Control board decision - DuetWifi or Smoothieboard...??

I've been doing a lot of reading and I'm struggling with which control board I should purchase for my new 300x300x400 CoreXY machine.

I've narrowed it down to the SmoothieBoard or the Duet Wifi, they seem to be very well suited for larger machines - not saying there aren't other out there.

The Smoothie has some nice features from what I can find that I really like:
- the ability to solder in a new driver in case of failure (not sure of the failure rate here, in any board)
- ease of setup (opinions may vary) for newbies
- 24v support
- Ethernet
- Optional voltage regulator, no need for a PC
- support for multiple print heads

The Duet Wifi:
- 24v support
- 256 microstepping
- RepRap firmware

What about safety, particularly in the regard to fire? With 3D printers needing to run for long runs, are there any built in features of either that can help put my mind at ease?

I know I've missed a lot of features of both boards... Does anyone have any experience with both boards that could comment on their experiences?

Any input would be greatly appreciated.

I saw a YouTube on Tom Sanladerer doing a review on the Smoothieboard but he didn't so one on the DuetWifi, has he and I just can't find it? Has anyone done a solid review on it?

ampapa,

The smoothieware running on a smoothieboard has some safety features built into the temperature control logic. If it takes too long to reach the set temperature, the board will shut off the heaters. If the temperature difference between the set temperature and actual temperature get too high it shuts down. see: [forum.smoothieware.org] Of course, those safety features can only work if the controller board is functional, and there are scenarios where that may not be the case.

---

Ultra MegaMax Dominator 3D printer: [drmrehorst.blogspot.com]

All the benefits you listed for Smoothie are also available on the DuetWiFi, except the direct ethernet connector. Duet also has thermal safety enabled by default.

I have used both boards and my personal preference is with the Duet. The on board web/print server makes the difference for me. To get similar functionality on Smoothie, you have to us something like OctoPrint. I know they are working on implementing a better web server, but not sure of the status.

Duet is also working on a new version based on the DuetWiFi, but with the ethernet connector.

Reprap Firmware that all you need to list in the +

I got a radds on this firmware and I wont change that firmware to anything else, simple to setup,modify lot of support and update ( not only to original duet user but also concurrent board like the radds that run the same firmware got support and update), its not something we can say from all firmware some refuse or are recalcitrant to help if that not they original board.

Maybe ask djdemon about the diff between the two board/firmware, he use a duet for a while now and recently move one controller to a smoothie firmware. he can probably tell more about them a least firmware side cause DJ do not use a original smoothie board.

Quote
ampapa
The Smoothie has some nice features from what I can find that I really like:

- the ability to solder in a new driver in case of failure (not sure of the failure rate here, in any board) - No different with the Duet, you can either solder in new drivers or connect additional or replacement drivers via the expansion connector
- ease of setup (opinions may vary) for newbies - Same for the Duet, plus you can edit the configuration files in the web interface, and there is a web-based tool for generating the configuration files at [configurator.reprapfirmware.org]. There is a learning curve to go through when switching to any new firmware.
- 24v support Same for the Duet
- Ethernet WiFi on the Duet WiFi, or Ethernet on the Duet Ethernet
- Optional voltage regulator, no need for a PC - Standard on the Duet, no need to buy it as an extra or fit it yourself
- support for multiple print heads Duet supports 2 hot ends and 2 extruder drives (same as Smoothieboard), plus up to 5 more hot ends and 5 more extruder drives using the optional expansion board

What about safety, particularly in the regard to fire? With 3D printers needing to run for long runs, are there any built in features of either that can help put my mind at ease? - Duet uses a FOPDT model of the heater to predict how it should respond, and will shut down the heater if something appears to be amiss.

I know I've missed a lot of features of both boards... Duet/RRF uses a high speed SD card interface, which allows it to support fast file upload to the SD card over WiFi or Ethernet. Also pressure advance. Also multiple independent X carriages. Smoothieware currently has better support for controlling machines other than 3D printers e.g. CNC machines and laser cutters.

I saw a YouTube on Tom Sanladerer doing a review on the Smoothieboard but he didn't so one on the DuetWifi, has he and I just can't find it? We gave him a free Duet WiFi last August and I think he did a very brief review of it, but I can't find it online.

