bed heater has an LED, then you can tell that PID is running because the LED will flash rapidly when the heater is at the correct temperature. Likewise the bed heater LED on the Duet will flash rapidly.

bot

That effect is definitely normal for an onyx bed without a finely-tuned PID loop.

I'm using an older firmware than you, so my PID values wouldn't work (I don't think), but I have my onyx bed tuned to hold within 0.1 degree, and no longer have any effects like this. The trick is that the PID needs to be aggressive and snappy in response to heat changes. If you look up "Tuning PID values," you should find some guides to manually tuning PID loops.

Also, my onyx bed is well insulated (both bottom and top, due to the layers of glass, aluminum, and PEI on top; cork on bottom). Insulating the bed seems to be part of the solution, too.

alansrobotlab

Yep, PID mode is definitely enabled.

I'm watching it print again, and the heated bed temp tends to overshoot and undershoot by up to a few degrees. PWM goes between 0% and 100%, and sometimes flashes quickly indicating some other duty cycle.

I have it hooked up to a beefier pc power supply and vdroop goes from 11.9 down to 11.0, and the artifacts are still there.
I'd say that the pid tuning parameters aren't right.

Is it time for me to switch to a meanwell dedicated 12v psu? Do they tend to hold their voltage better under load?
http://www.meanwell.com/productPdf.aspx?i=459

[[language]]
M307 H0
Heater 0 model: gain 84.8, time constant 515.0, dead time 4.2, max PWM 1.00, in use: yes, mode: PID
Setpoint change: P258.1, I0.50, D758.8
Load change: P258.1, I16.20, D758.8

bot

I use a CoolerMaster G750M 12V ATX psu. It sounds like your PID tuning is not aggressive enough.

My values for my Onyx bed are like this: M301 H0 P100 I1 D26 T2.0 S1.0 B1.0 W255

But I'm using a firmware from before a bug was fixed that would cause the values above to be changed. I forget how exactly, but doubled and halved in one way and another.

alansrobotlab

I just ran autotune again for H0, and reduced the max pwm to S0.6.
VMin still dipped as low as 10.5, but on average VCurrent seemed to stick in the 11.5ish range every time I checked it with M122.
The sample print still has the irregular surface, same as when I used my heavier duty 750watt psu.

So far swapping out PSUs and changing the scaling factor has no appreciable effect.

Does anyone run the MeanWell LRS-350-12? I'm hoping that it doesn't suffer from VDroop like these PC Power supplies do.

[[language]]
M307 H0
Heater 0 model: gain 101.5, time constant 553.3, dead time 7.8, max PWM 0.60, in use: yes, mode: PID
Setpoint change: P124.7, I0.23, D681.1
Load change: P124.7, I4.83, D681.1

dc42

@gestalt73:

Yep, PID mode is definitely enabled.

I'm watching it print again, and the heated bed temp tends to overshoot and undershoot by up to a few degrees. PWM goes between 0% and 100%, and sometimes flashes quickly indicating some other duty cycle.

I have it hooked up to a beefier pc power supply and vdroop goes from 11.9 down to 11.0, and the artifacts are still there.
I'd say that the pid tuning parameters aren't right.

Is it time for me to switch to a meanwell dedicated 12v psu? Do they tend to hold their voltage better under load?
http://www.meanwell.com/productPdf.aspx?i=459
[[language]]
M307 H0
Heater 0 model: gain 84.8, time constant 515.0, dead time 4.2, max PWM 1.00, in use: yes, mode: PID
Setpoint change: P258.1, I0.50, D758.8
Load change: P258.1, I16.20, D758.8

The main cause is that the PWM is cycling, which isn't right. The computed PID values are not working well. Try increasing the dead time parameter in the M307 H0 command to 10 or 15 seconds.

A MeanWell PSU should have much better regulation than the 12V rail of a PC PSU. But if you sort out the PID then the PC PSU may be adequate.

alansrobotlab

Hey dc42 & bot,

Thank you so much for your help with this. I think there are three things going on here:

The Rostock Max Onyx heated bed is a power hog
The Budget pc psu has a pretty extreme voltage drop while powering the onyx
The DuetWifi is sensitive to voltage drop in this configuration

The Meanwell LRS-350-12 is only $35 from most sources. If they have less vdroop than a pc power supply, I think that's what I need to try next.

Here's my stack of test prints. Overlit at an angle to highlight the surface finish as best I can. Some are a bit bent, that's just from me yanking them off the build plate and not straightening them out. They all printed perfectly flat with no curling or lifting.

