CNC Brain is ready! 6-axis Double Closed Loop Controller

[ger21]

Originally Posted by brccrb **Ger When power is off, it doesn't run (duh ) But with the Position Reference, you can power on a machine, move the axis back and forth a short distance, and be back of position within a click (5 micron in standard scales). Position Reference lets you use incremental scales as absolute encoders in recovering position. If you run 4+ axis, setup is made simple (an otherwise tough task since every bit of error is multiplied as it rotates).

So for Mikes example, recovering position after a power loss, he'll need to have expensive scales, correct? That's what I was trying to say, that he won't be able to do what he thinks just using standard encoders on the motors or screws.

[turmite]

Originally Posted by ger21 So for Mikes example, recovering position after a power loss, he'll need to have expensive scales, correct? That's what I was trying to say, that he won't be able to do what he thinks just using standard encoders on the motors or screws.

Hi Gerry,

I don't know how to start this post, so I'll just take a breath and jump in.

I have spent many hrs on the phone with Bruce since he first posted on the Zone way back when! I saw what I believed was a great potential for the small shop or home shop guy to be able to have the same kind of accuracy and repeatability that the guys like you for instance, have access to, on those 100,000 machines. It may or may not be my place to tell you what I know about Bruce, but I will give you a hint. He "ain't" average. Bruce, can I tell em?

The very first ph call I had with Bruce was in fact about the linear scales and the fact that on a machine like mine, they would be at least $20,000! I told him that would immediately put most potential customers out of the picture. He ask me if $75 would be too much to pay for a scale that would give .001" accuracy in about 40"? Well?

I cannot reveal all I know, but I have told Bruce to write an overview just for the Zone here, and get up asap. The reason is to answer questions like you have ask. I actually think you and I think a lot alike!

Yes, this is going to be a little more expensive than a typical build, but not by much. Size the motors right, make sure they are dual ended shafts, get the rotary encoders and make or buy when they become available the inexpensive linear scales and the the CNCBrain. The only real added cost will be the cost of the CNCBrain, and the encloders, and in my humble opinion, a small price to pay for what will be gained.

I used to be an advocate of the opend loop system that was sized right because of the cost and the simplicity. After seeing the amount of $$ that system has cost me in lost time and ruined parts, this step is going to be small change.

Mike

[brccrb]

Ger,
Actually, he can, with the Position Reference functionality.

This can be used an absolute encoder! Let me explain:

A single sensor is placed on the axis. Along the axis (at intervals like 0.25", 0.25" + 0.05", 0.25" + 0.1", etc ... or much larger spacing) are magnets (you can use optical other otherwise).

When it passes over three of these references, the position is found through a simple math trick (unique pattern). As his machine is moving, it constantly "re-references" the position.

So, with a couple of magnets and a rotary encoder, he can keep position and recover quickly after power is off.

Now, if you already have or plan to purchase scales with a reference position, then that is fine too ... more accurate ... and no magents to deal with.

There is also another option in creating linear axis encoders:
------------------------------------------------------------
Plastic gear racks, two small gears, and a rotary encoder.

The gear racks (fine tooth, straight, molded plastic, about $2 per foot) are made to fit end-to-end. This makes a linear track (mount under L angle to keep from getting dirty).

A rotary encoder (such as US Digital E2-1000) has a small hollow shaft put through it with a small diamter gear pressed on it. A shaft, with an identical gear slides into the hollow shaft letting the gears sit next to each other. A small spring on the other end makes the assembly a small backlash gear.

The gears, with tension from the spring, ride the track.

It is much more accurate than you first think. It isn't as accurate as glass scales, obviously. But it isn't bad at all.

Here is a picture of a rotary version (for 4th+ axis) to give you an idea ... think of this, just straight (this uses a backlash gear as the rack ... for linear the backlash gear would be on the encoder):






At 12in, it is within 0.001" on a rotary (which is obviously more delicate than a straight axis).


s/Bruce

Originally Posted by ger21 So for Mikes example, recovering position after a power loss, he'll need to have expensive scales, correct? That's what I was trying to say, that he won't be able to do what he thinks just using standard encoders on the motors or screws.

[sandefuj]

What about your CAM post processor? Almost all CAM programs make a post processor for MACH; do you think the BRAIN could use the MACH post processor?

