Direct Drive, Bowden, or Remote Drive… oh my!

  • administrators

    Since it seems there are a variety of people here and most if not all of the options I am considering are in use by members of this forum I figured I would ask this here.

    I currently have a single Flex3Drive on my CoreXY (quasi custom C-Bot derivative). It works pretty well, I have managed to tune the current to my steppers to allow it to spin fast enough to get some decent retraction speeds (about 30mm/s). I currently print mostly in PLA, PETG, ABS, or occasionally nylon but will be trying some more exotic filaments once I am done tuning this printer (ha ha ha ha yah like I am ever going to be done…)

    Question I have is this.

    Flex3Drive (with a custom housing re-designed to better work with the piezo sensor),
    Zesty Nimble,
    Titan Aero,
    Titan Bowden,
    BondTech BMG,
    BondTech QR (Direct or Bowden)

    Which would you choose for your own CoreXY, which have you used, what has your experience been? The cost is not really an issue for me, but I am considering going to dual extrusion sometime soon so keep that in mind when considering weight etc...

    The information out on the other forums is pretty jumbled/incoherent/laden with bias and I am hoping that maybe this group has a little bit more practical and/or scientific experience.

    Also... for the Flex3 I am thinking that changing my stepper might allow me to get even faster speeds since with a 40:1 gear ratio I don't particularly need torque. Does anyone have any feedback on if that would make a difference, or what specs I should be looking for to allow for the highest rotor speeds without losing accuracy?


  • I've used flex3drive on a corexy quite happily. Use a small stepper, low current like 500ma, low acceleration and jerk and then nudge the jerk/acceleration up until it stalls instead of retracting. I've got a Nimble it works too, but it's basically the same thing, in a slightly different shape with slightly different gear ratio but tuning it is the same process.

    I have titan aero on my corexy now. If you use a 20mm pancake nema 17 stepper and fix it to the x carriage using 2 of the motor bolts you have a very simple and not that heavy solution but it's still more mass than a remote drive. But super fast retractions and direct extruder performance. And I've a piezo plate design to sit between the carriage and motor which works.

    Titan bowden if you do it with a pivoting/swinging extruder or a second gantry with it on like Ian/Deckingman for a bowden tube below 20cm it will perform to all intents and purposes like the direct titan but with less mass to move around.

    I have not tried a bondtech but would like to get all the stickers in my extruder sticker book so if you buy one and its not for you let me know.

  • I'm running a Zesty Nimble on my D-bot. Can confirm, I can do pretty fast retractions (40-50 mm/s) with my pancake stepper+really low current (0.5-0.7 A, depending on the cable). It's a massive upgrade over a bowden and you can also easily fit 2 of them for an E3D Chimera type.

    I'd consider upgrading to the Titan Aqua for that sweet heated chamber performance. IDK on Bondtech, Titan seems more versatile and has enough grip for basically anything. You could use dual gears if you want to get a switching extruder (like Prometheus or Prusa MM style). Some extra pushing power is a massive help for that.

  • administrators


    direct extruder performance

    About this… I have only ever used bowden (original TEVO Tarantula) and remote. How does direct compare to remote and is it worth the gantry weight?


    I can do pretty fast retractions (40-50 mm/s) with my pancake stepper+really low current (0.5-0.7 A, depending on the cable)

    Thanks! Can you provide what motor you are using, or the specs for it?

  • Hi MSquared,

    If your wanting to increase printing speed using the F3D this should be as simple as turning up your print speed (and nozzle temp) in slicer settings. If you need more assistance or tips for your use case, just ping me an email or pm and ill be glad to help you dial in. The very early days of Flex3Drive development included speeds in excess of 400mms with high accels and jerk used for extended torture testing on 3mm and 1.75mm versions, but I will add that in real world scenarios, such speeds are rare due to other machine limitations.


  • administrators

    Hey J! Long time no speak ha ha. How does one get in touch with you these days? I sent you an email to the email I had for you around Christmas but not sure if you got it 🙂 I may be wanting some more PrintBite!

    I was talking more about retraction speed as the limiting factor when using an extruder that is as geared as the F3D is that moving the filament at high speeds (ie 30mm/s+) requires spinning the stepper quite fast. By my calculations its about 1350-2000 RPM (3500-4500 steps/mm * 30mm/s) / 3200steps/rev * 60 seconds). That is quite fast for a stepper in my experience and so the challenge in using a remote drive has been getting the stepper to spin that quick without stalling. Using lower current and/or smaller (i.e. pancake) steppers seems to mitigate this somewhat.

  • We advise this stepper for the Nimble:

    It should work just as well on your extruder and will enable faster retractions for you. The difference in gear ratio (30:1 to 40:1) will not make any difference.

  • @MSquared:


    direct extruder performance

    About this… I have only ever used bowden (original TEVO Tarantula) and remote. How does direct compare to remote and is it worth the gantry weight?

    It's basically the same, remote extruders are direct. The only thing they can't do is retract at 9000 acceleration at 60mm/s. Direct titan is simpler and cheaper. But practically the print quality is going to be the same up to a certain speed.

    As for whether its worth the gantry weight depends on how fast you print. If you print slower then direct makes sense. If you want to print faster then remote makes sense.

    But unless you can really optimise your remote extruder when your printing a model with a lot of retracts like Voronai or Eiffle tower, the slower, much more frequent retractions will add time to your print.

    But with both remote systems I could push the retractions up in acceleration and speed much higher than the default settings.

  • administrators

    Worth pointing out that direct and RDD extruders need much less retraction than Bowden extruders. So even though an RDD extruder may not be able to match the top retraction speed of a direct or Bowden extruder, it doesn't need to.

