Motion Controller Designed for OpenPnP

1420 views
Skip to first unread message

Lance Endres

unread,
May 23, 2021, 1:36:28 PM5/23/21
to OpenPnP
It seems to me that the specifics of a pick-and-place machine warrants its own motion control card.  The combination of high accuracy and high speed is unique to pick-and-place machines.  And the number of motors being driven is higher than most applications.  Jarosław Karwik has already done some really promising work:
But in the end, all those cards were designed as general cards or even aimed more towards things like 3D printers.  And I think we are all aware that pick-and-place machines come with their own specific challenges.

My friend and I were discussing what that controller should look like and here's what we came up with:
  • Mother and daughterboard pair (to allow remote mounting of daughterboard in the head)
  • Microprocessor capable of 3-order motion
  • On board USB hub, only 1 connection required to the computer
  • Capable of driving up to NEMA 23 motors for X and Y-axes
  • Able to reconfigure controller without having to compile new firmware (like Smoothieware)
  • Built-in vacuum sensors
  • Built-in protection against solenoids
  • Capable of running up to 4 nozzles
Besides the hardware challenges of pick-and-place machines, I think having our own control card would bring additional benefits:
  • The firmware could be pick-and-place specific.  No need to have a bunch of extra stuff related to 3D printers or laser cutters cluttering things up.  Updates could happen faster.
  • Tighter integration with OpenPnP could make setup much easier for new users/builders.
What other major features should this motion controller have?  It's my understanding that the Duet3 has a web interface for configuration, would that make setup easier for new users/builders?

I am interested to hear any discussion points people have about this topic.  Are there additional features required?  Do you think such a controller card is warranted?  I feel certain it will happen at some point.  Are we ready for it now?

Lance



Arthur Wolf

unread,
May 23, 2021, 1:52:11 PM5/23/21
to ope...@googlegroups.com
On Sun, May 23, 2021 at 7:36 PM Lance Endres <lance...@gmail.com> wrote:
It seems to me that the specifics of a pick-and-place machine warrants its own motion control card.

Just for reference/if you are curious about the opposed idea to this:

The *whole point* of the Smoothie project (smoothieware.org, currently recommended for OpenPNP by the project I believe), is to have a system that is *not* dedicated to a machine type in particular (the way things were before Smoothie was first started years ago), but instead to make a system that is designed from the ground up to support different machine types.

The reason why, is that instead of having 10 people working on a laser controller, and 15 other people working on a pick and place controller ( and 30 for 3D printing, and 20 for CNC milling, etc etc, the list is *actually* long ), you instead have *all* of those people all working together on a single thing.

So when some improvement to the acceleration code, or to the electronic components for the endstops, is added to the project, it doesn't just benefit *one* machine type, say the laser, but it benefits *all* of the machines.

The goal here, is that work is optimized, there is less duplication of labor.

This was the intended goal of the project at the beginning, and it has work beautifully since, this hasn't been just a theoretical thing, in practice this pooling of ressources/efforts has actually massively functionned.

Of course, anyone can do anything, and I'm sure there are benefits to making a "more dedicated" controller for a given machine type.

Actually, you should probably read the history of this mailing list, the idea of making a controller dedicated controller just for the OpenPNP project isn't new, and is something I've seen many conversations/projects about, I sort of get the feeling we get a new "openpnp-specific" controller design maybe once a year or so around here.

So talking about "pooling efforts", maybe take a look at what people have already been doing, I'm sure there's a lot of their work you could re-use/extend/benefit from.

And maybe it would even make more sense to contribute to existing boards/projects/efforts, rather than re-inventing the wheel starting from scratch.


  The combination of high accuracy and high speed is unique to pick-and-place machines.  And the number of motors being driven is higher than most applications.  Jarosław Karwik has already done some really promising work:
But in the end, all those cards were designed as general cards or even aimed more towards things like 3D printers.

