Servo drive which is better pwm or step dir

More
14 Sep 2020 02:36 #182048 by Smarttowers
I am getting ready to retrofit a bostomatic bd18-2 and trying to get my ducks in a row.

I recently figured out the motors were driven with pwm drives after purchasing some step direction drives.

Now I'm trying to figure out the benefits of both options. Will the speed of one be better than the other? Will the accuracy be the same(this I assume would be).

Really appreciate any input on this as if it's truly one sided i may resell the drives and get pwm ones.

Btw i purchased www.ebay.com.au/itm/322460615556
And am looking at pico pwm drives if I should switch.

Please Log in or Create an account to join the conversation.

More
14 Sep 2020 15:24 #182096 by jmelson
OK, the Pico Systems PWM system has the control loop in the PC. This allows easy tuning through LinuxCNC, you make a move and immediately can see a graph of real-time following error, try an adjustment and do it again, until the error is good enough for you.

I don't know how you tune those machmotion drives, but it may involve additional Windows computers hooked to the drive.

With out PWM system, LinuxCNC is always aware of the following error, and a sliding scale of error tolerance is used, so more error is allowed for rapid traverse moves than cutting moves. Most of the less intelligent drives don't do this.

Jon

Please Log in or Create an account to join the conversation.

More
14 Sep 2020 15:49 #182098 by Smarttowers
The mach drives only use the + from the encoder could i use the negative as input to linuxcnc for a similar result?

Is there speed lost by using step/dir vs pwm? Just trying to understand if one is clearly a better solution.

Please Log in or Create an account to join the conversation.

More
14 Sep 2020 16:10 #182099 by jmelson

Smarttowers wrote: The mach drives only use the + from the encoder could i use the negative as input to linuxcnc for a similar result?

Well, if the mach drives have some delay in them due to the microprocessor, that could make the servo loop unstable.
What would you read the encoder into? If you want to use the software encoder component in LinuxCNC, it is limited to the base thread rate, and could easily lose counts. Also, connecting an encoder to two different systems could lead to a lot of trouble do to the current pulses from the servo drive.

Is there speed lost by using step/dir vs pwm? Just trying to understand if one is clearly a better solution.

How are you going to send steps to the mach drive? If you would use software-generated steps, then there certainly is a speed limit. Depending on CPU, that could be anywhere fro 20K to 50K steps/second. If using a hardware step generator, then this is less of a problem, 300K to millions of steps/second, depending on what unit is generating the steps.

With PWM, the limit is the motor supply voltage.

Jon

Please Log in or Create an account to join the conversation.

More
14 Sep 2020 16:22 #182101 by Smarttowers
I'm still deciding on the BOB but leaning strongly towards a mesa card or 2. Would any of their cards solve the step/dir speed issue and be able to read the encoder properly? The encoders are 5000 ppr quadrature. Would that mean sending 20k steps is only one revolution?

I have dual spindles on the machine so will need to find a solution which can handle 2 vfd.

Please Log in or Create an account to join the conversation.

More
14 Sep 2020 17:03 #182108 by jmelson

Smarttowers wrote: I'm still deciding on the BOB but leaning strongly towards a mesa card or 2. Would any of their cards solve the step/dir speed issue and be able to read the encoder properly? The encoders are 5000 ppr quadrature. Would that mean sending 20k steps is only one revolution?

5000 "pulse/rev" would generallly mean 20,000 quadrature counts/rev, so yes. In addition to Mesa, there is also Pico Systems (me) and our PWM controller system and PWM amps. These will relieve the speed issue, the computer requests a position report, computes a new velocity and sends that to the PWM generator. The encoder feeds back ONLY to the PWM controller. If you are concerned about noise issues, you can put a differential reciever next to the controller board to read both the + and - signals from the encoder fro better signal integrity.

I have dual spindles on the machine so will need to find a solution which can handle 2 vfd.

Well, if the spindles are supposed to run at the same speed, then you only need 2 run command signals to run just one or both. You can feed the same analog speed control to both. Some unused M code could be used to make the spindle selection.

Jon

Please Log in or Create an account to join the conversation.

More
14 Sep 2020 17:12 #182110 by Smarttowers
The spindles are different. One 8k rpm low speed spindle and one 10k-40k high speed spindle. Machine was originally very well spec'd back in the mid 90's. It has a atc for the low speed spindle. I am still figuring out all the required I/O and such. I appreciate all your assistance and wish i had done more research before buying the amps. I'm seriously considering either repurposing those amps or reselling them for pwm control.

Please Log in or Create an account to join the conversation.

More
14 Sep 2020 17:17 #182112 by gaston48
I tested Machdrives, it is very good! Pico pwm does not handle speed servoing and linuxcnc's servo thread pid loop is not as fast as that of machdrives to control speed in addition to position.

Please Log in or Create an account to join the conversation.

More
14 Sep 2020 17:18 #182113 by rodw

jmelson wrote:

Smarttowers wrote: I have dual spindles on the machine so will need to find a solution which can handle 2 vfd.

Well, if the spindles are supposed to run at the same speed, then you only need 2 run command signals to run just one or both. You can feed the same analog speed control to both. Some unused M code could be used to make the spindle selection.

Jon


Linuxcnc V2.8 and above now supports multi spindle out of the box so there is no need to create any M codes. See:
linuxcnc.org/docs/devel/html/gcode/m-code.html#mcode:m3-m4-m5

Please Log in or Create an account to join the conversation.

Moderators: PCWjmelson
Time to create page: 0.182 seconds
Powered by Kunena Forum