Speed Limit

[mikemwa53]

What are the speed limits for the different print speeds in Slic3r and the Printrbot?
After getting my Plus dialed in for size accuracy ( within .003" ) I thought I would check out the speed limit. I started out with everything at 100, did a print and then went 50 higher. I kept going up to 500 and it is still printing. I'm sure this isn't really happening. Then I went back down to 200 and it looks like it is printing at about the same speed.
Are there limits to how fast I can go? Is there any way to see how fast it is actually going?

[HaniC]

Two part reply:

A) speed of x, y, z and e is capped in firmware. I believe Marlin defaults are 450mm/s for x and y.

B) more importantly these printers do not simply go from 0 to the defined speed. They accelerate to the set speed. Changing the speed from say 10mm/s to 80mm/s you will notice the difference because the print has enough distance to accelerate to the given speed. The default acceleration in Marlin is 3000mm/s^2. This means that given you max build size, you can only attain a maximum speed based on the distance you have available and your acceleration setting.

[jr57k]

great answer BTW!

[mikemwa53]

I went from a small piece I was printing to a flat 3" square. How can I tell where to set the speeds where I will be getting the max out of them as they are?

[FalconFour]

There are a number of issues you'd need to overcome with speed. First, the motor stepping frequency causing electromagentic noise and loss within the motor cables. Since they're not signal wires, they carry the full power needed to drive the motors. So when they quickly alternate back and forth, between different poles, switching polarity, etc., it's quite an EMI crapfest that causes the wires to negate each others' signals at higher speeds. So you lose steps (slipping easier), lose torque, and sometimes the motor just grinds if it goes too fast. The default firmware settings essentially have no limits - 3000mm/s (as I believe I've seen default) is about 6.5 MPH, so let's picture your printhead moving at the speed of a bicycle for a split second. The problem is that the firmware settings are tied to "speed" instead of "steps per second" as they should be. Speed could physically change, if the motors have different gearing... but all the X/Y/Z/E motors have the same rotational-speed limitations. Seems outright foolish that Marlin would tie the limits to axis speed instead of motor rotation speed.

Second, the extruder can't keep up with high speeds of anything. So if you're trying to blaze a quick straight line of plastic, you can forget it! The limit you can get is when you raise Z and disable motors, heat up the extruder, then see how fast you can turn the gear by hand without stripping. That's how fast the thing can ever print - and even then, it's a crapshoot that the faster you print, the less likely the layers will be to stick to each other, as the extruded plastic won't have enough heat stored to be able to melt and fuse to the underlying layer.

Third, the biggest problem with the Printrbot design, is X/Y jerk. When you go faster, it tries to apply that to everything. So when you go around the teeth of a motor or some narrow infill, the machine snaps back and forth on each segment, vibrating the machine and causing belts to slip, motors to skip, etc. Inertia is a bitch. I found some ways to minimize that jerk by having it slow down (further) around small segments: viewtopic.php?f=28&t=1116 - but you really can't (shouldn't) do much about the small segments.

Slower = better for most cases. Much as I'd hate to say... otherwise you'll probably just end up crapping out junk prints.

[Mooselake]

I'm sure you remember the 200mm/sec (12000mm/min) guys here .

Did you do an M203 and up the max speeds? I'm at the wrong computer and don't recall what the defaults are, but IIRC they were pretty low. M503 (just type M503 in the command window) will tell you what the current settings are). They're somewhat arbitrarily set, remember Marlin isn't aimed at just Printrbot's and I don't believe the defaults were adjusted in the Printrbot specific version. They're also in mm/second (mm/minute vs mm/sec is a frequent confusion issued) so multiply by 60 to get mm/min (the feedrate on G commands is mm/min). You're limited by the speed of the slowest axis, but thinking about how X, Y,Z, and the extruder limits would interact is beyond my insufficiently caffeinated brain.

Marlin claims a 30kHz step rate, or around 460 mm/sec with the pbot's X and Y around 65 steps/mm. Like all marketing claims that 30kHz is probably overstated, but 20k should be pretty doable (around 300 mm/s X,Y, 10 mm/s Z, 30mm/sec Z)). That's still audio range, the stepper wiring shouldn't be an issue at that low a frequency. See CL1's microstepping discussions elsewhere, too. If you ever get to where Marlin's a limit check out linuxcnc.org - I get a 50kHz step rate (way faster than the hardware) on the machine that drives my CNC router.

Jerk (basically how fast you can yank an axis sideways while it's moving?) seems to be a big topic in research circles, but the discussions and papers are from the academic crowd. They rapidly degrade into math that's beyond my 40+ year old memory of all those classes I took in ee school. FF4 probably understands it a lot better than I do.

