- FineLine Saturn 4x2 with ClearPath, UC300-ETH / MB2

[1Jumper10]

Ok. I posted my VFD settings back in my build thread. I'll find them and link to them.

[1Jumper10]

http://www.cnczone.com/forums/cnc-wo...e-forum-4.html

See post #77. There's more info there, including some settings in UCCNC. Here's the VFD parameter settings I used:


F002 ACCELERATION TIME 1.7
F003 DECELERATION TIME 1.5 (BRAKING RESISTOR INSTALLED)

A002 RUN COMMAND SOURCE (USER PREFERENCE) 03:MODBUS <--- only for controlling with the laptop attached. It will be different for UCCNC control
A003 BASE FREQUENCY 400
A004 MAX FREQUENCY 400
A041 TORQUE BOOST 01: (AUTOMATIC)
A044 V/F CHARACTERISTIC CURVE 03: (SENSORLESS VECTOR)
A045 V/F GAIN 50
A082 AVR VOLTAGE SELECT 02: (220)
A097 ACCELERATION CURVE 00: (LINEAR)
A098 DECELERATION CURVE 00: (LINEAR)
B012 LEVEL OF ELECTRONIC THERMAL PROTECTION 50
B021 OVERLOAD RESTRICTION OPERATION MODE 00: (DISABLED)
B082 START FREQ 6
B083 CARRIER FREQ 7.7
B090 DYNAMIC BRAKING RATIO 10
B095 DYNAMIC BRAKING CONTROL 02: ENABLE ALWAYS
C028 AM TERMINAL SELECTION (USER PREFERENCE) 01: OUTPUT CURRENT
C064 HEATSINK OVERHEAT WARNING 75

H GROUP PARAMETERS ARE SET BY AUTO TUNING AND WILL BE DIFFERENT FOR DIFFERENT APPLICATIONS! I LIST MINE HERE ONLY FOR REFERENCE. THEY ARE SET AUTOMATICALLY AFTER PARAMETER H004:

H002 MOTOR CONSTANT SELECTION 02: AUTO TUNED DATA
H004 MOTOR POLES SETTINGS 00: (2 POLES)
H030 MOTOR CONSTANT R1 (AUTO TUNED DATA) .845
H031 MOTOR CONSTANT R2 (AUTO TUNED DATA) .487
H032 MOTOR CONSTANT L (AUTO TUNED DATA) 5.75
H033 MOTOR CONSTANT IO (AUTO TUNED DATA) 4.96
H034 MOTOR CONSTANT J (AUTO TUNED DATA) .012

[DDgitfiddle]

Thank you! I'll dive into this a little more, but I do see one question right off...It looks like you are using a braking resistor. I had not planned to do that. I assume that the only real disadvantage is that the spindle will take longer to slow down?

Thanks,
Robert

[1Jumper10]

Yes. You'll want to increase parameter F003 to a greater value. 4-6 seconds would probably be fine. I guess if your not doing a lot of stopping and starting with a heavy load you don't even need it, the Hitachi will handle it fine. It was pretty cheap and I thought it was good protection for the VFD so I added one.

Sent from my Pixel XL using Tapatalk

[ger21]

I bought a braking resistor for my Huanyang, and then found out that they left out the braking resistor circuitry to save money.
But here's what I found, and I'd imagine the Hitachi would work the same.
If you set a decel rate faster than the VFD can do, it'll just stop as fast as it can.
I have my decel set to 2 seconds. At 8000 rpm, it stops in about 2-3 seconds (I've never actually timed it.) But at 24,000, it takes about 5-6 seconds to stop. I would say most people should not need it.
My reason for getting one is that I am running two spindles from one VFD, switching between them with contactors. I can't make the switch until the spindle stops spinning, or I'll blow the VFD. So I wanted it to stop as fast as possible.

[DDgitfiddle]

OK. I'll look into motor stopping a little more. There is also a DC voltage that it looks like it can use for stopping but that heats things up. These VFDs are intended for alot of different uses that include alot of inertia. The router spindle is about as low inertia as you could expect, so slowing it down shouldn't be an issue.

