Dynomotion

[seapacer2]

Hello Tom,

Out in the garage again tonight, trying to understand what I'm looking at and how to resolve.
I performed these steps in Test 1:

• Opened Kmotion
• Opened multiple windows within Kmotion. (Config/Flash, C Program, Axis, Step Response, Digital I/O)
• Once all the windows were opened, I I set all channells to 0, "X" Axis.
• I pressed the servo "enable" button
• I pressed the "step " button and watched the graph.
• See the first attachment for the results.

Test 2

• Cleared old graph
• I pressed the servo "enable" button
• I pressed the "move" button
• See the second attachment for the results.\

Where do I need to go from here?

Thanks, Mark

[TomKerekes]

Hi Mark,

Follow the steps described here. You already did #1 and #2 but not #3.

After Step #3 check that the servo is working. Move motor by hand and does it resist and attempt to move back?

KanalogInitialPID.mot wasn't loaded so the Axis wasn't configured for your type of system. The Output Mode was set to No Output, so there was no Output and of course no movement - red trace flatlined.

Use the Move button only on the Step Response Screen.

HTH

[seapacer2]

Tom,

I performed Test 3 as per the instructions. Yes, the shaft pulley shows signs of resistance when trying to turn by hand. When you slightly turn the pulley CW, the pulley wants to try to turn CCW. So, I believe it passed Test 3.

Thanks,
Mark

[seapacer2]

I believe we are starting to make some progress.

[TomKerekes]

Hi Mark,

Yes. Congratulations! Your axis is servoing. That's a major milestone

Now you are ready to do a Move and check performance and begin Tuning. Somehow previously the KanalogInitialPID.mot settings were lost. Note now the Output Mode should be DAC Servo.

[seapacer2]

Tom,

Thanks for the guidance! I have a long day tomorrow, going to continue this tomorrow night, need to drive to the other side of the state to pick up a truck that I'm sending to auction.

Will pick up where I left off, I believe turning will be done in the Bode Plot window, is that correct?

Thanks, Mark

[TomKerekes]

Hi Mark,

Actually the Bode Plot Screen is an advanced tool that can give Users some insight into the system dynamics. Most Users are able to tune their system without it. Tune doing Moves on the Step Response Screen.

You might describe in your own words your understanding of the Plots and what they show. Command (blue), Output (green), and Position (red).

[seapacer2]

Good evening Tom,

Back at the turning process, I went back to read a few posts to try to compare my results vs some others. In my words the Blue would be the gas pedal (Command), the green would be a fuel injector (solenoid) adding fuel into the cylinder (output) and the red would be the position of the vehicle moving (Position).

So far, I have run a couple of tests utilizing the Step Response Screen, these are the steps every time.
- Press the "Disable" button
- Press the "Zero" button.
- Press the "Enable" button for the appropriate channel
- Press the "Move" button.

I first increased the Output to 2047
Then I increased the error from 200 to 100,000.
Then I increased the integrator from 200 to 400.
Then I increased the P from .2 to 1.00.

Here are my results, Am I supposed to be making the changes in the Step Response Screen? The graphs are shown in the order I ran them in.

I think the green should mirror the blue somewhat and the red should move even and smooth across the screen?

Am I going in the right direction?

Thanks,
Mark

[TomKerekes]

Hi Mark,

In my words the Blue would be the gas pedal (Command)

Not really. The Command (blue) is what we are requesting the Axis to do. The Trajectory. Where the Axis should be at each moment in time. You might think of it as the curves in the road, posted speed limits, Stop Signs, etc.

The Output (green) is more like the gas pedal. The Servo parameters are like the Driver of a Car. It looks at the road (blue) and tries to follow it by changing the gas pedal (green).

Correct the red is the Position of the vehicle vs time. It records what actually happened.

The basic idea is to get the red plot to closely overlay the blue plot. This shows it actually did what it was supposed to.

Except you plotted Velocity rather than Position. So with that plot type the red and blue plots are the desired and actual speed not position.

Please plot Command, Position, Output Command vs Time. Actually if you save the raw data and also post it then we can plot the data however we want as well as zoom in rather than only have the one way you plotted it.

