Mesa/Pico/PMDX

I apologize for rehasing a topic that has been beaten to death but I've done a ton of reading and haven't found the answer to a very specific question.

Will a Mesa/Pico card provide better performance than a PMDX-126 BoB in terms of either step speed or reliability?

Other people building the design I've chosen go for PMDX/Gecko/Steppers/Mach almost exclusively. I want to go with LinuxCNC because I'm a linux guy but I'll stick with Gecko G203v drivers, steppers, and the rest of the gantry system.

You might want to look here...
www.linuxcnc.org/index.php/english/forum...en-using-a-mesa-card

But first things first. Start with the computer you are going to use as test the latency. This will determine the maximum step rate you can use.
Then go to your hardware, can you achieve the step rate you need to go the speeds you want at the micro stepping rate / gearing you are going to use?

Also look here...
linuxcnc.org/docs/html/man/man9/stepgen.9.html
wiki.linuxcnc.org/cgi-bin/wiki.pl?Tweaki...ftwareStepGeneration
www.mesanet.com/fpgacardinfo.html

Rick G

Perfect! Thanks!

Tom wrote:

I apologize for rehasing a topic that has been beaten to death but I've done a ton of reading and haven't found the answer to a very specific question.

Will a Mesa/Pico card provide better performance than a PMDX-126 BoB in terms of either step speed or reliability?

Reliability? No, probably not, except in a vary narrow sense. Step speed? Yes, no question!
Our (Pico) stepper controller can go to 300,000 steps/second, and give much smoother
trainst of pulses than with software step generation. So, you can probably go about
10 X faster with an external step generator. Also, due to the smoother step pulse
timing, there is much less likelihood of the motors stalling or skipping steps.

Other people building the design I've chosen go for PMDX/Gecko/Steppers/Mach almost exclusively. I want to go with LinuxCNC because I'm a linux guy but I'll stick with Gecko G203v drivers, steppers, and the rest of the gantry system.

Gecko drives are VERY good! But, due to the 10 X microstepping, they need 2000 pulses/rev.
Especially with routers, but also with other machines where you want fast rapid feed rates, this
requires a LOT of step pulses/second. 300,000 steps/sec divided by 2000 gives 150 RPS
or 9000 RPM, so the motors become the limiting factor, you'll never run out of step rate.

Jon

Thank you, Jon. Yes, I'm new to this so your spoon feeding is really appreciated. I have a few more questions that have arisen from these answers.

How is the Pico USC different from a smooth stepper? I thought the LinuxCNC crowd was against any sort of external timing control? This question is just to help me understand. I am actively seeking out a solution that works and, right now, that appears to be a Pico USC.

It appears the Pico USC connects to a parallel port. Would it help the I/O rate to use a Mesa 6i25 E-I/O card (or similar) to drive the Pico? Perhaps not needed?

I was considering using Gecko G213v drivers, instead of G203v drivers, so I could increase the rapids. The CNC router isn't done so I have no experience with what will be needed. Perhaps the G213v drivers aren't necessary, if the Pico USC can drive 10x faster than directly connecting to a BoB?

One more question. What is the best way for a Canadian, like myself, to obtain a Pico USC (and Gecko G203v/G213v drivers, for that matter)? Do you ship via USPS?

I should note that there are at least 4 commonly used hardware step generation solutions for
Linuxcnc. These all do the same thing the the SmoothStepper does: offload high speed
step generation to separate hardware or a processor.

Pico Systems
FPGA based, parallel port interface

Mesa Electronics
FPGA based, PCI/PCIE/Parallel port/Ethernet interface

BBB
Beagle Bone Black uses its built in PRU (Programmable Realtime Unit)
for high speed step generation.

General Mechatronics
FPGA based, PCI interface


All of these systems have the advantage of generating fast and low jitter pulse streams
which get the most out of the stepmotor/drive system.

There are a couple more that I have heard about:

Pluto FPGA card
FPGA based parallel interface

RT-8P8C
Processor based Ethernet interface

Let's see if I have this correct.

I could source a Pico Universal Stepper Controller and 4 x Gecko G203v/G213v (has 2 * X-axis steppers). The Pico could drive both x-axis steppers by simply connecting x step and direction to both G213v drivers.

It appears the Pico USC could control the speed of my Hitachi M12VC router with a spindle DAC board and a SuperPID.

The USC appears to have inputs for home, limit, and e-stop switches.


OK. What about a pendant? Could that be done with a second EPP or would it make sense to go for a Mesa 6i25 so I have plenty of I/O? As I understand it, USB pendants are not currently supported?

One more thing.

It appears the Pico USC can take encoder input. What does it do with that input? Does it just e-stop the machine when it detects a problem?

My steppers (HobbyCNC NEMA23) are single shaft so they won't accept an encoder but I should be able to add an encoder to the gantry rack and have the same thing. Correct? Is there any way I could use two encoders on the x-axis to make sure the two steppers on the x-axis do not go out of sync and rack the gantry?

Tom wrote:

Thank you, Jon. Yes, I'm new to this so your spoon feeding is really appreciated. I have a few more questions that have arisen from these answers.

How is the Pico USC different from a smooth stepper? I thought the LinuxCNC crowd was against any sort of external timing control? This question is just to help me understand. I am actively seeking out a solution that works and, right now, that appears to be a Pico USC.

I'm not an expert on the smooth stepper, but I believe that it gets buffers of moves from
Mach, and then sends the moves out under its own control, so Mach is running a second
or more ahead of where the machine currently is.

The USC is interrogated for position 1000 times a second by LinuxCNC, then the PID
routine computes the best velocity for the most precise following, and sends a new
velocity, 1000 times a second. So, LinuxCNC knows where the machine is 1000
times a second, and can jump in and modify things at any moment. If the machine
actually has encoders, then it is in control of the machine's ACTUAL position, not
just the position computed from counting the steps as they go out.

It appears the Pico USC connects to a parallel port. Would it help the I/O rate to use a Mesa 6i25 E-I/O card (or similar) to drive the Pico? Perhaps not needed?

No, it is still able to check position and update 1000 or even 2000 times a second,
even using the parallel port.

I was considering using Gecko G213v drivers, instead of G203v drivers, so I could increase the rapids. The CNC router isn't done so I have no experience with what will be needed. Perhaps the G213v drivers aren't necessary, if the Pico USC can drive 10x faster than directly connecting to a BoB?

No, I would avoid the step multipler on the G210 or G213. It introduces a number of
esoteric problems in trying to get around the slow software step pulse rate limitation,
which of course you won't have with an external step generator.

One more question. What is the best way for a Canadian, like myself, to obtain a Pico USC (and Gecko G203v/G213v drivers, for that matter)? Do you ship via USPS?

I can't help with the Gecko drives, but Pico Systems gladly accepts Canadian orders, and
can ship USPS or FedEx. FedEx ground is not insanely expensive, and the USPS is
quite reasonable shipping from the US to Canada. (FedEx express IS awfully expensive
going to Canada.) You can check the price and shipping on our web store at
pico-systems.com, but prices will be in US $. (I've got Euros and GB Pounds
set up, though.)

Jon