Marlin Motion Related Configuration.h Settings for MakerFarm i3v

[Mjolinor]

Velocity is a vector, speed is a scalar. There is no direction attached to speed so it is possible to maintain constant speed and travel in a circle, it is not possible to maintain constant velocity and travel in a circle.

[AbuMaia]

It seems to me a moot point, as none of our axes travel in a circle.

[Mjolinor]

It seems to me a moot point, as none of our axes travel in a circle.

Sorry, can't see the relevance of what your printer does when talking about technical explanations of what acceleration is, enlighten me.

My print head moves in circles or parts there of quite frequently, so frequently in fact that they included the G2 and G3 Gcodes specifically for it.

[printbus]

As the thread starter, I'm hoping we don't have to continue the discussion on speed vs velocity, angular acceleration, etc. Mjolinor makes a technically valid point, but I believe the context of the comment goes beyond the current scope of the thread. Everything being discussed in this thread has been, and at least in my foreseeable future posts, will continue to be predominantly in the context of linear movements of the printer axes.

As Abumaia points out, an axis can only move linearly. It is the move planner that decides how to accomodate the possibly complex gcode commands as linear movements in X, Y, Z and the extruder, based on the axis constraints it has been given to work with. I am sure it is this three dimensional planning that forces the complexity into the move planning code. Ignoring the totality of that complexity in this thread is likely over-simplifying things, but I don't believe it is a flawed approach. Once we have some fundamentals covered, it may very well be interesting to know more about how Marlin applies constraints for the independent axes into move planning for more complex objects.

My plan has been to better understand the printer mechanics in a simple linear sense where I can easily visualize, demonstrate, see and feel the differences settings can make. Based on what I learned, I then froze the printer with some new settings so I could see whether they negatively affect the printing of real objects. This is where I'm at now. I would have rather been just trying settings someone else offered, but as of yet that option hasn't been made available. Personally, that's what I find the most frustrating about having to dig into the details.

Under the realm of possibility that there's just a difference between the US and UK on layman context regarding the term speed, I did go back through the thread and adjust how and where I had used it. I'm hoping I got it close enough and we can just leave it at that.

[printbus]

LINEAR CONCENTRIC SQUARES TEST
This brings me to the last of the non-printing demonstration gcode I have currently thrown together. I plan to run the printer with some new settings for a while, perhaps adjusting to the need for some new settings in the slicer. There may very well be better tests, but this was straightforward to put together from the demonstrations already prepared. One thing I like about it is you know exactly what movements the printer should be doing, making it easier to anticipate what to watch for.

OVERVIEW
The demonstration involves moving the printer axes in a path outlining a stack of concentric squares, from 180mm to 2mm width. Between each square, the Z-axis is increased by 0.2mm to mimic a layer shift. Three passes at the stack of squares is made. The first pass moves the X-carriage and Y-bed at a high travel rate. The second pass is identical except the X-carriage and Y-bed are moved at printing feed rates. The third pass continues with the printing feed rates, but adds in extruder movements mimicking printing and also retractions around the layer shift points. Including homing and a few seconds of pause between each pass, the demonstration runs for around three minutes.

!! REMOVE FILAMENT FROM THE EXTRUDER BEFORE RUNNING THE DEMONSTRATION GCODE !!

Two versions of the demonstration gcode are provided. File squares_oem.gcode uses parameters from the MakerFarm distribution for the 8 and 10 inch i3v printers. The same values might also apply for the 12-inch printer; I haven't downloaded the 12-inch firmware to compare settings. File squares_fast.gcode is identical except for changes I've made to DEFAULT_MAX_FEEDRATE, DEFAULT_MAX_ACCELERATION, DEFAULT_ACCELERATION, DEFAULT_ZJERK, and DEFAULT_EJERK. Both files assume a 250 mm/sec travel rate, 100 mm/sec printing feed rate, Z feed rate of 2mm/sec, and retraction rate of 15 mm/sec. Make sure your printer is configured for a DEFAULT_MAX_FEEDRATE of at least {250,250,2,15} before running the gcode, since gcode commands can't increase above what the printer is set to. You can verify this by checking Vmax x, Vmax y, Vmax z, and Vmax e on the i3v LCD.