Once you understand the plot then you should try to interpret what it shows happened. In Plots #2 and #3 I see the green plot going to ~ 200 DAC counts for over 0.5 seconds yet the red plot remains at 0 forever. This is like pushing down the gas pedal in a car and the speedometer remains at 0. Can you see how that would be very strange? Either the car was disabled or the car moved but the speedometer is broken. Its also helpful for you to tell us what actually happened. We are not there and can not see or hear the things you do. Did the motor move at all in tests #2 and #3 and also virtually nothing in tests #4 and #5?

In test #1 there seemed to be a violent acceleration as soon as the Move enabled the axis but way before it was even supposed to start to move. The green plot quickly goes to max (pedal to the metal). The axis accelerates at a high rate then in a fraction of a second coasts to a stop (even with the pedal still to the metal) so I assume the axis might have faulted or something. Which may be why all the other tests did nothing. I think this would indicate something in the axis is backwards. An axis with positive feedback instead of negative feedback causes a runaway effect. Think of it like reversing the steering wheel in a car. As soon as you get slightly off and try to make a correction it makes it worse instead of better and things rapidly escalate. However you claim you tested the servo and it corrected properly. So either it wasn't actually making corrections or something was changed since. OutputGain can be used to reverse the corrections.

[seapacer2]

Tom,

Last night was not a good night, I could not get any movement whatsoever, no matter what I tried. Tonight, I have some movement, I believe it isn't what we want yet but at least some progress., I think? What parameter do I need to look at to smooth out the movement? I have tried a lot of different things, now it is getting confusing.

See the attached same file. Hopefully this will help.

Thanks, Mark

[seapacer2]

Tom,

If i heard you right, you want me to mirror the blue and red lines, correct?

See if this one is better, see the attached file.

Thanks,

Mark

[TomKerekes]

Hi Mark,

If i heard you right, you want me to mirror the blue and red lines, correct?

Yes often once a system is well tuned they match so closely it hard to see any difference. We then plot Error which is the difference. But you aren't there yet.

I don't see any attached files.

[seapacer2]

It exceeds the max limits of the forum,

[seapacer2]

Any suggestions

[seapacer2]

Tom,

Let me know if this works?

[TomKerekes]

Hi Mark,

Zipping worked. The first Plot is unstable. The 2nd is moving but very Jerky. Here is the 2nd Plot:



You didn't tell us what your settings were or what you tried so we can't really help. Always post your Step Response, Config/Flash, and IIR Filters Screens for each Plot.

Also what is the resolution of your Axis? How many counts per inch?

[seapacer2]

Tom,

See the attached file, and screen shot. The only value I changed for test 3 was the "P" value from .2 to .75. This is the closest I've gotten the red to mirror the blue, but still a little jerky.

Thanks,

Mark

[TomKerekes]

Hi Mark,

You didn't tell us the resolution or show the IIR Screen.

The motion one direction is much different than the other. This usually indicates some mechanical issue as the software/control is completely symmetrical. Can you see or feel any mechanical issues?

After fixing any mechanical issue the next steps would be to increase the D Gain (typically around 10X the P Gain) to the maximum possible before going unstable then reducing 20-50% then increasing the P gain to the maximum possible before going unstable. And repeating the process.

[seapacer2]

Tom,

I will show the I/R Screen next time, can you please explain resolution, I'm not sure it I understand?

This may answer your question on mechanical, I have always been advised that I need to have the belt disconnected when testing, would you advise me to install and adjust the belt before trying to perform any more testing?

Thanks,

Mark

[TomKerekes]

Hi Mark,

Resolution is the number of counts per inch. If you don't know how to calculate it theoretically based on lead screw pitch, encoder resolution, gear ratios and so forth make a big move and record the number of counts moved and divide by the measured distance in inches moved.

You will need to tune with the system in its final configuration eventually as the characteristics will change when moving more masses with friction and so forth. Of course when you connect the belts and such you will need to be careful and be able to hit EStop if anything goes wrong. Normally when actually driving an axis more mechanical problems come in to play such as backlash, friction, binding, gravity, and so forth. So it is surprising there is so much asymmetry if you are only moving the motors unconnected from everything. Does the motor spin freely in both directions? Are these brush motors? What type of amplifiers do you have?