---

Large delta printer [miscsolutions.wordpress.com], E3D tool changer, Robotdigg SCARA printer, Crane Quad and Ormerod

Disclosure: I design Duet electronics and work on RepRapFirmware, [duet3d.com].

I've got a re-arm (smoothieware) and a duetwifi. The duetwifi is a much better board although a genuine smoothieboard might well be better than a re-arm and ramps. RRF is easy to use and very versatile and the support via David and the duetwifi forum is the best I've had on any product. Development is active and receptive to users requests.

If a duet board is within your budget I really can't see that there's anything that beats it right now.

I have no financial interest in any controller board.

As for fire safety it's a big problem, see the safety forum here for some ideas. I put together a smoke alarm shutoff device which might help. I have radio linked smoke alarms with one above my printers. A co2 fire extinguisher handy. Powder based fireball type things might be a good idea if you want to save your house but probably wreck your printer.

Edited 1 time(s). Last edit at 03/27/2017 05:45PM by DjDemonD.

---

Simon Khoury

Co-founder of [www.precisionpiezo.co.uk] Accurate, repeatable, versatile Z-Probes
Published:Inventions

One major difference that swayed my decision, is whether the firmware supports different LCD panels.

Last time I checked the Duet only supports the PanelDue touchscreen, which is expensive, and doesn't have a rotary knob. The Smoothie board supports many different types of panels, including ones with rotary knobs.

Something else to consider.

Quote
nebbian
One major difference that swayed my decision, is whether the firmware supports different LCD panels.

Last time I checked the Duet only supports the PanelDue touchscreen, which is expensive, and doesn't have a rotary knob. The Smoothie board supports many different types of panels, including ones with rotary knobs.

Something else to consider.

The PanelDue is touchscreen so doesn't need a rotary knob. It also displays a lot more information than the basic screens Marlin and the others support.

This is my board, so take anything I say with a grain of salt, and check it for yourself, but here are the things Smoothie users say they like most about it ( and tell me about it very loudly ), in particular relative to other systems :

* Documentation. We clearly have the most complete and most extensive one around. We put an insane amount of work into it
* Modularity. We make it easy to add things/do new things. Check out for example the switch module : smoothieware.org/switch
* Features. We have many many ( mostly small ) features other boards do not have. You'll have to go through the documentation to discover it all, but for many users it makes a huge difference.
* Speed. Obviously lightning fast compared to 8-bit. Compared to Duet, I'm yet to see anyone able to show a measurable advantage of one over the other, and we've seen several users compare print quality side by side and show no difference at all.
* Development. We have literally hundreds of volunteers working on various aspects of the project. A big part of those are working on v2 so their work isn't very visible yet, but even if you take them aside, what's left is still by far the most actively developed system around.
* Reasonable design. What I mean by that : several board designers chose to move to the DRV8825 drivers because it did 1/32 microstepping, and we refused to. It turned out *many* users had huge problems with those drivers. While it unfortunately didn't hurt the sales of the people who weren't careful, it still hurt the users. We care about this sort of stuff. We now, later, have 1/32 microstepping, but using another, better driver. We do this for everything. The reason we don't have 1/256 drivers, and don't have wifi, isn't because we didn't do the work : we did. It's because we weren't able to get it to work well enough for our users to be happy with it in the end ( 1/256 drivers tend to loose too much torque, and ESP wifi tends to be too unreliable ). We do the same thing for firmware dev by the way : we have many features we are working on, but have not released yet because we are waiting for those to have clean and reliable code.

There have been comments on Smoothie's web interface. Smoothie comes with a minimalistic interface that allows you to control the machine, upload and play files etc, but it's very simple. This is because that "default" interface is stored in "flash" ( inside the firmware itself ) so it has to be small so it fits there.
However, the board also has a 4GB SD card, and you can store web interfaces on that. And in that case there is no size limit, so users are able to make much better interfaces. And they have.
So, you can a much better web interface for Smoothieboard, simply by copying a few files to the SD card. For instructions see : [smoothieware.org]

Edited 1 time(s). Last edit at 03/28/2017 08:50AM by arthurwolf.

