On Sun, Sep 8, 2013 at 5:55 PM, Tom Kerekes <tk@...> wrote:

 

Hi Hugh,

The Axis Channel assignments sound reasonable.

We have a C program called AutoPhaseFind.c that you should run first to check your connections and find parameters regarding your brushless motor.

For your Z motor change the defines in the program to:

#define PWM_CHAN 14            // which pair of PWM channels used
#define ENCODER_CHAN 13        // which encoder we are connected to
#define ENCODER_GAIN 1         // Set to -1 if desired to reverse axis direction
#define AMPLITUDE 10           // Set how hard to drive the coils pwm counts
#define Z_BIT_NUMBER 102    // What bit the Z index is connected to
#define CLAMP_VOLTAGE 85      // volts - set a few volts higher than your supply
#define AXIS_CHAN 3           // Axis channel to be used and configured

The AutoPhaseFind.c should rotate the motor by energizing the phases (much like a stepping motor) and measure the encoder motion and Index Pulse and print a report.

After this is all successful you can modify the HomeBrushlessSnap4.c program.  You will want to change it to use KFLOP Axis Channel 3 so change all the references of ch4 to ch3.

SnapAmp0 has PWM Outputs 8 9 10 11

SnapAmp1 has PWM Outputs 12 13 14 15

A brushless motor requires a pair of PWMs

Since your Z motor will be connected to the "B" side of SnapAmp1 you will set Output0 = 14 (which uses PWM14 and PWM15)

SnapAmp0 has Encoder Inputs 8 9 10 11

SnapAmp1 has Encoder Inputs 12 13 14 15

Since you are connecting the encoder to the 2nd input on SnapAmp1 you will set Input0 = 13

The AutoPhaseFind.c as well as a properly configured HomeBrushlessSnap4.c program should contain everything to configure everything needed to perform the operation from a default power on condition.

Sorry it is so confusing.  We probably have over used the term "channel".  There Axis channels which is a basically a software entity that completely defines and handles a motor Axis.  KFLOP has 8 of these available.  Then there are Device Channels.  Devices can be Encoders, PWMs, DACs, ADCs, Step/Dir Generator, etc...  Individual devices are typically hard wired to specific pins.  An Axis Channel is configured to use specific devices.  This KFLOP Functional Diagram may help:

http://www.dynomotion.com/Help/SchematicsKFLOP/KFLOPFunctionalDiagram.htm

HTH

Regards

TK

Group: DynoMotion	Message: 8298	From: Tom Kerekes	Date: 9/9/2013
Subject: Re: Getting started [1 Attachment]