I looked through your parameters, and came up with a few questions:

A045 - You set the V/F Gain to 50. Was this based on testing or was it based on the calculation from the Hitachi manual?
B012 - You set this to 50 rather than 100. Are you just being conservative?
H004 - I need to check my spindle motor, but I'm not sure I know how many poles it has. 2 is probably correct, but I will have to take a look.

Thanks,
Robert

[1Jumper10]

OK. I'll look into motor stopping a little more. There is also a DC voltage that it looks like it can use for stopping but that heats things up. These VFDs are intended for alot of different uses that include alot of inertia. The router spindle is about as low inertia as you could expect, so slowing it down shouldn't be an issue.

I looked through your parameters, and came up with a few questions:

A045 - You set the V/F Gain to 50. Was this based on testing or was it based on the calculation from the Hitachi manual?
B012 - You set this to 50 rather than 100. Are you just being conservative?
H004 - I need to check my spindle motor, but I'm not sure I know how many poles it has. 2 is probably correct, but I will have to take a look.

Thanks,
Robert

I experimented a little with DC braking but the braking resistor seemed to work much better so I disabled it; A051-00.
A045 - I arrived at this value through testing as I did with most of them. Its been awhile but as I recall, this value isnt critical. I believe I could hear a difference in the sound of the spindle as this value was changed. It sounded best at around 50.
B012 - Yes, just being conservative. I dont expect to ever drive things hard enough for this to come in to play. I chose a reasonable value.
H004 - I'm almost positive you have a 2 pole motor. If its like every other 2.2kw spindle its a 2 pole.

There were other variables I experimented with but the ones I listed were the only one's I changed. You can dive down the rabbit hole and get lost looking at the couple hundred parameters that can be changed but Auto-Tuning is your friend. I experimented by monitoring the motor current, sound, and its performance. But in the end, the Auto Tuning feature achieved the best performance by far.

[DDgitfiddle]

I managed to find some time today to get the spindle running. Thanks 1Jumper10 for the advice to use auto tune and the ProDrive Next software for configuration. That worked out pretty well, after I did a little fiddling around (aka learning).

The instructions for getting ProDrive Next running are a little ambiguous. You need to install the driver first, but then you need to attach the VFD to the computer via USB to get the driver REALLY installed before you try installing the application.

I still need to work on all of the I/O for the drive, but I was able to spin the spindle with manual commands, and it sounded pretty smooth to me. I also need to optimize the accel. and decel. values and look into the DC voltage braking a little more. I don't currently see a huge need for super fast accel/decel, but a little faster than 10 seconds (the default) is probably not hard.

The I/O has been causing me a little trouble because I want to use the drive's Gate Suppress feature with my safety relay. I haven't been able to get the settings to take no matter how I try it. I just figured out tonight that there are a couple of microswitches that I need to flip to enable the inputs and outputs. Hopefully I will have a chance to work on that later this week.

Once the VFD is talking to UCCNC and the safety relay, I will start focusing on the ClearPath motors and getting them tuned.

I also want to put out an APB for Nate at FineLine...I have been trying to contact him about a few questions but haven't had any luck.

Thanks,
Robert

[1Jumper10]

Before I auto tuned and was trying to get optimum performance by manual tuning, I had a lot of over current trips during accel and decel. You should easily be able to accelerate to max speed in ~4 seconds without tripping the over current alarm. After auto tuning, I got mine up to 1 second to accel and 1 second to stop. When I was done playing I set both to 1.7 seconds. Seemed reasonable and plenty fast.

Just a little info on the VFD that you might have ran across: There are 2 safety features that you can use to prevent the VFD from starting before the software when the UC300/MB2 is first powered on.
1) - You can connect the M3 FWD/ run command and the M4 REV/Run command to the VFD even though you may not ever run it in reverse. That way when the MB2 powers up and all Outputs go active, the VFD will see a FWD run command and a REV run command. When the VFD sees these conflicting commands its is disabled until the condition is corrected.

