Steppers vs Servos in 3D Printers: An Engineering Analysis

[00:00:00]These are three electric motors, and the names  that have been taught to you are brushed DC motor, brushless DC motor, and stepper motor. But  just for a second, open your mind to this: There is no spoon. Everything is  fish. And these motors are all the same - they’re all permanent magnet DC motors. But then why do we use this one for 3D printers, this one for cheap toys and this one for  percussive “maintenance” of oil refineries? Well, historical reasons, but also, cost, complexity  and performance. The one that right now makes the least sense to me is why we still  use steppers for 3D printers. I’ve been experimenting with BLDC servo setups, because  for Caroline, steppers were easy to use for me, but I’m really running into their limits. In  this video, maybe we can explore together why, for 3D printers, nobody aside from  Bambu, for their extruder heads, seems to think it’s worth making that next  step from ancient stepper motors to proper servo drives. Though actually, steppers might  not be as bad of a choice as you’d think.

[00:01:13]But you know what else is the next  step up? Today’s sponsor, UGREEN NAS!

[00:01:17]We’ve all got data that we really want to  hang onto - photos, documents, work files, huge 3D scan projects - but just buying larger  and larger devices or relying on cloud services is super expensive, and sending everything  through your internet most of the time isn’t even fast enough, so the logical solution to  keep your data safe is a NAS for your home.

[00:01:40]That’s what I’ve been using for the longest time  now, and it is about time that someone like UGREEN starts mixing up the market. This is the UGREEN  NAS DXP4800 Plus, four bays for hard drives, two for NVMe SSDs for extra storage or caching,  a 12th-gen Intel 5-core processor with expandable RAM and 10G ethernet. Best of all, there are  no artificial limitations, like requiring you to use specific manufacturer-branded  hard drives. Who would do such a thing?

[00:02:12]The software on this is really important,  and the DXP4800 Plus runs UGOS PRO, super simple to set up, it has user management,  built-in backup, and a photo manager with AI-based search. Those are the exact things  that I need in a NAS. If you need more, Docker and VM support are also built in, which  is perfect for running Home Assistant, Frigate or anything else you might need in a home lab. You can access your NAS anywhere anytime and directly share files and  folders with a simple link.

[00:02:42]The UGREEN NAS series is now available  directly from UGREEN or through Amazon, and do check out their discounts  and bundle options, like grabbing a UGREEN DC uninterruptible power supply with the  DXP4800 Plus. Links in the description, with up to 25% off on October 7th and 8th, and thanks  again to UGREEN for sponsoring this segment.

[00:03:04]Now, I've had a look through some forums on  the internet, and there seem to be a lot of misconceptions about what a servo is and how it  relates to stepper motors, brushless DC motors, and brushed motors. Any of these motors can be  a servo as long as it has a way of measuring its position and using that as feedback  for how the motor is controlled. And once you have that position feedback, you can also  use it to control the motor more efficiently.

[00:03:30]But what does that mean? For all intents and  purposes, a permanent magnet DC motor is always a combination of a set of electromagnets and a  permanent magnet. Sometimes the magnet spins, sometimes, the coils that make up the  electromagnet, when you get both of them spinning, something has gone wrong. The hardest part is turning the electromagnet on and off at the right  times, and with the right polarity, and how you need to do that is also the  biggest difference between these motors.

[00:04:00]Stepper motors exist because they could  be driven with a super simple electrical signal. This would be a 4 steps per revolution  motor, real ones have 200 or 400, but it’s the same thing just repeated over and over. Unipolar motors, which are the 5-wire types that we don’t really use anymore,  can be driven by tying the common lead to a supply voltage and then switching on  each of the 4 “half-coils” in sequence.

[00:04:27]To spin the motor in the opposite direction,  you would simply run that sequence in reverse, and that entire control scheme is something  you can do with simple NPN transistors, or even relays if you really wanted to. Each step of that sequence is a full step, but you could also do half steps by turning on  two sequential electromagnets at the same time, essentially overlapping the magnetic  fields and creating an in-between position.

[00:04:50]Today’s “hybrid” type stepper motors rely on  the fact that nowadays, we can build much more complex and more powerful drivers that can  easily reverse the current through each coil, so now we’re using 100% of the coil instead of  just half at best. That makes it more efficient.

[00:05:07]On top of that, microstepping logic and  the current control that modern drivers can do has much improved, even  in just the last couple of years.

