Duet 3 demo at TCT

Phaedrux

Currently the DWC runs on the same CPU on the Duet that also handles everything else the printer has to control. It used to run on the even slower wifi chip. And while the Duet is far faster than the old school 8bit controllers like the Arduino, it's still nothing compared to even a $5USD Raspberry Pi Zero W single board computer.

Arduino 2560 = 16Mhz 8bit ATmega2560 with 256Kb ram
Duet 2 = 120MHz 32bit ARM Cortex-M4 with 128Kb ram and 512kb flash
Duet 3 = 300Mhz ARM??
Raspberry Pi Zero W = 1Ghz with 512Mb ram
Raspberry Pi 3B+ = 1.4 Ghz 64bit quad core ARM Cortex-A53 with 1Gb ram

The Duet electronics are freed up to function exclusively as the realtime hardware controller, possibly raising the ceiling on step rates. The single board computer like the Raspberry Pi can run a full modern Linux operating system which opens the door to much richer web application features than even the current DWC can provide. The easiest comparison is Octoprint. It has the same basic functions as the DWC, but is extensible with plugins to allow all sorts of other features and integrations.

Here's a few examples of some possible improved functionality this could open up.

Gcode realtime preview and time simulation estimates in a fraction of the time.
statistics tracking of successful prints and how long they took
database support to allow tracking and display of metadata for gcode files
thumbnail renders of the actual model in addition to the name
push notifications to a mobile device using various messaging services to let you know if a print completes or halts
USB webcam support and easy time lapses
local slicing of STLs to eliminate the need for a seperate PC
detailed logging and graphing of real time data
automatic firmware updates or package management
whatever else plugin developers can come up with

https://octoprint.org/ for more information. They offer a prebuilt sd card image that you put on a card and boot the Pi with and you're off to the races. It's geared towards the average user to to able to get up and running easily, but I could see how such a platform would be attractive to OEMs as well since they could customize the experience and extend functionality themselves to differentiate their offering from competitors.

Now I'm not knocking the DWC in the least. It does exactly what it needs to do and does it very well. But it is limited by the hardware it has to run on. A Pi3 is about $20 USD. So quite affordable and allows for a whole other level of capabilities.

I should clarify that I'm not hoping for a replacement of the DWC with octoprint. Rather I'm excited to see what can be done to expand the capabilities of the DWC given improved hardware. Though I could see a way in which octoprint could be customized to become a duet branded version to look and function much like the DWC does now with the added option of plugins. Pure speculation on my part.

deckingman

Thanks for that. Very informative.

dc42

Here's a slightly better video that I took this morning:

https://www.youtube.com/watch?v=LaNYeQiZK9M

More info at http://www.duet3d.com/duet3.

Dougal1957

This post is deleted!

Dougal1957

This post is deleted!

Mangy_Dog

Great meeting you David, the duet3 nearly made me blub as I just got the 2...

Nearly... It's an amazing bit of kit but certainly more than I ever need on my machine...

I hope I didn't seem strange hanging about a bit... I only wanted to have a chat with you but then Joseph came, I didn't want to interrupt so I waited patiently

Tct certainly was big... Tbh I got there at 10 thanks to traffic and Tbh... Its so big I didn't quite get to see everything there...

But I certainly got some very good insites to some new ideas....

Did you or anyone see the 6? Axis 3d printer? Normally you see those robot arms on high end cnc machines... I kinda had an idea that it would be possible a little while ago... I'm glad to see someone managed to make it...

Just at my rest stop half way home.. It's 830 traffic out of brummyland was hell... Not to mention a little scary on a motorbike with killer mad lorry drivers all over you and satnav was getting confused about what lane to take.... Anyway I'm alive... Had a great time thanks.

gtj0

@Phaedrux: I wrote almost the exact same thing in my survey response before I saw your post on the SBC option. Too bad I can't upvote your post more than once. Given that this is targeted as an advanced platform, an SBC is the way to go. Let the custom electronics concentrate on the things you can't do with a general computing platform and let the SBC concentrate on the things that aren't hardware dependent. As you alluded to in your last bullet item, a side benefit is that using a mainstream Linux environment opens up development to a wider audience and may allow @dc42 and @T3P3Tony more time to concentrate on the hardware and firmware that would still run on the Duet3. I'd definitely volunteer to assist on the software side. Another advantage is that the user could supply the "SBC" should they want more power than a Pi.

