[jsheerin]

I had a look too. I am almost tempted to buy one, then sell it back.

My recommendation would be to find a cheap one if you do this. I've seen plenty of expensive ones sit on ebay for a long time. I've typically paid $200 or less for a 10s or 5s motor, and less than $100 for anything smaller, but many of mine I've bought for significantly less.

If using an incremental encoder there needs to be an alignment step... If there are Hall or Fanuc signals then a much more satisfactory option is to start using Hall/Fanuc (for immediate full power) and then switch to using the encoder for sinusoidal commutation (on drives that support it)...

Do you know of a way to do this other than have the extra 4 inputs per motor for the hall signals? If I need the extra inputs, would I go with an extra 7i47 or something like that?

It has never actually been connected to a Fanuc motor. It has been tested in all configurations with the Mesa 8i20 and 7i39 (and, I think, with a generic Baldor BLDC drive)

Well that sounds much more positive than what could have been the case. As far as I know the only part of a Fanuc motor that is non standard are the hall signals, so it sounds like I could go without the halls entirely without too much pain. Edit - on the Fanuc motors I have, which all have incremental encoders. On the absolute and serial encoders, I doubt it's as simple.

Yes, that ought to work. I would be tempted to also try the Fanuc + Encoder to Hall to AMC combination too. That would be 7 inputs and 4 outputs (3 x Hall + PWM for speed control) Which I think fits into a parport.

The AMC drive I have is a B25A20: http://www.a-m-c.com/download/datasheet/b25a20.pdf So this takes +/-10V analog input, hall signals and a tach signal, but the tach signal is not required depending on the operating mode. So to confirm I would feed 3 encoder signals (a,b,z) and 4 grey code signals from the motor in to the parallel port and 3 hall signals and pwm out to the drive? Then I would need to convert pwm to analog, maybe just using a lowpass filter and some external electronics to amplify the signal? If so I will start playing around with that, as I'd like to get my motors tested and if I can do it without buying anything else at the moment, all the better. If you have anything specific you'd like to see to help validate your code, just let me know. Thanks for all the help with this!

Btw, my AMC drive only does trapezoidal commutation. This would be lower performance than the Mesa or Granite drives with sinusoidal commutation, right? I believe I read in one of your emc wiki articles that trapezoidal has torque ripple while sinusoidal does not.

[andypugh]

So to confirm I would feed 3 encoder signals (a,b,z) and 4 grey code signals from the motor in to the parallel port and 3 hall signals and pwm out to the drive? Then I would need to convert pwm to analog, maybe just using a lowpass filter and some external electronics to amplify the signal?

To do +/-10V you will need a two-pin pwmgen ( modes 1 or 2 LinuxCNC.org - PWMGEN ) in some sort of half-a-H-bridge. Easy enough if the drive has +10V and -10V pins to use. Otherwise there are DC-DC converters that can create the voltages (for example, this one makes +/-12V from a 5V supply Buy DC DC Converters RTD-0512 SMD isolated DC-DC,+/-12V 2W Recom RTD-0512 online from RS for next day delivery. ), then it is a case of low-pass filtering the PWM through an opto-isolator for each leg.

Btw, my AMC drive only does trapezoidal commutation. This would be lower performance than the Mesa or Granite drives with sinusoidal commutation, right?

Very slightly, it isn't a huge performance difference, but it is a lot quieter.

For the test, I wouldn't bother starting a GUI, you can set up the required components in a "simple" HAL file, and there is less to deal with when figuring it out.
I will try to find the time to create a sample file, but it would look something like this (set up the realtime threads, load the parallel port driver, pwm generator, bldc and encoder counter, add realtime components to thread, connect port pins to components). It will look a bit "hacker-ish" but the configuration you are proposing isn't really the sort of things the setup wizards were designed for.

[ebrewste]

The AMC drive I have is a B25A20: http://www.a-m-c.com/download/datasheet/b25a20.pdf So this takes +/-10V analog input, hall signals and a tach signal, but the tach signal is not required depending on the operating mode. So to confirm I would feed 3 encoder signals (a,b,z) and 4 grey code signals from the motor in to the parallel port and 3 hall signals and pwm out to the drive? Then I would need to convert pwm to analog, maybe just using a lowpass filter and some external electronics to amplify the signal?