I use the Duet WiFi, so my opinion is based only on using that one board.
Features? To be honest, I think the Duet is better, for ease of use. Upload the file and print. No need for cables, octoprint or anything else.

What makes the Duet the ultimate board for me is the support. David and Tony are fast to respond, know what they are talking about and will help if they can. Really really impressive. Especially if your fear the learning curve, these guys and the other users on the forum will be a great help.

As you can probably tell, I am a big fan of the Duet.

Edited 1 time(s). Last edit at 03/28/2017 08:45AM by Lykle.

Thanks for all the great input!

There's one thing for sure, these are both great boards and I think I can't go wrong with either of them.

The fact that the owner/developers of the products are chiming in on open threads shows the dedication to their products and the technology, thanks for that!

A couple of follow up questions if I may..

- In regard to steppers and resolution, how large of an impact does the 256 vs 32 steps actually make in the print and for that matter in the noise level? For a CNC maybe the steps are not as large an issue about a certain point?
- Does the Duet allow for direct soldering on the main board in the event of a driver failure?

ArthurWolf, I agree, the Smoothie documentation appears to be more robust and laid out from a quick glance.

I hope I win the Smoothie on "OpenBuilds"... that will solve my dilemma.

ampapa,

Quote
ampapa
- In regard to steppers and resolution, how large of an impact does the 256 vs 32 steps actually make in the print and for that matter in the noise level? For a CNC maybe the steps are not as large an issue about a certain point?

The general idea is : passed a certain micro-stepping level, you are not gaining any more precision, *but* by lowering the amplitude and increasing the frequency of the pulses, it reduces noise.

1/32 is already pretty quiet though, and you don't really gain accuracy by going above, so many consider 1/32 the "sweet spot". 1/256 *is* quieter though.

On a CNC, you might want to go higher than 1/32 because it helps with resonance, but on 3D printers that doesn't enter into account.

Quote
ampapa
I hope I win the Smoothie on "OpenBuilds"... that will solve my dilemma.

Note if you are using the board to do something new and innovative nobody has done before, and even more if that thing would result in new code for the smoothie firmware, we do give *a lot* of free boards away to people who do cool things with them. We also give money to contributors when we can/when it helps push the project forward. We really really are doing this for the community and for the reprap project.

Quote
arthurwolf
This is my board, so take anything I say with a grain of salt, and check it for yourself, but here are the things Smoothie users say they like most about it ( and tell me about it very loudly ), in particular relative to other systems :

I'll just pick up on a couple of things you wrote...

Quote
arthurwolf
1/256 drivers tend to loose too much torque....

Arthur, I'm surprised that you've fallen for that myth. Beyond half stepping, the torque per unit angular error is constant regardless of microstepping. Higher microstepping beyond about x16 doesn't increase positioning accuracy but it does make the motors a lot quieter - although it's hard to tell the difference between x128 and x256.

Quote
arthurwolf
..and ESP wifi tends to be too unreliable

For the vast majority of Duet WiFi users, the ESP gives a reliable connection, with file upload speeds that exceed 1Mbyte/sec in some cases. A few users with a congested local WiFi environment do experience difficulties, and for those people and others for whom WiFi is less practical than wired Ethernet, we are about to start shipping the Duet Ethernet. We'd like to have an external antenna option for the Duer WiFi, but sadly certification issue make that impractical.

---

Large delta printer [miscsolutions.wordpress.com], E3D tool changer, Robotdigg SCARA printer, Crane Quad and Ormerod

Disclosure: I design Duet electronics and work on RepRapFirmware, [duet3d.com].

Quote
dc42
Arthur, I'm surprised that you've fallen for that myth. Beyond half stepping, the torque per unit angular error is constant regardless of microstepping. Higher microstepping beyond about x16 doesn't increase positioning accuracy but it does make the motors a lot quieter - although it's hard to tell the difference between x128 and x256.

This is the first time ever I hear this, after hearing a lot of extremely competent people telling the opposite.

3 minutes of googling shows only evidence for torque loss from credible sources ( including some selling microstepping, so having an interest in saying it's fine, but still saying torque loss ) and none for it "being a myth".

