For nerds only: Raspberry Pi for EMC2 controller?

[mpack]

Hi Thomas,
When I first heard of the RPi it did sound remarkably similar to the processor and devtool evaluation kits I've used for years. For example the old EB30 Atmel ARM7 eval board I played with in... 1997?

The problem with eval boards is that they are often available only in limited numbers, especially per person, and as the name implies they are sold at big discounts (maybe below cost) to encourage adoption of some processor in hw designs. The people who make them would be less than happy if they were disappearing off the shelves with no prospect of processor sales: and not enough sales of the boards themselves to offset mfring costs.

The RPi is as cheap or cheaper and should not have these political problems. My only concern is that I don't know how they can make the RPi this cheap!

ps. Got a big shock yesterday. Was Googling for info on the RPi and discovered that one of it's leading lights - Pete Lomas - is someone I know. My company owns shares in his company, we use them to design our boards, but sadly the RPi is a separate charitable hobby of his, so I get no shares from that... bummer :-)

[mpack]

Keep in mind, if a custom linux distro could be compiled which was stripped down (more so then the Ubuntu based EMC2 package) to a bare bones setup I think the hardware would be more then sufficient.

I would be wary of that approach. First, it would require a lot of Linux expertise to decide what can go and what needs to stay. Second, if you start with a working OS and take bits out, chances are you'll have a non-working OS.

What I had in mind was working from the opposite direction. Start with the CNC app: lets assume it's EMC for arguments sake. We take the source code and find out where all the OS calls are - and it's possible that the devs already isolated these in a few modules (I would). In which case we implement those calls more or less directly on the RPi hardware - we would have no OS to speak of at all.

Each unnecessary frill would complicate the task. For example TCP networking would require a full network stack and drivers for the Ethernet chip. Ditto USB. So, we need to think carefully before adding these on some whim. If the RPi can control the motors perfectly using simple PIOs then there is absolutely no need to introduce something more complicated.

[BanduraMaker]

I have heard of G-code, that is exactly what I had in mind as the script, though undoubtedly there will be other options. An ARM processor is easily capable of implementing a g-code interpreter that works in real time.

I'm sure it can, I'm just pointing out that the g-code interpreter/trajectory planner implementation is a non-trivial task.

Do people need a display in the workshop to tell you what the CNC widget is up to?

Absolutely. You might be able to get away with a few lines of text but a full blast display would be much better.

What I had in mind was working from the opposite direction. Start with the CNC app: lets assume it's EMC for arguments sake. We take the source code and find out where all the OS calls are - and it's possible that the devs already isolated these in a few modules (I would). In which case we implement those calls more or less directly on the RPi hardware - we would have no OS to speak of at all.

If the RPi can control the motors perfectly using simple PIOs then there is absolutely no need to introduce something more complicated.

Now this sounds like a pretty good approach. There's definitely no reason to include a smooth stepper in this processes if there's already adequate I/O on the board....but, that has yet to be proven.

You might check out the Dynomotion (kflop) site to get some ideas.

[Tkamsker]

Mpack:
As i wrote if you want avoid the OS even if it is already ported. Then have a look at the ramps 1.4 stuff
I think adding a little screen to see orientation does the Job.

The ti. 3530 is also base of beagle board so i expect this boards to Stay.
Thomas

[ihavenofish]

Now this sounds like a pretty good approach. There's definitely no reason to include a smooth stepper in this processes if there's already adequate I/O on the board....but, that has yet to be proven.

You might check out the Dynomotion (kflop) site to get some ideas.

the pi has 4 gpio pins. 1 will be used for a real time clock. so your left with 3.

that is not adequate for cnc, unless you use them as a serial connection to an external board which would need development. at that point though to me youve defeated the purpose by making a complicated expensive system out of an underpowered PC.

the only things the rpi has going for it are size and cost. if you eliminate one of those factors, you may as well buy a more fully functional system.

alternatively you can go the other way, and use an arduino, which is better suited to be the "middle man" between a pc and a drive.

[KuchateK]

No offense, but I think some of you have no idea what are you taking about. Or you are simply insane

Gcode was invented long time ago. In 1950s. First machine was running from vacuum tubes and relays!

16mhz 8 bit AVR microcontroller with 2 kilobytes of ram can run gcode interpreter for a CNC machine. It has 2 kilobytes of ram.
Grbl Gcode Interpreter - Contraptor
http://www.youtube.com/watch?v=mHmDhi2u2BE]GRBLshield CNC Milling Machine Conversion - YouTube
http://www.youtube.com/watch?v=QDKlZyB5RpE]Zen Toolworks CNC Controlled by Arduino/GRBL - YouTube

We are talking about 700 MHz 32 bit ARM chip with 256 megabytes of ram. 300mhz Pentium equivalent. Turbo CNC can run on 66mhz 486 and 4 megs of ram.

Parallel port must die for CNC the same way it died for printers. It is unreliable, it requires big old junk PC. It was great, but it starts to be outdated technology. Currently it exists pretty much only on second hand market. It is like floppy drive in some older industrial machines. Still works, but is it good? We can replace big, unreliable, noisy, delicate (dust and vibration) computers with something nice and small that won't require maintenance to keep it running.

