Recipe for speed - Fast 3D printer Motion system

[@eddietheengineer]

Thanks for the shoutout! Passing 3000mm/s is such a fun milestone 😄 I am doing some more research on driver tuning and will have a video out in the future!

[@jeremyglover5541]

Back EMF is what you are talking about in the motor current. 'A motor has coils turning inside magnetic fields, and a coil turning inside a magnetic field induces an emf. This emf, known as the back emf, acts against the applied voltage that's causing the motor to spin in the first place, and reduces the current flowing through the coils of the motor.' We go to great lengths to minimize/cancel this in the coils in speaker drivers.

[@SianaGearz]

Thanks for pushing the limits. It might be seen as a "useless" endeavour just like space exploration and F1 racing, but i do expect that insight that gets generated in the process is transferred into better faster more reliable prints for everybody eventually.

[@andybrice2711]

I don't think it's a frivolous endeavour at all. I think it's entirely practical. Fast 3D printers allow ideas to be prototyped more quickly. And they allow complex parts to be bulk-manufactured with minimal space requirements and tooling costs. Speed is not the most important factor for every 3D-printing application. But for some projects, it's exactly what you need.

[@GatlingNG]

Every technological hobbyist-specialist should have overviews like this. Great information in a super convenient format. Thanks !

[@jeremyglover5541]

yes, to expand on the 12v vs 5v drive. your mosfet is a switch that is controlled/opened by the gate voltage (measured from Gate to source, or 'VGS Voltage from Gate to Source) to allow current to flow from drain to source (well, the electron holes will flow from source to drain) when you switch with 12V, the gate is biased on harder, so in effect the channel is opened wider so more current can flow. This means the drain to source channel presents a lower resistance, meaning it creates less losses (seen as heat) and a more linear transfer characteristic. with 5v operation, the channel is constricted, creating higher resistance and you can use ohms law to calculate how much extra heat (for the same load current) needs to be dissipated in the mosfet when it isnt completely open.

[@BH4x0r]

that depends, good Logic mosfets will switch as fast (and hard) or maybe faster on 5V than a generic 20VGS fet on 12V, granted the resistance isn't too high and thus the current is high enough to deal with the gate capacitance fast enough (especially in PWM situations), what also matters is the RDS(On), package type, and ultimately PCB design, you can have a fet thats heatsinked and driven well enough to handle 50+A continuously but that doesn't mean shit if the traces and other connectors overheat at 30A, same for the wiring going to the heaters

[@jeremyglover5541]

@@BH4x0r but we arent talking about logic level fets. most of your logic level low RDSON fets that will be triggered by even 1.2V and are designed for server, or CPU power supplies, dont take kindly to switching 48V and having at least that, ie more VDS voltage ... and with the amount of back EMF and other ripple here, you arent going to want to be using anything too twitchy that switches with too low a voltage. Most of the fets you are talking about are designed for max ~20-25VGS. Mello wont be using anything too speccy here, with this budget.

[@BH4x0r]

@@jeremyglover5541 none of these boards are rated to switch over 24V, the 48V goes solely to the stepper drivers and capacitors around it, the rest is made to run on 12V to 24V, same with other brands like BTT, so 50VDS or even 40VDS mosfets should be more than plenty to switch resistive loads (i.e. hotend, and other heaters)

[@jeremyglover5541]

@@BH4x0r If I put the device package exposed pad 8-SON + LFPAK56 package for example into the digikey parametric search on standard single SMD MOSFET (not GAN, SiC or anything speccy) my VGS voltage choices get MUCH smaller. The internal regulator of the 5160 is kinda lame as well and the datasheet specifies an external driver for higher voltage switching, as the internal reg cant handle the voltage drop and gets too hot. Of course there are exceptions, but we were starting with the proviso of talking about a driver that performs better with higher gate voltage, so I explained the mechanism for that. If iit was OK switching with 1.2V and had a very low gate threshold, obviously 5V would already be enough and we wouldnt even be having this conversation. The topic far too broad to cover all options in a post. regarding PCB layout/implementation being king, you wont get any argument from me, the PCB in question is nowhere near optimised, particularly thermally. With proper layout for 5160 and well chosen fets, water cooling shouldnt even be part of the conversation. The PCB alone will be enough with nice, low RDSON parts and a thermally optimised, 2oz copper PCB, with enough space/copper devoted to the Fet drains/power pads. These parts sink their heat to the PCB, not through the plastic package to the top of the part ...