There are other, OpenPNP-dedicated boards around, you should search this mailing list for them (and I'm sure their authors would be glad to talk to you about them).

  And I think we are all aware that pick-and-place machines come with their own specific challenges.

My friend and I were discussing what that controller should look like and here's what we came up with:
  • Mother and daughterboard pair (to allow remote mounting of daughterboard in the head)
  • Microprocessor capable of 3-order motion
  • On board USB hub, only 1 connection required to the computer
  • Capable of driving up to NEMA 23 motors for X and Y-axes
  • Able to reconfigure controller without having to compile new firmware (like Smoothieware)
  • Built-in vacuum sensors
  • Built-in protection against solenoids
  • Capable of running up to 4 nozzles
Besides the hardware challenges of pick-and-place machines, I think having our own control card would bring additional benefits:
  • The firmware could be pick-and-place specific.  No need to have a bunch of extra stuff related to 3D printers or laser cutters cluttering things up.  Updates could happen faster.
  • Tighter integration with OpenPnP could make setup much easier for new users/builders.
What other major features should this motion controller have?  It's my understanding that the Duet3 has a web interface for configuration, would that make setup easier for new users/builders?

I am interested to hear any discussion points people have about this topic.  Are there additional features required?  Do you think such a controller card is warranted?  I feel certain it will happen at some point.  Are we ready for it now?

Lance



--
You received this message because you are subscribed to the Google Groups "OpenPnP" group.
To unsubscribe from this group and stop receiving emails from it, send an email to openpnp+u...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/openpnp/1260cb7e-863c-45c2-9e16-7fde92e81611n%40googlegroups.com.


--
勇気とユーモア

dc42

unread,
May 23, 2021, 4:33:00 PM5/23/21
to OpenPnP
If you want to mass produce PnP machines, then of course a dedicated board makes sense to reduce costs and perhaps simplify wiring. However, unless you are making at least a thousand machines, the costs of development and small production runs make a dedicated solution expensive, and a general purpose board such as Duet or Smoothieboard will be more economical, even if it offers some functionality that you don't need.

Looking at your wish list:

  • Mother and daughterboard pair (to allow remote mounting of daughterboard in the head)
  • Microprocessor capable of 3-order motion
  • On board USB hub, only 1 connection required to the computer
  • Capable of driving up to NEMA 23 motors for X and Y-axes
  • Able to reconfigure controller without having to compile new firmware (like Smoothieware)
  • Built-in vacuum sensors
  • Built-in protection against solenoids
  • Capable of running up to 4 nozzles
Duet 3 MB6HC provides all of those, except for the on-board USB hub and built-in vacuum sensors (would you really want the vacuum sensors on the main board anyway?). Duet 3 Mini likewise, except for the ability to drive Nema 23 motors.

I guess you might want to make a new head-mounted board to drive multiple small motors, to more easily handle a head with 4 nozzles. The options currently available from Duet3D are the Tool Board 1LC, which has just one motor driver; and the EXP3HC, which has three high-current ones and is therefore quite large. You could design a new head-mounted board based on the Duet3D EXP3HC expansion board design, but use TMC2209 drivers to reduce cost and space, and leave off any other parts that you don't need.

Or perhaps the 4-nozzle solution you had in mind was an IDEXY configuration, with two heads moving independently in XY having two nozzles each? IDEXY 3D printers already exist (using Duets); maybe OpenPnP can support this arrangement too?

Bear in mind that now is not a good time to be designing new electronics that uses microcontrollers or FPGAs, because these are getting very hard to source. Lead times for many parts are around 1 year. You might be able to buy a few MCUs for prototyping, only to find that none are available when you want to manufacture.

Arthur Wolf

unread,
May 23, 2021, 4:38:03 PM5/23/21
to ope...@googlegroups.com
On Sun, May 23, 2021 at 10:33 PM dc42 <dcro...@duet3d.com> wrote:
If you want to mass produce PnP machines, then of course a dedicated board makes sense to reduce costs and perhaps simplify wiring. However, unless you are making at least a thousand machines, the costs of development and small production runs make a dedicated solution expensive, and a general purpose board such as Duet or Smoothieboard will be more economical, even if it offers some functionality that you don't need.