Don't let all the reasons why it won't work discourage you. Bump up the max speeds and play with the acceleration and jerk values (one a time, of course) and see what your one of a kind Pbot will do! If you have a way to video your bot (with a ruler somewhere in the background) and then single step through the individual frames it would be real interesting to see just how fast it's really going.

All you risk is a pile of blobs in the saved for filabot box, maybe vibrating your bot onto the floor, and a massive explosion of flying bot fragments.

Kirk

Edit: fix min/sec mistake

[mikemwa53]

I'm at work now but when I get home later or tomorrow morning I will try to absorb the last couple posts. I don't quite understand all that but am very interested in it. I'm getting pretty decent prints and like to push limits to see what they will do before it flies across the room . I too think these machines are capable of more. As far as I know I have it going as fast as it will go with things as they are. I just need to be able to change things to push it farther. Thanks.

[drawcut]

FYI, simple way to think about it:
Velocity (speed) is the rate of change of position.
Acceleration is the rate of change of velocity.
Jerk is the rate of change of acceleration.


ETA: Jerks are bad.

[Mooselake]

[drawcut]

I really haven't had much time yet to play with the accel and jerk values. I'd start with FalconFour's post in Showcase that he links to above. Also this blog post has some good info as well:

[Mooselake]

Here's the default values from PB's Marlin, other than the calibration numbers for my bot:

>>>m503
SENDING:M503
echo:Steps per unit:
echo: M92 X61.00 Y61.90 Z2305.00 E590.00
echo:Maximum feedrates (mm/s):
echo: M203 X60.00 Y60.00 Z2.00 E14.00
echo:Maximum Acceleration (mm/s2):
echo: M201 X2000 Y2000 Z30 E10000
echo:Acceleration: S=acceleration, T=retract acceleration
echo: M204 S3000.00 T3000.00
echo:Advanced variables: S=Min feedrate (mm/s), T=Min travel feedrate (mm/s), B=minimum segment time (ms), X=maximum xY jerk (mm/s), Z=maximum Z jerk (mm/s)
echo: M205 S0.00 T0.00 B20000 X20.00 Z0.40 E5.00
echo:PID settings:
echo: M301 P22.20 I1.08 D114.00

Kirk

[Mooselake]

Anybody had the time to experiment with speeds, acceleration, and jerk? What have you found?

KIrk

[mdfast1]

I have not messed with them yet, but I did want to modify the XY Jerk setting. Please correct me if i'm wrong but to modify the XY JERK to 10mm/sec I could enter: M205 X10 then entering M500. Has anyone got any recommendations regarding a good XY Jerk setting?

[PxT]

[Marcus]

Does anyone have found good & faster settings so far?
I got everything dialed in a bit better now on my LC, but I get the impression I could speed things up a little without too much quality problems. I am unsure on how to modify the settings though, as just increasing the speed values would probably do no good without changing some of the acceleration and extrusion settings?

[plexus]

[plexus]

Ok I figured it out. it was nice to use google to search for this and find the first link this thread on printrbottalk. i merged my thread below into this one...

ok so there are max speeds set in the firmware. if you do M501 you get M203 X60.00 Y60.00 Z2.00 E14.00

so for my configs, anything above 60 was being ignored! this explains why i was thinking i wasn't getting anything faster than a moderate speed. even though i set the parameters in slic3r.

so i opened up the throttle by manually entering M203 X200.00 Y20.00 Z5.00 E14.00

I experimented with Z and found 5 to be the fastest I could go up. now with my fast configs... ITS FAST! wow...

i am printing a small cube now to test it. its printing very fast at around 150-170mm/s. print quality is down but at least now i can get the fast speeds.

you have to put the M203 command in your g-code header. be careful with this as it will make your PB print very quickly which may result in undesirable results or even permanent damage. know what you are doing.

[mikemwa53]

Cool!! Thanks Plexus. That's what I've been looking for. I'll see what I can do on my Bot. If you find any more tips and tricks keep them coming.

[plexus]

I am uploading 3 videos as we speak. they are with the defaults and sped up to full speed. once i have the links i'll post them. fast is FAST!

[mikemwa53]

Can you also list your Gcode?

[plexus]

[scantrontb]

it's probably in the firmware code, hidden like the high temp hard-coded limit of 250C, it won't print out to any kind of info dump, but it's there lurking in the code none-the-less... also same thing for the X,Y,and Z hard-coded limits of 205mm (and on some bots, only 190mm i think it was) again, not able to be shown via an "M" command, but it will prevent your hypothetical super-duper QUAD-Bot from doing anything more than a regular bots stock build-volume... unless you hack the firmware and recompile it with the new larger limits...

[plexus]

[mikemwa53]

WOW! This really makes a difference with speed. It's what I've been looking for for a long time. I've been able to bump up the speeds without any adverse effects except that the corners stick out more than normal. Not sure why. Thanks Plexus.