[00:05:16]A brushless DC motor works in much the same  way. The big difference is that instead of two separate coils or two pairs, it uses  three coils that are wired up to each other, and by powering one of the coils, you're also  partially powering the other two so that they can all work in tandem. Now, you can totally run  a brushless DC motor with a fixed step sequence, but just like hybrid stepper motors, the way  they are built today is optimized to work with a modern driver setup instead of catering to the  specific needs of some outdated control system.

[00:05:52]But in the end, the windings and how they’re  wired up is just a means to an end, and that ultimately is creating a magnetic field that  forces the rotor into the position we want it in. Whether it’s two coils, or three,  or eleven, doesn’t really matter when you have a modern controller that run runs a  variant of FOC, field-oriented-control, where instead of going through a fixed step  sequence or current tables, it continuously does the math for how it needs to align and  combine the field from the electromagnets relative to the magnet in the rotor so that you  get the torque or positional output you want.

[00:06:33]I’ve been using simpleFOC to run these  BLDCs, and it can just as easily run a stepper motor if you give it four outputs  instead of three. To the FOC algorithm, these are all the same. SimpleFOC is based on  the Ardunio runtime and you can actually tie it into your standard step and direction signals. You can even get little bolt-on driver boards that directly sit on a stepper motor and  run it in FOC mode, instead of just stepping through pre-calculated amounts of current like a  traditional driver. But these ones from Makerbase are running proprietary firmware that you can’t  replace, so I wouldn’t really recommend them.

[00:07:11]I assume FOC is also what Bambu uses for the  BLDC extruder in their printers. There are a bunch of benefits to the FOC drive mode, and  also some downsides, and we’ll cover those later.

[00:07:24]Side note: How do stepper motor get their steps?  Because they have 200 or 400 steps per revolution, and that is an insane amount if, for example, you  wanted to get that many physical steps from a BLDC with individual magnets. You’d need 200 magnets! How steppers do it is by just using one, big magnet, packing it into an  iron core, and giving that ridges.

[00:07:51]Because that iron core is so good at guiding  the magnetic field, whenever you get an air gap, it’s like that field isn’t even there. So the  way that the ridges in the rotor and the stator line up is so that you first get one pole of the  rotor lined up, then you turn it a little bit, and suddenly the other pole is lined up. It’s  like having hundreds of individual, virtual magnets in there, and it’s what gives steppers  such a fine resolution, even with just one magnet But probably the worst one of these  motors, but also, in some cases, the best one, is the brushed DC motor. The  basic principle is the same as the other two, but instead of needing a step sequence or  some complicated external control loop, you just give it straight DC power, and it  has a built-in slip ring contact that more or less accurately sends power to the correct  internal coils. If you reverse the polarity of your power source, it spins the other way. It trades efficiency, performance and fine-grained control for ease of use - and that’s why it’s so  commonly used as a servo drive. Because there, you have an external control loop anyway, the lack  of fine-grained control doesn’t matter as much, and it simply becomes an incredibly cost-effective  way of producing torque. But brushed motors have a quite limited lifetime, especially when you  strain them with harsh direction changes, and because they typically can’t deliver  as much torque as BLDC or stepper motors, applications like RC servos need to heavily  gear them down, which introduces backlash. So not the best candidates for our application,  but for something like an inkjet printer, they’ve become the de-facto standard solution and  are being combined with optical encoders—either the long strips of finely graded transparent  plastic film that will be labeled "do not touch", or rotating types that have an optical sensor  attached to the motor itself. What turns a motor into a servo is just that: a way of getting  positional feedback, and then using that to adjust how the motor is driven. It's a closed-loop  control system where the thing that's controlling it not only knows what it is putting into the  motor but also what is coming out of the motor.

[00:10:22]The kind of encoder that simpleFOC or platforms  like Odrive typically uses is a magnetic encoder, where you attach a magnet directly to the motor  shaft and a separate chip then measures the exact rotational position. So, like any servo, this  allows the system to work in closed-loop mode. If there is an external disturbance, a closed-loop  system can detect that and compensate for it, where an open-loop system, like typical  stepper motor setups, would just lose a couple of steps and never realize it. But this kind of magnetic encoder is also what enables the FOC drive mode. Typical BLDC  speed controllers like for RC cars and drones, already make an educated guess for where the rotor  is at any time, simply speaking, by measuring how much the motor is acting as a generator at any  time. This only really works at higher RPMs, and for robotics and 3D printers, which are  robotics, that low-speed control is pretty important. Also happens to be something that  stepper motors naturally are really good at.

