Why is this so hard and expensive? I view this question as a huge business
opportunity. The cost and time involved in setting up a CNC system is MUCH
more than it should be. Fix that problem and you'll be rich, The cost
could be reduced by maybe a factor of 8x. Customers would be happy to see
a 4X reduction and you pocket the difference.
I am in my other hobby building robots. It is just like CNC, really except
more axis and the motions are not pre-programed.
So on my desk right now are four DC motors with quadrature encoders. They
are literally strapped to a bread board. I'm driving them with about $15
worth of electronics. I'm doing motion planning and using ultrasonic
"limit switches" It needn't be expensive. Machine tools electronics
tend to be expensive only because people say "What the heck it's a $40,000
mill what does $2,000 worth of electronics add to the total cost?" But
when you look at what you really need, a micro controller and some 10 cent
MOSFET switches it comes to under $20.
Part of the cost comes from the fact that we (especially those building up
systems from LinuxCNC/Machinekit) are using a set of lego blocks that don't
match. Some blocks of square bumps and round hole and they don't lock
together so we spend time making adapters. For example your computer
makes step and direction pulses but a bipolar stepping motor wants to see
voltages on it's four lead wires. Why not have the computer computer the
That said if you can bill customers at an hourly rate it is well worth
$2,000 or more to get the project moved forward and working by a few days.
But a hobbyist is not turning away one hour of billable time for every
hour he fusses with his mill.
So, if this is a hobby and some part is expensive, skip it. There are
always lower cost DIY alternatives. You can trade cost for time. Or in
my case I care a lot about size rune weight and power
An example. Decoding quadrate signals. All of the ARM chips made by ST
Micro have quadrature decoding hardware on-chip. Typically four of more
sets. These will completely solve the problem in hardware and the chips
cost under $3, or $13 if you buy them on a usable PCB that has herders, USB
connectors and so on. Mesa boards cost about $100 give or take but you can
buy the same FPGA on a generic brand board for $13. What happens is that
when it is marketed as a machine tool component the price goes up by a
factor of at least 8X.
One should be able to build an integrated machine controller for about $150
that does visualization and controls four motor axis. The visualization
and controls could live on a iPad or phone and the real time control on a
small uP or FPGA. A modern tablet or phone is more then enough to run
Machine kit or the like and a $3 ARM M4 chip can do real-time enough for a
multi-axis industrial machine.
Such a product would be "disruptive" in the marketplace.
An example of someone who is set to make a mint in the 3D printer industry
has come up with a way to reduce the cost and complexity by about 4X.
First, the controller is a normal cell phone that you lay on your desk face
up. The printer is stacked over the phone. The communication path is the
light on the cellphone's screen. An LCD screen is a VERY high bandwidth
channel. and moving the "brains" out or your product saves the cost if
providing the "brains". Some other cost saving features are using just
one motor rather then the normal three motors and making the unit from
injection molded plastic rather than machined metal. They can retail tiny
size printer for about $100.
Point is that there is a huge opportunity every thine you see something
that is more complex and expensive than it needs to be. Building a CNC
machine is 100% full into this category.
Uber of course was another example. Why place a complex taxi meter in a
taxi cab when GPS software can compute the distance and time? Why hire a
dispatch office when electronic messaging can do the job with no humans.
Why even buy taxi cabs when every one of your employees already owns a car?
Why handle cash and credit cards when we can do electronic billing? They
got the cost down to nearly zero and they pocket most of the savings.
Post by Gene Heskett Post by Jon Elson Post by Gene Heskett
What x86 board, suitably small but not that outpriced yudoo thingy,
will run linuxcnc well AND can do the spi thing?
Cheers, Gene Heskett
Well, the various versions of the Beagle Bone run the
Machinekit fork of LinuxCNC fairly well.
The Bone has a lot more I/O available, including hardware
I don't know any X86 boards that have built-in SPI available
The killer in the beaglebone soup is the cost of the "capes". You've over
$200 plus psu's etc before it can turn 2 motors at the same time. I
don't begrudge the guys trying to make enough to buy their beer, but a
7i90 for a bit over $60, simply blows most of the capes I've looked up
into the next drainage.
The spi bus out of the pi runs at 32 megabaud, sending 4 byte packets 4
bytes at a time with an extra stop bit between 8 bit bytes. Thats the
equ of a 32bit word at 4 megahertz. The parport would have to run at 8
megabytes/second to match that amount of data in that same amount of
time. I think... At any rate, that's an order of magnitude faster than
the parport can legally do it, it just does not have the ability to
deliver the slew rates down the cable that would need. 10% of whats
needed is all that is legally available Does linuxcnc have the ability
to work over that slow a com channel? Seems like that would be pushing
Post by Jon Elson
-------- Check out the vibrant tech community on one of the world's
most engaging tech sites, Slashdot.org! http://sdm.link/slashdot
Emc-users mailing list
Cheers, Gene Heskett
soap, ballot, jury, and ammo. Please use in that order."
-Ed Howdershelt (Author)
Genes Web page <http://geneslinuxbox.net:6309/gene>
Check out the vibrant tech community on one of the world's most
engaging tech sites, Slashdot.org! http://sdm.link/slashdot
Emc-users mailing list
Redondo Beach, California