[wd5gnr]

Thanks to Plexus, I was printing last night and a policeman pulled my printer over and wrote it a ticket. Great work and very helpful since I am printing a bunch of cal pieces and now they take much less time.

I have a feeling this is going to increase the frequency of lube events though.

Oh, you ought to pull that post, put a disclaimer on it that you should do it after you are pretty well "tuned up" and make it sticky.

[SolidVis]

[plexus]

[RetireeJay]

This is a fascinating discussion. Along with Plexus, I assumed that the Y-table mass would be less than the Extruder with its motor, but I took actual measurements while building my kit and found it's not the case. The Extruder with its carriage (i.e. everything that the X-axis motor must move) weighs about 1.1 pounds. On the other hand, the Y-axis table for a Printrbot Plus with the heated bed and a pane of glass weighs about 2.2 pounds. The glass alone accounts for 0.8 pounds of that, so aluminum sounds like a very good alternative.

From a mechanical engineering point of view, you have a mass coupled to an actuator. The less mass you have, and the "stiffer" your actuator, the faster you can move the mass around. The actuation force comes from a stepper motor, delivered through a toothed belt, both of which have a finite "stiffness" and therefore there are mechanical limits to how fast you can move - or more precisely how fast you can accelerate (change the speed or direction of motion). Plexus has shown that the default values in the software are significantly less than optimum for a Printrbot Plus.

On my "to-do" list is to take my smartphone and tape it down onto the Y-table. Then I'll use a vibration analysis app to measure the actual acceleration and jerk. Mechanical engineers often test this kind of system with an accelerometer by tapping it lightly with a hammer. Not enough to make the motor slip a step, but just enough to excite the natural vibrational frequencies of the system.

[RetireeJay]

Another point is that the wooden bed on a Printrbot Plus seems to have much more wood than absolutely necessary. It extends 4" beyond the heater, when less than one inch would do. And if you are using glass or aluminum on top, and cork or silicone under the heater, then the wood underneath the stack doesn't have to be solid. You could drill out several "lightening holes" and reduce the mass of wood.
The big problem for me is: If I shorten the bed then I must drill new holes for attaching the clamps that hold the mounting rods - and the distance between these holes needs to be VERY precise in order to keep the rods exactly parallel.

[scantrontb]

[gyronictonic]

I think the reason why the wooden bed is longer than it should be is because the extruder is not truly center. The way its mounted makes it about 4" off center so the extra length from wooden bed has to make up for that offset distance. If you were to change the X frame to a horizontal rod setup like most mendel machines, the wooden bed would be a lot smaller.

I have a feeling that cutting holes to lighten the mass of the Y carriage is going to compromise the structural integrity of the wood bed. The belt applies a lot of stress to the bed and it may creep and deform. Would need figure a way to apply some sort of additional support at the right spot without increasing the mass. The X carriage mass can be lighten by switching to a bowden setup.

I'm interested about that vibration test. The new slic3r added a new but untested feature that gives you the ability to fiddle with the frequency or vibration. I can also perform this test and give you the data.

[SolidVis]

[plexus]

[MegaMoonMan]

As a side note, if you use repetier-host you can change most (if not all) of the firmware values under EEPROM settings and avoid command lines altogether. I've been using it for months and I've never had to use one command line code. Set the new value, reboot printer, and it's using the new values.

Adjusting that acceleration setting from 2000 down to 700 really helps with bumpy corners. I also upped my max speed from 60 to 100 without much quality degradation.

[plexus]

[halley]

[holmes4]

Why would a lighter motor help on the X axis? The only motion of the X motor is in the Z direction.

[plexus]

[RetireeJay]

To close the loop on my earlier posts, I have now attempted to measure the Y-axis motion with the accelerometer built into my smartphone. I say "attempted" because it didn't really work out. The time interval between data points is 20ms, which is not fast enough to really see what's going on. Also, the acceleration is relatively low relative to full scale, so there's a resolution problem there, too.

I did modify the bed. I left a solid inch of wood all the way around the original perimeter, and I also left a solid inch-wide "beam" of wood between the two belt anchor points. I made sure that the corners and a substantial part of each edge of the heated PC board were supported. But in the areas not "protected" by these rules, I drilled holes varying from 1" to 2-1/2" in diameter. Mostly, the holes are not connected or overlapping, so the effect is vaguely like a "truss bridge" structure. You see things like this in the frames of airplanes. Overall, I took about 1/2 pound of wood away from the assembly. Perhaps not terribly significant on an assembly that starts out at well over 2 pounds (including the silicone pad below the heater and the glass above), but little things can add up.
BedWithHoles_sm.jpg

By the way, while it is true that the mass of the bed increases as you build your part, the contribution to the mass in terms of dynamics will generally be negligible. Most of the parts I expect to make will weigh a fraction of an ounce. An entire roll of filament is only one pound, right?