[@jeremyglover5541]

@@BH4x0r fair call on the 48v going to the motor. like I said though, if you put 1.2V-1.8V drive into parametric search, your options pretty much drop to +/-20. Now that might be ok, but I wouldnt choose those parts for this job personally.

[@TheNamelessOne12357]

Last part was the most important. I used to design and build fast robot gantry that grabs, moves and drops 100kg of load at 2 m/s with 5K acceleration. I used largest NEMA34 servo-stepper motors at 80V power supply, even had to make separate braking circut to burn deceleration energy into heat. When this beast accelerated it was really scary, although i managed not to crash it at all.

[@ezyfnef]

That sounds like a lot of fun. What do you do?

[@rexxx927]

I used super lube for 3-4 years now on printers its great I may have even shared this with you way back when, on heated chambers above 75C you will need krytox 200 series grease but skate board butter works good to for those without a chamber or it will run from the heat and just dry up cheers Great job explaining the deal on mach Chiken speeds and control keep it up!

[@ldomotorsjason3488]

Haha, I love the idea to get some alcohol to calm down your heart beat from the high speed running. Let's GO VzBot!

[@hereisanaccount]

Really glad to see you doing this longer format. I certainly learned a few things. You answered questions I didn't know I had. Looking forward to whatever comes next.

[@noanyobiseniss7462]

Looking forward to Eddy's Video!

[@christoskaragiannis7973]

My routine: I see Vez uploads, i like the video, then I watch it :D

[@umitbaba5690]

Insane! I have a quastion please. For safety, using closed loop circuit in addition to prevent lose step or crash would it work?

[@twanheijkoop6753]

Haven't seen any 48v closed loop stepper kits. MksServo42C pcb seems to be a nice piece of kit but I'm not sure how it'll stack up in raw power to external 5160s

[@eugene3d875]

Thank you for walking through your process! Rum and PTFE!

[@T0tenkampf]

you pronounced cohones in pure Tejano style man, great guide btw...I will be using this info on my VzRat hybrid

[@tadhgd350]

I've been running my 2804s on 2.382A for a couple hundred hours (specific current because of tmc tuning) inside a 90+C chamber and they're still working great. when I talked to Jason he said 2.8A even should work

[@ldomotorsjason3488]

The current on LDO drawing or specs is RMS current, so no need to redue to 0.7 of it. Just make it crazy 🤣

[@tadhgd350]

@@ldomotorsjason3488 yeah my power supply can’t deliver 2.8A to each motor😂 only reason I run em lower

[@ballbous]

@@ldomotorsjason3488 😍

[@Vez3D]

​@LDOMOTORS wait what?? Thought it was peak!!! Wow ...ill crank even more😂

[@ldomotorsjason3488]

@@Vez3D yes, try it; if driver no problem. Some guys even run 4A on 2504AC in short time and no any issue. When you have water cooling, more possible……

[@HardCoil]

Thanks for the part about the oil. I have been thinking WAY to much about this, since it's wrong on paper, but if it works in you application, it'll be fine for just about everybody else :)

[@kilianlindlbauer8277]

247 printing had the same issue with mcu processing speed. Albert was using a manta e3 ez with a stm32gb0 or gb1 (cant remember) like all the other manta boards. It has only 64 MHz. This mcu is also used in canbus boards and other low duty boards like the skr e3 series. Even though the processing speed doesn't matter for klipper as there are no calculations to do but you still need a fast mcu for the step pulses. The super 8 pro uses most likely a stm32h7xx series mcu like the skr3 (i have a skr3, thats why i know that) and has 550MHz. About the grease, thats my experience: i first cleaned the disassembled rail (even the balls where in their own container), cleaned with brake cleaner, so it really gets rid of any dirt or leftover machining oils and reassembled with ptfe grease. I lather got skipped steps while doing the z tilt with 48mm moons at 1,4 amps. Redid the same process but used regular bearing grease that i use for wheel bearing in a car (the translucent yellow stuff) and had 0 issues since then. Some have great success with ptfe lubricants, others like me don't, so it really a matter of luck i guess