2) - There is a setting I believe its B13 (don't have my manual available at the moment) that disables the VFD if it receives a RUN command within a few seconds of first being powered up. It generates an error message and disables the VFD. The alarm clears when the RUN command goes away. This would work pretty well with UCCNC - UC300ETH- MB2 to prevent the spindle from uncontrolled operation when you first turn on the power (when all outputs on the MB2 go active) and before the software takes control. This is the feature I'm going to use on mine. I'm not using any sort of safety relay.

Lastly, I haven't seen Nate but I'll keep my eyes open

[DDgitfiddle]

I'm planning to just reduce the accel and decel times until I start having issues, then back it off a little.

Thanks for the info for keeping the VFD from starting on power up. I think I have this covered with the safety relay set up, assuming I get the Gate Suppress feature working on the VFD. I have the E-stop button, the Charge Pump on the MB2, and the OSSD (e-stop circuit on the MB2) all in series feeding the safety relay. The OSSD monitors all of the ClearPath HLFB outputs and will be reporting that the motors are enabled/not error. I used the MB2 OVERRIDE input with a push button to allow everything to be reset. I've confirmed (many times by accident) that the charge pump doesn't turn on until UCCNC is running. The safety relay won't reset until the e-stop circuit is closed. The safety relay controls the Gate Suppress inputs on the VFD (GS1 and GS2, via contacts 3 and 4), so the spindle motor can't be run until everything is running. The Gate Suppress function on the WJ200 series is certified by Hitachi to meet industrial standards. That was worth the extra cost on the VFD for me to avoid having to pull the plug on the VFD on e-stop.

I really like having the safety relay controlling all the moving parts. It was a relatively inexpensive ($85 surplus) piece of insurance that was designed to comply with safety standards for industrial equipment. I haven't gone through the necessary hazards analysis to say that this system would meet any standard, but I am comfortable that it will behave very consistently. It also prevents the startup behavior on the UC300ETH/MB2 combo from creating any unsafe conditions. Besides, I get a really satisfying KATHUNK when the main contactor closes when everything gets powered up. There have been a number of other ways described on this site that will also result in a safe system, this was just the path that I was most familiar with and comfortable with.

The only thing that I wish I could have achieved was to be able to stop the ClearPath motors without having to kill their power. Cutting the Enable signal to those should technically work, but Teknic does not certify that this meets the industrial standards. Maybe I'm being a little too picky for a hobby machine, but I'm pretty attached to my fingers, so it is worth spending a little extra for peace of mind.

-Robert

[DDgitfiddle]

I've had pretty limited time to work on things this week...I hate it when my day job gets in the way of my hobbies.

I've mostly been working on getting the UC300ETH/MB2 to talk to the Hitachi VFD. I connected the C motor step output (CS+/CS-) to the EA and L pins on the VFD as 1Jumper10 did. I used the NO output on relay 2 from the MB2 as the M3 signal from UCCNC to actually switch on the spindle. After a few things not quite working, I found the errors in the VFD settings and was able to turn the spindle on from UCCNC.

I have the speed running theoretically right, but the numbers I had to use in UCCNC don't make a lot of sense to me. Like 1Jumper10, I ended up using 62.5 steps per rotation in order for the UCCNC spindle speed setting to scale correctly as RPM. The VFD supposedly scales 32kHz to the maximum frequency setting, which is 400Hz. 24000 RPM is 400Hz (24000/60= 400), so I would expect to see a factor of 4/3 in there somewhere (32000/24000). It might have something to do with the axis C settings in UCCNC. At any rate, it works, so I'm not complaining. I do need to try to check the actual spindle speed to confirm it, though.

I also bumped the accel and decel times to 2 seconds each. I didn't find a point where the VFD gave me any errors, but that seemed about fast enough. I didn't turn on any of the DC voltage braking or add a braking resistor. I think the inertia is low enough that the VFD can handle it. It won't be continuously speeding up and slowing down, so good enough as is.

I forgot to copy my VFD settings, but I will post the changes from default values with some explanation in the next couple of days.

Thanks again for the help 1Jumper10! It made things a lot easier!