Looking at your wish list:
  • Mother and daughterboard pair (to allow remote mounting of daughterboard in the head)
  • Microprocessor capable of 3-order motion
  • On board USB hub, only 1 connection required to the computer
  • Capable of driving up to NEMA 23 motors for X and Y-axes
  • Able to reconfigure controller without having to compile new firmware (like Smoothieware)
  • Built-in vacuum sensors
  • Built-in protection against solenoids
  • Capable of running up to 4 nozzles
Duet 3 MB6HC provides all of those, except for the on-board USB hub and built-in vacuum sensors (would you really want the vacuum sensors on the main board anyway?). Duet 3 Mini likewise, except for the ability to drive Nema 23 motors.

I guess you might want to make a new head-mounted board to drive multiple small motors, to more easily handle a head with 4 nozzles. The options currently available from Duet3D are the Tool Board 1LC, which has just one motor driver; and the EXP3HC, which has three high-current ones and is therefore quite large. You could design a new head-mounted board based on the Duet3D EXP3HC expansion board design, but use TMC2209 drivers to reduce cost and space, and leave off any other parts that you don't need.

Or perhaps the 4-nozzle solution you had in mind was an IDEXY configuration, with two heads moving independently in XY having two nozzles each? IDEXY 3D printers already exist (using Duets); maybe OpenPnP can support this arrangement too?

Bear in mind that now is not a good time to be designing new electronics that uses microcontrollers or FPGAs, because these are getting very hard to source. Lead times for many parts are around 1 year.

I'll confirm that strongly: we are re-designing Smoothie v2 from the LPC4330 to the STM32H7 for this exact reason: nearly one year lead time on the component we initially designed around, and it's taking us only a few months to redesign around the new chip and port the firmware (both of which are nearly done now).
This was quite the crazy realization, when we were nearly ready to start shipping the board, and were arranging production of the last prototype run, as we tried getting the components for those last prototypes, realizing some components had lead-times counted in months, and others had up to 5000% price markups...
Quite the crazy situation for board design lately.

You might be able to buy a few MCUs for prototyping, only to find that none are available when you want to manufacture.

On Sunday, 23 May 2021 at 18:36:28 UTC+1 lance...@gmail.com wrote:
It seems to me that the specifics of a pick-and-place machine warrants its own motion control card.  The combination of high accuracy and high speed is unique to pick-and-place machines.  And the number of motors being driven is higher than most applications.  Jarosław Karwik has already done some really promising work:
But in the end, all those cards were designed as general cards or even aimed more towards things like 3D printers.  And I think we are all aware that pick-and-place machines come with their own specific challenges.

My friend and I were discussing what that controller should look like and here's what we came up with:
  • Mother and daughterboard pair (to allow remote mounting of daughterboard in the head)
  • Microprocessor capable of 3-order motion
  • On board USB hub, only 1 connection required to the computer
  • Capable of driving up to NEMA 23 motors for X and Y-axes
  • Able to reconfigure controller without having to compile new firmware (like Smoothieware)
  • Built-in vacuum sensors
  • Built-in protection against solenoids
  • Capable of running up to 4 nozzles
Besides the hardware challenges of pick-and-place machines, I think having our own control card would bring additional benefits:
  • The firmware could be pick-and-place specific.  No need to have a bunch of extra stuff related to 3D printers or laser cutters cluttering things up.  Updates could happen faster.
  • Tighter integration with OpenPnP could make setup much easier for new users/builders.
What other major features should this motion controller have?  It's my understanding that the Duet3 has a web interface for configuration, would that make setup easier for new users/builders?