@dc42 Great work!

A Former User

@dc42
Wow looks very promising. My compliment (as far as a young man can judge - but I thought every compliment counts )

Since I was so busy with the hardware of the big printer I am building for those poeple here I missed the survey (argh!) but maybe it is not too late I hope:

separate "DC-IN" only for motor with higher voltage rating e.g. 36 or 48V to sustain higher torque of steppers with those faster positioning moves -> User can start with 1 power-supply like now: just make a jumper/bridge from that general +&- to the motor +&- (everybody has 2 thick cables at home for that I assume). That shouldn´t add much cost?
( maybe a "cheap" closed-loop option (maybe some code from "ServoStrap"/"uStepper"/"aus3d.com.au/electrical/closed-loop" could be recycled?) on the long run, though if I would have my higher voltage chances are that even on very big printers with high speed no step loss is likely and so the "closed-loop" is more an academic thing I guess - EDIT: a day later I think even more with a higher voltage in for motors this is not needed at the moment because "done right" the board would need a non volatile memory to store the position e.g. in case of a power out and this would be so much work and also added cost that with really high-A (which the new board seems to have already) & V on the motor-side + a little bit "pretensioned" axis on the printer that are stable also after a power-out closed-loop is not really needed)

Yeah I know that at some point things get too much and it is also an art to throw stuff out and keep it simple , and I am well aware that the issues posted here by me are niche-issues so -> they surely should go on one of the daughter-boards to keep the price of the "main"-board low, eventually on that big stepper-daughter-board with the genral I/O-expansion, so I could hook-up my big steppers only for x & y there and drive em with big voltage leaving 2 of the stepper-drivers of the main-board blank for something in the future . I like the fact that the duet-2-mainboard is still reasonable priced for it´s features and I hope the expensive stuff is going onto the optional daughter-boards.
(But building this huge printer here for the folks those things I would dream to have... hope you understand)

One last general idea that is only a cost-thing but maybe worth calculating: Most developers of terminal-blocks like "ece.com.tw", "wago", "phoenix-contact" have now terminal-blocks for pcb´s that allow simple "push-in" or "lever-actuation". This could reduce setup-time because cables just need to be cut off, unisolated (crimped - some even not that anymore) and easily put into the port (no special orientation of the cable/crimp is necessary when pushed in). Some of them are even better regarding vibration since screw´s tend to get loose with vibration over time. I know this is a cost-issue, but also it wouldn´t be needed at all ports I guess.

A Former User

@dc42 said in Duet 3 demo at TCT:

One you won't have heard of because Trinamic hasn't announced it yet. Supports all the usual features including stealthchop, stall detection, interpolation to x256 microstepping from any lower value etc. Not sure of the maximum peak current until we do thermal tests but expected to be between 4 and 4.5A. In the demo we're running the Nema 34 motors at 3A and although we are fan cooling the board, we probably don't need to.

Wow would this support full-step with *256 interpolation for a z-axis-stepper to use most energy-efficient the "natural" magnetic field of full-step + combining it with the least possibility for a step-loss since the angle for the "next stable position" is the biggest in that driving-mode (with the risk of step-loss because of ringing is gone because of the interpolation while moving)? (yeah I know this is a theoretical question but as long as the printer-axis are not closed-loop this is a "black-out-scenario" proof setup because if power is off the stepper is already at a stable position assuming the mechanics is built in a way the z-axis can´t move by it´s own weight)

dc42

@lb said in Duet 3 demo at TCT:

Wow would this support full-step with *256 interpolation for a z-axis-stepper to use most energy-efficient the "natural" magnetic field of full-step + combining it with the least possibility for a step-loss since the angle for the "next stable position" is the biggest in that driving-mode (with the risk of step-loss because of ringing is gone because of the interpolation while moving)? (yeah I know this is a theoretical question but as long as the printer-axis are not closed-loop this is a "black-out-scenario" proof setup because if power is off the stepper is already at a stable position assuming the mechanics is built in a way the z-axis can´t move by it´s own weight)

No, because the driver will still use microstepping.