The demonstration starts with a home-all, with all pertinent parameters set to what are known to be safe parameters. This way if you alter settings later in the gcode that mess up printer movements, the next pass of the gcode will restore anything needed for the homing to function properly. After the home all, gcode commands are used to set up certain demonstration parameters; those would be the parameters you could change for tailored rates you want to try. In the pass including extruder movements, changes to XY travel rate, XY print speed, Z feed rate, and retraction/replenish speeds would require editing hardcoded values in gcode commands throughout that pass.

OBSERVATIONS
I first ran the squares test with travel rate limited to the 160mm/sec rate determined earlier to not lead to X and Y belt flutter. I found that to seem woefully slow for a travel rate used in printing, and quickly raised it to 200 mm/sec and ultimately to the 250mm/sec rate from the MakerFarm defaults. The 250 mm/sec rate is pretty snappy. Although it is high from the reprap wiki notion of achievable step rates, I have the perception that a lot of people have kept the 250 mm/sec rate and no one has reported any problem with it. Interestingly, when I went to the 250 mm/sec travel rate, I no longer had any of the belt flutter that earlier drove me to the 160mm/sec rate. In going back and verifying settings, I'm convinced that flutter had to be an artifact of previously overdriving the X and Y motors.

In observing how the printer accelerated from each corner of a square, I thought it was interesting to see how the printer would get to the printing rate of 100mm/sec fairly quickly. For the acceleration to a 250 mm/sec travel rate, however, I offer two opinions. First, even on my 8-inch bed the printer took a lot of time and distance to accelerate up to the 250 mm/sec travel rate. As the demonstration moves to the smaller and smaller squares, you see how the high travel rate is never achieved on many of the moves. Second, the sound of the acceleration makes it seem like the motors are straining under load, and slow to get to speed. These are step motors and that isn't the case, but it sounds that way.

I like the 15mm/sec extruder max feed rate being applied to retraction and replenish. It seems snappy enough. I've been using that on all of my prints for the last couple of weeks, and it seems to work fine.

TAILORED SETTINGS IN THE FAST DEMONSTRATION FILE
I eventually settled on 750 mm/sec per second as values for X and Y in DEFAULT_MAX_ACCELERATION. That seems to be a great help in getting X and Y moving at the travel rate quickly, at least quick enough that the full travel rate should now be applied to a significant percentage of travel moves. At a glance, this may seem like a reduction from the values of 1000 in the MakerFarm DEFAULT_MAX_ACCELERATION, but with DEFAULT_ACCELERATION at 500, those higher settings were being clipped to 500. To benefit from the higher settings in DEFAULT_MAX_ACCELERATION, DEFAULT_ACCELERATION was also changed from 500 to 750. I could have just increased DEFAULT_ACCELERATION alone, but I just don't like propagating the confusion of having a setting that ultimately gets limited by another, so I set them to the same value.

Z axis movements at the OEM settings seem slow, and have the same motor-under-load sound. Earlier testing had determined the Z motors I have are limiting me to a Z feed rate of 2 mm/sec. To speed up Z adjustments, I changed the Z term in DEFAULT_MAX_ACCELERATION from 5 to the 500 mm/sec per second originally used for X and Y. I maybe wouldn't have had to go that high to get the benefit I was looking for, but I've seen no issue with the high value. Z layer shifts are now very crisp and quick. I admit I've been rethinking the weight I've got strapped onto the Z-rods in the two aluminum shaft couplers - at some point they may be viewed as a limiting factor in Z motor acceleration. The extruder was left at the 500 mm/sec per second value in DEFAULT_MAX_ACCELERATION and DEFAULT_RETRACT_ACCELERATION.

Based on earlier experimenting, I increased DEFAULT_ZJERK from the 0.4 mm/sec MakerFarm default to 10mm/sec, and DEFAULT_EJERK from 5 mm/sec to 10 mm/Sec. DEFAULT_XYJERK was left at 20 mm/sec since I didn't want to increase the risk of XY corners ringing due to jerk.