[brccrb]

Sandefuj,
Excellent question! Most excellent! Yes, we can. The G-Code was based on the RS274 standard ... when it didn't cover a topic, used the "CNC Machining Handbook" by James Madison from Industrial Press, Inc.

There are features that are unique to the Brain (mainly for use in script), so those are unique.

We've tested the post processor from MadCAM for both the ISO-G and Mach, which both worked fine.

Since the interpreter is open (in the GUI source code), we can tailor it to support multiple formats (which is the intent ... where possible).

If you have a particular post processor you want to use, send us an example.

**Now, we can have some fun with this interpreter a bit (the CNC Brain deals in absolute positions and all the messy motion control math ... so the interpreter just takes G-Code and gives the Brain final positions ... simplifies everything greatly!!!). Nothing is stopping us from taking it to another level.

The only limit is the imagination!

For example, nothing is stopping someone from adding to the interpreter more functionality, such as inserting conditional statements written in basic (or other language) directly into the G-Code itself.

There is nothing stopping us from turning the GUI into a true Kinematic system.

In the CNC Brain, we are adding support for elliptical arcs and eventually b-splines. So, pretty soon you can describe with one line what 10,000 lines of G-Code takes now to do. All of this is available to the interpreter as a "do this" command.

For example (one of my favorites ... it will be apparent the elliptical and b-spline support): http://www.safeguardrobotics.com/for...osts.asp?TID=8

s/Bruce

Originally Posted by sandefuj What about your CAM post processor? Almost all CAM programs make a post processor for MACH; do you think the BRAIN could use the MACH post processor?

[Khalid]

The more i knew the more i amazed
I wish to be a beta tester of this software for my new machine

[Khalid]

3 days ago i can't think we can get a cheap software...I always think of Mach3 that ART introduced to us ...He did us a great favour,indeed...

[brccrb]

Khalid,
Yes, we all owe Art (and Brian who continues his legacy) a big "Thanks". He's a pioneer.

Hopefully, we can continue in Art's spirit of "innovation for the rest of us" in leveling the playing field for everyone to create. That is our whole purpose, the next level.

I want to clear up one point (sorry if I've been vague), the CNC Brain is a hardware device. It come with free software and an SDK. The CNC Brain is alot like a game engine (a physics engine), but for CNC Machines

But again, you've said what we all have thought, and said to each other, for so long, "Thanks Art".

s/Bruce

Originally Posted by Khalid 3 days ago i can't think we can get a cheap software...I always think of Mach3 that ART introduced to us ...He did us a great favour,indeed...

[Zathras]

Bruce,

I PM'd a few comments to you earlier that included this question below. I'll post it for others to see, where it should be.

There is one main thing(s) I would like to ask about. There is (so far) no information on how to interface the hardware such as the motor drivers.
The stepper drivers are trivial, no confusion there. However, I intend later to use servos on my hexapod. How does the CNCBrain interface, say with, a Gecko servo controller? In particular, now do we have three servo loops to deal with? What would happen with the F/B from the motor to Gecko with respect to the overall control of an actuator? Would the inner Gecko servo loop "fight" the CNCBrain under some circumstances? How do you tune the CNCBrain servo loop or is it automatic/unnecessary?
Can the CNCBrain interface directly with a servo amplifier or is it strictly step/dir output?
I assume you have heavily based it on FPGA so could the hardware be reconfigured (assuming enough pins) to produce desired control signals?

Very intriguing and unique product, I'm impressed. Looking forward to more from Safeguard Robotics!

Keep up the great work Bruce.

Regards,
Mark

P.S.
Thanks for the Hexapod tute you promised me! Grunt Grunt

[brccrb]

Mark,
Yes, excellent question! Sorry for not being more clear.

And the "Grunt grunt" has me in laughing tears ... whew! ... ok, (for everyone's reference ...http://www.safeguardrobotics.com/for...osts.asp?TID=8)

OK, the output of the CNC Brain is currently Step/Dir (will be adding more in the future for directly attaching to a servo amp).

The Brain wants, if you have it, the same encoder values that comes from your servo that also goes to your servo driver (the encoder on your servo). It wants to know what the servo is doing.

The Brain doesn't try to fight with the servo controller. It sees the "big picture" while the servo controller is only looking local. So what the Brain does is when it sees a problem beginning, it coordinates all the signals to compensate ... this include accelerating/decelerating as needed ... which in the case of servo controller, lowering/raising the bar of what it needs to meet.

