Welcome to LinuxCNC.org

Hi

Forewords: I am total beginner in CNC-world so my questions may be a little dumb.

For another project I designed a controller board that has ethernet-connection and stepper motor controllers + several digital I/Os.

For that same project, I am going to build a cnc router (to make parts) and I started wondering if it would
be possible to control cnc-machine from EMC over TCP-connection?

New PCs do not have parallel ports any more and I have had my share of add-on card mess.

I wonder if it would be possible, instead of creating stepper control signals over parallel port, send control over TCP?
In the controller board end it would be very easy to run motors based on the commands received over TCP and likewise easy to read limit switches,
encoders etc. and pass that info back to EMC.

I did search this forum and google but either this is something nobody has done or I was using wrong keywords.

-MJ-

I would say TCPIP is not a good match for EMC because it is nowhere near real-time
Possibly with a dedicated Ethernet card on the host (or MB with two Ethernet interfaces) and UDP you might get passible performance.

I think better choices are RTNET (which uses standard Ethernet hardware) or EtherCAT (which uses standard hardware on the host but special slave hardware) Note, all Ethernet options require a dedicated Ethernet connection, so you still may need add-on cards...

New controls with Ethernet servo drives using CIP Motion and CIP SYNC protocols are coming out (IEEE1588 based technology).
They promise sub-millisecond servo loop update time using regular Ethernet connection and even faster by using switches with QoS
It would be nice to have something similar for EMC2

Is CIP is really real time? the fact that CIP needs IEEE 1588 and timestamp based protocols to skip
over late packets seems to say even its promoters don't trust its real time abilities.

CIP as it was introduced years ago is not a real time, but widely used for I/O control.

CIP SYNC is relatively new variation of CIP and it can deliver timestamped data with sub-microsecond accuracy, I saw and used CIP SYNC enabled devices to get time synced accross the network

CIP MOTION is very new, later this month Rockwell Automation (www.ab.com) will be releasing new version 18 of ControlLogix that will have CIP motion implemented, I beleive drives with etherenet interface are out already.
As far as I understand, control and drive sync'ed their clocks first using CIPSYNC, and timestamped data will be coming over UDP multicast with periodic clock corrections. This technicsally will replace SERCOS interface

While what was said about CIP sync is true, there is one potential problem that I know of. When it comes to rigid tapping or some similar operation all the axis must be coordinated to the spindle position. Even If the spindle position encoder is connected directly to EMC2, and not over the CIP network, I don't know it this will be too limiting. Of course one could always tap slower and it would work, but we need to figure out what maximum delay and jitter can be tolerated. The other problem with CIP sync is that ODVA does not give the standard away for free. It costs $1000 for a copy. On the plus side, CIP sync uses standard ethernet hardware unlike etherCAT which requires a special MAC. Many suppliers are now offering ethernet with IEEE 1588. The best way to implement IEEE 1588 is to use hardware in the phy for the time stamp.

While I am using EMC2 for DIY hobby project only, my day-time job directly involved in PLCs and CIP I/O
With millions Ethernet/IP I/O modules with CIP implementation it is the network of the choice for many integrators
I am not sure about $1000 for spec, but I know it is not free.
Using standard Ethernet hardware and TCP/IP stack is a huge advantage, it makes it very affordable

I can't speak for CIP Sync and CIP motion - I did not have too much exposure yet.
It's been 8 years since I used SERCOS based controls, CNCs and drives, but I can't see why ridit tapping issue you described would be different with SERCOS?
Why Spindle position feedback can't be sent back to the host via Ethernet?
If they advertising up to 25 coordinated axis (I heard even larger number) why spindle suddenly becomes a problem?

Anyway, it is harder and harder to find a PC with a reliable parallel port today
I am looking forward and hope that someone will add CIP to EMC2
This will allow access to CIP based digital and analog I/O widely available
At that point servo/stepper controllers can be build to inteface with motors.
Steppes would be very simple to implement as they don't have any feedback in most of cases.

Rigid tapping using CIP sync motion is different as I will explain.

The way CIP sync motion works is that all of the axis on the CIP network are first synchronized using the 1588 time sync protocol. After time synchronization, the motion planner (EMC2 in our case) will then send commands to the drives at a periodic rate similar to the way it is done with other protocols except in addition to a position, velocity or torque command, a time value is associated with the command so that the drive executes the command at the correct time allowing many axis to be perfectly coordinated. The update rate to the axis is still relatively fast with CIP sync, 1 mSec for example, but the commands arrive enough in advance of the required commanded time so that all the axis will be in exact coordination. The motion planner knows the capabilities of each drive it is commanding and will never send a motion profile that an axis cannot follow.

For the case of rigid tapping the motion planner must read the spindle position and then command the axis to move accordingly. The delay between the spindle encoder read and the new position command must be short for this to work. The benefit of CIP sync is lost unless the spindle position is predicted for a time in the near future.

Now I did not say that rigid tapping could not be done, I just said that this could be a problem. So I did some more research. Let us assume that the position commands are sent to the drives every 1 mSec and the drives can respond before the next update period, then we need to determine how far off the axis can be in 1 mSec. If the spindle is at 500 RPM or 8.333 Rev/sec the spindle will move .00833 revolutions in 1 mSec. If the thread pitch is 13 TPI or .0796 in/thread then the tap position would be off by .00064 inches in 1 mSec. This seems to be in the ball park. If not, then the spindle needs to run slower, the thread pitch needs to be finer or the spindle position needs to be predicted 1 mSec in advance.

Now the more difficult part will be finding someone to write a CIP sync driver for EMC2 and someone to make an affordable CIP sync drive.

Good discussion,

We have put together a small work group for EMC2, retrofitting and general motion control, the original I dea was to use EMC2 as a training and teaching aid for the staff on a PPMC and then move to an EttherCat system ie..BeckHoff or SoftServo etc....

After reading this discussion there may be something more substantial that we can contribute...

My experience is as a technician and end user....(i am also financing this venture). By March 1st we will have three systems dedicated to using EMC2, two ME's for I/O controls, one programmer and myself...programmer, coordinator and entrpeneur.

Our end goal is to become automation and industrial control profesionals. CNC machines is the quickest and most readily available technology we can work on.

Test and data aquisition hardware and software is allready in place.
Our shop space is about 700sq feet.
John Elson from Pico System is just 100miles up the road.
We have a good inventory of mechanical hardware and all the tools a fab shop could use.
One retrofit in progress
One orginal design in the books

We would be excited to work on ethernet capabilities of EMC2....

We would need a partner/mentor who is knowledgable in the mentioned technologies and we could cary out the grunt work.

Let me know.