IMO, these few changes in settings leads to an i3v that both looks fast and sounds fast. Combined with adjusting the stepper drivers in real use and not by voltage reading, the printer sounds like an entirely different machine. From the 160 mm/sec travel rate I entered the testing with, the settings in the fast file led to a time reduction of about 30 seconds in the demonstration. The difference between the OEM settings file and the fast settings file is about 15 seconds on my printer. That's not a scalable number that means much, but it does demonstrate the changes have led to a faster printer.

In attempting to push test prints to a 100mm/sec print speed with these settings, I have ran into some issues, primarily with extrusion being light for short segments in solid infill layers like on the bottom or the top. This did lead me to reduce the XY acceleration term from 1000 to 750 mm/sec per second, but I also simultaneously increased my extruder temp and put more attention on what the actual movement speeds were on perimeters and solid layers. I think I've just gotten by with sloppier settings on earlier prints when I haven't been pushing the printer (I'm typically slicing with Cura Engine and Repetier-Host, with all speeds proportionally adjusted through one integrated speed slider control). I also haven't ruled out the possibility that the test issues have been with the filament or the extruder stepper driver setting. To further assess viability of printing at 100 mm/sec, I'm likely going to have to look at the rate of filament volume the hot end can properly handle. I also need to validate the Y motor driver setting and Y axis acceleration term with more weight on the Y-bed than I've had in the few test prints completed.

FOLLOWUP COMMENT: The print issue I was having was evidently the stepper driver adjustment. I must have been skipping a few steps on the extruder motor. I haven't seen the extrusion problem after adjusting the extruder driver upwards a bit more. It is amazing how much quieter the printer runs with the readjusted driver settings. It's also important to note that my "run hot" type motors also now hardly warm up at all.

[printbus]

PRINTBUS SETTINGS RECAP

SLICER -
Travel rate: 250 mm/sec or 15,000 mm/minute
Print rate: 100 mm/sec max, perimeters and solid layers adjusted to suit
Retraction: 15 mm/sec

MARLIN -
HOMING_FEEDRATE {100*60, 100*60,2*60,0} (but I have also modified Z homing code in marlin_main.cpp)
DEFAULT_MAX_FEEDRATE {250,250,2,15}
DEFAULT_MAX_ACCELERATION {750,750,500,500}
DEFAULT_ACCELERATION 750
DEFAULT_RETRACT_ACCELERATION 500
DEFAULT_XYJERK 20
DEFAULT_ZJERK 10
DEFAULT_EJERK 10
MANUAL_FEEDRATE {100*60, 100*60, 2*60, 5*60} (in configuration_adv.h)

Print area is equipped with a print cooler capable of providing the airflow necessary to print PLA without minimum layer time restrictions on most prints.

FOLLOWUP COMMENT: It is determined later that the 5mm/sec value for E in MANUAL_FEEDRATE is about at the limit for a sustained extrusion using 1.75mm filament. For the same amount of extruded volume, 3mm filament feeds slower. 1.5 to 1.7 mm/sec would be a more appropriate value for 3mm filament.

[printbus]

ESTIMATING A FEED RATE LIMIT ON EXTRUDER E-STEPS CALIBRATION

The background information leading to this has been puzzling. Some people have reported inconsistent results if they run at the Pronterface default of 100 mm/min feed rate during extruder calibration, yet I've known I could do calibration at the 300 mm/min default in Repetier-Host. But I've also realized my results also get more consistent if I back off from that. And then there's the "wonky" extrusion I see when I try to extrude into free-air above a 5 mm/sec feed rate, which correlates to 300 mm/min. As it turns out, all of these are likely tied to size of the filament being used and a notion of extrusion capacity for the hot ends. In other words, just how much volume of raw filament can the hot end melt and extrude?

