Why "Print Speed" Doesn't Matter Much

[Ryan Carlyle]

People around here treat print speed like it's an *ahem* manhood length contest. But for small prints, the number you enter for "print speed" is pretty close to irrelevant. The parameter you put in the slicer barely does anything once you hit a certain point. Unless you're doing big, straight-line prints, the only thing that really matters is your firmware acceleration settings. 

This is just like the speedometer in a car. It doesn't matter if the dashboard says your jalopy can go 180 mph if the engine sucks. And even a Ferrari can't get much speed if you're only going to the end of the driveway. What matters is how fast you accelerate, times the length of straight road you have to get up to speed (and then slow back down again).

It doesn't matter what speed you tell the slicer to use, if your model doesn't include long and straight extrusion paths for the nozzle to get up to speed.

To prove the point, I printed some 20mm calibration cubes today with my R2x. I set all speed parameters to the same value on each cube (shells, infill, tool moves). All extrusion paths were linear and at 90 degree angles to the box, so all plastic was extruded on a 20mm long straight line. This is a nice controlled test.

Print parameters:

Sliced in Makerware

All prints started after printer pre-heated

20x20x10mm calibration box at 100% linear infill -- 4,000 mm3 extruded volume

0.1mm layer height and 0.4mm nozzle

Firmware acceleration settings about 10% more conservative than Makerbot's defaults 

Theoretical print speeds calculated from the plastic volumes and travel speeds (NOT the slicer estimate)

Print speed setting results:

Slicer: 50 mm/s

Theoretical print time: 33 minutes 

Actual print time: 37 minutes

Average print speed: 45 mm/s

Acceleration added ~11% to the non-accelerated print time

Slicer: 100 mm/s

Theoretical print time: 17 minutes

Actual print time: 27 minutes

Average print speed: 62 mm/s

Slicer: 150 mm/s

Theoretical print time: 11 minutes

Actual print time: 26 minutes

Average print speed: 64 mm/s

Slicer: 200 mm/s

Theoretical print time: 8 minutes

Actual print time: 26 minutes

Average print speed: 64 mm/s

I could tell the printer to run at a 10,000 mm/s and it would still finish in 26 minutes. It never reaches any speed over ~105mm/s on this print, so there's simply no point in telling it to move any faster than that. 

Yes, this is all for a small part with 20mm long features. But this applies to any size print if it has complex shell geometry. Infill pattern matters enormously too -- the Makerware default hexagon is quite slow because the extruder has to turn 60 degree corners over and over again. (Cat and shark fills are just god-awful.) Even linear fill needs to be aligned with the part shape to really be fast. Most people never tweak the infill rotation settings but they're very important to both print speed and part strength. One time I cut print time in half just by changing fill orientation.

If the nozzle is turning corners, jumping around the print, or printing any type of infill other than orientation-optimized "linear," then you are probably spending most of your print time in an acceleration-limited state. 

What you should take away from this is:

• The mechanical quality of different printers mostly affects how conservative or aggressive you can make the acceleration settings and still have good print finish. "Print speed" is mostly a red herring.
• Almost all marketing hype about high print speeds is useless, if not outright deceptive. You can't trust any impressive print speed figure unless you can see the test print geometry AND know the firmware acceleration settings.
• Fine-tuning your firmware acceleration and print geometry is usually much more important to your true print speed (and sometimes quality) than slicer speed settings, over perhaps ~80mm/s or so. 
• If you really want faster prints but can't increase layer height, use linear infill aligned with the longest axis of the part. That way your nozzle actually has enough "runway" to hit top speed once in a while. 

[Dan Newman]

> *Print speed setting results:*
>
> *Slicer: 50 mm/s*

> Theoretical print time: 33 minutes

> *Actual print time: 37 minutes*
> *Average print speed: 45 mm/s*

> Acceleration added ~11% to the non-accelerated print time
>

> *Slicer:* *100 mm/s*

> Theoretical print time: 17 minutes

> *Actual print time: 27 minutes*
> *Average print speed: 62 mm/s*
>
> *Slicer:* *150 mm/s*

> Theoretical print time: 11 minutes

> *Actual print time: 26 minutes*
> *Average print speed: 64 mm/s*
>
> *Slicer:** 200** mm/s*

> Theoretical print time: 8 minutes

> *Actual print time: 26 minutes*
> *Average print speed: 64 mm/s*

And, as you would expect, I see the same sort of results
if I run things through Sailfish's simulator. (I have
a utility called sailtime that spits out the total print
time. It's quite accurate.)

Dan

[Steve Johnstone]

Great write up Ryan,

I always suspected but never understood why.

Cheers Steve

[David Kessner]

But you're talking about "fast" print speeds.  Going to the opposite extreme, slow print speeds, does make a difference.  Not only does it increase print time, but it also usually improves print quality.  I just wanted to point that out, just in case someone didn't read beyond the headline or first paragraph of your post.

Other than that, great post!  

-David Kessner

[Joe Larson (aka Cymon)]

As I read your post, Ryan, I kept thinking "duh." But I guess there are folks out there who don't know these things. Even so, the day I turned on acceleration I was blown away at how casually fast the printer was suddenly going. It really makes a difference.

On Wednesday, April 30, 2014 6:50:30 AM UTC-6, David Kessner wrote:

But you're talking about "fast" print speeds.  Going to the opposite extreme, slow print speeds, does make a difference.  Not only does it increase print time, but it also usually improves print quality.  I just wanted to point that out, just in case someone didn't read beyond the headline or first paragraph of your post.

Not necessarily so. In fact faster print speed can actually make corner ringing less visible and there are other advantages, particular with printing ABS. Of course ABS like's it HOT but it also likes it fast.

[Ryan Carlyle]

What printing very slow does is get you out of the range where acceleration is entirely controlling print speeds. Then your somewhat-aggressive default acceleration settings aren't allowed to damage your print quality. And it makes sure the firmware motion planner is always able to keep up with the necessary motion path computation without having to drop anything.

One option (in theory) is to have two sets of acceleration settings -- one aggressive for fast prints and one conservative for high-detail prints. Then you could just leave the slicer print speed settings high and always get the optimum speed/quality tradeoff.

This would be like driving your car really aggressively to get your wife to the hospital when she's in labor, but then driving very gently when taking the new baby home. Same car, same speed limit signs, different acceleration.

On the other hand, this relies on the acceleration planner to always do a perfect job, which it doesn't (due to processor power limits mostly). Firmware doesn't know the slicer's intent, has a puny processor, and it can only see so far ahead. So an even better solution is an acceleration-aware slicer. Put the vector math onto the desktop machine! Unfortunately, attempts at this like Makerware's doDynamicSlowdown function haven't done a great job at this to date.

[Steve77]

Ryan:

Thank you for the test data.

Have you run any tests with different acceleration settings in Sailfish to determine appropriate settings?

Do you just set Sailfish acceleration to one value, or do you change it depending on the specific part being printed? What acceleration value(s) are you using?

[Ryan Carlyle]

I started with the default MBI R2x settings, but lowered them by about 20%. But I'm mostly printing rough mechanical parts where ringing isn't a big deal. Sailfish defaults are a lot lower than that. In particular, I believe Sailfish used a much lower number for the high-mass Y axis.

Having two extruders on the gantry means you should probably run acceleration settings ~50% lower for the R2x than the Rep2, to get the same print quality. In theory you CAN use two different sets of settings for high-quality and low-quality prints, but I've been too lazy to do it so far.