I am interested to hear any discussion points people have about this topic.  Are there additional features required?  Do you think such a controller card is warranted?  I feel certain it will happen at some point.  Are we ready for it now?

Lance



--
You received this message because you are subscribed to the Google Groups "OpenPnP" group.
To unsubscribe from this group and stop receiving emails from it, send an email to openpnp+u...@googlegroups.com.


--
勇気とユーモア

Jarosław Karwik

unread,
May 24, 2021, 2:11:17 AM5/24/21
to OpenPnP
Yeah, it is crazy -  especially with components. These days even highest priced STM32H7 are hard to get...
It does not help that people started buying components like never before - I have myself bought vital parts for next 12 months of my production. So the shortages will last, but not forever (saturation of stocks will happen at some pint). I can already see some light for the next year.

In my case the controller is part of commercial project - I need my own board with some specific features ( I design electronics for industrial ink printers).
When doing it - I have realized that the design is pretty good match for pick and place machines :-) and I decided to share.
It still does not work - mainly because I have to finish first one high priority design for my biggest client.

About the features - I would say you need ethernet ( even if you keep USB).  There are too many cases USB is troublesome, while ethernet works flawlessly.
In my design I use ethernet to master board ( with possibility to use USB), remaining boards are connected via high speed CAN - FD .

betzt...@gmail.com

unread,
May 24, 2021, 4:20:20 PM5/24/21
to OpenPnP
I can relate highly to this thread. My certified designs had the STM32's in it go out of stock until 2022, so I had to redesign and purchase a years worth of a different, yet available STM32, two days later those chips went out of stock world-wide!! 

Crazy times....

Peter.

Arthur Wolf

unread,
May 24, 2021, 4:34:12 PM5/24/21
to ope...@googlegroups.com
On Mon, May 24, 2021 at 10:20 PM betzt...@gmail.com <betzt...@gmail.com> wrote:
I can relate highly to this thread. My certified designs had the STM32's in it go out of stock until 2022, so I had to redesign and purchase a years worth of a different, yet available STM32, two days later those chips went out of stock world-wide!! 

Which chip was it?


Crazy times....

Peter.

--
You received this message because you are subscribed to the Google Groups "OpenPnP" group.
To unsubscribe from this group and stop receiving emails from it, send an email to openpnp+u...@googlegroups.com.


--
勇気とユーモア

Betz Technik Industries Ltd.

unread,
May 24, 2021, 4:35:15 PM5/24/21
to ope...@googlegroups.com

 

It was the STM32F411 in LQFP, now is the STM32L072 in LQFP.

 

Jarosław Karwik

unread,
May 24, 2021, 4:35:59 PM5/24/21
to ope...@googlegroups.com
I have similiar issue with STM32F105RCT6

phil barrett

unread,
May 25, 2021, 10:38:51 AM5/25/21
to OpenPnP
I'm new to the group but have been making motion controllers for some time now. My most recent offering is a 5 Axis motion controller based on the Teensy 4.1 (iMXRT1062 at 600 mhz) using grblHAL (a 32-bit extension to Grbl that runs on 13 different microcontrollers). Supports up to 5 external Relays/TTL outputs. It has I2C and UART headers that allow for expansion. Capable of driving steppers at up to 400 kHz.  I also have several I2C daughter cards in the pipeline - one is an isolated 12 bit ADC (vac sensor?) and another is a 16 bit I/O expansion card. Adding additional ADC channels is fairly easy.  The grblHAL developer is also adding PnP extensions  and is motivated to add additional ones to support the OpenPnP community.

I have several follow on boards that also run grblHAL and am interested in creating a board that meets the needs of the OpenPnP community so I am following these discussions closely.

Harjit Singh

unread,
May 26, 2021, 6:20:16 AM5/26/21
to ope...@googlegroups.com
Please consider incorporating Pololu style 16 pin dual inline stepper motor driver board support.

It may have been your project for a carrier board for the Teensy 4.0 but that used Marlin. If so, why the change to Grbl?