Robert

[1Jumper10]

Glad I could help. If you figure out the wierd UCCNC settings values let me know. I'll do the same.

Sent from my Pixel XL using Tapatalk

[pontoonman]

I've had pretty limited time to work on things this week...I hate it when my day job gets in the way of my hobbies.

I've mostly been working on getting the UC300ETH/MB2 to talk to the Hitachi VFD. I connected the C motor step output (CS+/CS-) to the EA and L pins on the VFD as 1Jumper10 did. I used the NO output on relay 2 from the MB2 as the M3 signal from UCCNC to actually switch on the spindle. After a few things not quite working, I found the errors in the VFD settings and was able to turn the spindle on from UCCNC.

I have the speed running theoretically right, but the numbers I had to use in UCCNC don't make a lot of sense to me. Like 1Jumper10, I ended up using 62.5 steps per rotation in order for the UCCNC spindle speed setting to scale correctly as RPM. The VFD supposedly scales 32kHz to the maximum frequency setting, which is 400Hz. 24000 RPM is 400Hz (24000/60= 400), so I would expect to see a factor of 4/3 in there somewhere (32000/24000). It might have something to do with the axis C settings in UCCNC. At any rate, it works, so I'm not complaining. I do need to try to check the actual spindle speed to confirm it, though.

I also bumped the accel and decel times to 2 seconds each. I didn't find a point where the VFD gave me any errors, but that seemed about fast enough. I didn't turn on any of the DC voltage braking or add a braking resistor. I think the inertia is low enough that the VFD can handle it. It won't be continuously speeding up and slowing down, so good enough as is.

I forgot to copy my VFD settings, but I will post the changes from default values with some explanation in the next couple of days.

Thanks again for the help 1Jumper10! It made things a lot easier!

Robert

I use Hitachi WJ200 VFD's and have tried the pulse train input method using UCCNC with good results. It's not documented very well, but if you look at parameter P055 you will see that it sets the max frequency for the pulse train. The spec sheet max as you state is 32k Hz, but the default max (P055) is 25k Hz. So, 25000/24000 = 1.04167 x 60 = 62.5

P055 Pulse Train Frequency Scale Default value: 25.0 Range: 1.0 ... 32.0

There are a few other pulse train parameters: P056, P057 & P058. I didn't adjust any of these (used default values).

Frank

[DDgitfiddle]

I use Hitachi WJ200 VFD's and have tried the pulse train input method using UCCNC with good results. It's not documented very well, but if you look at parameter P055 you will see that it sets the max frequency for the pulse train. The spec sheet max as you state is 32k Hz, but the default max (P055) is 25k Hz. So, 25000/24000 = 1.04167 x 60 = 62.5

P055 Pulse Train Frequency Scale Default value: 25.0 Range: 1.0 ... 32.0

There are a few other pulse train parameters: P056, P057 & P058. I didn't adjust any of these (used default values).

Frank

Eureka!! That actually makes alot of sense. UCCNC probably assumes that the motor is a stepper, so using 60 steps per rev would effectively convert 24000 RPM into a 24000 Hz output pulse train. Setting P055 to 25000Hz in the Hitachi VFD introduces the slight error that I have corrected in UCCNC with the 62.5 step/rev setting.

So in reality, the best way to set all of this up would be to set P055 to 24000, and set the steps per rev in UCCNC to 60. At least then the math makes sense.

Thanks for clearing up the mystery Frank!

-Robert

[1Jumper10]

Yeah, Thanks Frank! I'll make the same change here shortly.

[DDgitfiddle]

Time has been a little scarce over the last week, so progress has been a little limited. Major things that happened:

- Finished up the control box and tidied up the wiring. In theory that is all finished except for making the final connection between the spindle and the VFD. I'm not sure how long to cut the cable yet, so that will happen a little later.



- Mounted the control box to the machine frame. I still need to put the door back on, but it is solid.