[@Andre_M_3D]

Nice. Learned a lot. Merci cheers from Canada 🇨🇦

[@YouvBeenThumped]

The man has all the monitors now

[@andybrice2711]

More monitors = more success. Wall Street understands this.

[@jeremyglover5541]

In pro and high end Audio, when we drive in a bridged or balanced/differential configuration, we actively drive the output with + (VCC+) /- (VEE-) voltages, instead of single ended with + and GND, we get what is called Common mode cancellation (or specifically 'common mode rejection'). because the 2 outputs are equal and opposite in phase and any noise, or distortion (basically error) that is common to both outputs is cancelled when they are summed at the output. in a perfect world they would be cancelled completely, but in the real world we have a factor called the common mode rejection ratio (measured in dB) that explains the factor of that reduction in error. I expect you are seeing something similar with the AWD system, as many of the forces will be created in a ... similar, if not equal amount by each side of the axis, as it has a motor creating this vibration/resonance from either end and some amount will be cancelled in the middle.

[@dad0033]

Should try a build with a duet3 6hc , the driver are TMC 5160 and he mono support 48v , and up to 6 amp for steppers motor , my board an older version so him stuck at 32v max

[@Dave-wv9vc]

Vez I'd love to see how fast you can do a quality benchy run. Thanks for the video!

[@phylippusvonhohenheim3232]

I loved the video Vez, it was a great learning experience, I could see that you have a lot of experience with "go crazy", but I have some questions. What is the maximum speed to print every day? Without breaking records or testing the machine's limits, what's the fastest you can go without taking too much risk but still impressing onlookers? please tell me the print flow in mm3 and motors speed. Greetings from Brazil

[@christianmontagx8461]

As always: No guts no glory :) Well done. 😉

[@tsstn]

Hey VEZ thanks for showing us all this awesome work you have been doing. Also, just to be helpful and not intended to be rude in any way. Oscilloscope is pronounced -- AH SILL OH SCOPE You are not the first to wonder how to pronounce that correctly and won't be the last.

[@Vez3D]

It sounds better in french 😂

[@SianaGearz]

I have never met a person who pronounces the whole word though :D

[@simonlaker2139]

Cheers for the info. We are blessed for your input to the printing community. Can't wait to see what you come up with next.

[@_Xantras_]

When you said you’re keeping the printer as it is, does it mean the design won’t change anymore ? Also I guess that means you’re not gonna try servos after all ?

[@xenon2008]

i love your humor❤

[@regun2434]

Hmm good advices.. 6 motor delta@48V should be fun 🤔

[@MakerViking]

Really interesting stuff. Will definitely utilize some of it on my printers.

[@harold2718]

Hey Vez have you ever looked into 3 phase steppers (supposedly the torque drops off less quickly at high speeds)

[@Fregaa]

Can you talk a little bit about the servo42c board? Did you tested it? Looks like this aproach reduces motor heat, increase torque and max rpm, but I dont see people talking about it

[@valdolandar534]

I wonder what is the average speed to print the `` benchy '' to scale 1: 1 with the highest speed but on time with better quality. I feel that these speeds in the maximum tests have exceeded the limits of the impressions, but it is almost like car velocity tests for records in a straight desert or in a circuit, it is not very clear if it is a car or a rocket. Then returning to the initial question it would be good to know if there are speed limits to printing due to the nature of the material and no longer to the printer movement.

[@jannejcperme3238]

Is Vez considering making a budget bedslinger printer? Just asking, i know that CoreXY is superior to Cartesian type but some of us are fans of bedslingers 🤪

[@Drsen57]

So the kit on mellow's store... how close is that to exactly your vz330 right now?