[machinedesign.com]
[www.geckodrive.com]
[www.micromo.com]

Quote
dc42
For the vast majority of Duet WiFi users, the ESP gives a reliable connection, with file upload speeds that exceed 1Mbyte/sec in some cases. A few users with a congested local WiFi environment do experience difficulties, and for those people and others for whom WiFi is less practical than wired Ethernet, we are about to start shipping the Duet Ethernet. We'd like to have an external antenna option for the Duer WiFi, but sadly certification issue make that impractical.

It's really a matter of choosing how reliable is reliable enough, not everybody is going to have the same standards there.

In our case, we go with ethernet by default ( which is perfectly reliable ), and users who want Wifi can just plug a $20 wifi/ethernet bridge into that.

We'll be working on a wifi extension board for v2, but we are not sure what we'll use for it.

Here's why people get confused about torque and microstepping.

The torque of a stepper motor is the holding torque of the motor (which depends on the motor design and the drive current) times the sine of the lag angle (the angle between where the motor wants to be and where it actually is, scaled such that 4 full steps = 360 degrees).

Let's suppose that the motor is under no load so it is where it wants to be. Then you hold it in​ position and send it enough steps to command it to a position (say) 1/16 of a full step away. It doesn't matter whether you send it a single 1/16 microstep or sixteen 1/256 microstep, the resulting current in the motor will be exactly the same and so will be the torque produced.

Unfortunately, too many people focus on tables showing the torque per microstep, when what actually matters is the torque per unit angular shaft error. What those tables tell you is that as you increase microstepping, you reach a point at which you can no longer rely on the motor to position itself to the nearest microstep. In practice this means that on most 3D printers, increasing microstepping beyond about x16 doesn't improve positional accuracy.

HTH David

Edited 1 time(s). Last edit at 03/28/2017 04:11PM by dc42.

---

Large delta printer [miscsolutions.wordpress.com], E3D tool changer, Robotdigg SCARA printer, Crane Quad and Ormerod

Disclosure: I design Duet electronics and work on RepRapFirmware, [duet3d.com].

Quote
dc42
Here's why people get confused about torque and microstepping.

Actually, I already know everything you've just said. No confusion here : Still.

From what I understand, it's your explanation that is a misconception ( or rather an oversimplification ).

Yes, as microstepping increases, torque *per step* decreases to the point where you aren't getting more accuracy. But it can actually get to the point where you get *even less* accuracy. It just depends on torque.

This explains the problem pretty well : [www.micromo.com]

I quote :

Quote
Micromo
« The consequence is that if the load torque plus the motor’s friction and detent torque is greater than the
incremental torque of a microstep, successive microsteps will have to be realized until the accumulated torque
exceeds the load torque plus the motor’s friction and detent torque.
Simply stated, taking a microstep does not mean the motor will actually move. If reversing direction is desired, a
significant number of microsteps may be needed before movement occurs. That’s because the motor shaft torque
must be decremented from whatever positive value it has to a negative value that will have sufficient torque to
cause motion in the negative direction. »

This means that depending on the torque you are dealing with, there is a certain level above which you are actually *loosing* accuracy.

And not just "we lost some accuracy and gained some, it balances so it wasn't worth it increasing microstepping" ( which you are talking about ), but it can get to "we should actually be using lower microstepping here as we'd end up with better precision that way".

I first learned about this from people who professionally use microstepping drivers in cnc routers and pick and place machines, and for whom this is very critical. They actually measure the torque, do the math and have to choose the ideal microstepping for a given machine/application.

This has been measured and is pretty well known by the pros. I don't think we have much data on this for 3D printers since 1/256 microstepping for 3D printing is pretty recent, but I'm pretty sure at some point somebody will measure this and we'll have data to show it's the same for 3D printing as it is for other industries ( no reason it wouldn't be ).

From experience, I'm pretty sure the folks in the deltabot google group will be the first to provide data for this They love doing that sort of stuff.

Edited 1 time(s). Last edit at 03/28/2017 04:30PM by arthurwolf.

Quote
arthurwolf
...