- Assembled one of the motors to a motor drive. This was a little more involved than I had hoped. The main problem was the part that the spring tension adjustment screw threads into. It is mounted to the main plate of the drive with a flat head screw. No real problem there, except that the head of the screw sits up high enough to contact the spring adjustment screw. When my machine was built, everything was sort of forced together, so the threads on the spring adjustment screws got a little mangled. I managed to get everything together with just enough clearance but it is slightly hack. I ended up chasing everything with a tap and cutting partial threads in the top of the flat head screw. The root of the problem is that there isn't enough clearance, and the countersunk hole for the flat head screw has the wrong angle for the screw. It all works, but it's not ideal.

- Mounted the first drive to the y-axis so I could try to tune the motor. I greased up the linear bearings and the rack with some NLGI 2 grease and mounted about 63 lbs. to the top of the linear bearing plate (gantry mount plate) using a 5/16" threaded rod threaded into one of the existing holes in the plate. I weighed the gantry at very close to 125lbs, so 63lb is close to half of that. I fired up the ClearPath software and everything connected to the motors as I would expect. I jogged the motor a little just to confirm that it would work. (I will post a quick video later this week).

I then ran the ClearPath tuning program. Everything ran as I would expect. The motor moved around quite a bit. Unfortunately, when I jog the motor after tuning, the performance looks pretty poor. The acceleration is a little jerky no matter how fast I set it, and when it stops, it hunts for position for a second or so. Clearly not tuned right. I saved off the profile, then tried re-tuning a couple of different times. The gains are very different in each of the profiles, so it seems like the program is having a tough time converging for some reason. It seems like the controller is too "soft" and is not maintaining position correctly. I'm reluctant to put the gantry on until I can demonstrate that the motor is tuned properly, just because it is a minor pain.

So I think my next step is to connect the motor to UCCNC and see if that pulse train works any better since it will run over digital I/O instead of USB. If that doesn't improve things, I will reach out to Teknic and see if they have any ideas. If that doesn't work, I will call in a friend who is a controls engineer.

-Robert

[DDgitfiddle]

Catching up on posting some promised information. First the settings for the VFD. I am running a Hitachi WJ200-022SF with a 80mm, 2.2kW Chinese air-cooled spindle. The spindle is rated for 220V, 400Hz operation (24000 RPM). The settings below are the only things that I had to change. I've added some notes to each section to explain the purpose of the settings in my system.


F002: Acceleration time (1) = 2F003: Deceleration time (1) = 2
// I set the F002 and F003 settings based on what seemed like a decent start and stop time without tripping anything. Not really scientific. //


A001: Frequency source = 06 -(Pulse train input)
A002: Run command source = 01 -(Control terminal)
// These two settings are needed in order to use the Step output (regular motor output on the MB2). A001 tells the VFD to accept the pulse train to set the spindle speed. A002 tells the VFD to accept a discrete input for starting and stopping the spindle. //


A003: Base frequency = 400
A004: Maximum frequency = 400
// A003 and A004 are set based on the spindle ratings. The WJ200 series max out at 400 Hz, though. //


A041: Torque boost select = 01 -(Automatic torque boost)
// This setting is based on 1Jumper10's recommendation. Not clear to me that it will have a huge effect unless really pushing the system. //


A044: V/f characteristic curve = 03 -(Sensorless vector (SLV))
// This setting allows the VFD to "sense" the speed of the motor and control it. It also is useful when using the pulse train input from the MB2 and does a nice job of holding spindle speed. I believe this uses back EMF from the motor to sense speed. Not as accurate as an encoder, but not bad for a spindle. //


A082: AVR voltage select = 02 -(220)
// based on motor rated voltage//


A097: Acceleration curve selection = 00 -(linear)
A098: Deceleration curve selection = 00 -(linear)
// sets a simple linear acceleration and deceleration profile. Doesn't have a major impact on the spindle.//


b012: Level of electronic thermal = 50
// A value from 1-100%. This is used in calculating the theoretical heating on the motor to help avoid overheating. Setting this at 50% is pretty conservative. //


b031: Software lock mode selection = 10 -(High level access including B031)
// I had to set this to 10 to gain access to all of the settings I needed on the keypad. Once I get this all settings tested with real-world operation, I will change this to 02. //


b037: Function code display restriction = 00 -(Full display)
// I had to set this to 00 to be able to see all of the settings I needed on the keypad. //



