Crazy New Idea

[Mariss Freimanis]

Hi,

OK, after a couple of years crying, I've gotten over the G100 thing. I'm ready to go after another nutty controller idea.:-)

Imagine you have a contract to build 1,000 gizmos. You write up the G-code, test it and it works like it's supposed to. Could be something you do on a router, mill or a lathe. Maybe it's even a CNC sewing machine; it doesn't matter. What matters is you need to build a lot of them.

After you proof your CNC program, (could be Mach3, EMC2, WinCNC, Smoothstepper, whatever), you insert an SD memory card, plug your CNC program step and direction output connector and press the "Record" button on a Super G540. The Super G540 records your CNC program 4 axis step and direction outputs for programs as long as 24 hours.

Once done recording, you unplug your PC from the Super G540. The PC isn't needed anymore. Press the "Play" button and it faithfully plays back your recorded CNC program without a computer attached. Remove the finished part, put in new material and press "Play" again. Repeat 1,000 times.

The program is all saved on the SD card. You have an existing product that has a re-order from a year ago? Plug in that product's SD card and press "Play". Off it goes. You have 10 machines to turn out 1,000 gizmos? Each one runs a copy of the same SD card. Divide and conquer.

Think of it as an i-pod mp3 player except it plays back CNC programs instead of music. Just like you have music play lists for your i-pod you could have CNC program "play lists". Just like music, it doesn't care who the "artist" is, it doesn't care what CNC program generated the step and direction outputs; it can record anything, even MPG signals. The "song" can be a quarter second long or 24 hours long.:-)

What do you think? The idea stinks or it has legs?

Mariss

P.S. I didn't lay out the obvious like file structure, canned Home routines or any of the other details because it would have interfered with describing the core of the idea.

Similar Threads:

• Crazy Idea
• Build Thread- Crazy Idea - 'X-Saw'
• New crazy idea....

[TOTALLYRC]

Hi Mariss, I see this as a competitor to a plc for many applications.
A couple of things jump out at me. If you need to make a change down the road, you would have to record the "song" again. Not the end of the world but if it is several hours long, it would be tedious at best.

Once you have it recorded, I would assume that the SD card can be copied by a computer, this way if you need ten cards you don't have to record it ten times.

I think that in its current format as I understand the way it works it would have limited use amongst the diy CNC crowd. I do realize that you are not only in the hobby business but are spread out into many other fields.

It also looks like you would have no way for accounting for tool wear on a high tolerance part.

Of course one of the enterprising people here will write a program that will convert G-code directly to the file format that you are using and that will save the recording time and will make editing a file easier than rerecording.

As a way to run a CNC sewing machine or other repetitious task were there is not a tool wear problem it has merit but I would check to see how it is already done.


Not trying to rain on your parade but you did ask for ideas.

I too was disappointed that the G100 didn't live up to its potential but it is what it is.
It would have been a world beater if it had worked.


Mike.

[KIMFAB]

I would also be a little leery of the longevity of the data on the SD card.
I know they are reliable but I would want some kind of error checking routine and probably a hard backup.
I think it would work great tho for something like laser engraving or something that would repeat on a regular basis.

[Geof]

Haven't you just invented a solid state "paper tape reader".

I seem to recall back in the good old days a G-code program was stored in the form of numerous holes on a long strip of paper tape; which could be read over and over again.

Mind you a paper tape that took 24 hours to read would need reeeaaallllllyyy big tape reels.

[dertsap]

i think that it's a great idea for repetitive applications and would probably compete well against companies such as omron who have the stand alone servo controls that run on g code and are used in factory automation or what have you . But I kind of have some doubt how well it will fly in the cnc world , there are a number of variables in machining that need to be addressed such as setting the occasional d comp , tool height setting , and other things . I think that most diy systems are hobby systems that generally see 1 offs or a few parts at a time , there are those out there who are using their setup to pump out parts but I think those numbers are far less than the other .
the sky is the limit as far as the possibilities in using that type of setup , and that would probably be ideal for a machine setup that has been whirl pooling in my head . its for a simple part that we make thousands of and the last op is painful to watch , I have the mechanical design in my head but its a matter of making it work with the right components . I've recently talked with a friend about the stand alone g code controller , he sells omron and other automation equipment and the stuff ain't cheap but in the big scheme of things it's still a small price to pay
a stepper setup would be far cheaper but is there a way to ensure stability under certain loads or applications , is there a way to setup feedback similar as what you can with servo's

[Chris D]

Interesting, but not for CNC applications. You would be recording GPIOs as well as step and direction information only based on your description. That would be fine for something that runs without needing adjustments. CNC machines utilize offsets to allow small adjustments to be made to the tool path to maintian part accuracy. For something crude like a plasma or possible a wood router, you might be able to get away without any adjustments. For precision machining, forget it. Pick and place type machines would be fine too I believe.

Furthermore, a re-run of an older program stored on an SD disk has another problem. The tool lengths would have to be identical to that originally used in the setup. The parts relative location on the machine (from the home position) would also have to be identical. If there isn't any way to adjust for these, the device would be very good at making a 1000 scrap parts :-(

In the right application though I don't see why it wouldn't work - just not precision machining.

