Messages

11

General Mach Discussion / Re: Physical buttons for plasma

« on: November 26, 2016, 06:59:41 PM »

I would expect a servo to handle those kinds of moves, if tuned well. Could it be a case of the short/sharp moves, combined with a bit instability in the tuning causing the problem?

Could you try reducing the max current/acceleration?
You obviously don't want to reduce it too much that it's not fast enough to respond, but dropping it slightly might be enough to take the edge of the moves which are causing the heating.

12

General Mach Discussion / Re: Question on Plasma THC control to users of this facility

« on: November 26, 2016, 06:50:16 PM »

Just had a quick scan through the thread on mig-welding. Are you the Adrian H who happened to sell a rapid cycle Harrison mill a number of years ago?

The basic older TB drivers are not great. They're ok if you don't push them near their limits, but lots of implementations have woeful heatsinking, which compound their limitations. For not much more money, I'd buy some newer individual TB6600 drives (search ebay for TB6600 - there are plenty options), which use far better chips, and would allow you to run a higher voltage power supply. You also get the benefit that if you do blow one up, you only need to replace that one.
For plasma, you need speed, and the main way to get speed from steppers is by increasing the power supply voltage.

13

General Mach Discussion / Re: Robot part loader for CNC Lathe

« on: November 26, 2016, 07:04:41 AM »

A basic KFlop is a mix of 3.3V and 5V, but the various add-on boards can give 24V IO.

I like this video for demonstrating just what the KFlop is capable of doing - https://www.youtube.com/watch?v=MSONYPXe3bE
You could control that via Mach just by providing basic 6 axis g-code, as all the kinematic calcs are handled by the KFlop.

Dynomotion have also fairly recently added 5-axes gimbal support, where all you need to do is set the radiuses, and the KFlop can handle all the kinematics and compensate for tool offsets - https://www.youtube.com/watch?v=R7KF9tNmQFA (skip to about 2:40 to see the simulation) or https://www.youtube.com/watch?v=HSnjGD2edGE shows a real machine.

14

General Mach Discussion / Re: Robot part loader for CNC Lathe

« on: November 25, 2016, 08:33:45 PM »

AFAIK ESS doesn't support index homing itself. What you would normally do is move the servo until the home switch is triggered, slowly back off until it releases, then activate the servo drives internal homing and wait until the drive signals it's homed.

How much variety will there be in the parts?
My first thought would of been a PLC, however you may want to look at a KFlop?
A KFlop would allow you a lot more flexibility in terms of programming (think HiCON/Galil), and gives you options to run standalone (you can program into the KFlop's flash memory what you want it to do and have it run automatically when powered on), or with a custom screen to control what you want done (could be industrial touch screen with buttons so operators are limited to what can be controlled, or even physical buttons).

However, without more details about what you're wanting to do, it's hard to give advise.

15

General Mach Discussion / Re: New cnc machine running but....

« on: November 13, 2016, 07:48:46 AM »

Those TB6600 drives you linked to are OK. The good thing about separate drives, is you can replace/upgrade relatively easily in future.

I've got a couple of those TB6600 drives on the desk for when I get around to retrofitting an engraving machine, and from the research I done, they're a major improvement over the older all in one TB6560 boards. The main thing is make sure you're not pushing them too near their limits.

The big killer of the older chips was people running them pretty much on their voltage limit, along with often poor board design, meant a slight voltage spike took them overvoltage frying them, or they simply overheated.

16

General Mach Discussion / Re: Steps aren't consistent positive to negative

« on: November 11, 2016, 06:50:19 PM »

Could be a direction setup time issue.

Depending on the drives, after the direction has been changed, they may need a certain amount of delay before the next step pulse. If this was the problem, then the drive can move one step in the wrong direction before then moving in the correct direction, so it ends up two steps of the desired position, and it is possible for it to only affect the direction change in one direction.

You could try flipping the Step Active High/Low option, or increasing the Dir Pulse setting on the motor tuning page.

17

CS-Lab / Re: IP-A and Servo drives, voltage matching...

« on: November 10, 2016, 04:24:49 AM »

Ideally the tuning move should reach the maximum speed the machine will run at, however you should gradually work up to that when tuning, just to avoid causing any prolonged wild oscillations if a change in PID settings makes things badly unstable.

KFlop lets you do two different types of moves. A block step move, and a move limited by the Velocity, Acceleration, and Jerk settings.
The block step is an instantaneous command (i.e. Acceleration and Jerk settings are ignored) to move the requested position amount, which you use to determine just how quickly the system can accelerate, and how quickly it can handle peaks in command, but it's brutal on the machine.

The VAJ limited move is more like a move you'd command during G-Code running, which is what you use for normal tuning.

What you're looking for when tuning, is for when things become unstable, then backing of a bit.
Ideally after sudden change in requested position/speed, there should be a short peak in output, which should quickly settle down with minimal oscillation.
If you have no peak, then the output isn't being driven hard enough, and you're following error will most likely jump up until things catch up.
If you have a peak, that is followed by endless oscillations, then best case scenario the following error can be pretty low, but the servo will be continually oscillating to stay on position, and in the worst case, the servo will be wildly oscillating with a huge swing in following error.
You're aiming for a compromise between the two.

And I've just remembered there's a guide on the Dynomotion wiki - http://www.dynomotion.com/wiki/index.php?title=Main_Page#Axes_Servo_Tuning_and_Trajectory_Planner
Some of it won't apply, but the underlying process will be very similar.