In other words, it tries to work with the servo controller, but can see what might happen and begin correcting the problem across all axis so it doesn't loose position.

For tuning, it just wants a basic premis: Acceleration and Top speed. Think of this as driving someone else's car. It works basically the same as your car, but it behaves a little different (more/less mass, v8, v4, etc. ... like a hint "hey, the clutch sticks"). After that, the processing takes over. In the axis setup tab, there is an option for "Stepper" or "Servo".

Does that make sense? Sometimes I'm standing too close to the chalkboard to see the question.

** Next, yes, you have a good insight. Yes, the device can be rewired remotely. But, the MPP Framework ... uh, this is going to be a little odd ... the MPP is what generates the matrix, writes the code, and even generates the netlist to build the PCBs. It can variably span enough silicon to do the job ... yes even the 60+ to process video ... and can process and exchange data between the 140+ processors every clock (yes, I know it typically takes three clocks to exchange data even between two processors, let alone process it, but it is ... unique).

For some quick history, I built my first parallel processing computer in 1983 .. two Motorola 6809E, over clocked at a whopping 0.6Mhz running a hacked version of OS9. Since then, I've done a "few" things with parallel processing.

** The CNC Brain is a sincere creation. The idea is to help everyone with the interface between the virtual world of the computer and the real world.

The CNC Brain has a tough job. It isn't expecting to run on a 20,000lb rigid piece of equipment with 480V servos (that would be SO much easier). It expects to be running on something as small as a desktop mill and try to give it the performance of a high priced machine.

s/Bruce

Originally Posted by Zathras Bruce, I PM'd a few comments to you earlier that included this question below. I'll post it for others to see, where it should be. There is one main thing(s) I would like to ask about. There is (so far) no information on how to interface the hardware such as the motor drivers. The stepper drivers are trivial, no confusion there. However, I intend later to use servos on my hexapod. How does the CNCBrain interface, say with, a Gecko servo controller? In particular, now do we have three servo loops to deal with? What would happen with the F/B from the motor to Gecko with respect to the overall control of an actuator? Would the inner Gecko servo loop "fight" the CNCBrain under some circumstances? How do you tune the CNCBrain servo loop or is it automatic/unnecessary? Can the CNCBrain interface directly with a servo amplifier or is it strictly step/dir output? I assume you have heavily based it on FPGA so could the hardware be reconfigured (assuming enough pins) to produce desired control signals? Very intriguing and unique product, I'm impressed. Looking forward to more from Safeguard Robotics! Keep up the great work Bruce. Regards, Mark P.S. Thanks for the Hexapod tute you promised me! Grunt Grunt

[Switcher]

brccrb, For example, nothing is stopping someone from adding to the interpreter more functionality, such as inserting conditional statements written in basic (or other language) directly into the G-Code itself.

So I could run statements like (mixed in with my g-code):

FOR
(some code)
ENDFOR

IF
(some code)
ENDIF

If thats the case, looks like your running neck & neck with Siemens 840D, thats a very big deal to me!

Anyone that has never ran Siemens, conditional statements make cnc so much fun.

---------------------------------------------------------------------

As far as your SDK goes, I could see a bunch of options for small OEM to write conversational programs, I run one on a 5-axis cnc toolgrinder, from Schutte, It's cool because I can tweak any of the g-code, thanks to Siemens.



.

[brccrb]

To be filed under (near) future enhancements:

Basic function would be a minor change to the interpreter. Would that be something you would want??

The way the CNC Brain was built makes this sort of thing easy. We can keep it compatible with the current G-Code standards so CAM post-processors aren't a problem.

Also adding elliptical arcs and b-splines. Not sure what G number to give it (suggestions taken). With this, we would be setup to take files, like Rhino, and use the lines as tool paths (in other words, Draw the tool path in Rhino from line segments, the Brain would read the file directly and use it like G-Code).

Thoughts?

Originally Posted by Switcher So I could run statements like (mixed in with my g-code): FOR (some code) ENDFOR IF (some code) ENDIF If thats the case, looks like your running neck & neck with Siemens 840D, thats a very big deal to me! Anyone that has never ran Siemens, conditional statements make cnc so much fun. --------------------------------------------------------------------- As far as your SDK goes, I could see a bunch of options for small OEM to write conversational programs, I run one on a 5-axis cnc toolgrinder, from Schutte, It's cool because I can tweak any of the g-code, thanks to Siemens. .