Chris

[Andre' B]

I see problems with changing out broken and dull tools, making offset changes to keep the parts in tolerance etc..

Most of the time when a tool needs changing we wait until tool just in front of it is running then turn on the op stop. When the machine stops with the dead tool in the spindle flip into handle mode remove the tool, put the new tool in the spindle enter the new offset (measured off line) flip back to run single block, push cycle start a few times until you are sure the tool is going where it is suppose to turn off single block and let it go.

Edit:
Reading Chris's message, if you have 4 or more router and P&P type machines it could maybe make sense. Below some number of machines it would likely be cheaper to just have a computer at each one.

[Andre' B]

If you want a new project.
I think you should look at a module that plugs in between the computer serial port and one of your step/direction drives and is intended to run an indexer. Where the CNC can send comands out the serial port tell in the indexer what to do. Examples of commerial units would be Haas, Yuasa etc..
http://www.yuasa-intl.com/images/pro...rsInderxer.pdf

Some units you can program a number of steps into and then an M code output at the CNC (a set of relay contacts) is used to tell the indexer to move to the next step, each step would send the finish signal back when the move was started or when it is ended depending on how it was programmed.
This type will work with CNC machines that do not have a working serial port.

[Geof]

Question for you naysayers:

Is my "paper tape" analogy incorrect?

I am not a computer or software wizard so maybe I am misunderstanding the concept but it seems to me all that is being recorded are a bunch of coordinate positions, and speeds and feeds, which is what the holes in a paper tape defined. So how did NC machines reading tapes handle offsets and things like that.

[ger21]

Question for you naysayers:

Is my "paper tape" analogy incorrect?

I think so.

Isn't this how the Carvewright works? Just checked, and yes, it is.
http://www.carvewright.com/cms/machine

You'd need to write a simple app to convert G-code into your format for the SD card. So just open the g-code file, and send it to the SD card.

However, you'd likely need jogging abilities, and probably some sort of display like the Carvewright uses.

[ger21]

And how about getting the stepper-servo drive done first.

[Andre' B]

Question for you naysayers:

Is my "paper tape" analogy incorrect?

I am not a computer or software wizard so maybe I am misunderstanding the concept but it seems to me all that is being recorded are a bunch of coordinate positions, and speeds and feeds, which is what the holes in a paper tape defined. So how did NC machines reading tapes handle offsets and things like that.

Paper tape NC machines were before my time, but I assume the offsets were added to the commanded positions. Note there was still something like a computer translating the g code into motion commands.

What Mariss is taking about is recording each encoder step and the time at which it is to take place. That is a lot more info then was on those paper tapes.
I do not think he would be encoding feed rates, but you could do some compression on the recorded data (not unlike that done for music in an I-Pod) and get such info. In effect converting it back to something like G code.

[MrWild]

Killer robotics app. The baking/cooking biz is full of robots that do this very thing. Automakers have welding bots that do repetitious tasks. I know of one Modine intercooler manufacturing application that this would work well on. This could well be a "Build it, and they will come." It might not fly with CNC machinists, but there are a LOT of applications it would fly on.

[Al_The_Man]

Question for you naysayers:

Is my "paper tape" analogy incorrect?

.

The early NC machines did not read the whole tape in as did the later versions that had memory storage to load the P.T. into.
NC had no real memory so the commands were entered and ran a line at a time.
The tape was stepped through.
But in some machines, tool offsets could still be stored to offset a particular T code.
So I guess, yes your analogy is the same to some degree.
Al.

[Chris D]

Paper tape analogy - yes and no, but mostly no.

Marris is talking about recording the step and direction pulses as delivered by some sort of application which is connected to a stepper drive. So, all he would be recording is the step frequency and direction setting. It would not be recording the raw motion control data such as G-codes.

The stepper drive (the G540 in his example) does not do any motion translation etc. It only accepts step and direction signals which it then converts to a step on the servo or stepper motor. The step signals are nothing more than a very short pulse that is precisly timed to create a velocity. The direction signal is either on or off to select the direction of motion. The stepper drive never sees anything like a number or letter etc.

A paper tape actually contained all of the commands in a CNC program. Each character (letter or number) was encoded into a very simple binary format (the series of holes across the tape). On NC machines, there was no memory for the program to reside within, so the paper tape provided each instruction one-by-one. Offsets were "added" to the motions as the CNC commands were translated in realtime.

Around the early 1980's, memory was added to controls and the generic name changed from NC (Numerical Control) to CNC (Computer Numerical Control). At that time, paper tape was still used to load the program into the memory. Once there, the CNC control would execute the program from memory as it does today. Memory wasn't so big in those days though so sometimes you still had to run from tape.

RS-232 (serial communications) wasn't applied to CNC controls till the mid 1980s. With RS-232 ports on CNC machines, we could then use a PC to transmit the CNC program into the memory which was essentially an electronic version of what the paper tape was doing.