[@Vez3D]

Its the exact same thing..except im running the watercooling

[@coolstuffhere1215]

love your content keep up the good work

[@rothlive]

Very good video thank you. Unfortunately, the Stepper motors are sold out.😪

[@feliperojas8527]

Could I connect those Mellow Fly external drivers to a BTT Octopus Pro without compatibility issues? Anyone have tried it?

[@Vez3D]

Yes they are compatible

[@controlflow89]

Hi, Vez! Great video, thanks for sharing some secrets! Have you ever considered trying Ultimaker kinematics on rails with 4 motors? It's simpler comparing to CoreXY, shorter belt paths, easier to use 10mm/12mm belts to increase shaper frequency, symmetric axes (no additional mass on Y), less motor rpm to reach the same speed... It has it's flaws, but in terms of possible speed/simplicity/the quality you can reach... it seems superior over CoreXY, at least for me. I really wonder how crazy cross-gantry printer may look like if you add some Vez insanity to it :)

[@Vez3D]

Like a K3 from Annex? Super nice indeed. I still have not seen a IS graph from one of those though. I wonder how it compares go my results

[@controlflow89]

@@Vez3D yeah, like K3. It uses rails w/o any kind of support (all rails are designed to be supported), this may result in bad results, but I believe 3D printing loads should be acceptable...

[@C1Rob1990]

​​@@controlflow89 so like k1/k2... Also the rails on k3 are at the max acceptable length for being unbacked ;)

[@Ugaux]

Question about x axis weight. Why did you not mention using carbon fiber rods with polymer bushings ? You could also save many grams and it could be cheaper, right ?

[@Vez3D]

You mean like igus?

[@Ugaux]

@@Vez3D Yes exactly! I have seen some people using them. And the Bambu Lab X1 actually has them too.

[@drog]

It's just me or the audio in this vid is absolutely terrible? Watching it in headphones and the right channel is whispering with a slight delay

[@tadhgd350]

Seems fine to me

[@REDxFROG]

Also has constant background noise

[@drog]

Double checked with a phone speaker (mono), noticeable background noise is still there

[@hereisanaccount]

I heard it too. Sounds like the right channel has audio from the camera mic. I turned my volume way down and it was less noticeable. Mistakes in editing happen sometimes.

[@Vez3D]

Mistake on the camera sorry and I tried to remove noise as much as possible in editing. 😂 the mic was not fully inserted in the cam so it used the cam mic. Sorry about it

[@velaastro904]

I use way oil that’s normally used for the ways on lathes and mills it has a low viscosity but enough to stick to vertical surfaces

[@slonktonkster9680]

most helpful comment

[@unharmonix]

Where's the link to Eddy the Engineer?

[@airheadbit1984]

Kind of like Moore's law, you seem to be doubling numbers every year. Next will be 'chilled' water cooling then moving on to liquid nitrogen cryocooler - Keep it up, having the part before we print it is the goal. Update: Kraken 550Mhz dual core, 4 - 8amp each TMC5160, refrigerated liquid cooling, 60 volt 600 watt X and Y axis. os·cil·lo·scope

[@botvinnik_]

Just a native spanish speaker note: that 'cojones' was pretty well pronounced. Regards!

[@BeLikeLong]

Was the setting put in the description for the drivers?

[@chicoxiba]

very cool.. but for me, no advantage to print fast if you loose in layer adhesion; When you print your most important functional parts in asa for example, how fast do you print vez? Anyways.. you are an inspiration.

[@Vez3D]

Depends on the part.. 250-300... but sometimes 4-500mms

[@cyrikdnamps]

Do we need to update our Electrical panel house ?

[@jpsaila]

Where are the TMC settings/params he mentions in the vid - nothing much available in the printer.cfg files on the github

[@personalview7388]

Why not to move to custom CAN boards with drivers placed near motors? Such way you fully avoid slow STM microcontroller. And it'll be separate microcontroller on each board.