I quote :

Quote
Micromo
« The consequence is that if the load torque plus the motor’s friction and detent torque is greater than the
incremental torque of a microstep, successive microsteps will have to be realized until the accumulated torque
exceeds the load torque plus the motor’s friction and detent torque.
Simply stated, taking a microstep does not mean the motor will actually move. If reversing direction is desired, a
significant number of microsteps may be needed before movement occurs. That’s because the motor shaft torque
must be decremented from whatever positive value it has to a negative value that will have sufficient torque to
cause motion in the negative direction. »

I agree with all of that.

Quote

This means that depending on the torque you are dealing with, there is a certain level above which you are actually *loosing* accuracy.

But IMO that doesn't follow. What I think you are saying is that when you move a motor slowly, if it moves on every microstep then you will get more even steps than if it takes two or more to build up enough torque to make it move. But that assumes that if it moves on every microstep then the steps will be of equal size. That won't necessarily be true unless the static friction is much higher than the dynamic friction and the rest of the load.

I can see that in CNC applications where you are trying to move a tool slowly one step at a time and machining to a precision of a few microns, knowing that the tool will move on every microstep might be desirable. But in a 3D printer, the motion is generally faster and continuous, so static friction in only important when changing direction.

If the effects of different microstepping beyond x16 are visible anywhere, I think it is likely to be in the Z axis of a Delta printer. Although when 0.9deg motors are used, the moire pattern is already hard to see.

---

Large delta printer [miscsolutions.wordpress.com], E3D tool changer, Robotdigg SCARA printer, Crane Quad and Ormerod

Disclosure: I design Duet electronics and work on RepRapFirmware, [duet3d.com].

Quote
dc42
I can see that in CNC applications where you are trying to move a tool slowly one step at a time and machining to a precision of a few microns, knowing that the tool will move on every microstep might be desirable. But in a 3D printer, the motion is generally faster and continuous, so static friction in only important when changing direction.

I'm not going to be able to give any math, or any references here so I don't think we can go very far, but I mentioned discussing this with folks who make machines where they have to take this into account, and some of those machines were pick and place machines ( not bulky cnc routers ). In those cases the torque came from accelerating/decelerating at high speeds ( which we do on 3D printers ), and microstepping would cause issues when changing directions.

When microstepping was too high on those setups, the machine wouldn't go all the way to the end of it's movement on direction changes, but would move the other way around once caught up, so position wasn't lost. This was small but measurable, and was a problem for machine accuracy.

As I said : 1/256 for 3D printers is pretty recent. It's surely going to get much more common ( no matter if it's better or not, it's a bigger number so people are going to want it. Universal Reprap rule ). I'm sure with time, we'll have more users of it, so more chances to get reports of issues if these problems I mention actually are relevant to 3D printing ( assuming users are capable of pinpointing the problem to microstepping, which isn't a sure thing ).

Back in the day when I was working on clean room automation, I remember instrumenting motors with accelerometers and measuring the vibrations in a belt drive system at different levels of microstepping, up to 1/256 microstepping. The difference between 1/8th stepping and 1/16th stepping was incredibly small, and we couldn't measure any difference beyond 1/16th microstepping. What was a much larger factor was the quality of the stepper motor and whether it was designed specifically for microstepping. Turns out motor manufacturers can modify the internal tooth profile to increase the holding torque, but to the detriment of the microstepping accuracy. Personally, I don't worry about anything beyond 1/16th microstepping.

Arthur's replies to my own in this thread caused me to pause for thought and prompted me to run some tests. As this has drifted seriously off-topic, I've started a new thread here [forums.reprap.org].

@LoboCNC, in a 3D printer there is a big difference in the noise level when you switch from x16 to x256 microstepping. In other applications such as CNC machines, the reduction in noise may not be significant. Edit: I take your point about the importance of tooth profile for microstepping accuracy. Sadly I've never seen this mentioned in a stepper motor data sheet. I read somewhere that Nema 17 motors are typically optimised for accuracy and larger motors for holding torque, but i don't know whether that is true.

Edited 2 time(s). Last edit at 03/30/2017 03:10AM by dc42.

---

Large delta printer [miscsolutions.wordpress.com], E3D tool changer, Robotdigg SCARA printer, Crane Quad and Ormerod

Disclosure: I design Duet electronics and work on RepRapFirmware, [duet3d.com].