Thanks.

Mike Menci

unread,
May 26, 2021, 7:36:24 AM5/26/21
to OpenPnP
and Bills Github !
mike

Arthur Wolf

unread,
May 26, 2021, 9:51:13 AM5/26/21
to ope...@googlegroups.com
On Wed, May 26, 2021 at 12:20 PM Harjit Singh <zeet...@gmail.com> wrote:
Please consider incorporating Pololu style 16 pin dual inline stepper motor driver board support.

There are very good reasons why these have progressively gotten out of fashion over the years in the Reprap and other Open-Source CNC communities, despite the carrier boards themselves seeing significant evolution/improvement over time.
Despite that betterment/progress, fewer and fewer boards support those driver boards.

The reason is, these boards are a terrible idea.
They were initially used on the RAMPs boards at the very beginning of Reprap, simply because they were the only easy thing available, and the fastest way to a working prototype/system.

But they pretty much go against the way drivers are *designed* to be used. Drivers are supposed to be cooled by the PCB below them, and those drivers simply do not have enough PCB surface to properly cool the driver from the heat usage generates.
This is why drivers rated for 2A use, when installed on Pololu carriers, have rated maximum current abilities of 1.4A typically.

They also severely limit the ability to use the more advanced features of systems like the Trinamic drivers, and create a complete mess of standards/compatibility due to trying to use the socket in situations it's really not designed for.

All in all, these are not a good idea, and it's a good thing they are falling out of favor.

Directly integrating the drivers on the board is the way the drivers are *designed* to be used, and it's a very superior way to do things.

phil barrett

unread,
May 26, 2021, 10:59:27 AM5/26/21
to OpenPnP
While the Stepstick style boards have lots of limitations as other have noted, they do meet a need. If you have an application that doesn't need a lot of current for the stepper motors, they work fine.  Especially with the newer TMC drivers.  I would not recommend them for steppers that need over 1A.  But there are lots of cases where stepsticks are a perfectly valid solution. They certainly have increased the number of motion controllers in the world and allowed price points that are not possible with external or integrated drivers.

That said, I generally do not recommend them. The low cost is alluring but if one builds a system with stepsticks and then realizes they need more power or speed, they wind up wasting time and money.  And they maybe have to get a new system case because they need a lot more room for external drivers. 

phil barrett

unread,
May 26, 2021, 11:05:00 AM5/26/21
to OpenPnP
|  It may have been your project for a carrier board for the Teensy 4.0 but that used Marlin. If so, why the change to Grbl?
No, that was not mine.  I have only used grblHAL with the Teensy 4.1. My board targets CNC machines like routers, mills, lathes, laser cutters/engravers, foam cutters and similar.

Arthur Wolf

unread,
May 26, 2021, 11:08:52 AM5/26/21
to ope...@googlegroups.com
I definitely do not deny their (stepstick/pololu) usefulness historically.

I want to note though, on the price side of things, it's generally a misconception that they are advantageous.
All in all, it costs less to solder/integrate a driver directly to your controller, than to have to pay for :
* Female headers
* Male headers
* Assembly of above (they are not SMT, that means extra manual assembly costs)
* Soldering of above (same)
* Extra PCB processing costs
* Inability to design a full-SMT board (further price increase for the design)

All of which direct integration allows you to save the cost of.




--
勇気とユーモア

bert shivaan

unread,
May 26, 2021, 11:19:55 AM5/26/21
to OpenPnP
Until I blow a drive, then how do I fix it?

I have a control right now using 3 "step stick" style drives and 2  Other high power drives. I found it just fine to grab step/dir from the header where the step stick would have been. I can agree about the cost of the headers, But if thats really an issue offer the boards sans headers. 

One could simply direct solder their step stick to the control board, and still be able to remove it when I blow the driver.

Just my 10 cents

Betz Technik Industries Ltd.

unread,
May 26, 2021, 11:22:58 AM5/26/21
to ope...@googlegroups.com

 

“Until I blow a drive, then how do I fix it?”

