[brccrb]

**Switcher,
This is THE MAN. He beat us to compiling on the Free Microsoft Version. He's been teaching us what is possible!

**CNCOR,
The Control Panel is undergoing some amazing changes. Will probably insert a V1.6.0.4 this week with the new layout while we prepare the analog boards for production.

This new layout wasn't designed by nerds (thank goodness ), but by professional machinists. We just connected the dots and did as we were told.

Yes, we are working on a Mach plugin. You need to have what you are used to as the learning curve is as important as the machine itself!

For the Control Panel though, we are already talking with some people to make plugins (what we've seen so far is amazing stuff!). You will be able to view, try, and buy all from within the Control Panel.

Anyone interested in creating their own wizards for sale, just contact us. We'll help with the framework to promote your product. The source code is open that you will need to interface (so you know what it going on behind the scenes), but your plugin stays private.


--The CNC Brain is available as a 3"x3" board with a single 100pin Hirose connector. This is the OEM version. The 5V power is supplied across the Hirose connector. The USB plugs into the board. This should fit your robotic project nicely

For even smaller robotics applications, it might be possible to purchase chips that can be "chained" for several hundred axis syncronization. For that, we'd need to talk about your application offline ... give us a call .

--THIS IS WAY OFF TOPIC-- but you might find this sort of thing of interest ... from about 7 years ago ... injection mold compatible ... skin, tendons, ligaments, and an artificial muscle (that unfortunately overheats keeping it in the lab for now )

[Al_The_Man]

--The CNC Brain is available as a 3"x3" board with a single 100pin Hirose connector. This is the OEM version. The 5V power is supplied across the Hirose connector. The USB plugs into the board. This should fit your robotic project nicely

The main beef I had with the system suppliers that I purchased from was their efforts to miniaturize their controllers by making their I/O connectors smaller and smaller.
I realize it is probabally cost effective for them, but my point was that machine tools don't really get any smaller, why miniaturize the connector?!
From the likes of Mitsubishi/Fanuc using micro/mini D sockets, and Galil using their 100pin breakout connector. At some point, there has to be a larger break-out point, why not make it sooner, not later down stream?
The micro/mini connectors often require either expensive tooling or high priced pre-made cables.
IMO, offering the much cheaper more practical, standard 'D' or IDC ribbon, allows the direct hookup of items like standard Opto cards and common IDC to terminal breakout's.
I believe for industrial applications, this would be a definite asset.
Al.

[brccrb]

**Al, the Man,
The reason we picked the Hirose for the OEM version was the option for PCB mounting (that actually worked well) and flat ribbon cable (to break out into whatever you wanted). We made sure the power also went across the same port so you didn't have more than one plug.

Here is a link to Digikey for the same plug used on the CNC Brain board (the OEM version is the same small 3"x3" Brain board by itself).

http://search.digikey.com/scripts/Dk...name=H10554-ND

At the bottom of the page, there is a list of mating connectors for both PCB and ribbon.

The OEM version is intended to be used in creating 3rd party conversion "kits", where the Brain board is mounted as a daughter to someone's custom board or on the end of a ribbon cable. The "kit" creators know the machines far better we ever will (we crunch numbers, that is our specialty, they know the machines).

What do you think? It isn't perfect (wish I had your knowledge ), but it seemed logical at design time.

s/Bruce

[Ch_Irawan]

So, if I understand you right, the website needs more videos.

Just one more video please: "CNC Brain in action"

It won't be too difficult right? There is a motor driven machine already...
Understandable if it still in development then a not so perfect action will be ok.

Anyway this thread titled "CNC Brain is Ready!" in "Product Announcements & Manufacturer News" board, and it's already for sale at $499, advertised in this forum.

Sorry if I push you to much - no mean to offend, just very-very curious - like the others perhaps.

Peace, Ichan.

[cncor]

I like the closed loop idea, but one thing I do not understand.

When using a rotary encoder on the axis of stepper motors, you will not be
able to correct backlash or none linearity etc.

For this you would need feedback from your X/Y table. Linear position encoders for instance.
But when using the table position as feedback, you will create a highly complex system. Some sort of PID control will be needed to make it stable.
You want to have a system that reacts fast, but without overshoot. Some parameter values will need to be calculated and to do that, you need to measure the frequency response of the complete system. This is far beyond the possibilities of an avarage CNC user.

How does CNCBrain handles this?

[brccrb]

**Ch_Irawan,
No problem there. Have to machine some parts for a machine this week anyway. Easy enough to shoot (don't expect Hollywood style ).

Actually, these parts would make a good show as a few of them are bearing housings. That would nicely show off the circular interpoltions!

Great idea!

And please, no problem asking questions. We've not been as concerned with the "presentation" as with covering the bases. When we get to a "production version", then we'll get "fancy" .

**CNCOR,
The best sensor configuration is something that is independent of the mechanics for measuring true axis position (glass scale or other linear measurement).