[00:11:31]But by knowing the exact position of the  rotor, the driver can precisely control each of the coils to run the motor as smoothly  and as efficiently as possible. For example, if the motor is already where we want it to be,  then we can just stop sending it current entirely.

[00:11:50]Not so with stepper motors: We just constantly  energize them, because we don’t know how much we need at any time. That ends up being  too much when the motor isn’t taxed much, and might be too little for example, in rapid  accelerations, or when the print is curling up and starts interfering with the nozzle. It’s  like you’re always pegging the throttle in your car because you don’t know if you’re going up  a hill right now. This does mean that a stepper motor will constantly pull the rotor towards  the ideal position, while a servo relies on the software control loop to initiate that correction.  But steppers aren’t infinitely rigid, either, the act sort of like a spring and only produce  maximum torque just before they lose a step, and software loops can be tuned to better  suit a system. At the expense of needing quite a bit of computing performance. Theoretically, even with stepper motors, 3D printers can make estimates about how much  torque they need in every specific situation by looking at how hard they’re accelerating, and  dynamically ramping up or down the current, but I don’t know how much, or even at all  that is being used yet. But definitely, the ability to make predictions and  use them for things like input shaping is a major contribution to why open-loop  steppers still work so well for 3D printers.

[00:13:21]And under normal circumstances, the CoreXY  motors on a 3D printer will only need to move the tool head, which ends up being a pretty  well defined and very predictable use case, so needing to compensate for things outside  of your control isn’t something that these motors typically need to do.A robot arm,  though, where you're interfacing with things outside of your control, like handling  physical objects, that definitely is something that is not exactly predictable, and there, a  closed-loop servo setup is very much needed.

[00:13:56]So it totally makes sense that the first motor  that Bambu has replaced with a BLDC servo is the extruder. Filament is the one thing inside of  a 3D printer that is unpredictable. It might be harder to extrude, it might be easier to extrude,  or it might have some contamination in it that suddenly spikes the force required to push it  through the nozzle. When you add feedback to any motion system, you can almost always use much  smaller motors and get a more reliable system.

[00:14:26]Because you’re now only sending as little current  through the motor as possible, which keeps it from unnecessarily heating up, you gain headroom to  now drive it harder when you actually need it.

[00:14:37]And if that still isn’t enough, you can just make  up for it as soon as possible. Lagging behind for a fraction of a second is much less of an issue  on the extruder than it is for the CoreXY motors.

[00:14:55]But that weight reduction on the toolhead is also  where it matters the most. When your tool head is lighter, you're also reducing the requirements for  how rigid the linear guides need to be. When you have lighter linear guides, you're reducing  the torque requirement for the drive motors of your core XY system. So it's not just the  direct weight-saving in the motor, but it's an exponential weight saving in everything that  needs to support that motor weight as well. The two core XY drive motors are completely static  within the printer, so except for the inertia of the rotor, saving weight there would really not  make a difference in overall performance and cost.

[00:15:37]And again, they don't see unpredictable  loads as often as the extruder motor.

[00:15:42]But would it be cheaper? That’s something I  don’t have a definitive answer to. Right now, this BLDC setup, which I would put at at least  comparable with a NEMA17 drive, cost me about 40 bucks in parts, while a NEMA17 and a driver are  closer to 15 bucks. Of course, those prices come down at volume, and the motor and electronics  can most likely be engineered to be cheaper, but the magnetic encoder would need to be replaced  with something a little more cost-effective. Not impossible, but I think more work than most 3D  printer companies would be willing to take on.

[00:16:28]Now, I know that there is always a "golden goose"  to chase where you can make things better and improve them, and it's something that I often have  to actively stop myself from doing. At some point, you reach "good enough" and the extra effort  spent on optimizing something just isn't worth it anymore. In the case of stepper motors in 3D  printers, we might just have reached that point.

[00:16:49]With a bit of smart software, stepper motors  manage to achieve incredible performance to a point where the other parts of a 3D printer are  now a bottleneck, and switching to BLDC servos just because they are “better”, might not even  improve the product that is the entire 3D printer.

[00:17:07]I mean, it’s worth a try, right? These would fit  on a Voron, but I didn’t just want to make a video going all “WOW I put BLDCs on a VORON”, “I’m so  innovative”, because there are good reasons why it might actually not be a good idea. As always,  I invite you to discuss that in the comments.

[00:17:26]Thank you all for your support,  thank you for watching, by the way, if you’re at Formnext next month, say hi, I’ll  be wandering the showfloor, hunting for content.

[00:17:34]Until then, take care of yourselves, keep  on making, and I’ll see you in the next one.