Controller Advice Needed

[joeavaerage]

Hi routalot,

I'm amazed at the $2800 figure.The Dell Optiplex that controls my hobby machine cost about 1.4% of that and it includes the monitor.I suppose the desire to stay with Windows is what has made Microsoft the most valuable company in the world.

Are you being deliberately obtuse? That figure includes three 750w Ethercat servos/drives/cables......does that change your calculation?

Craig

[joeavaerage]

Hi,
this is the link to the $2800 Mach4 Ethercat solution, it includes a Vital Systems Ethercat board, Mach4 license and 3 x 750W Ethercat servos for $2895:

https://www.automationtechnologiesin...-3-axis-110vac

To be honest I don't think this is the best Ethercat solution, I mean all you are doing is replacing a Step/Dir board like an ESS with a Vital Systems Ethercat board.

This link is more what I had in mind. Kingstar and Interval Zero have collaborated to prouduce a realtime core for a Windows machine which therefore requires no external board whatever.
This kit includes a refurbished PC (inc monitor), Mach4 license, RTX64 runtime license and three Ethercat 23 size closed loop steppers for $1958. Given the hardware content and the licence suite its probably
not bad value.

https://www.automationtechnologiesin...ema-23-cnc-kit

As you know I'm not a fan of closed loop steppers otherwise I would have considered this myself for my new build. Ethercat has advantages but for hobby machines its a wash....and therefore stuck with Step/Dir.

Craig

[joeavaerage]

Hi,

What about if you decide on servos though?
I figure for 4 servo motors (proper AC ones). You're looking at 16 inputs for encoder and index points, and 5 to 6 inputs each for fault signals.

Where did you get that idea?

Modern AC servos have a drive which means the loop is closed by the drive, ie there is no need to wire the encoder back to LinuxCNC. In this circumstance an AC
servo is much like an open loop stepper, it requires only a Step/Dir input signal pair, an Enable input, and an optional ALARM output. Your proposed Mesa board
would handle AC servos just as easily as it will handle steppers.

If you wish to use old school DC servos with the encoder connected to LInuxCNC so that LinuxCNC closes the loop then, yes, you will require more inputs than your
proposed board.....but these old servos are a step back into history....why would you bother?

Craig

[routalot]

Hi routalot,


Are you being deliberately obtuse? That figure includes three 750w Ethercat servos/drives/cables......does that change your calculation?

Craig

Well if I include my steppers and drivers the figure goes up to about 9%.

[joeavaerage]

Hi,
if we assume $750 each for the Ethercat servos, and I base that on what I would expect to pay for750W Delta Ethercat servos, that leaves $2895-$2100=$645, less
the Mach4 license $645-$200=$445 for the Vital Systems Ethercat controller, and that sounds about right for Vital Systems. The estimate of the price of the servos
has a major effect on this calculation.

As I have said an external Ethercat board is not my prefered solution, the Kingstar approach requires zero external hardware to enact Ethercat. At this time Kingstar
do not offer the Kingstar/RTX64 runtime license as a separate item, only bundled with the kits, so its difficult to estimate what they charge for the license fee alone.
You would be right to claim that any license fee, lets say $400, would be $400 more than a LinuxCNC solution..... but in the scheme of things is about half the cost of
one servo alone.

Craig

[Sray]

ISo many options its hard to keep up with what Mesa offers.

It is overwhelming to someone new to the CNC world, that is for sure.

What about if you decide on servos though?
I figure for 4 servo motors (proper AC ones). You're looking at 16 inputs for encoder and index points, and 5 to 6 inputs each for fault signals.
This is why I suggest to look ahead. If the ultimate plan is for servos eventually then save yourself money and get the capable electronics now. Then you only have to change a few things later on rather than everything.

If the plan is just to stay with steppers on said machine then something like a Mesa 7i96e or 7i76e is prob plenty.