18

CS-Lab / Re: IP-A and Servo drives, voltage matching...

« on: November 09, 2016, 06:19:56 PM »

The default drive tuning for the selected motor should be pretty good on modern digital drives. I know trying to tune my drives didn't make much difference to the end result. Depending on what mode you're running the drives in, depends on just how much the drive tuning affects thing. Torque mode at most you'll get some input filtering, Speed mode you'll also get a bit more tuning to ensure the speed is stable, and Step/Dir mode the full tuning in the drive is used.
What CS-Labs will be referring to is older amplifiers, where you had to manually tune the amplifier to match the motor, otherwise you ended up trying to tune the PID loop with an already unstable motor/amplifier.

I'd try adjusting the PID loop in the IP-A as per the instructions above, and see if you can improve things.
I'm not sure how you test moves with the IP-A, but the KFlop software gives you the option to quickly edit the settings, then command a back and forth move, at which point a plot is displayed, so you can see almost instantly how things have changed. It is wise to have your hand over the E-Stop though, just in case you trigger some big oscillations (been there, got the medal, and even managed to overheat the motor!)

19

CS-Lab / Re: IP-A and Servo drives, voltage matching...

« on: November 09, 2016, 04:47:06 AM »

Just found a relevant post from Tom Kerekes (The man behind Dynomotion), from the yahoo group explaining a bit about tuning-

Unfortunately Servo Tuning is complex. There isn't really a simple - this parameter causes this problem situation.  Everything somewhat interacts and needs to work together to get the desired result.  Every system is different and has different requirements.

But basically and simplistically:

P Gain simply applies an Output Drive Proportional to error (how far the system is away from the Desired Target Destination).  This is usually the primary source of corrective feedback.  ie used for: "If way off to the right side of the road turn way left"

I Gain increases or decreases the Output Drive at a rate proportional to error (Output is proportional to all previous errors accumulated).  Corrects small persistent errors.  ie used for: "If slightly off to the right keep gradually rotating the steering wheel to the left until centered"

D Gain applies an Output Drive proportional to the rate of approach to the target.  ie used for: "Rapidly approaching a stop sign - apply brakes until speed is reduced"

PID values will influence how the system responds dynamically to errors.  Filters will also influence the dynamics of how the system responds and offer additional and more flexibility than PID values.  It is very difficult to intuitively predict how a change to PID+Filters will affect the dynamic system behavior when thinking in the time domain.   Working in the frequency domain is much more intuitive.  There is a considerable learning curve involved which is beyond most Users but is really the only way to have an understanding of the dynamics of a system.  It involves a two step process.  #1 one must understand how the shape of the Magnitude/Phase plot (Bode Plot) of the system will determine its response. #2 one must understand how PID+Filters will affect the Bode Plot.  For example a Bode Plot with a High Bandwidth and with large Gain and Phase Margins will correct errors quickly and be a stable system.  A Low pass Filter will reduce the amplitude of higher frequencies and introduce phase lag at higher frequencies.   Integrator Gain will increase the magnitude of low frequencies and add phase lag.  D gain will increase the magnitude of higher frequencies and add phase lead. 

This is all classical control theory.  There are numerous resources on-line
http://www.google.com/search?q=pid+tuning+tutorial

Regarding Max Limits: in general when a system is operating normally Max Limits (Errors, Output, Integrator, Following) should have no limiting effect and should have no influence on the dynamic response.  Generally the Max Limits are designed to help in abnormal situations in order to avoid an overly violent system response to the abnormal situation.  For example Max Integrator reduces a problem referred to as Integrator Wind-up.  If you take a well behaved and well tuned system (with some I Gain) and force the motor off of target a small amount and hold it off for a long time the Integrator will accumulate a huge amount of error.  If you suddenly release the motor shaft the system will respond with a violent overshoot.  Max Integrator can be used to make the integrator stop accumulating error at some level.

The order of tuning operations are usually:
#1 start with P gain
#2 add D gain (2nd order low pass filter 1000Hz Q1.4 will reduce output spikes)
#3 repeat #1 and #2 a few times
#4 add small amount of I gain

HTH
Good luck
Regards
TK

20

CS-Lab / Re: IP-A and Servo drives, voltage matching...

« on: November 09, 2016, 04:43:49 AM »

Dave, the motors you're using don't have any tach signal, they only have encoders. If things weren't wired correctly, then the drive would trigger a fault.

Your following error does sound pretty high. My Lathe can manage sub 50 on both axis, but it took a bit of manual tuning to get there, with some tips thrown in from somebody who really knows about tuning.

Essentially you keep increasing P until things go unstable, back it of slightly, increase D until things go unstable, then back it of slightly.
Repeat the above a few times, until you find the limit of stability.
Then apply some I, but once you apply I, you may have to decrease PD, as I makes things unstable.
Auto-tuning will usually get reasonable performance, but it rarely works well with unbalanced axis i.e. vertical Z axis.

There could also be a problem with the controller fighting against any inbuilt filtering in the servo drives, which is something I hit when trying to run my Kinco servos in Torque mode. Torque mode should in theory give you better performance, but tuning is far more critical.
I'm not sure what tuning/plotting capabilities CS-Labs have, but KFlop comes with some very good tools so you can generate various plots, and see exactly what is happening. Being able to see the output oscillating, along with the following error under different conditions makes tuning far easier.