C003: Input [3] function = 77 -(GS1:GS1 input)
C004: Input [4] function = 78 -(GS2:GS2 input)

C021: Output [11] function = 62 -(EDM:STO (Safe Torque Off) Performance Monitor(Output terminal 11 only))
C026: Alarm relay function = 62 -(EDM:STO (Safe Torque Off) Performance Monitor(Output terminal 11 only))
// These settings enable the "Gate Suppress" function of this VFD. This is an ISO compliant method of cutting motor power without dropping the input power to the VFD. It avoids needing a contactor for the VFD. It does require two signals, to work however. I am using two contacts on my safety relay. C021 and C026 are technically redundant. They are the confirmation signals that indicate that the Gate Suppress function has been been activated. My safety relay has a feedback input that Is connected to this signal. I am using the relay output on the VFD instead of the solid state output because it is easier to wire.



H002: Motor constant selection = 02 -(Auto tuned data)
H004: Motor poles setting = 00 -(2P)
H030: Motor constant R1 (Auto tuned data) = 0.845
H031: Motor constant R2 (Auto tuned data) = 0.487
H032: Motor constant L (Auto tuned data) = 5.75
H033: Motor constant I0 (Auto tuned data) = 4.96
H034: Motor constant J (Auto tuned data) = 0.012
//H002 allows for auto tuning the spindle which works great for the spindle. H004 is motor specific. The rest of these settings are the result of auto-tuning and are specific to my spindle. I have provided these in case someone has a similar spindle and needs a point of comparison.

[Richmaple]

Super useful info, and thanks for sharing! I'm super curious what you learn about the clearpath tuning as well. I've used them on other applications and have not had the problem you mentioned. But those were also their position and velocity models, not their step and direction ones.

Sent from my SAMSUNG-SM-G920A using Tapatalk

[DDgitfiddle]

So I had a very successful afternoon tuning my first ClearPath motor. I tried to auto-tune it a few different times, but couldn't get it to work. I did contact Teknic, and got an initial response, but not a lot of of help. I recognize that they are in the business of trying to sell large quantities of motors into production systems, so supporting a hobby CNC builder is not high priority. So no criticism there.

I recognized that the ClearPath motors can also be manually tuned, but the online documentation from ClearPath doesn't really discuss that in detail. I did a Google search and came across a 4 year old document that described a process for manually tuning the motors. It references an older revision of the MSP software.

http://www.motionsolutions.net/store...ote%20V1.3.pdf

I also found a non-published key stroke combination (Shift-Cntl-P) that brings up the manual tuning screens that are used in this document. In general, I found that these instructions work with the ClearPath MSP software version 1.2.103 that I downloaded from Teknic's website.

Before I continue, I feel obligated to provide a few warnings. The MSP software gives you access to the gains that control the stability of the servo. If these values are changed too drastically, the servo could go unstable pretty quickly. This can lead to personal injury or damage to the motors or other hardware. The process causes things to move, and at times I found the movement to be pretty abrupt, to the point of being concerned. A main example is the initial gain tuning where a "square wave" input is provided. I had to tighten up the anti-backlash spring on the FineLine drive to prevent the gear from loosing contact with the rack due to the very high torque applied during the move. I found that none of the gains were so sensitive to changes that the system would snap from semi-safe to unsafe with very small changes. Changes of 20-50% GENERALLY were OK, but I was careful to increase gain values slowly, watching for changes along the way. That said, your system may be different from mine. My tuning process involved bolting 63 lbs of metal to the axis being driven by a NEMA 34 motor (CPM-SDSK-3421S-ELN) with a max torque capability close to 1100 in-lb, so unless your system is very similar, don't use these gains as a reference. I hope to provide a process that people can follow if needed.