And you are very right about needing a control loop and PIDs have a tendency to overshoot/undershoot badly.

So let's start with the loops (keep in mind all the parts below work in parallel with each other, not sequential ... that might seem a little weird):

Axis:
-Each axis has its own set of motion equations that are redefined by an AutoTuner as the machine runs in closed/double closed loop (too complex otherwise). Their job is to create a "theoretical" equation for the motion of an axis.

-Each axis also has its own "motion adjustment" RLS (recursive least of squares ... again, AutoTuners ... used for capping the motion equation) for position correction while in motion. This helps in the "how do I correct from here to there at this speed under this stress" at an axis level.

-Each axis has its own "non-motion" RLS (recursive least of squares ... again, AutoTuners). This might seem a little odd, but there are separate behaviors for when an axis is moving than when an axis is not. The non-moving correction is intended dampen motion and warrants the added complexity.

-Each axis also has a dynamic backlash calculator. Under different stresses, a machine can have different amounts of backlash (cutting wax doesn't exactly stress a machine ) and can have different amount of backlash at different positions on the table. Unfortunately, while there are some predictive attributes, raw "sledge hammer" processing is needed.

Motor:
-In double closed loop, the sensor on the motor provides information that cannot be attained from the axis sensor alone. The sensor information isn't directly related to the position (the motor vibrates, isn't always in position, etc.). As stress on the machine increases, there are more steps used to move the same distance. To handle this mess, the double closed loop uses statistics to profile motor behavior and works with the dynamic backlash calculator. In the end, the goal is detect "pre-error" conditions and begin correction before changing the axis position path.

Planner:
-Now, the planner has a tough job. Any motion may include one or more axis, so it has to figure out how to interpolate all the motion, define the acceleration boundaries, and help any axis compensate for error in near real time (not to mention deal with mixed open/closed/double closed, stepper/servo axis issues). For that, it has its own set of accel/decel parameters it must create based across all axis (and their equations) involved in a motion and how to get them to the end point at the same time considering the different behaviors of each axis.

-As the machine is corrected, the plan is updated. Something as simple as accel/decel is a dynamic consideration as the speed of the motion may change during the motion ... remember, it has to end on the exact spot across all axis with different profiles so the time required to stop changes as correction occurs (throw in constant velocity issues in here as well).

-The real kicker is that the planner must do this task constantly with consideration of the current state. Raw "sledge hammer" processing is needed.

------So, that is the simplified "30,000 ft view" version of what is going on inside the Brain. The most confusing part is all the major and minor parts work at the same time in parallel.

Now, I'd like to say everything is perfect, but nothing is perfect. We are constantly working on the AutoTuner algorithms improving them, which make up the bulk of our work.

You see, this isn't a single machine with a fixed configuration (that would be so nice ). The Brain supports the "Ebay Configuration" which can have anything matched with anything. All has to work together in harmony.

Surprisingly, it is going pretty well.

Great question! Hope it was answered in enough detail. Any lower detail and you get into the "how" of the Brain.

**All,
One other point that was brought up the other day: "Signal Quality". Any system can have an amount of jitter in the calculation of the interpolation. In integer based systems there is no mechanism to handle partial pulse train motions.

The Brain uses true HUGE floating point signal generation. Also, the signals for each axis are generated by independent processors giving very fine resolution (unshared processing power). Between these two aspects, the signal transistion from one interpolation to the next is very fine.


s/Bruce

[PoolQ]

Just a thought. Is there an input for a spindle index pulse? Is there a use/need for one? Does it matter if someone uses pulley changes for speed instead of a VFD?

[brccrb]

**PoolQ,
Yes, the spindle has a single pulse input (A side only) and an AB input.

This is used to report the spindle RPM (also used to control the PWM for a VFD).

The AB input is used for fine resolution in threading and rigid taping.

Even if your machine doesn't have a VFD, there are some features that can be useful (such as the BitSaver ... slows down the axis if the spindle slows).

Great question!

s/Bruce

[Al_The_Man]

Does it matter if someone uses pulley changes for speed instead of a VFD?

This usually comes into play where the motor speed is commanded by Sxxx M3/4 in a program, the motor command has to be adjusted accordingly to what gear or speed is selected.
In a machine that has a manual change gearbox or belts, then normally there is an input to the control via a manual switch that simply indicates the gear.
Typically a Program Stop is programed and an operator message is displayed to inform the operator what gear to select and also indicate what switch position to turn to, the controller then scales the analogue output accordingly, based on the switch position.
In the case of a automatic gear box, the gear shift mechanism relays this to the controller automatically by limit/sensor switch on the gear shifter, and the output is again scaled.
Obviously the second method is less prone to operator error.
Al.

[brccrb]

**Al,
We have screens in the V1.0.6.4 that have the support for the gear changes. The other versions don't (yes, we were ... h'hum ... informed ... we were told what was needed, how it should look, and how it should work ... then went off and fixed it like good little nerds ... that had been educated the old way (chair) ).

s/Bruce

[jariou]

Bruce,

I discovered your product last week. I read the whole thread over a few days to have a clearer idea of exactly what it is.

