[Windscreen]

Hello all:

As you can tell from my post count, I'm new around here. A mechanical engineer by training, but not scared by electronics or computer programming. However, I'm sure I'll still ask some stupid questions.

I've been researching motion control systems for an RF45 class mill I will be purchasing shortly. My goal is to select components to get the best rapids for the best value. The direction I'm considering is Mach3 talking to a KFLOP with a pair of SnapAmps, which would be controlling 3 axis (for now, perhaps a rotary table later) BLDC servos. I like the ability to add on glass scales for dual loop feedback. The machine will have P3 ball screws, so backlash compensation would be mostly for machine deflection and belt stretch, I think.

I'm looking at the Keling NEMA35 BLDC motors, and probably the CUI AMT102 encoders from the same people
DC Brushless Motor KL34BLS-125 ( single shaft) | Automation Technology Inc

So, questions:
1 - Is there anything fundamentally flawed with this direction? Will this motor work well for servo positioning duties?
2 - Am I asking for problems by putting a pair of 660 W rated motors on a single SnapAmp?
3 - What sort of power supply should I spec for this? Are there any issues around using two power supplies in parallel?
4 - At what rate does the KFLOP talk to Mach? If I get a usb mpg that talks to Mach, will there be noticeable latency, or is there a better solution?

Thanks,
-Steve

[Al_The_Man]

I'm looking at the Keling NEMA35 BLDC motors, and probably the CUI AMT102 encoders from the same people

BLDC motors normally require Commutation pulses, which normally now originate on the encoder, unless the older Hall type come on installed on the motor.
If you need the commutation signals the CUI102 does not have, you can get the commutation version, but they have to be programmed on the motor, I have experimented with them and found them lacking and non consistent, intermittent motor reversal or stutter etc.
Support on them did not exist at the time, either.
You may be better off looking at traditional optical encoder from US Digital etc.
Edit:
Looking at the motor sheet they appear to have the older hall sensors on the motor.
Al.

[Windscreen]

I'm under the impression that with a kflop-SnapAmp combo, the hall sensors on the motor simply won't be hooked up, and the kflop-SnapAmp combo will hand commutation duties with input from the motor encoder I put on the back and feed to kflop.
Pinouts and (limited) description of operation here:
SnapAmp Connectors

So, back to my questions. Are there certain specifications that make a BLDC motor less appropriate for motion control than others? How appropriate is the Keling I linked to? This may be a better question for the general equipment forum than the Dynomotion forum.

How do you size the drive and motor combo? The drive is rated at 1000 W, the motor rotor power is shown as 660 W, but if you use the peak current value of 55 A at 48 VDC, that's 2.6 kW!! If you use the typical assumed motor efficiency of 75 - 80%, then we're talking about 825 - 880 W, electrical.

Thanks for any help!

-Steve

[TomKerekes]

Hi Steve,

Regarding Brushless commutation with SnapAmp: Yes the hall signals are not required because SnapAmp can perform Sinusoidal Commutation from the Encoder Position. But an Index Pulse is usually required to set the initial Commutation offset.

I'm not a big fan of CUI AMT102 capacitive encoders. They have much higher noise level than an optical encoder. Optical encoders usually have a noise level much less than their resolution. Capacitive (or magnetic) encoders often have a noise level higher than their resolution. One user captured some data that we plotted to analyze a "jiggle" problem. You can see the noise level on page 5 of the attached report. It is as high as +/-10 encoder counts at the 8192 counts/rev setting. However the noise is at high enough frequency that it tends to average out. So these encoders can work fine for some applications. Just something you should be aware of. I prefer optical encoders.

Regarding your questions:

1 - Is there anything fundamentally flawed with this direction? Will this motor work well for servo positioning duties?

I think it could work. The only thing I can see is that it is a little low voltage which means it takes higher current to get the same torque than a motor rated at 80V would require.

2 - Am I asking for problems by putting a pair of 660 W rated motors on a single SnapAmp?

No. SnapAmp is rated for 1000W continuous each Brushless Motor

3 - What sort of power supply should I spec for this? Are there any issues around using two power supplies in parallel?

Depends on the Power that you need. Typically your Acceleration Settings determines the required power. If you don't have enough power then the system will usually still work until you try to increase acceleration. SnapAmp has a nice feature to plot Supply Voltage while moving and accelerating.

Connecting two regulated supplies in parallel can be a problem if they each want to control the voltage to different values and fight each other. It would probably be better to use a separate supply for each motor (or each SnapAmp).