The top two were printed with the heated bed off. To my eyes, they're perfect
The next two going down were printed with a tuned heated bed, D=15.0. The're very good, but there are fewer, wider lines that are faint but still there. These prints are pretty good
The bottom 5 are various prints using either bangbang or wrong pid for the heated bed.

bot

It is definitely looking better much better. Good work.

dc42

@gestalt73:

Hey dc42 & bot,

Thank you so much for your help with this. I think there are three things going on here:

The Rostock Max Onyx heated bed is a power hog

The Budget pc psu has a pretty extreme voltage drop while powering the onyx

The DuetWifi is sensitive to voltage drop in this configuration

It's not the Duet WiFi that's sensitive to the voltage drop, it's the hot end heater. The power produced by a heater is proportional to the square of the voltage applied to it. When the bed heater turns on and as a result the VIN voltage drops from 12V to 10.6V, the hot end heater power drops by 22%. This is a large drop and will cause a significant drop in the nozzle temperature, which the PID will take a few seconds to correct for (the dead time of a hot end is typically around 5 seconds, so the PID can't react much faster than that). This temperature fluctuation will affect the viscosity of the filament, which in turns affects extrusion width and back pressure.

alansrobotlab

Alright, I think we can put this one to bed.

tldr; Wrong gauge wire used for hotend heater. Getting perfect prints after using 18gauge stranded instead of what was originally used.

Truth be told, I have a client build that I'm troubleshooting, and as part of the process I upgraded him to the DuetWifi. I love this board!

dc42, your comment prodded me to examine the hot end setup, and I noticed earlier that the gauge of wire used for the hotend seemed a bit smallish. I just rigged up an alternative wiring harness using a heavier gauge wire, and the irregular surface finish all but disappeared.

I'm still concerned about the vdrop though, so I'm going to try out a dedicated 12v power supply before I hand the machine back over to its owner.

Thanks everyone for your help!

dc42

I was using a PC PSU on one of my printers for a while and I also had a problem with voltage drop. Switching to a Corsair CX430M PSU mostly solved it. However, the bed heater on that printer only draws about 10A; your client's may be drawing more.

T3P3Tony

One of the common issues with PC PSUs is voltage drop on the 12V rail when loaded, adding sufficient load onto the 3.3V and 5Vs often significantly reduces this drop on the 12V rail.

For more details see:

http://reprap.org/wiki/PC_Power_Supply#Minimum_Load
http://hydraraptor.blogspot.co.uk/2012/12/mendel90-updates.html (10R on 5V and 4R7 on 3.3V rails)

bot

Also to keep in mind, some ATX PSUs now come with "Zero load" rating – meaning they are designed to work as we use them, without having to add load to the 5v or 3v rails. see, for example: http://www.coolermaster.com/powersupply/modular-gm-series/g750m/

Haswell C6/C7 support & zero load operation

I don't know how the zero load operation relates to haswell c6/c7 state support, but there ya go. Also notice the single 12v rail – helpful when hacking the psu.

T3P3Tony

Ahh that's good to know. I imagine the cheaper PSUs don't have this feature or did not a couple of years ago when I was still using PC power supplies.

dc42

My guess is that PSUs with zero load rating have independent regulation on the 5V and 12V rails - and therefore better regulation of both - because I think it would be difficult to achieve a zero load capability with a single regulation feedback signal for both rails.

alansrobotlab

Thanks guys for all your help.

I think I'm sold on these dedicated 12v power supplies. They're only $35-$40.

Found a 12v power supply locally, wanted a meanwell but couldn't find one on short notice.
http://www.microcenter.com/product/446051/12V_360W_Dedicated_Power_Supply_for_3D_Printer_HEAVY_DUTY

This power supply only sees a vdrop of 0.5 volts under full load. Tuned to 12.1 idle, drops to 11.6 volts.

There's only two challenges with this type of power supply. No power switch and a really really loud fan.

I hacked on a 100mm silenx case fan and I found these to install to get a proper power switch.
http://www.ebay.com/itm/like/142086311320?lpid=82&chn=ps&ul_noapp=true

I've got his printer for another week for burn in and testing. If it all checks out I think I see some power supply upgrades in the near future for my printers.

dc42

I use a fanless 12V 300W PSU to power one of my printers, and I printed a cover for it to safeguard the wiring and fit a switch. See http://www.thingiverse.com/thing:441081.

I think the Meanwell PSUs have thermostatically-controlled fans.

T3P3Tony

We use similar power supplies but 24V on the Kossels and I have another in the BigBox. David's Cover is a good idea.