Raspberry Pi has 8 GPIO pins, SPI, I2C and serial port on the header pins.
RPi Low-level peripherals - eLinux.org

It is enough for 3 axis machine. With extra I/O port expanders on SPI or I2C you can have a lot more pins for spindle/coolant power, buttons, safety features, probes, MPGs, tool changers, LCD screens for DRO, etc...

I think Raspberry PI has great potential for hobby CNC controllers. Because it is cheap there is great interest and there will be thousands of them. CNC controllers will grow on this platform the same way they did on Arduino.

Arduino is the controller for RepRap. 3 axis stepper driven machine.

In my opinion Raspberry will take place of PC and smoothstepper type cards for complete solution. It has enough power to run gcode interpreter and user interface and it has GPIO to run motor drivers and other electronics pretty much directly.

Linux is not good for real time? I bet it is a matter of time to port changes and extensions they have for EMC (LinuxCNC) on PC Linux to Linux running on Raspberry.

There are real time OSes for ARM chips.
FreeRTOS - Wikipedia, the free encyclopedia

Imagine something like this for your CNC machine:
Introducing the MakerBot Gen 4 Interface Board Kit v1.1! « MakerBot Industries
$100 seems about right. A lot cheaper and a lot more portable than a PC.

But I think Raspberry PI will be able to do a lot more. This hardware has a lot more power than even higher end CNC requires.

"Industry standard" Mach 3 requires 1GHz CPU and 512 megs of ram. Remember that it is running on Windows, so we have networking, multimedia and a lot of other crap eating those resources.

Even if I'm wrong in the end we can always just add FPGA (smoothstepper equivalent) to those GPIO pins on Raspberry. It still will be cheaper and better solution than any x86 PC. I will invest my time and money to work on it.

[Tkamsker]

Kuchatek :
I was exactely referring to Kind of makerbot.
As i work with cnc and now own a 3d printer i think you need an little Controller for the steppers servos relais toolchanger what ever so a Lot of gpio are required
but fancy screen on an Mill in the basement whatfor.
I can Control my printer even wireless from an android phone ...
So i will Investor my time there lets see where i get till june
Cu
Thomas

[Tkamsker]

Kuchatek :
I was exactely referring to Kind of makerbot.
As i work with cnc and now own a 3d printer i think you need an little Controller for the steppers servos relais toolchanger what ever so a Lot of gpio are required
but fancy screen on an Mill in the basement whatfor.
I can Control my printer even wireless from an android phone ...
So i will Investor my time there lets see where i get till june
Cu
Thomas

[ihavenofish]

Raspberry Pi has 8 GPIO pins, SPI, I2C and serial port on the header pins.

8?

why was i thinking 4? its actually up to 17 it seems, if you dont need uart, spi, etc. one pin even supports pwm which could possibly run a spindle (not sure if it mees he needs of a vfd).

with all that in mind, then yes, screw the external board. treat the gpio as if it was the pc's parallel, and youll probably have as just a much power as any other pc setup, for $30. the speed of the gpio seems adequate as well.

now if only i could GET one of these boards!

[BanduraMaker]

No offense, but I think some of you have no idea what are you taking about. Or you are simply insane

If you don't mean to offend, don't use offensive terms.

Conceptually, a stand alone replacement for Mach 3 and/or EMC2 at a cheaper price is truly an appealing thing. However, if it's a step backwards in terms of features and possibly performance, much of its appeal is lost, at least for users such as myself.

I'm a Mach 3 user and while it leaves many things to desire, it does have a very rich feature set and is quite flexible.

Consider that a SmoothStepper is about $150, A 1GHz laptop is about $100 and Mach 3 is about $150. If you can out perform Mach 3 in terms of outright performance and feature set (including display) for <$400, you've got something that would be of interest to me.

If on the other hand, you make something for $100 that has a crippled and poorly implemented G-code interpreter, lack of flexibility and a lack of features (e.g. spindle control, macros etc.) then you may have something of interest to some but certainly not to advanced users.

[ihavenofish]

i think what he was trying to say is the rpi is actually quite powerful enough on its own.

after a little more refreshed reading, im inclined to thing hes right. for some reason i confused it with another board is was looking at with only 4gpio, 2 in 2 out. the rpi sees configurable enough to come close to matching the io on the traditional par port, with a theoretical clock speed that would not be a downgrade vs mach3's 100khz.

so, a rpi, a realtime clock and a break out board in a tiny box for $100-$200 would be quite viable.

the hitch now is someone needs to get software that works with it... whether its emc, mach3, or something from the ground up.

maybe its time to implement one of my ancient ideas... a web browser based UI that talks back to a nice simple motion planner and pulse generator.

couple that with xzeros new tiny low cost stepper drives and you might have something awesome and under $400 all in.

[BanduraMaker]

i think what he was trying to say is the rpi is actually quite powerful enough on its own.

Was anyone really disagreeing?