The Extruder with its carriage weighs just over 1 pound, less than half the Y-table. So I wonder if we can optimize the speed of the system by using different maximum speeds and different acceleration rates for X and Y?

[plexus]

I am doing some experiments. I did a run of my speed settings from 2-8 (previous post) with XY acceleration set to 1000 and XY jerk to 20. i was noticing that as the speeds increased, i was getting bulky corners. so i created a version of each speed gcode file from 5-8 with acceleration at 9000 and XY jerk at 50, basically set to the maximums.

Of course this sped up the prints. for example, the speed 8 print with the slower acceleration clocked in at 9:59 but with fast acceleration 7:18 (about 25% faster with the fast acceleration). i will work out a new matrix using the fast acceleration times and see if there is more linearity in the fast end.

but one thing i am noticing is that there is much less corner bulk with fast acceleration! make sense though, right?

also... with acceleration at 1000 there is still more corner bulk than with acceleration at 9000 even on the slower prints. i only went down to my #5 setting though. but #5 with slow acceleration (1000) produces WORSE quality than at fast acceleration (9000).

this kind of blows away the idea that slowing your acceleration down produces better quality. frankly i think slowing it down some is just good for the printer. so maybe 5000? im definitely not going to go as low as 1000 and those that say numbers line 400 are good, well, do a comparison and see if its any better.

[plexus]

[MegaMoonMan]

I was using Mr jaws as a test for the acceleration. I scaled it down to 37 mm total length for testing and 700 acceleration at 100 mm/s had less corner bumping than 2000 acceleration at the same speed.

Granted, I scaled it down to around .3 of normal, but at that scale 700 acceleration made it better than the default 2000. I print a lot of tiny parts so this helps me a lot. I'll try 1000 acceleration and see how that goes, total print times didn't vary at much (if at all) between tests.

[plexus]

It could be that there is no one ideal setting for accel and jerk for given speeds. or it could be variations in our printers. but i get better results slow to fast with faster acceleration. however 9000 is just too much at fast speeds. so i am using 4000 and 25 and that seems ok. maybe its then pointless to share these numbers as universal settings??

[SolidVis]

[pomonabill221]

This has been a fantastic thread!
ALOT of VERY hard work, alot of testing and GREAT documentation!

I am going to try some different settings like the ones posted here, to see if I can't get things printed a little faster with good results, also.
I am going to create several printer profiles in repetir so that I can select the speed vs. quality in the future.
Thank You for your time and effort!
Of course, I still need to get my prints to stick yet!

[pomonabill221]

BTW... in your videos, how do you get your bot to meow like a cat? (hehehe...it was comical to hear the cat!)

[pomonabill221]

I have created several print speed "profiles" using slicer/configure/print settings/speed and also added the custom G codes to set max speed, X,Y,Z stepper calibration values and acceleration as I did see the EEPROM values change a couple of times and don't know why it did that, so I wanted to make SURE I override the eeprom settings when I print.

They do show up in the generated Gcode after slicing so I know the calibrated values are there.
WOW, the fastest REALLY makes the printer MOVE!!! I didn't use any filament YET, but will time a couple of prints and see if the printer is really moving as fast as the profile is set for... speed isn't necessarily the goal, but speed AND quality are so we'll see!

Thanks alot for this thread! Very helpful!

[pomonabill221]

[pomonabill221]

Also, it took me a while to figure out where these speed settings were going, but I found that they are under slicer/configure/print settings/speed.
It would have been nice (NOT a sarcastic remark, please) to mention where these settings go for us noobies.

I also added start Gcodes to the slicer/configure/printer settings/custom gcode to force calibration, speed and acceleration.

WOW this is FFAASSSSTTTTTT now!!!

Only thing is, I don't get the meowing?!?!

Once again, great thread and detailed info!

[pomonabill221]

[violadavis]

The motor stepping frequency is actually very important. It causes electromagnetic noise and loss within the motor cables. Eventually it will lead the wires to negate each others' signals at higher speeds. FalconFour is absolutely right about all this.

_________________________

[RetireeJay]

[plexus]

[thawkins]

[Mooselake]

It's only running at 20MHz if they've recently changed the microcontroller from an AT90USB1286 to an AT90USB1287. Since PB hasn't made any announcement (per the usual quick web search) then I'd be suspicious about that having happened.

Without looking (phone...) I seem to remember that Marlin has a configuration option for clock speed. If so, should be pretty easy to check.

A while back I did some more speed testing that appeared to indicate my X axis was step rate limited to around 15K s/s. For comparison, the BBB will do around (iirc) 200K with it's PRUs, and it has 3D printer firmware (of uncertain quality, haven't tried it).

Kirk