4 - At what rate does the KFLOP talk to Mach? If I get a usb mpg that talks to Mach, will there be noticeable latency, or is there a better solution?

We don't support MPG's connected to the PC because of the Windows latency. It is better to have an MPG connected to KFLOP directly for guaranteed instant response.

HTH
Regards

[Windscreen]

Tom:

Thank you very much for that great info. I didn't read the spec page close enough to see the 1000 W rating is per side of the SnapAmp. I have a couple more questions for you.

1 - What is the maximum frequency the differential input can take per channel from the encoder (or are resources shared between both channels of each quadrature). I want to make sure I don't end up with an encoder count x motor speed that is too much for the motion controller to handle.

2 - Do you have any suggestions for places to get 80 V BLDC servos that are reasonably priced (or cheaply priced, even ).

Thanks,
-Steve

[TomKerekes]

Hi Steve,

What is the maximum frequency the differential input can take per channel from the encoder (or are resources shared between both channels of each quadrature). I want to make sure I don't end up with an encoder count x motor speed that is too much for the motion controller to handle.

KFLOP can receive 1 million quadrature transitions (encoder counts) per second on all 8 encoder inputs simultaneously.

2 - Do you have any suggestions for places to get 80 V BLDC servos that are reasonably priced (or cheaply priced, even ).

No sorry I don't

Regards

[Windscreen]

KFLOP can receive 1 million quadrature transitions (encoder counts) per second on all 8 encoder inputs simultaneously.

Forgive the next newbie question, but what does that mean, in terms of maximum pulse rate? Since we need to establish the phasing of the the two encoder channels, it would seem the standard x2 per Nyquist would be inadequate, assuming both input A/D are on the same clock. I'm guessing something more like 4x the pulse rate is the minimum. Is there a rule of thumb for this?

I doubt I'll be pulsing that fast, but the engineering mind loves to learn!

Thanks again, Tom!
-Steve

[TomKerekes]

Hi Steve,

It doesn't seem like we are talking about the same thing. The encoder is (normally) a digital device that outputs digital square waves. There is no A/D involved anywhere. The A and B signals are 90 degrees out of phase (quadrature). For example a typical encoder might have 1000 lines/rev printed on it. One full rotation provides 1000 full square wave cycles for both A and B. There are 4 times as many A or B signal transitions that can be counted to measure position. This would result in 4000 quadrature counts per rev.




Because KFLOP can count 1 million quadrature counts per second the maximum rotational rate would be:

1,000,000 / 4000 x 60 = 15,000 RPM

For more info see here.

HTH
Regards

[Windscreen]

Tom:

I understand all that, and perhaps A/D was the wrong term to use. Perhaps logic gate or similar would have been a better term. So, in the case of your differential input, a difference of >0.2 V triggers the gate, logic, or what ever you call it to high. I'm thinking of the system from the stand point of sampling, and oversampling, and perhaps that is wrong. Does KFLOP poll the differential input 1M times a second, or does it make a time stamp when the input goes high, or something else? If it is something similar to a time stamp, then I can see where my thinking of it like a digital acquisition system and worrying about aliasing is all wrong.

Thanks,
-Steve

[Windscreen]

Hmm, I think what I wrote, polling 1M times a second or time stamping on a state change is effectively the same thing. I'm sure there is some digital trickery going on that I've not experienced before, so I'll leave it at accepting 1M quadrature counts per second happens, somehow.

-Steve

[TomKerekes]

Hi Steve,

That is correct. If you would like to know some details: The A B Signals are sampled digitally (as a simple high or low) by KFLOP's FPGA at 16.67 MHz. The signals are passed through a programmable digital filter with a default value of 7 (a signal must be sampled 7 consecutive times as a changed level before considered to be a valid change). The basic rule with quadrature is that only one signal may change at a time. A noise glitch that results in both signals changing at the same time can result in an error (loss or gain of counts). The filtered forward/backward quadrature transitions are counted by an 8-bit digital up/down counter. Every 90us the KFLOP DSP reads the counter and accumulates the position into a 64-bit value for the servo calculations. At max 1M counts per second the FPGA counter will count +/- 90 counts. An 8-bit counter will overflow at +/-127 counts so the max count rate of 1M is somewhat conservative.

HTH
Regards

[Windscreen]

Thanks for the education, Tom! This is way deeper into digital circuits than I've ever ventured.

-Steve