Attachments : HomeBrushlessSnap3_V2-8298.c Hi Hugh, It looks like the AutoPhaseFind.c program is sometimes missing the index pulse.  That is why the Count/rev is inconsistent.  AutoPhaseFind.c looks for the index pulse every 1ms.  There is a AutoPhaseFindFast.c that looks for the Index pulse every Thread Time Slice (180us) and rotates slower.  You might use that instead.  I changed the HomeBrushless3.c program to check faster see attached. But I think the result is sufficient to determine your parameters.  The two important things are counts/cycle and Commutation offset.  The counts/cycle are probably exactly -2048.  The report is rounding to -2050 which is incorrect.  Use -2048 to compute the inverse distance per cycle. -0.00048828125 A small error will cause the commutation to drift off more and more each rev which wouldn't normally be a problem until you move a few revs.  But you missed the minus sign so that might be the problem. The ticking sound is because of an over current.  If SnapAmp detects an over current it shuts the output off for about 1 sec. (and the fault light on SnapAmp will go on).  Once something goes wrong the servo error will be large and thereafter each time the amplifier turns back on will usually generate another over current. You didn't describe what happens.  The motor should first rotate slowly to find the index mark and set the commutation before the servo enables.  Does this occur? Regards TK Group: DynoMotion Message: 8314 From: Hugh Sontag Date: 9/11/2013 Subject: Re: Getting started [1 Attachment] Hi Tom, I ran the "HomeBrushlessSnap3_V2.c" successfully when the motor is sitting unconnected mechanically, and it appears to servo, in that it can't be moved. But when I connected the motor to the Z axis mechanism and I run "HomeBrushlessSnap3_V2.c", the motor slowly moves about 1/8 rev CCW, then moves fast back to where it started, and then repeats. I increased the motor current to 13 (18 A), but it still doesn't rotate the motor to find the index pulse. I was able to set up some PID parameters and run a "Move" successfully. However, a "Step" always faulted, so that the data generated didn't seem to be useful. Since then, I've turned off the KFLOP and turned it back on, and it won't do that now. --> How can I get "HomeBrushlessSnap3_V2.c" to run successfully when it's connected to the Z axis mechanism? I also have a question about commutation offset. It seems that the commutation offset should be calculated from the result of finding the index pulse. The calculation in "HomeBrushlessSnap3_V2.c", however, was originally 3*8000/2.0/12.0 * 1.02 I changed that to fix it at -341, which is what "AutoPhaseFind.c" calculated. However, that can't be right, because my encoder has 2048 counts per revolution, not 4000 counts. Also, the counts per revolution used to generate -341 was 2050, not 2048. --> What is the correct way to calculate commutation offset in "HomeBrushlessSnap3_V2.c" ? Thanks, Hugh On Mon, Sep 9, 2013 at 2:03 PM, Tom Kerekes <tk@...> wrote:   [Attachment(s) from Tom Kerekes included below] Hi Hugh, It looks like the AutoPhaseFind.c program is sometimes missing the index pulse.  That is why the Count/rev is inconsistent.  AutoPhaseFind.c looks for the index pulse every 1ms.  There is a AutoPhaseFindFast.c that looks for the Index pulse every Thread Time Slice (180us) and rotates slower.  You might use that instead.  I changed the HomeBrushless3.c program to check faster see attached. But I think the result is sufficient to determine your parameters.  The two important things are counts/cycle and Commutation offset.  The counts/cycle are probably exactly -2048.  The report is rounding to -2050 which is incorrect.  Use -2048 to compute the inverse distance per cycle. -0.00048828125 A small error will cause the commutation to drift off more and more each rev which wouldn't normally be a problem until you move a few revs.  But you missed the minus sign so that might be the problem. The ticking sound is because of an over current.  If SnapAmp detects an over current it shuts the output off for about 1 sec. (and the fault light on SnapAmp will go on).  Once something goes wrong the servo error will be large and thereafter each time the amplifier turns back on will usually generate another over current. You didn't describe what happens.  The motor should first rotate slowly to find the index mark and set the commutation before the servo enables.  Does this occur? Regards TK Group: DynoMotion Message: 8320 From: Tom Kerekes Date: 9/11/2013 Subject: Re: Getting started Hi Hugh, I think you need to increase the energizing current to have enough torque to overcome some resistance when mechanically connected.  Without enough current the motor will miss-step much like a stepper motor stalls, but with much bigger "steps".  The C Code: A=10.0f;   // set coil current amplitude PWM units determines the amount of sinusoidal current that will be used.  You can look at the Analog Screen to see what the current is.  You should use enough current to have enough torque to move reliably, but not too much to over heat your motor.  Again this mode of movement (before the commutation has been figured out) is like a stepper that just burns heat to hold the rotor at a certain position. The change you made to increase the allowed peak current won't have any effect.  That doesn't have any effect on the relatively small sinusoidal drive to the motor coils. It is more for safety protection in case of a short or some other abnormal reason where the current spikes to a huge value.   Within microseconds of exceeding the peak current setting SnapAmp will fault and turn off the outputs for ~ 1second. The Servo and a "Move" will only work after the commutation phasing has been properly set by the HomeBrushlessSnap3.c program.   After turning off KFLOP you will need to re-phase find. That calculation for commutation was based on some other motor.  It is just a comment.  Delete or ignore it. I think the Auto Phase find value of -341 is probably correct.  