I have had two Panucatt boards that run Smoothie. I have one Duet Wifi. The only thing I don't like about the Duet is the price. If I had it all to do again, I would have two Duets. I added a PT100 sensor in about twenty minutes. The DuetWifi forum support is courteous and prompt.

The Panucatt support is dismal at best. They still have not returned my Azteeg X5 v 1.0 board that I sent in for repair. It has been months. The heated bed channel died the second time I used it. My experience with the Smoothie support of the PT100 sensor was spotty and confusing. I did figure it out after several hours, and using a nightly build. Normally the documentation for Smoothie is quite good.

The Duet Wifi has a great interface, so I don't need an Octopi. File transfers are very speedy, even over wifi. It's a great board.

Quote
Traveler0612
The Panucatt support is dismal at best.

I just want to be clear : Smoothieboard's support is universally acclaimed as really really good. I have spent every morning for the past 4 years answering Smoothie questions/writing documentation. Just want to say that so there's no confusion between the two here.

Quote
Traveler0612
My experience with the Smoothie support of the PT100 sensor was spotty and confusing. I did figure it out after several hours, and using a nightly build. Normally the documentation for Smoothie is quite good.

This probably happened only because you were trying to use PT100 before the feature was officially fully integrated ( nothing wrong with that, it's just more trouble being an early tester ).
The feature is now fully integrated and documented.
Smoothie takes longer than most other systems to integrate new features because we are extremely careful of how we integrate things and of the quality of the code, which leads us today to having a really really nice looking codebase that's still easy to extend despite having lots of things in it.

Edited 1 time(s). Last edit at 03/30/2017 05:29AM by arthurwolf.

I'm definitely not qualified to comment on the capability of the drivers and the stepping motion but I think from what I've read that anything above 16X is going to impact noise and not accuracy, I hope that's not an over simplification of the technical discussion above. Torque and holding power I think would be more critical in a CNC application and less in a printer application but that's just a observation on my behalf.

I am also curious about the support of the software and the code development since this will be on an "Core XY" / "H Bot" configuration. With my limited knowledge of the software is the "Core XY" design inherently supported in both boards?

Is Smoothieware closed source?

I'm pretty sure the Duet is RRF but I thought I read somewhere that there was a forked version for "Core XY" support?

ampapa,

Quote
ampapa
Is Smoothieware closed source?

Wait what ? Smoothieware is extremely open-source, it has always been and will always be. It's created by a community of hudreds of volunteers around the world.
The documentation and the hardware are fully Open-Source too.

About H-bot, it's fully supported by Smoothieware, and I would be very surprised if it wasn't by RRF as it's fairly easy to do.

I was going to write a long post but instead I'll say:

Smoothieware - no issue with you guys you've made a great firmware and are supporting a lot of users who didn't buy your boards, but use compatible hardware some of which is not proper open source hardware.

Panucatt - need to up their game especially with re-arm. Even if they published a full table of pin definitions so that users could use the Smoothieware documentation to configure their device they would be doing better. I've got mine working and I'm quite pleased with it but we've been working here to support people who have been struggling with it, when its Panucatt's job (not ours or Smoothieware's).

One thing I would say for duet board (especially the Duetwifi) is that the firmware and board are a package, as it the support and its very much a comprehensive package. If there were multiple compatible duet boards, some not great quality, in the wild then that support would be harder to provide, but currently there are not.

---

Simon Khoury

Co-founder of [www.precisionpiezo.co.uk] Accurate, repeatable, versatile Z-Probes
Published:Inventions

Quote
ampapa
I'm definitely not qualified to comment on the capability of the drivers and the stepping motion but I think from what I've read that anything above 16X is going to impact noise and not accuracy

I think that is typically correct. On my delta, if I increase the motor current to 1.5A (the motors are rated at 1.68A) then I can just feel the effect of individual 1/32 microsteps (0.025um vertical movement) on how hard nozzle grips a piece of paper. But the mechanics has a backlash of about two 1/32 microsteps, so in practice it's not going to increase the precision. At the 1A motor current I normally use, it feels as though I don't always get movement on individual microsteps.

If you want to improve precision, use 0.9deg motors instead of 1.8deg and make sure your mechanics has low backlash.