I, like switcher, think that the open SDK is the most exciting part. I have downloaded it and installed it on two of my machines and I will be exploring it as time allows in the coming weeks.

The reason I am writing now is to point out what seems to me like something that may become a problem in the software installation. When starting the BrainPanel, the demo machine configuration is created under the installation folder. You might consider changing this to create the machine configuration files under the user documents area.

On one of the machines I installed it, I run as a user without administrative privileges. When installing, I am prompted to provide administrator credentials to copy the files over. After the install, the default user permissions on the install folders are Read, Execute and List but not write.

I suspect that in some industrial or business environment, this would be the default modus operandi. Some administrator installs the application, users use it and their own configuration should be saved under their own file area.

It may be that I can change that code myself to make it behave like this for me. I am not sure, I did not explore the code much yet. However, I explored it enough to see that the path used to copy those configuration files to is built up using the DevicePath property of the CNCBrain activex control. My guess is that that control's behavior is not under my control.

This is not major, just a little stomper possibly for someone with similar configuration to mine.

By the way, as mentioned before, VB Express will build the project so will SharpDevelop at an open source .Net languages IDE. I think that for someone without any VB.Net experience it might be a simpler and perhaps more productive environment to use. Not trying to start a war here, just letting people know of another very good option.

Sorry if I have been a little long winded for my first post.

Congratulations. I am looking forward to follow more of the development of our device in the next weeks and months.

[brccrb]

**jariou,
You made some EXCELLENT points!

The original intent for installing the program directory was for ease in debugging. When a Brain is plugged in, a directory of the serial number is created with the configuration file for the Brain and the Control Panel. This also has any script files, tools, offsets, etc..

It makes it easy to find when on the phone with someone debugging a problem as they typically zip and email the directory. Here, we run a modified Brain that lets us use their serial number.

When in the "documents" section, its harder to find. Will definitely consider the point ... just need to think about how to make debugging easier ... would probably be better to have a menu option inside the Control Panel to automatically send the config.

Like the info about SharpDevelop. Glad to hear it compiles there as well.

Thanks for the suggestions and kind words. Much appreciated!

Keep up the great work!

s/Bruce

[rokag3]

Sorry,
I start to become a little bit suspicious now.
Unless your work it's only virtual (so with no value) You should have a prototype working with some backslash and 2 encoders in closed loop using a servo controller like gecko to test your work

Nobody is asking you to make a superproduction like Bollywood but just a little video of some of your trials (if any)
Lucien

[brccrb]

**rokag,
That was intended to be a joke. Bit of dry humor.

We've spoken with people who said they watched all of them, which our reply is usually, "Wow! Not sure we could sit through all of them"

Sorry if the humor didn't come through.

s/Bruce

[rokag3]

Sorry again

I am not a believer, but just a very primitive sapiens sapiens so please do not loose your time to reply to my mail just send some video and it will prove everything
Lucien

[turmite]

Hi Bruce,

Well I don't want to start anything here, but I am a believer, in more ways than one!

I started to post this at 3am this morning but now will do.

At around 2:30 am this morning it became evident to me once more why I need closed loop control. My machine lost steps, and merrily kept running along like nothing was wrong. I lost nearly all of 48 hrs of programming (yes I know I should have proofed it), but due to a deadline caused by many things, I could not.

I have watched your videos, have sopken with you on the phone and know full well the box works. I will call you later about shipping my little pink brain!

Mike

[alipavsky]

Bruce,
Will the impending Mach3 plug-in be a direct connection between the Brain hardware and Mach3, completely bypassing all of your software and closed-loop functionality? Or will this be a situation where the closed-loop functionality is still happening under the hood, sandwiched between the Brain and Mach3?

Thanks,
Andrey

[cncor]

Bruce,

Can you tell me what the power consumption of the CNCBrain is?

The fact that it contains massive parrallel processing (150 parrallel processors at 50MHz?) implies a very high power consumption.
But seeing the small PCB without any heatsink and the closed enclosure implies a low power consumption .
So wat is the aktuel power consumpiton?

[turmite]

Bruce,
Will the impending Mach3 plug-in be a direct connection between the Brain hardware and Mach3, completely bypassing all of your software and closed-loop functionality? Or will this be a situation where the closed-loop functionality is still happening under the hood, sandwiched between the Brain and Mach3?

Thanks,
Andrey

Hi Andrey,

I'm not Bruce but I can answer that one for you. The closing of the loop happens inside the Brain so the interface, whether Mach or Control Panel, is a GUI to allow us mere humans the ability to use the machine!

Mike

[Milo]

Hi Andrey,
To add to Turmite's post. Think of the CNC-Brain as a breakout board that connects to the PC through a USB port with built in intelligences. Bruce correct me if I'm wrong, It would process information from Mach 3 the same way it would from the Control panel.

Milo