I made another MP3DP…..kinda

New Home Forum Milled Printed 3D Printer -MP3DP Hardware Development – MP3DP I made another MP3DP…..kinda

Tagged: ,

This topic contains 160 replies, has 14 voices, and was last updated by  Barry 3 months, 1 week ago.

Viewing 30 posts - 121 through 150 (of 161 total)
  • Author
    Posts
  • #51545

    Ryan
    Keymaster

    So stuck in this rabbit hole. New part, test the angle/jerk from 90-~180°. Learning new things each test but not getting closer to any concrete accel and jerk numbers.

    MVIMG_20180122_1840112

    #51553

    geodave
    Participant

    If you really want to get adventurous you can try the klippy firmware.  I did not play around with it enough to get any better prints than with Marlin, but there seems to be quite a few folks getting better & faster prints with it.  That firmware will slowly accelerate and decelerate.  If you can get the settings right it should get rid of those rings.

    https://github.com/KevinOConnor/klipper

    #51555

    Barry
    Participant

    I wish 3D printers were this fast!

     

    #51563

    Ryan
    Keymaster

    geodave – klipper has wonderfully laid out motion equations. https://github.com/KevinOConnor/klipper/blob/master/docs/Kinematics.md

    I have to try klipper, just can’t ditch the low cost option yet.

     

    Barry – My printer at 120 looks pretty crazy!

     

    This whole tuning thing comes down to, why the heck doesn’t it print perfect at slow accelerations (bulging corners). Klipper seems to be saying (among other things) pressure advance needs to be there, makes sense. Marlin has linear advance, before I give up I am going to try it. As is the corners are not very good, they look okay, but the negative acceleration side is extra strong and the positive acceleration side is extremely weak. Then Jerk should actually only be used as a look ahead tuning device for curved paths.

    #51571

    Jeffeb3
    Participant

    This whole tuning thing comes down to, why the heck doesn’t it print perfect at slow accelerations (bulging corners).

    Yeah. It’s hard enough since there are three variables to tune at once. It’s even hard if these have a “right” setting, with each extreme looking terrible. I keep thinking about it as a “conservative” and “compromised” scale. Extrusion is the hard part of it. The linear advance looks good. The fact that klipper and Marlin are using different physics models indicates that it’s probably not as simple as the motion. The pressure (klipper) theory makes a lot of sense too.

    I still think using jerk to smooth corners is not the right choice. It’s nice that it does, but the extrusion is causing the problem and jerk still seems to me to be an optimization. I would really like it if acceleration could be tuned with jerk off. But I’m not doing the work :). (((My phone keeps putting the word “Jeff” instead of “jerk”. I wonder what it knows that I don’t))).

    #51582

    Ryan
    Keymaster

    HA dam phones, no more Freudian slips, Machine learning, knows too much.

    Jerk is the only one I can get to really do anything but it honestly seems to just be on or off, the only variance is at what point on the subtle curve it triggers. The extremes don’t do anything 3.5 to 4.5 seems to be the valid range.

    I have spent too many hours, something isn’t right but the motion equations are so convoluted spread out over 6 different files there is no way to double check it. It took two years to figure out Jerk didn’t do anything, now that it works (kinda) something else is busted. There is a huge margin where the settings don;t do anything. At the extremes it works but in between something else is clamping it. I can only think to check marlins direct output and plot each point speed/time.

    I just checked Prusa’s build and it doesn’t even have the same motion going on. His config is not even remotely similar. There is even a bunch of settings called Prusa’s magic.

    As it is the prints are as good as I can get them for the speed (up from 35 to 100!), I just want to look into the advance settings to make better corners and small details.

    #51587

    Jeffeb3
    Participant

    Linear advance seems like a good place to go. The unfortunate thing is that you’ll have to tune the k factor when you change temps or filament. But it seems easy enough to tune (I say. Having never tried).

    IF you try that, and it works, then it would be worth seeing if you can get decent corners with low low jerk. Too much jerk and it will have ringing in corners, no matter what Accel you choose.

    Can you tell if the 100is actually being reached? Are the print times changing? Is it worse at 120?

    #51588

    Jeffeb3
    Participant

    W.r.t. prusa magic: it really seems like it shouldn’t be that hard, but I guess if some magic salt helps…

    #51594

    Ryan
    Keymaster

    Can you tell if the 100is actually being reached? Are the print times changing? Is it worse at 120?

    Yeah it gets there quick. 120 is too fast to test with the parts are too soft at a single wall so I had to do most testing at 55-75.

    I can see the one setting that explains Jerk in the firmware, but not the actual equation. it just says speed with the jerk offset applied. leads me to believe it is just “zero to jerk equals no accel, then accelerate starting from this speed to the final speed if the final sped can be reached before negative accel” Then there is a the look ahead that says “from vector 1 to vector 2 is the speed change more than the jerk value per axis, if so accel.”

    I think the problem might be in the extrude jerk, accel settings. It really might be clamping it.

    After shipping I will look into it deeper.

    #51616

    Jeffeb3
    Participant

    “zero to jerk equals no accel, then accelerate starting from this speed to the final speed if the final sped can be reached before negative accel” Then there is a the look ahead that says “from vector 1 to vector 2 is the speed change more than the jerk value per axis, if so accel.”