It's not clear what if anything the software does with it, but Repetier-Host even has a setting for this - max volume per second, with a units of cubic millimeters per second. The comment on the setting says one should be able to reach 12 mm^3/sec. I've seen 10mm^3/sec mentioned as a typical capacity in various web posts, with others encompassing a range from as low as 8 to as high as 18. I haven't heard or found of a value specific to any hexagon hot end, what yet specific to a 0.4mm nozzle version of it like I have. As with a lot of other things on the printer, it was enlightening to compare these numbers to the volume of filament necessary to support things like the extruder calibration.

Let's assume a possibly optimistic goal of 12 mm^3/sec as the maximum amount of filament we can melt and extrude. The input feed rate that correlates to this would depend on the filament diameter. The area of a circle is pi*r^2. The volume of a cylinder like our filament feed is simply the area * length. For perfectly round and properly dimensioned 1.75mm diameter filament, the area of the filament end is about 2.4mm^2. Calculating over a one second window for a volume of 12mm^3, the cylinder length of filament would be (volume)/(area), or (12 mm^3)/(2.4mm^2), or 5mm. Hmmm. Right at the 5mm/sec point where I've noticed my free-air filament extrusion getting "wonky". And 5mm/sec extends to the 300mm/minute threshold I've seen in the calibration extrusion. This suggests that a 5mm/sec feed rate while extruding seems to be at about the limit or slightly beyond the limit of my 1.75 mm hot end and 0.4mm nozzle.

How about 3mm filament? Area of the filament end is again pi*r^2, or just over 7 mm^2, about three times the area of 1.75mm filament. The cylinder of filament to be extruded in one second for volume of 12 mm^3 would be (12 mm^3) /(7 mm^2), or about 1.7 mm. 1.7mm/sec is just about the same as 100 mm/minute, where some people have been reporting issues in the calibration. Were they people with 3mm feeds and not 1.75? IDK - I should have paid more attention, but that's sounding pretty likely.

Maybe as a conservative view, extruder calibration should be conducted at something like half these rates, or 50mm/min for 3mm filament and 150mm/min for 1.75 mm filament.

WHAT ABOUT OTHER CONSIDERATIONS ?

Should the E term in DEFAULT_MAX_FEEDRATE be changed to reflect this? I don't think so. DEFAULT_MAX_FEEDRATE should likely stay at the mechanical limit of the motor drive in support of retraction and replenish. We don't have to factor melting new filament for those movements. It's just obvious now that we can't expect to be extruding at that rate, or at least not for very long.

So the capacity of the hot end to melt filament will likely be a limit on printing speed. All you'd have to do is factor the volume of the extrusion (most places say to just calculate volume as extrusion width * layer thickness and ignore factoring rounding of the corners), and divide it into the hot end flow rate capacity to come up with a print speed limit. The key to this would be in having a solid number for the flow rate capacity. Variables I've read others mention include the nozzle size and temp, specifics of the filament (material, color), and effectiveness of the thermal junction between the aluminum block (the hot end heat reservoir) and the nozzle. One source declared that the higher surface area to volume ratio for 1.75mm filament allows it to support a higher hot end throughput than 3mm. I can logically expand this list to include more possible variables like the quality of the thermal connection between the cartridge heater and the aluminum block, whether or not the aluminum block is insulated, and if so, the quality of that insulation. The approach used to cool the hot end heatsink could make a difference (air volume & direction, nature of insulation vs. airgap above the heat break, etc.). The accuracy of the hot end thermistor...

Results from any volume testing I did would likely be of little value to anyone else. Thinking I had cross-threaded my hex hot end aluminum block, I replaced it with a different one that is 20mm square instead of the hexagon's standard 16mm block. This likely changed the characteristics of the heat reservoir formed by the block. Since the block is larger, I've replaced the cartridge heater with one 20mm long; this heater seems faster (more effective) than the original one. I used heatsink compound on the cartridge to block connection; I may have even put a small amount on the nozzle threads. The aluminum block is insulated in two layers of kapton, with portions covered by three or four layers. The hot end shroud has been customized to remove the bottom plate and to improve the amount of air forced over the hot end heatsink. I almost always print with a print cooler running. Most days I'm using 1.68mm diameter filament - shy of the 1.75mm "standard". All of these factors will likely matter.