 

In the context of this group, it really is not difficult to replace a small surface mounted chip...

 

 

Arthur Wolf

unread,
May 26, 2021, 11:23:26 AM5/26/21
to ope...@googlegroups.com
On Wed, May 26, 2021 at 5:19 PM bert shivaan <bert.s...@gmail.com> wrote:
Until I blow a drive, then how do I fix it?

That's 2008 thinking :)

I'll replace your board.
It's so incredibly rare, it's no issue at all.

Stepstick drivers sometimes burn. On-board drivers never do (or close to never) in comparison.

I expect most other board manufacturers would do the same and just replace the entire board for free.

Modern boards have diodes protecting the drivers making it close to impossible to destroy them.
Even if you did manage to destroy a driver for a board the manufacturer doesn't want to replace for you, wiring an external driver is extremely easy and extremely cheap.

ma...@makr.zone

unread,
May 26, 2021, 11:43:14 AM5/26/21
to ope...@googlegroups.com

Is it easy to blow one of the more modern ICs today? I've made many mistakes and never blew one from outside connections.

However, I did blow one, when I inserted a stick the wrong way around. Luckily the board survived. So in my book there is no real "for dummies like me"-advantage to have sticks. ;-)

It would be nice though, for a board to break out the step/dir pins etc. and to be able to disable the on-board driver (if it matters), so one can attach external drivers. There are valid reasons to do so (cabling, special purpose drivers, motor integrated drivers, etc.).

_Mark

Arthur Wolf

unread,
May 26, 2021, 11:47:20 AM5/26/21
to ope...@googlegroups.com
On Wed, May 26, 2021 at 5:43 PM ma...@makr.zone <ma...@makr.zone> wrote:

Is it easy to blow one of the more modern ICs today? I've made many mistakes and never blew one from outside connections.

However, I did blow one, when I inserted a stick the wrong way around. Luckily the board survived. So in my book there is no real "for dummies like me"-advantage to have sticks. ;-)

It would be nice though, for a board to break out the step/dir pins etc.


All the modern controller boards I can think of, as far as I know, do break those out.
I would strongly recommend Lance/anyone else thinking of designing a new controller system, to also break those off.
Alternatively, most controller boards break out "extra" GPIO, and will allow you to use *those* for step/dir pins for any of the axes (some might require you edit the source code for this, but at least they allow you to do it *some* way).



--
勇気とユーモア

Sairfan

unread,
May 26, 2021, 11:21:54 PM5/26/21
to OpenPnP
I would strongly recommend to add Ethernet support as USB has lots of limits, Stepper motor drivers should be flexible to choose like even if some wants to use closed loop steppers, board should be open to new experiments like highly customizable. give lots of headers you can put on board jumper that can be cut off when needed and headers could be solder when needed, i believe this could be helpful because lots of people out there want to work on oipen pnp and such highly customizable board will also greatly helpful not only quickly setup a machine but for those as well who want to move forward and improve the machine. 

Phil Barrett

unread,
May 26, 2021, 11:28:16 PM5/26/21
to ope...@googlegroups.com

I would strongly recommend to add Ethernet support as USB has lots of limits,
Definitely. I have ethernet on my board and recommend it over USB.  Galvanic isolation as well as high levels of EMI resistance make it head and shoulders better than USB. It does add cost (though not that much) but the price is worth it. 

Jason von Nieda

unread,
May 26, 2021, 11:39:02 PM5/26/21
to ope...@googlegroups.com
Hi Phil,

Thanks for posting about this. I'd never heard of the board or of grblHAL and both sound really interesting!

Thanks,
Jason


Eric Norton

unread,
May 27, 2021, 12:11:12 AM5/27/21
to ope...@googlegroups.com
I figured I'd jump right in and add my two cents to this conversation so here goes...