  • Regarding the zesty nimble, I asked for a custom mount for my nimble and they had one for me a few hours later. Support is also worth taking into consideration

  • @dc42:

    Worth pointing out that direct and RDD extruders need much less retraction than Bowden extruders. So even though an RDD extruder may not be able to match the top retraction speed of a direct or Bowden extruder, it doesn't need to.

    I agree with the first statement. But the second is a conflation in that an RRD needs the same retraction amount but cannot do it as fast as a direct extruder. It's fast enough for most I agree but put a direct titan against a RRD printing a model with dense retractions like those I mentioned above, and the increase you can achieve in print and travel speeds with RRD would be counteracted by the time taken to do the retractions.

    Printing a much less retraction dense model the RRD wins everytime its going to allow way higher speeds without inaccuracy or artefacts.

  • administrators

    Thanks guys 🙂

    What got me thinking about it was a comment I saw from Michael on sublime layers mentioning that he didn't feel that his remote systems were in the same class as the Titan Aero or the BondTech BMG. That surprised me enough to ask some questions and get others opinions. Your answers more or less confirmed what I already thought.

    For reference I am currently using a Bushang 17HD40005-22B that I scavenged off my Tarantula, which while not ideal is much better than the Kysan 1124092 that I was using prior. With my current tuning I am getting about 30mm/s retractions (firmware set to 40/1800/120 @ 700mA) up to about 4mm before the motor stalls out; since I started using pressure advance most of the time I am getting away with 1 - 2mm of retraction. I plan on tuning it more and getting a more appropriate motor.

    Which leads me to my follow up question(s) hopefully Dave or someone can clear this up for me, what makes a shorter stepper better at spinning quickly? They all seem to be low current / high inductance which is more or less the opposite of what you want every place else.

    What is the difference between maximum instantaneous speed changes and accelerations both in scientific as well as practical (print speed) terms. I suppose this question isn't strictly related to extruders as honestly I would love to know for other axis as well.

    Here is my config for those that want to pick it apart.

    M350 E16 I0                               ; Configure microstepping without interpolation
    M350 X16 Y16 Z16 I1                       ; Configure microstepping with interpolation
    M92 X100 Y100 Z800 E4390.344              ; Set steps per mm
    M566 X900 Y900 Z12 E40                    ; Set maximum instantaneous speed changes (mm/min)
    M203 X12000 Y12000 Z180 E1800             ; Set maximum speeds (mm/min)
    M201 X1500 Y1500 Z250 E120                ; Set accelerations (mm/s^2)
    M906 X600 Y600 Z600 E700               	  ; Set motor currents (mA)

  • the OMC short ones are not particularly high inductance but the main advantage is that there is much less mass in the Rotor to access decel which is what improves the velocity response and I am currently running with 1mm retract on mine and don't use Pressure advance I let KissSlicers preload algorithms do the work. 4mm retracts on a RDD is far to much and the reason that too high a retract speed might not be advantageous is that it tends to break the filament rather than retract the melted portion which can lead to ooze.

    Edit for typo's

  • administrators

    I too use KissSlicer but my personal experience is that pressure advance has given me better results. Using the tuning wizards in KISS my "best" setting for preload was essentially 0; however I was able to tune pressure advance to produce very consistent extrusion. So I just run my PA tuning script on every filament at my average temp for that filament and then load my tuned PA command into the MATL GCODE tab in KISS. My machine prefix script sends this command, along with whatever acceleration settings I have set for the printer to RRF before it starts each print.

    I agree 4mm is too much, was mostly posting that for reference, however you may have just explained why on my ooze tests I started seeing stringing again at about 3mm retraction LOL.

  • Isn't stringing more to do with too low travel speed coupled with too high hotend temp?

  • administrators

    With the short retraction distances of RDD extruders, the peak retraction speed is likely be acceleration limited. So a high ratio of torque to rotor inertia is desirable.

  • @MSquared:

    Thanks! Can you provide what motor you are using, or the specs for it? - that's exactly the one I'm using. Smaller one is better because you don't need all that torque with the gearing on RDD, and the lighter rotor is easier to move.

  • The peak retraction speed is very much acceleration limited and I'm am quite dubious they achieve the speeds people think they do. Easy enough to test send M207 Fx and vary their retraction increasing gradually the speed. At some point the motor will cease to turn any faster. At this stage you have to nudge the acceleration up to get it to go faster an you may stall. This presumes firmware retract is in use.

    David if running at speed 200% are retracts and unretracts performed faster also? If so you'll need to leave yourself some headroom when tuning the extruder, in case you want to go a bit faster when printing.

  • administrators

    If running at 200% speed then retracts/unretracts specified in the GCode using G0/G1 commands will be speeded up, but not firmware retraction commands.

  • Hi Msquared,

    I perhaps misread your question, (synapses misfiring??). taking it as being one about print speed, as opposed to retraction speed (thought it odd). Regarding email, send again (or use contact form on website) and I should receive it, there were some (unusual) site issues toward end of last year.

    re motors: Whilst motor spec is important, with the wrong order of tuning steps, an "optimal motor" will perform far below a less optimal motor tuned correctly. I have had large (long can) NEMA17s performing just as well as small NEMA14s and most others that are in between. Avoid very high inductance motors though, and use only 1.8 degree step angle type. 24v system voltage can be beneficial. Of greater importance however is the tuning. This goes from choice of microstepping, motor current and jerk set to a minimum, then dialling in on accels (in fw) and speed (in slicer).

    This is simplfied but preferring to keep short on here. Ping me a message and I can assist in greater depth.



Looks like your connection to Duet3D was lost, please wait while we try to reconnect.