You could use an encoder with hall emulation and get rid of the all the wiring and programming of your hall emulation scheme. You would have to figure out how to mount it, though. These are available at Digi-key.
http://products.cui.com/CUI_AMT303-V...df?fileID=1496

Also, a low pass, works, but make sure not to make it too low. You bleed a lot of phase from low passes (bad for performance). If you use a low pass system like that, I would use your drive in velocity (not torque mode) because it is less sensitive to the phase loss that torque mode.

Btw, my AMC drive only does trapezoidal commutation. This would be lower performance than the Mesa or Granite drives with sinusoidal commutation, right? I believe I read in one of your emc wiki articles that trapezoidal has torque ripple while sinusoidal does not.

Sinusoidal is nice, but it isn't absolutely required, especially with bigger motors. While the torque constant ripple is real, it is +/- (1 - cos(30)) = 1 - 0.866 = 0.134 = 13.4% ripple, which is not the end of the world. With low inductance motors, you can sometimes hear clicking on the transition between the phases, which can be an annoyance.

[andypugh]

You could use an encoder with hall emulation and get rid of the all the wiring and programming of your hall emulation scheme. You would have to figure out how to mount it, though. These are available at Digi-key.
http://products.cui.com/CUI_AMT303-V...df?fileID=1496

That looks like it could be a good solution in a lot of cases.
One possible niggle is that not all motors produce the same Hall sensor pattern, and not all drives want the same pattern. I worked out that there are 47 unique possible hall commutation patterns (I am not sure that they have all ever been used). One slightly wierd use of my software component would be to translate between motors with one pattern and drives expecting another.
(I have some Slo-Syn motors and some Bodine drives that really won't work together without conversion)

[jsheerin]

Thanks for all the info guys.

I'd rather avoid buying new encoders if possible. Those are $90 each and have half the resolution or less than the encoders that are already on the motors. Also, I believe if I went with the 8i20 drives, I'd still have to have the inputs to get the hall signals into the system before sending them over the serial port to the drive.

The AMC drive does have +/-10V outputs that can source 3mA, so it sounds like that could work for converting pwm to analog. However I am struggling to find an explanation of a half bridge that I can understand how to use to generate an analog signal, so if you know of a circuit or explanation I could reference, that would be much appreciated.

[andypugh]

However I am struggling to find an explanation of a half bridge that I can understand how to use to generate an analog signal, so if you know of a circuit or explanation I could reference, that would be much appreciated.

Possibly something like this, but doubled-up? (Note that I couldn't find a correct block for the opto-isolator, so ignore the 15k resistors and the fact that the LED is upside-down.

From Gibbs

[jsheerin]

You mean like this?

[andypugh]

Yes.
I have no idea if it will work and the behaviour with the PWM inputs floating might be worth considering.
Electronics is not my strong point.

[jsheerin]

Simulating it in spice it looks like it would work. Setting the inputs to pulse at 0 to 5V relative to ground produces a swing of +10 to -10V at the output relative to ground. However the current draw is too high to use the +/-10V available from my AMC amp as it is, so I might have to figure out something different. Electronics is not exactly my strong point either.

[andypugh]

However the current draw is too high to use the +/-10V available from my AMC amp as it is, so I might have to figure out something different.

That should just mean using bigger resistors and smaller capacitors.

Don't get too carried away, though. I can't find a parport config with the right number of pins. We need 6 inputs (F1, F2, F3, F4, EA, EB) and 5 outputs (H1, H2, H3, PU, PD). The available modes are 5in-12out or 13in-4out.

So, there is a choice of using Fanuc signals full time, and converting to Hall (likely to be lumpy, 6 into 16 doesn't go) or using the encoder when bldc gets a reference and only having a single PWM channel.

I will PM you a HAL file for the former a bit later on.

[ebrewste]

That should just mean using bigger resistors and smaller capacitors.

That is true. Also, don't set the low pass corner frequency below 750 Hz or so or you will limit servo performance.

[jsheerin]

I think the attached circuit should work (lowpass at ~1600Hz). I ordered parts for this which should show up this week. Andy got me setup with a test hal file for running servos, so now I just have to clear some bench space and get everything wired up so I can test it all out.

[ebrewste]

I think the attached circuit should work (lowpass at ~1600Hz). I ordered parts for this which should show up this week. Andy got me setup with a test hal file for running servos, so now I just have to clear some bench space and get everything wired up so I can test it all out.

Looks good. What is the PWM period?