So first, I powered up the system, started up the MSP software and connected a USB cable from the motor to the computer. I pressed (Shift-Cntl-P) to access the tuning parameters screen and expanded the performance monitoring graphs. Following the procedure in the Tuning App Notes linked above, I pressed (Shift F1) to have the program set itself up for initial Kv tuning. This will rotate the motor back and forth at about a 0.5 second cycle with a square wave input. Pressing (Shift F1) sets all the gains to zero, so there will be no motion at first, and this is normal. As the Kv gain is increased, you will see and hear the motor start to move back and forth, and the graph will start to improve. The example in the Tuning App Notes is pretty clean and nice looking. My system was not so clean. I also had to increase the Kv gain to over 200,000 to get close to acceptable performance. The machine was shaking pretty badly because the motion is very abrupt. I tried to minimize the time doing this because it didn't seem too healthy for the mechanism. I also turned down the commanded motion amplitude and increased the period to reduce the wear and tear. At the end, this is what I had:



I then proceeded to increase the Kp-Out, Ki-Out, and Knv as prescribed in the Tuning App Notes. All of this process involves the square wave movement and is very jarring, so I tried to turn off the movement when making changes to reduce wear and tear.

Once this is done, it is time to define a ramped motion for setting the remaining gains. This is much less jarring because the motion speed ramps up and down linearly. The Tuning App Notes don't recommend settings because they would vary greatly based on motor and mechanical system. The criteria is to try to get to 80-90% of torque and/or phase voltage. This is where I realized just how over powered these motors likely are for this application. I was unable to meet this level even after setting the motors to 1000 RPM (close to their maximum rated speed) and using a 12000 RPM/s acceleration. That speed is equivalent to 981 in/min which is probably WAY faster than this machine will actually run. So my guidance here is to set the motor speed close to its maximum and really try to bump up the acceleration as high as you can. This looked like a pretty aggressive move to me. I then started adjusting the gains as recommended in the Tuning App Notes: Kfa, then Kfv, then Kfj. The Ki-Out gain is set to zero for this process. At the beginning, the performance looked like this:



Be careful to scale the graph range to something reasonable so that you don't fool yourself into thinking that the settings are either really good or really bad. Use the cursors on the graph to actually measure the number of counts of error. Also keep the resolution of your motor in mind. This motor has a 6400 pulse per revolution resolution, so 5 or 6 pulses is a pretty small fraction of the motion. The standard resolution motors have 800 pulses, so 5 or 6 is a more significant error.

When tuning the Kfa, Kfv, and Kfj gains, I tried to keep an eye on the distance between the highest and lowest points on the graph by measuring them with the cursor. I was trying to minimize the difference between them and also make sure that it was evenly spaced across the zero line on the graph (e.g. +14 counts above the line and -14 counts below the line). I played with these three gains for a while, but the best I was able to do was +/-14 counts.

I re-applied the Ki-Out gain and rechecked performance. Things improved a little. I then went back and tried adjusting all the gains a little to see if I could improve the performance. To my pleasant surprise, I was able to increase a few gains and get the error down to about +/-5 counts. That is less than 0.0008" error on this system, compared to the 0.00015" resolution. Not bad. Here is the final graph, which does loosely resemble the graph in the Tuning App Notes:



After my initial auto-tuning, I found that the axis felt "soft." When I pushed on it, I was able to move the axis by about 0.08-0.125 inches before it got stiff enough to not be able to push. After the manual tuning, it is MUCH more rigid. I don't know how much it moves, but the belt drive only moves slightly. When jogging the axis, there was a noticeably jerky motion with the auto-tune. It now moves smoothly as I would expect. I compared the gains, and there is a large difference in most of the gains, especially the Kv gain which is different by a factor of 100x. My hope is that because I tuned a pretty fast move that the slower moves will have an easier time and perform at least as well.

I still don't know why the auto-tuning algorithm didn't work well. It may have something to do with springiness in the drive due to the anti-backlash spring or inaccuracies in the rack and pinion drive causing ripple in the torque transfer. You can see from the graphs above that the general performance is not as clean as the Teknic examples. That implies that the mechanical system is behaving in a more complex way.

This is a long post, so I will try to add another post later to explain the effect of the various gains on the performance. Hopefully this will help some others with ClearPath motor tuning in the future.