The only reason that full steps give more torque is that when using full stepping, both phases are always driven at the full current.

NoSkillzEngineer

Can I please get my hands on this? Can I be a BETA tester or overpay for an early prototype board? I need this in my life!

I am a CNC machinist and I am pushing to truly combine thermoplastic Additive Manufacturing with industrial manufacturing equipment such as Haas, Mazak, Fanuc, CNC mills and lathes. There is no denying that the Duet has reigned supreme as far as the most advanced 3D printer controller.

In recent months/years mills and laser cutters have been easier to run with the Duet boards.

The more I learn about manufacturing, robotics, CNC programming, and I'm just starting to take an interest in PLC programming because the company I work for builds machines for other companies and is a big integrator for ABB Robotics, the more I realize just how similar so many industrial machines really are, as far as electronic control is concerned (oversimplified, of course). Steppers, brushless, brushed, geared down for specific speed, with and without encoders, et cetera, et cetera. Even seeing and controlling motors with encoders as individual "steps" so it's controlled nearly identically as a stepper (over simplified, I know). It's fascinating just how similar things become the deeper you go.

Are there hardware and firmware files for this? I REALLY want to play with this. :]

TLAS

Any chance we could get more details on that SBC interface? Also, Wifi and an Ethernet port (ideally with that extendable antenna) would be amazing and solve the vast majority of the problems I've been having with the Duet.

Also, I'd love to have a REST API interface for the Duet for future Duet Web Control expansion. It would also enable a lot more in the IoT world. It's possible the Duet Web Control already has this - if so, can you share some details?

T3P3Tony

We will be able to give more details about the Duet 3 to SBC interface in the next month or too. Most of what is requested is easy to implement on the SBC (e.g. REST API, all the network interface choices)

We are a fairly long way to having a beta. What we showed at the TCT was a concept. Once we get closer to prototype production we will be announcing more details.

brunofporto

@t3p3tony I am planning to use the Duet3 with a really large scale FarmBot as my younger brother plans to work with it as his final engineering project - how Duet3 will handle "exotic" (non FDM/CNC/Laser) systems?

T3P3Tony

@brunofporto sounds good, however it's likely that a farm bot would work great with Duet 2 unless you planned on having lots of heads!

brunofporto

@t3p3tony The issue is the size. We would like to design a polar system, with a Delta as "fine" tool manipulator to compensate for the imprecision of a 50m radius woobly thing

So a duet3 main would control the big polar system, pumps, feeders, etc. Then a add-on card would control the local precision tool.

sigxcpu

@phaedrux said in Duet 3 demo at TCT:

@deckingman

Currently the DWC runs on the same CPU on the Duet that also handles everything else the printer has to control. It used to run on the even slower wifi chip.

Duet 2 = 120MHz 32-bit, ESP8266 = 160 MHz 32-bit.
I think moving the web task to the main Duet 2 chip was because there is no ESP8266 in Duet 2 Ethernet.
And in terms of performance, that is a mistake, because the main chip has realtime tasks to do while ESP8266 is mostly idle now.

dc42

No, the main reason we moved the web server to the main processor is because it was too hard to debug the web server on the ESP, and because of the closed source SDK for the ESP we could never be sure whether CPU time used by the web server was impacting on the WiFi performance. Also it was taking too long to upload DWC to the ESP during manufacturing.

Like most users, we run the ESP at 80MHz, not 160MHz.

percar

Sounds really exciting. Do you have a timeline for when beta or first retail will be available?
I would like to be on the list of Beta Testers if it is possible

T3P3Tony

@percar we want to have some initial production boards in Q2 2019.