    Not sure what the problem is. That seems to match with my understanding. The reason things are worse with lower jerk is because the extruder is being controlled linearly through a bunch of stop and go in your curve. So the extrusion is even, even though the speeds are not. But you want the speeds to be uneven, that’s the only way you’ll get detail at high speed.

    #51617

    Ryan
    Keymaster

    I’m good with all of it, I just feel there is no way to really dial it in. The range for the Jerk was easy, high enough to get the smooth curves, as low as possible to keep print times as low as possible. Acceleration really comes down to how the machine sounds, you can here it banging around when it accelerates too fast. I just wanted a easy analytical way to do it. Wam bam thank you print test. I think I settled on 1600 accel, 5 Jerk, I dialed in the Z axis a bit more as well. Print times are a bit longer 15 min/8hours at my slow print times. Tomorrow I will re-slice with higher speeds and see how it comes out.

    Still haven’t figured out the mini rambo lcd. I tried the archim fix (mem page size) and adjusting the memory timings….Above my pay grade but I am going to give it one more look before I post it to github, for the brutal help.

    #51619

    Barry
    Participant

    These parts are soooo cool looking.

    #51629

    Ryan
    Keymaster

    Adding another or just updating the current one?

    #51634

    Barry
    Participant

    I actually haven’t built one of your printers yet.  I’ve turned some hardboard into vague parts shaped like your first version, but didn’t finish it.  I havent decided if I’ll finish this one either or if it just gives me something to tinker with.  I think i have all the parts printed now though.  I have a spare seeme hot end I could use too.  I have to make another wine bottle holder for our ikea shelves this weekend, I’ll probably cut out the wood parts then.  I get the feeling I’m going to have a second 3d printer sooner or later…  ?

    #51639

    Ryan
    Keymaster

    Do you happen to remember the direct /remote drive extruder you pointed out while we were in line for tacos? I would love to see if they are still at it and how the prices are. It had the flexshaft drive thing.

    #51640

    Ryan
    Keymaster
    #51646

    Barry
    Participant

    The Nimble ?

    That’s the one!

    #51700

    Bill
    Participant

    Dang, $100 each… Really good idea though. I could see six of them driving into a six way diamond head giving me all the colors I’d ever want. 🙂

    #51705

    Barry
    Participant

    http://www.sublimelayers.com/2017/05/everything-you-wanted-to-know-about.html

    I think I posted this before in the MRRF thread, but here’s a really good review on it.

    #51723

    Jeffeb3
    Participant

    Wow, 30:1 ratio. I wonder how much backlash there is.

    Do you have one Barry?

    I’m assuming this would still need a hotend (and a mount). It seems like a great compromise between Bowden and direct.

    #51733

    Barry
    Participant

    Nope, I made a flying extruder for my rostock.  Yea, it doesn’t come with a hot end, but they post all their adapters on thingiverse.

    #51802

    Ryan
    Keymaster

    AHHHHHHHHH dammit. I finally have the LCD working as of this morning. Yesterday’s updates from the main Marlin 2.0 buggered up the rest of the system now…….grrrrr.  I tried to roll back the main updates but just made a mess of things so I will wait for it to get sorted, probably just a missing comma or something.

    Dang it I had tests to run, MP3DP V1’s to update….

    Side note, I did some driver tuning tests….There are definitely too low of settings, at about a 0.5V on the drv’s I was getting nasty textures, ramp them all back up and bam smooth sailing. My two mini-rambo printers were printing sooo much nicer, it spurred this whole tuning fiasco. The digipots really make this easy but the ramps are holding there own again after some driver tuning. Can’t remember why I had turned them down in the first place, bad bearings causing hot steppers I think.

    #51804

    Barry
    Participant
    #51807

    Ryan
    Keymaster

    HAHAHAHAHAH

    #51823

    Ryan
    Keymaster

    I keep refreshing the page to see if the updates have been fixed, and came across this thread. And in that thread the is a gcode analyzer!!!!!

    This is how you can visualize your tuning. So awesome. http://www.gcodeanalyser.com/

     

    I figured out how to revert the two bad commits, hopefully I didn’t screw anything up for the fix when it happens. Look for some MP3DP firmware updates shorty for both versions.

    #51829

    Bill
    Participant

    Look for some MP3DP firmware updates shorty for both versions.

    Along with new suggestions for default speed, acceleration and jerk? B)

    #51830

    Ryan
    Keymaster

    So far I have 1600 and 10. If I remember right. The file is on my github. I will probably revisit it one more time now that we have the visualizer. I have printed a few things now with the new settings and it only sounds funky doing quick infill so I think bringing it back down to 5 jerk fixes that. But the prints look great just spend a little more time checking it.

    Looking at the visualizer I think i was right my moves are being limited by the extruder so I want to sort through it.

    Thinkyhead found another issue with the planner, last segments are getting disconnected for some reason.

    #51834

    Jeffeb3
    Participant

    Great visualizer and great issue thread. No one is that thorough for free on the internet. Almost no one.

    #53804

    Marshall
    Participant

    I’m not sure where this part goes.20180223_081031

    #53812

    Ryan
    Keymaster

    That is a cap to optionally cover the X axis stepper.

Viewing 30 posts - 121 through 150 (of 161 total)

You must be logged in to reply to this topic.