Quote
ampapa
I'm pretty sure the Duet is RRF but I thought I read somewhere that there was a forked version for "Core XY" support?

The de facto RRF master is the one in my github repository, which is a fork of the original version by RepRapPro (development of that version ceased 2 years ago). Chrishamm also maintains a fork, which is closely aligned with mine. We've both had CoreXY/HBot support for a long time, I forget how long but well over a year.

---

Large delta printer [miscsolutions.wordpress.com], E3D tool changer, Robotdigg SCARA printer, Crane Quad and Ormerod

Disclosure: I design Duet electronics and work on RepRapFirmware, [duet3d.com].

Quote
dc42
Arthur's replies to my own in this thread caused me to pause for thought and prompted me to run some tests. As this has drifted seriously off-topic, I've started a new thread here [forums.reprap.org].

@LoboCNC, in a 3D printer there is a big difference in the noise level when you switch from x16 to x256 microstepping. In other applications such as CNC machines, the reduction in noise may not be significant. Edit: I take your point about the importance of tooth profile for microstepping accuracy. Sadly I've never seen this mentioned in a stepper motor data sheet. I read somewhere that Nema 17 motors are typically optimised for accuracy and larger motors for holding torque, but i don't know whether that is true.

I really respect your response here, so much that I logged in to just say that. (I lurk, I lurk.) Arthur I'm kind of confused how he would be off base (I'm not really sure how to phrase that, Arthur, I don't mean to be speaking not directly to you as well.)

However, I am very confident DC42 is correct, and arthur shows a bad interpretation of that app note. I am familiar with it myself. It is not meant to interpret that you truly lose accuracy in a real way. You "lose accuracy" because you can be off by the difference in the microsteps. With 4X microstepping being off by 1 or 2 means you're less accurate with respect to position and microstep command, BUT, in the case of 1X microstepping you are only accurate insofar as you could cannot ask for finer gradation, but you might actually want it.

Summary point: You only lose accuracy if you interpret accuracy to mean "is my position the sum of the number of steps I've taken?" But that is not accuracy for our application. We want to get as close of an output position as we put in, and microstepping gives you that better than not-microstepping.

Microstepping = can decrease accuracy if you define accuracy as "my real position vs K*steps I've taken)" . You do *not* lose accuracy when you define accuracy as what we care about "my desired location vs my position" Sorry I've said the same thing multiple times, I am realizing I've not done a perfect job.

The special drivers that interpolate 16th stepping to 256th are a little different. They use a running average I believe to determine the interpolation, and could be off a little. I am not saying they are, I simply haven't looked into how it works. But the interpolation running average thing is probably why they have problems with Marlin FW using double-steps to get faster steprate.

I think there should be a much stronger social norm to know the topic when implicitly arguing against someone else's firmware. I understand Arthur you're not a motor expert and you've put together an awesome project, but it seems like you're off base here. I don't really need to come to the defense of Duet, since Duet and Smoothie are both awesome boards, but I was a taken aback when Arthur was citing the app note incorrectly. I felt it's OK to post (...3months late) because I actually am a smoothie user and not duet user, so I feel it would be reasonably appropriate for me to comment.

Quote
dc42
Arthur's replies to my own in this thread caused me to pause for thought and prompted me to run some tests. As this has drifted seriously off-topic, I've started a new thread here [forums.reprap.org].

@LoboCNC, in a 3D printer there is a big difference in the noise level when you switch from x16 to x256 microstepping. In other applications such as CNC machines, the reduction in noise may not be significant. Edit: I take your point about the importance of tooth profile for microstepping accuracy. Sadly I've never seen this mentioned in a stepper motor data sheet. I read somewhere that Nema 17 motors are typically optimised for accuracy and larger motors for holding torque, but i don't know whether that is true.

DC42 I really respect your response here, so much that I logged in to just say that. (I lurk, I lurk.) Arthur I'm kind of confused by your response.

However, I am very confident DC42 is correct, and arthur shows a bad interpretation of that app note. I remember reading it a few weeks back when I was trying to relearn the information as well (I am likewise not a motor expert, just a mechanical student.) It is not meant to interpret that you truly lose accuracy in a real way. You "lose accuracy" because you can be off by the difference in the microsteps. With 4X microstepping being off by 1 or 2 means you're less accurate with respect to position and microstep command, BUT, in the case of 1X microstepping you are only accurate insofar as you could cannot ask for finer gradation, but you might actually want it.