Driver Modules - the driver modules if designed properly can handle up to 3 amps per pin on the header alone. If you design the driver board properly with four or more layers you can dissipate the heat by placing the IC and its components on the bottom side and add a suitable heatsink to the top exposed ENIG pad with sufficient vias to transfer the heat. I have designed several using the TMC2209 and the TMC2226 drivers with no performance degradation and thermal performance is well within tolerance.

USB/Ethernet - USB 2.0 is not a bad interface if used properly and understand how the protocol works to successfully transfer the data. You need it to be an isolated interface if attached to a machine as I have found ground loops cause all sorts of problems and especially with motors involved. Ethernet is a solid interface but again you need to isolate it from the machine.

ma...@makr.zone

unread,
May 27, 2021, 2:03:28 AM5/27/21
to ope...@googlegroups.com

Can Ethernet be made as fast as USB (low latency) on common MCUs?

_Mark

Jarosław Karwik

unread,
May 27, 2021, 2:18:16 AM5/27/21
to ope...@googlegroups.com
Both (usb and ethernet) have similiar latency - usually 2..3 task switches. This equals 2..3ms max. My loop driven solution on 200MHz CPU is about 0.5ms.

In hardware both solutions are DMA driven and quality of software decides the speed, altgough there is more overhead in case of processing ethernet frames. But it is still fraction of ms.

Niclas Hedhman

unread,
May 27, 2021, 2:18:16 AM5/27/21
to ope...@googlegroups.com
If you keep it on separate segments, and don't mix in other traffic (say run OpenPnp and watch YouTube meanwhile) Ethernet should be in sub-millisecond latency on all packets, which I assume is adequate. And of course it is not Ethernet itself that adds latency, it is the buffering in network cards and the Linux kernel. Anecdote; investment banks that have direct connectivity to stock exchange servers have dedicated/embedded hardware that does high frequency trading with ~10microsecond round trip and Linux kernel adds too much latency for this. 

Niclas

ma...@makr.zone

unread,
May 27, 2021, 2:48:00 AM5/27/21
to ope...@googlegroups.com

Then I guess Ethernet is definitely the better choice due to inherent isolation.

Otherwise an on-board USB isolator or other means of isolating from the machine rails, seems necessary, at least for a trouble-free setup with no special care about how many plugs away your computer is connected to the mains and how long the USB cable is. Just based on some very simple calculations with expected current pulses, bulk capacitor ESR and lead/connector resistances (not even taking inductances into account). A fast PnP is not a 3D printer.

_Mark

Niclas Hedhman

unread,
May 27, 2021, 3:20:52 AM5/27/21
to ope...@googlegroups.com

A lot will also depend on higher level protocols and how they are used. It is quite easy to not do the right thing, thanks to the packet nature of Ethernet, yet having TCP as a streaming programming model. It is less straightforward than a serial port.

Niclas





Markus Kreidl

unread,
May 27, 2021, 8:15:26 AM5/27/21
to ope...@googlegroups.com
Hi,

a while ago I was thinking about a new controller board too. My
approach was a bit more advanced using a xilinx zynq or xilinx Kria
K26 [1].

The plan was to generate stepper signals in the FPGA part and move
Real time Tasks to a bare metal Jailhouse hypervisor cell [2].
This means OpenPnP (in a Linux Cell) and controlling software could
run on the same hardware.
As soon as I have the Kria K26 Board Evalboard I will test image
processing capabilities.


just my two cents

Markus

[1] https://www.xilinx.com/products/som/kria/k26c-commercial.html
[2] https://github.com/siemens/jailhouse
> To view this discussion on the web visit https://groups.google.com/d/msgid/openpnp/CADmm%2BKcojTErW45Rrj%3D8SDWi79WLs8Qq6ACcSeKa4QCn1aagPg%40mail.gmail.com.

Juan-Antonio Søren E. Pedersen

unread,
May 27, 2021, 8:52:05 AM5/27/21