[@mattlott1113]

CAN boards = more weight on the toolhead. Weight = slow. That's why vzbot doesn't recommend those. We are shaving grams to get more speed.

[@personalview7388]

@@mattlott1113 If you account multiple wires drag instead of 4 small wires of CAN BUS, especially ones of real ordinary users it'll instantly change in reverse. For any real users of VzBot this gram shaving makes zero sense anyway. As hot end won't keep up. This is why we need proper scientific approach and step by step look on changes.

[@mattlott1113]

@@personalview7388 You clearly haven't done your research on vzbot. We use the highest flow hotend on the market. Don't take my word for it. Watch the videos on Vez's channel. It's all there. The "science" is in the proof. Weight is definitely the enemy of speed when you can extrude as fast as vzbot. If you disagree, you need to research more.

[@personalview7388]

@@mattlott1113 I don't care about Vez, same way as you don;t care for Formula 1 team if you service your car. I care about real people who make VzBot. And they will gain a lot going CAN way. 99% of them are not using CPAP cooling, 99% of them are not using such hotend. If you stop being obsessed with speed because of.. speed - it'll be better for all.

[@mattlott1113]

@@personalview7388 The whole point of vzbot is speed. Every printer project has it's own niche. We care about speed.

[@twanskys]

im surprised you dont have the metal y gantry spacers on the 235,,, i would have thought otherwise..

[@jazzboy]

Rum!!! Brilliant haha!

[@SuperKoe]

Nice, i already have point 10 done! alcohol.

[@BH4x0r]

We need an iMXRT1062 (Teensy 4.0) based board and Klipper Overclock support for it, with a heatsink they can run at 1GHz 😂, but they are 600MHz out of the box, imagine 1GHz tho Imo there should be more closed loop drivers, if the motors fail it should be able to detect it, and maybe stop the printer so nothing breaks that's why i think servo or closed loop steppers will always be superior

[@jeremyglover5541]

I would prefer something like the AMD/Xillinx 'Zynq' MCU/FPGA/SOC running Microblaze. kzbin.info/www/bejne/pXO5Xqt3lK93fKc

[@jeremyglover5541]

and yes, its overkill :) but this is a printer that defines overkill, does it not? That chipset and software suite with an FPGA combined with a CPU and fast AF memory and coms buss, allows you to essentially mimic any (many) types of hardware chipset. you can write software and run it as hardware; at hardware speeds. the hardware literally morphs into different circuits, allowing you to run specialised tasks VERY fast. the one chipset could run klipper on the CPU. perhaps accelerating some parts of the code in hardware and then instead of adding a chip, you form the fet gate array into a custom chip that works with the CPU and IO to achieve tasks that would be slower to emulate in software and combined with some external fets, mimic being an 5160 driver. you dont add the hardware, you literally change part of the chip into the specialized hardware you want.

[@BH4x0r]

@@jeremyglover5541 Well honestly the Super8 (non pro, which the Mellow board uses) has a 168MHz ARM CPU, the core in the iMXRT1062 can already execute more than the core in the Super8, at the same speed, so nevermind at 1GHz, and the steppers are already at the limit on the 48V supply aswell so to ever get good usage of that FPGA you'll need even more than 96V cuz even that would be handleable by the arm chip i mentioned

[@jeremyglover5541]

@@BH4x0r yes, sure, but we were talking about realtime DSP for servo drive and as my second comment above mentions, its a chameleon. This sort of thing could even look at realtime resonance correction, where new data on performance is coming in all the time and the servos react. if you want your board to have new functionality hardware wise, you program the code to describe the hardware in C and the code runs in hardware, as hardware. if you had the skills and tools, it would mean someone could come up with an idea and instead of everyone running out to buy the new add-on board, you distribute the code. like I said, its overkill, but I would prefer to spend money on a more powerful and flexible CPU/FPGA SOC on the mainboard that $50 here, $50 there, $100 here, 100 there every time the latest gimmick comes out.