-Robert

[DDgitfiddle]

I'm going to make an attempt at explaining what each of the gains in the ClearPath motors are doing in the system. The big caveat here is that I am over 20 years removed from the controls training that I had in college, so most of that knowledge has leaked out. If controls theory interests you, please dig deeper and look for more reliable sources. The Wikipedia entry for PID controller is actually not bad and has some nice graphical descriptions.

The ClearPath motor controllers are based on a pretty common PID control loop, but with some additional things added in to improve performance. PID stands for "Proportional, Integral, Derivative." These terms refer back to calculus, but they have a pretty understandable meaning. In the case of a CNC router, what we care about at any given time is the position of the axis. We want the actual position to be as close as possible to the commanded position. Any difference between the actual and commanded position is the error in position. At the same time, it is pretty easy to see that when we command the system to move from one point to another, we are also commanding an ideal velocity in addition to position. Extending this further, we can understand that we are also commanding an ideal acceleration for the axis, also. A PID loop simply looks at the error in position and velocity, multiplies each of those errors by a number (called a gain), adds those values together, and uses that number to update the voltage and current supplied to the motor. In a digital system, this is usually done a few hundred times a second.

So the three main gains in a PID loop are:

Proportional - This is the gain multiplied by the position error for the motor. Remember that position is measured by the motor encoder in this case.

Derivative - This is the gain multiplied by the error in motor velocity. Velocity is just the change in position over time. We don't directly measure this. Instead, the controller looks at the change in encoder position and divides that by the time since the last reading. Because the encoder has some fixed number of pulses (e.g. 6400 in the case of my motors), there is always a little error in this number.

Integral - This is the gain that is multiplied by the accumulated position error in the system. Here is how I think of it: Think about a sort of drag race between two cars where both cars accelerate to 60 MPH as fast as they can, but then have to stay at 60 MPH until they reach the finish line. One car accelerates faster than the other, so it pulls into the lead. Think of this car as the move command on a CNC router. The other car will lag behind all the way to the finish line. Once both cars are at 60 MPH, the gap will stay the same, but it will never get smaller. The integral term in a PID loop looks at the position error multiplied by the amount of time the error has been present and then multiplies that by the Integral gain. So the longer an error remains, the more effect the integral term has. In the car analogy, it is like allowing the slower car to exceed 60 MPH for a short time in order to catch up. The challenge here is that if this gain is too high, the system will over-shoot the target position (i.e. the slower car would pass the faster car).

So let's look at the gains in the ClearPath system:

Kv - This is the Derivative gain
Kp-Out - This is the Proportional gain.
Ki-Out - This is the Integral gain.
Kp-Zero, Ki-Zero - These appear to be non-linear gains. I don't honestly know what they do. It appears that they can be changed optionally, but with the settings I have, they are equal to the Kp-Out and Ki-Out values.

Knv - Teknic calls this the Inertia Matching Technology gain. I don't know exactly how this works, but my guess is that it is used in an equation that helps prevent the Integral term from over-shooting.

Kfv - This is a feed-forward gain based on velocity. The controller looks at how the position command is changing and uses this gain to adjust the command to the motor a little in advance. Obviously there is a limit to how much this term can do, but it helps to reduce the error.
Kfa - This is similar to Kfv, but it looks at how the velocity command is changing.
Kfj - This is similar to Kfv and Kfa but it looks at how the acceleration command is changing. The change in acceleration over time is called "jerk" (no, not like your boss). Fast changes in acceleration have high jerk. In my previous post where I describe the jarring nature of the square wave motion, I am talking about a high jerk situation. In that case the acceleration goes from essentially zero to very high instantaneously because the controller is trying to follow an instantaneous position change command.

So the tuning process that the ClearPath Tuning App Notes describes provides a fairly methodical way to set these gains while reducing the chance that the system will go into an unstable condition. It also focuses first on the primary gains that have the biggest impact on performance then uses the feed-forward terms to refine performance.

That's probably about all I am qualified to try to explain. If any controls experts are following this, please feel free to correct anything I have written.

-Robert