Could there be a relationship between hot end flow volume capacity and acceleration and jerk settings? Maybe, at least to a small extent. By increasing XYZ acceleration and going with a higher jerk allowance, we're reducing opportunities for "slack time" where the hot end might otherwise be able to restore heat lost to extrusion.

Yes, smarts built into the heater temperature control loop can deal with a lot of the variables - at least up to some extent. Regardless, I'm starting to see how there's more to pushing a printer to the limit than I realized.

HOMEWORK AND INDEPENDENT RESEARCH

For the free-air extruder e-steps calibration, what limit on feed rate do you get if you assume the hot end can only handle 10 mm^3/sec?

As I understand it, some slicers attempt to keep extrusions to something like 0.4mm while others accommodate a width setting. Do you know what yours is? Have you attempted to measure it on a thin-wall calibration print? Assume a simple rectangle extrusion of width * layer height, and calculate an estimated print speed that equates to the hot end volume limit of your choice.

A thicker layer or a wider extrusion on the first layer would lead to a lower limit on the print speed. Come up with another print speed estimate based on what you think you're getting for an extrusion on the first layer.

REFERENCES

http://www.extrudable.me/2013/04/18/...ty-and-limits/ - provides a good overview of the issue and demonstrates how feed volume capacity goes up with nozzle temperature. States that 8-10mm^3/sec is the limit for a 0.4mm nozzle. Note that the stated volume test results are high since tests were conducted with a 0.65mm nozzle - look at the data for an understanding of how temperatures, etc. affect volume flow rate, not as a baseline that you should strive for.

http://forums.reprap.org/read.php?1,226728 - says 10mm^3/sec is typical

http://umforum.ultimaker.com/index.p...der-extrusion/ - sort of a followup to the extrudable me reference. One comment states 10mm^3/sec is the limit for a 0.4mm nozzle. Comment #8 touches on the 1.75mm vs 3mm aspect.

FOLLOWUP COMMENT: I've noticed the MakerFarm FAQ page describes the hexagon hot end as being capable of 200 mm/sec print speeds. That seems high in comparison to the sense I'm getting regarding my hexagon hot end. Unfortunately, MakerFarm doesn't elaborate on the test conditions where that print speed is possible. If the test conditions are a 0.30 mm nozzle and a 0.10 mm layer height, 200 mm/sec could be reasonable because of the smaller extrusion volume involved.

[N5QM]

Do you expect that the extrusion feedrate could be increased by using a larger nozzle and/or wrapping the hot end in some insulated material minimizing the heat that is wasted?

Robert

[printbus]

Do you expect that the extrusion feedrate could be increased by using a larger nozzle and/or wrapping the hot end in some insulated material minimizing the heat that is wasted?

I was sort of hoping someone else would offer their opinion...

I assume we're talking about the extrusion feed rate that can be obtained during printing. To me, the limiting factor is the efficiency that the hot end can apply heat to and soften the filament being fed. In that regard, insulating the hot end should help (to some extent) since as you said, less heat would be lost to free air. As I understand it, new versions of the hex hot end come with an insulating jacket, perhaps for this reason. Benefits might be had through anything that improves the efficiency of heat transfer from the heat cartridge to the nozzle. The clamping style of the aluminum block on the E3D shows promise in this aspect. One could also add a bit of heat sink compound to the heat cartridge and possibly even the nozzle threads to improve heat transfer across the mechanical connections. Minimizing voltage loss to the cartridge heater or even raising the 12V supply up a bit might also help since more power would be dissipating in the cartridge heater.

I'd have to think about the nozzle diameter. My initial response is that I don't know that it would make a difference. It seems like it could be deceiving - it might, but would the larger diameter simply be allowing flow at a lower temperature and material that isn't as soft?

[N5QM]

I'd have to think about the nozzle diameter. My initial response is that I don't know that it would make a difference. It seems like it could be deceiving - it might, but would the larger diameter simply be allowing flow at a lower temperature and material that isn't as soft?

Interesting thought. I expect you are right, if the heater can't keep up with the demand then the larger nozzle may just extrude material that is not consistent over a larger print causing various other problems related to print quality.