Serial communications is utilized in two different methods today and both are referred to as DNC.

DNC #1 = Distributed Numerical Control
DNC #2 = Direct Numerical Control

Distributed numerical control refers to an architecture wherin you have many machines on the floor and a host computer sends a program into the control's memory for execution. In this architecture, the programs are almost always small enough to fit in the control's memory.

Direct numerical control is much more closely related to the early NC machines. The program is sent to the machine, but the control only accepts a few blocks of data then sends a "hold" signal to the host computer. As the CNC machine runs the program and the buffer empties, a "Go" signal is sent to the host and the program transmission resumes. Only a few block of code actually reside in the control at any point in time. The flow control codes are usually done with two methods

Hardware handshaking with RTS and CTS signals within the RS-232 port
Software handshaking with X-ON X-OFF data signals

You will often read or hear people talking about "Drip feeding" programs. This is actually Direct Numerical Control and as explained is very similar to using paper tape.

Not all CNC machines can perform direct numerical control. On Fanuc controls, the usual indication is if the Mode selector switch has a "TAPE" mode. Once the communication link is configured to access the serial port for "Tape mode", the machine runs very much like a very old NC machine.

Hope that clears things up a bit :-)

Chris

[ee_t]

The only advantage really is that "the PC isn't needed anymore". I don't think this matters to the majority of people who use a dedicated PC to control their CNC.

I do see what you're getting at, but IMO this is very niche (someone who has lots of DIY machines used for production).

And how about getting the stepper-servo drive done first.

Would it be overly optimistic to assume the stepper-servo is 100% complete and he has already moved on to a new project?

[joshendy]

Im thinking G540 with a ARM microprocessor behind the scene. LCD screen with mini fully programmable cnc controller interface (via usb), reads g-code via SD card or onboard memory.

No computer needed at all The super 'crazy insane' G540 edition, it's your 4 axis cnc controller and computer built into one. Id take one.

[BobWarfield]

The paper tape analogy may be apt, but it doesn't necessarily mean a successful product. You'd have to go see which machines and applications can run without UI like that. A lot of reasons have been brought up why most CNC isn't in that category. Paper tape hasn't exactly survived intact, even with more modern media.

But that does not mean there aren't still a lot of interesting fires to drive the fire truck to!

Start with the PC world. Parallel ports are going away and Windows is not a real time OS. We need a USB solution backed by a company with the stature of Gecko.

I'd go make a deal with the Smoothstepper guy. Get him on the payroll full time so there is no possibility of any flakiness down the road. Make sure the Mach guys are totally bought into it. Line up some EMC guys to do a driver while you're at it.

Start off with a standalone Gecko-ized Smoothstepper and follow up with a G540USB. Down the road if you want applications where a single PC controls lots of machines, upgrade Smoothstepper to sit on a network and get the Mach3 (or whatever) drivers to talk to a whole raft of them from a single Mach3 session.

That'd be one approach, and I'd wager it would pretty quickly take over because its cheap, its easy, and there is a good sized ecosystem supporting it already. The future is pretty well driving that way.

Then there is another kind of play. It's all about what language will be used to talk to these controllers. Recorded step/dir is just too low level. Smoothstepper is a notch up from that. There is an argument to be made to take g-code directly and interpret it. That doesn't mean we have to have UI or LCD screens on the little boxes. But, if we can feed them g-code and they can run it, our UI's on the PC are easy to write, and they can deal with the tool offsets and all the rest.

This is another path the future is driving to. There are already Arduino controllers in the Reprap community that can execute a subset of g-code. If you do the heavy lifting to run the g-code well on the controller, the UI is pretty easy to write too. And BTW, a subset would take you pretty far. No need for most of the language, just the basic motion stuff.

The nice thing about this model is g-code is a lot more universal than Smoothstepper's API's. There would be a much bigger ecosystem and market available in the long run. The short run pain is that the world wouldn't be ready for it the day it was released. It would be a slower build to a potentially bigger finish. But this is a pretty good sized software project to tackle. It probably needs 2 or 3 software engineers or one maniac who is insanely good.

Cheers,

BW

[rowbare]

Then there is another kind of play. It's all about what language will be used to talk to these controllers. Recorded step/dir is just too low level. Smoothstepper is a notch up from that. There is an argument to be made to take g-code directly and interpret it. That doesn't mean we have to have UI or LCD screens on the little boxes. But, if we can feed them g-code and they can run it, our UI's on the PC are easy to write, and they can deal with the tool offsets and all the rest.

That sounds a bit like CNC Brain doesn't it? I hope that Bruce finally gets some cycles to get back to development there. I remember Marris mentioning that he had spoken to Bruce but that there had been no follow up. Wouldn't something like a CNC Brain integrated with G250 drives (or a servo equivalent) be sweet!

bob

[WayneHill]

Possible uses would be:

Robotic Welding / Painting.

Is the 'record' gathering info from the step and direction of each axis?

Paper tape, Yes I remember punching paper tape program for a 1972 vintage control. Logged many long hours on both a Friden Flexowriter and a ASR-33 Teletype.