After you get things working you might try adjusting it slightly to see if you can find something better.  If you know how the brushless motor is constructed, wired, and how the encoder is mounted it can be possible to compute the commutation offset.  But it is very complicated.  The method AutoPhaseFind uses is more empirical.  If you energize the motor windings in a DC manner the rotor will naturally snap to a position that generates zero torque.  This is the worst "commutation" for that position.  If you were to force the rotor away from that position the torque will increase in a sinusoidal manner.  It will increase to a peak.  Then decrease. Then eventually go negative (that will cause it to "miss-step" and snap to the next pole).  The peak torque will occur 1/4 of a cycle away from the zero torque position.  So basically AutoPhaseFind finds the commutation that generates no torque at the index pulse position, and it figures out how many encoder counts the motor moves for one complete cycle, then offsets by 1/4 of a cycle.  BTW it has always somewhat amazed me that no matter how the motor phases are wired, or how the encoder phases are wired, this will always work. HTH Regards TK Group: DynoMotion Message: 8330 From: Hugh Sontag Date: 9/12/2013 Subject: Re: Getting started OK, I'll use the -341 that AutoPhaseFind.c found for commutation offset for now. Regarding your statement: "The peak torque will occur 1/4 of a cycle away from the zero torque position.  So basically AutoPhaseFind finds the commutation that generates no torque at the index pulse position, and it figures out how many encoder counts the motor moves for one complete cycle, then offsets by 1/4 of a cycle." Could you please tell me what you mean by "cycle" ? For a 3-phase brushless DC motor, how does this relate to one revolution of the motor? --- I assume that the commutation offset is fixed for a particular motor/encoder combination, basically it's the relationship between the motor's poles and the index of the encoder. Therefore, I'll need to find the commutation offset for each of the other 3 motors we have in our CNC router setup. I had thought that it might be advisable to find the commutation offset at startup, to allow for scenarios like changing out a motor/encoder combination, or if the encoder slipped.  It would probably be best to find the commutation offset, then to compare it to the the expected commutation offset, and issue a warning if they're substantially different, since if the encoder slipped on the shaft, it might slip again, and one would want to know that there is likely a problem. Thanks, Hugh On Wed, Sep 11, 2013 at 5:58 PM, Tom Kerekes <tk@...> wrote:   Hi Hugh, I think you need to increase the energizing current to have enough torque to overcome some resistance when mechanically connected.  Without enough current the motor will miss-step much like a stepper motor stalls, but with much bigger "steps".  The C Code: A=10.0f;   // set coil current amplitude PWM units determines the amount of sinusoidal current that will be used.  You can look at the Analog Screen to see what the current is.  You should use enough current to have enough torque to move reliably, but not too much to over heat your motor.  Again this mode of movement (before the commutation has been figured out) is like a stepper that just burns heat to hold the rotor at a certain position. The change you made to increase the allowed peak current won't have any effect.  That doesn't have any effect on the relatively small sinusoidal drive to the motor coils. It is more for safety protection in case of a short or some other abnormal reason where the current spikes to a huge value.   Within microseconds of exceeding the peak current setting SnapAmp will fault and turn off the outputs for ~ 1second. The Servo and a "Move" will only work after the commutation phasing has been properly set by the HomeBrushlessSnap3.c program.   After turning off KFLOP you will need to re-phase find. That calculation for commutation was based on some other motor.  It is just a comment.  Delete or ignore it. I think the Auto Phase find value of -341 is probably correct.  After you get things working you might try adjusting it slightly to see if you can find something better.  If you know how the brushless motor is constructed, wired, and how the encoder is mounted it can be possible to compute the commutation offset.  But it is very complicated.  The method AutoPhaseFind uses is more empirical.  If you energize the motor windings in a DC manner the rotor will naturally snap to a position that generates zero torque.  This is the worst "commutation" for that position.  If you were to force the rotor away from that position the torque will increase in a sinusoidal manner.  It will increase to a peak.  Then decrease. Then eventually go negative (that will cause it to "miss-step" and snap to the next pole).  The peak torque will occur 1/4 of a cycle away from the zero torque position.  So basically AutoPhaseFind finds the commutation that generates no torque at the index pulse position, and it figures out how many encoder counts the motor moves for one complete cycle, then offsets by 1/4 of a cycle.  BTW it has always somewhat amazed me that no matter how the motor phases are wired, or how the encoder phases are wired, this will always work. HTH Regards TK Group: DynoMotion Message: 8337 From: Tom Kerekes Date: 9/12/2013 Subject: Re: Getting started Hi Hugh, By a "cycle" I mean one complete sinusoidal cycle of the phases of the motor.  I think most 3 phase motors are usually have 4 poles which result in 2 cycles per revolution of the motor.  But it depends on the motor construction. To help understand this you might change the AutoPhaseFind Program to rotate very very slowly.  Then watch on the Analog Screen how the current in one of the phases varies up and down through a complete sinusoidal cycle.  At the same time watch the shaft rotation of the motor. HTH Regards TK