Summary point: You only lose accuracy if you interpret accuracy to mean "is my position the sum of the number of steps I've taken?" But that is not accuracy for our application. We want to get as close of an output position as we put in, and microstepping gives you that better than not-microstepping. If you want to get to point X, which is halfway between one step and a second step, only microstepping gives you a chance of getting to that point, even if it's not certain to go to that point.

Microstepping = can decrease accuracy if you define accuracy as "my real position vs K*steps I've taken)" . You do *not* lose accuracy when you define accuracy as what we care about "my desired location vs my position" Sorry I've said the same thing multiple times, I am realizing I've not done a perfect job.

The special drivers that interpolate 16th stepping to 256th are a little different. They use a running average I believe to determine the interpolation, and could be off a little. I am not saying they are, I simply haven't looked into how it works. But the interpolation running average thing is probably why they have problems with Marlin FW using double-steps to get faster steprate.

I think there should be a much stronger social norm to know the topic when implicitly arguing against someone else's firmware. I understand Arthur you're not a motor expert and you've put together an awesome project, but it seems like you're off base here. I don't really need to come to the defense of Duet, since Duet and Smoothie are both awesome boards, but I was a taken aback when Arthur was citing the app note incorrectly. I felt it's OK to post (...2months late) because I actually am a smoothie user and not duet user, so I feel it would be reasonably appropriate for me to comment.

Edited 3 time(s). Last edit at 06/19/2017 03:59PM by Tomek.

I think you're confusing resolution and accuracy ( which the app note actually adresses ). I've actually looked into this *a lot* a year or so ago when doing research for what type of driver to use for Smoothieboard v2, and I'm very confident I'm right on this, however it feels like every time I have a conversation about this, it takes an extreme amount of time to get people to understand what's going on.

What I'm trying to say is : I don't have the time to explain/have this conversation right now, v2 smoothie is picking up speed right now, and I'll have to wait until it's released and a bit more mature before I go back to arguing points/explaining things in forums.

Cheers.

Quote
arthurwolf

Quote
dc42
Here's why people get confused about torque and microstepping.

Actually, I already know everything you've just said. No confusion here : Still.

From what I understand, it's your explanation that is a misconception ( or rather an oversimplification ).

Yes, as microstepping increases, torque *per step* decreases to the point where you aren't getting more accuracy. But it can actually get to the point where you get *even less* accuracy. It just depends on torque.

This explains the problem pretty well : [www.micromo.com]

I quote :

Quote
Micromo
« The consequence is that if the load torque plus the motor’s friction and detent torque is greater than the
incremental torque of a microstep, successive microsteps will have to be realized until the accumulated torque
exceeds the load torque plus the motor’s friction and detent torque.
Simply stated, taking a microstep does not mean the motor will actually move. If reversing direction is desired, a
significant number of microsteps may be needed before movement occurs. That’s because the motor shaft torque
must be decremented from whatever positive value it has to a negative value that will have sufficient torque to
cause motion in the negative direction. »

This means that depending on the torque you are dealing with, there is a certain level above which you are actually *loosing* accuracy.

And not just "we lost some accuracy and gained some, it balances so it wasn't worth it increasing microstepping" ( which you are talking about ), but it can get to "we should actually be using lower microstepping here as we'd end up with better precision that way".

I first learned about this from people who professionally use microstepping drivers in cnc routers and pick and place machines, and for whom this is very critical. They actually measure the torque, do the math and have to choose the ideal microstepping for a given machine/application.

This has been measured and is pretty well known by the pros. I don't think we have much data on this for 3D printers since 1/256 microstepping for 3D printing is pretty recent, but I'm pretty sure at some point somebody will measure this and we'll have data to show it's the same for 3D printing as it is for other industries ( no reason it wouldn't be ).

From experience, I'm pretty sure the folks in the deltabot google group will be the first to provide data for this They love doing that sort of stuff.

This makes sense to me. At a point at which each microstep results in motion you are better off increasing microstepping to gain even more precision.