[jsheerin]

I don't know. I think I can set it, but Andy mentioned 5kHz could work. I found another article on pwm to analog that suggested a time constant for an RC filter of 50-100x the pwm period which would mean I'd need a significantly lower cutoff than what I have drawn, but I'm not too worried about that part. I have plenty of caps and resistors sitting around so I can measure the pwm frequency and behavior of the circuit on my scope and then change the filter pretty easily if necessary.

[jsheerin]

I've been digging through my cnc pile in the basement. I found my other AMC drives (I actually have 2 other ones). They're these: http://www.a-m-c.com/download/datasheet/bx15a20.pdf The difference on these other than a lower current limit is that they don't have a hall velocity mode - only current, open loop and tach velocity.

[jsheerin]

I took a bit of a detour this weekend on a fun little project. I designed and built myself a vacuum chuck for my router and then designed a pcb for the pwm to analog converter circuit above using Eagle, generated code using pcb-gcode, and milled it. I've wanted to do this for a long time - I've had the parts for the vacuum chuck for about 3 years but never got around to building it. I'm not sure if the board I made will work (some of my pads and vias are a bit on the small side), but it was fun and I learned some good stuff. Pics of the process below:

After I got everything designed, some dimensions written down, and the cam done, I cut some aluminum to size.


Next I routed pockets in a piece of mdf for some 1/4-20 nuts and drilled matching holes in the aluminum plate. The plate will get bolted to the mdf, and the mdf clamped to my router table. I'll surface the mdf to get a plane surface for the chuck to sit on.


Close up of the nut - I had to tap the nuts in with a hammer. It was a nice fit.


The aluminum mounted to the mdf base (already surfaced) on the router.


While the router was cutting the vacuum chuck pattern, I got busy with my drill press and turned this bar into swiss cheese. Then I attached some ball valves and hose fittings. This will be my vacuum manifold. My vacuum source is a venturi vacuum from joewoodworker.com run off my air compressor.


The finished vacuum chuck. After it was done on the router I drilled and tapped the holes to distribute vacuum to the different zones and mount the hose fittings. The o-rings drop into the rectangular outer channels.


The finished setup with some pcb material mounted on it. It works great! I made the edges 5.1" apart so I know how much to adjust my offsets when I flip the board to cut the second side.


The first side of the board being milled. I'm using a 60degree v-bit.


The board flipped and getting the second side milled. I put packing tape over the milled side of the board to get it to seal to the vacuum chuck. This worked okay, but was not the best. I had a small leak, but my vacuum pump could keep up with it and keep the vacuum above 20" of mercury.


The finished board, first side.


The second side. Now I'm waiting for some drill bits to try drilling it, but I might redesign it and cut it again in the meantime so it's easier to work with. In any case it was a fun weekend project (which also included learning Eagle and setting up pcb-gcode to generate the code).

[Khalid]

Very nice work on the Vacuum Chuck..How much cost incur including design and machining time..I see good improvement on CNC pcb..

[jsheerin]

Thanks Khalid. To design the vacuum chuck it took me an hour or so, and another few hours to do the cam and cut it out along with making the vacuum manifold. It also took me a few hours to learn eagle. I redesigned my board layout last night using bigger pads and vias and then generated new code. That took about an hour. The more I use the program, the faster I get with it. I watched a series of tutorials on youtube about how to use that got my going very quickly.

[llilrex]

nice work. I would like to mill PC boards on my mill. I have the carbide engraver bit for it. now I just need the mill

[jsheerin]

I bought a new motor a few weeks ago - a Fanuc 30F. I don't actually have data on this one, but I believe it's an older version of the 30S. I was updating some of the calculations in my design spreadsheet for the motor performance, and decided this would help me out on the x axis (plus I found a deal on ebay). This motor will actually give me less peak force than a 10S motor which was my previous plan due to a lower torque constant and my fixed upper current limit with whatever drive I go with, but it will give me much more continuous force due to the larger continuous current capacity. If you can't tell from the scale the pics, it's pretty large - about 80 pounds. The only problem was that since I didn't have any data for it, I didn't know what encoder it had. Turns out it's a serial A encoder. Bummer. I took off the encoder, and it sits in a round rebate in the end of the motor. There's an Oldham style coupling between the motor shaft and the encoder shaft. I looked at the encoders on some of my other Fanuc motors and they're a different mounting type (see pic of a 0S 2500ppr incremental encoder), so I'll probably just look for a standard incremental encoder on ebay to replace this one with.