[@jeremyglover5541]

@@BH4x0r like, the BACK emf correction mentioned by Simon above. That back EMF from the motors could be sensed in realtime and adjusted for, rather than making a blanket setting, measuring, then iterating for every tuning or material.

[@daliasprints9798]

Annex Hypernova will solve the step rate problem and let you use 256 microsteps at 3000 mm/s. 😁

[@triscuit_biscuit9489]

Where’d you get that shirt!?

[@Vez3D]

A guy in uk offered it

[@drfailbucket]

The vzbot is going to be the '2jz' among 3d printers 😂❤

[@MakerViking]

Nice!

[@kiiiddd]

You need to work out something with Nero so he can build a VzBot on stream

[@Rainmotorsports]

Nero has already said one is happening in a recent stream. VZBot and Annex and i forget what else but all you have to do is watch the next stream because every single stream someone asks even though he has answered several times.

[@girenloland]

great video. Insane. You have some interference in right channel in your mic setup. Like a slight echo noise

[@Vez3D]

Yeah I know...had mic issue in the video and only realized it arround the end 😅

[@leoberube3271]

Fait un speedboat benchy! Je suis sur que tu battrais ton record!

[@Jkauppa]

dont put the melting unit at the moving nozzle (ie minimal moving mass, preferably only a piezo movement to get stuff moving really fast/accurate)

[@Jkauppa]

yes you can print locally really accurately with piezo element movement, just like hard drives do micro movements

[@Jkauppa]

why so large size of the print head, you can strip down that radically

[@Jkauppa]

you can put most of the stuff on the static printed mount, not on the head, like filament pusher, and a nozzle melter can be electric peltier stack, or just resistance/induction heater with peltier cooler

[@Jkauppa]

the less you have on the print head, the less you need in the movement axis

[@Jkauppa]

why would you use circular motors for linear motion (homopolar dc linear actuator motor recommended, ie cross magnetic field to a brushless 90 degree wire, brushless railgun motor)

[@3D_Printing]

One Meter per second (approximately) ≈ 2.2369 miles per hour (approximately)

[@ИванИванов-о4ц9с]

Он скоро через телепорт голову двигати начнет😂... 🎉

[@kennybeaudoin]

tbnk mon chum tu travaille tard pour nous haha :P

[@hrmny_]

your microphone seems to be a bit broken at least the right side of the audio is

[@Vez3D]

Yup i know

[@the-matrix-has-you]

Try to increase baud rate on mcu in printer config I think it wss 250000 by default make it a million tell me how it goes your problem was mcu could not handle instructions that much so if you increase baud rate mcu can follow commands sent by Klipper I guess

[@hd-be7di]

Speed comparison: Ender 3 = Toyota Corolla Voron = Ford Mustang RatRig = Bugatti Veyron VZBot = Millennium Falcon with Class 0.4 Hyperdrive

[@Vez3D]

Hahaha

[@daliasprints9798]

But which one's fun to rice? 😂

[@Davor_3D_ZrAM]

I used PTFE drylube on linear rails on large wood and composite milling CNC because dust wasn't collecting on rails. Dust would create gunks with oil or grease.

[@daliasprints9798]

Supposedly PTFE is really bad for rails - causes the balls to slide rather than roll.

[@Davor_3D_ZrAM]

@@daliasprints9798 maybe it is, but 6 years 2 shifts still intact. Passed rail service life expectancy long ago. Carriages new every 2 years as regular maintenance because they are cheaper than 1 day out of service.

[@GOLD-ei3qh]

Am I the only one who hear demons whispering in the right ear?

[@Vez3D]

No ..its an issue i had with my mic

[@fiery_transition]

Yeah, I wish I could disable the right channel, I thought it was a joke at first lol

[@sir_gnome]

U cook?

[@sir_gnome]

cause damn ur recipe looking tasty.

[@jmsivori]

Pará, pará, pará!!! Yo como un bolud… traduciendo todo en Google para hablarte y vos hablas español??? 🤣😂😅😆😁😄

[@rexxx927]

Flash speeds are as important