Well as I am new to all this I am not up to speed on all the technical aspects (inputs/outputs/step frequencies/etc) of all the components and what is needed for each of them to work with each other. I definitely need to understand all the I/O requirements better. It seems every new venture I get into there is so much to learn and understand. Why can't it be simple? LOL

I am picking up bits and pieces as we go though.

Works excellent with my UC300eth- UCBB- UCCNC. .

Could you explain exactly how step frequencies work? You had mentioned earlier about how you needed more than the 400kHz that the UC products offer. I am not sure what that means.

[JohnZ]

Step frequencies would be the number of pulses needed to turn the motor the number of rotations for the desired speed.

Example:

200 step / rotation motor
1/8 micro step
1000 rpm max speed

For each full rotation of the motor, the controller needs to send the drive 200 * 8 = 1600 pulses.

To reach 1000 rpm, you would need to send 1,600,000 pulses.

Since there are 60 seconds per minute, 1,600,000 / 60 = 26,666 pulses per second or 26.66 kHz step frequency.

Hope that helps.

Thanks, John Z

Sent from my SM-G991U using Tapatalk

[joeavaerage]

Hi,
provided you don't go overboard with microstepping most hobby machines will run quite comfortably at pulse rates well under 100kHz.

Servos are different, one they tend to run much faster, 3000 rpm and more, and two, with high resolution encoders.

For example my Delta B2 servos have encoders that produce 160,000 counts per rev. Lets say I wanted them to run at 5000rpm, their abs max, then the pulse rate required:
160000 x 5000 /60=13.33Mhz!!! Very few controllers could mange such a pulse rate, and then there's the problem of trying to signal the drives with what amounts to
radio frequency signals.

All servos have 'electronic gearing' which allows you to manipulate the 'effective' encoder resolution up to the built in encoders max resolution. In my case I chose to 'set' the electronic gearing
to 5000 pulses per revolution. My servos are direct coupled to 5mm pitch ballscrews which results in a linear resolution of 0.001mm per pulse or 1um, a convenient number
and very adequate for my work.

The required pulse rate is:
5000 x 5000 / 60=416.66 kHz.

This is still a fast pulse rate, but well within the capability of my SmoothStepper (4Mhz) and would be easy with your Mesa board (10MHz) but a little too fast for the UC400 (400kHz).
The signaling rate (differential) to my Delta drives is 500kHz, and (single ended) 200kHz. So with differential signaling they run fine at 5000rpm.

The Mesa boards are very fast, second only, to my knowledge, to Gallil (22Mhz). If you have to ask what a Gallil is you can't afford one!

Craig

[dazp1976]

Hi,
Where did you get that idea?
Modern AC servos have a drive which means the loop is closed by the drive, ie there is no need to wire the encoder back to LinuxCNC. In this circumstance an AC
servo is much like an open loop stepper, it requires only a Step/Dir input signal pair, an Enable input, and an optional ALARM output. Your proposed Mesa board
would handle AC servos just as easily as it will handle steppers
If you wish to use old school DC servos with the encoder connected to LInuxCNC so that LinuxCNC closes the loop then, yes, you will require more inputs than your
proposed board.....but these old servos are a step back into history....why would you bother?
Craig

This is definately not DC. It's pretty recent.
It's a 30Amp 2phase/3phase A4 system. 1.8kw.
http://servo.xlichuan.com/Private/Pr...6713131150.pdf

6 digital out for fault to controller. I suppose I would get away with only using some of them. Or I could daisy chain through an isolator.
I'd need- S-RDY- ALM- ZSP- TLC for definate. So that's 4 controller inputs.
Encoder feedback is: 3-way high-speed pulse output, output signal form: 5V differential signal.
1-way Z signal single-ended output signal. My UCBB board can take these directly - (incrimental encoder)
That's A+A-, B+B-, Z+Z-, CZ. So that's 4 inputs to my controlllers high speed optocouplers.
So I don't need to use this feedback then?. What about the CZ for positioning?. (spindle use).

I admit I haven't fully tested all the features yet.
Haven't even looked at the RS485 interface either. (shakes head).

[dazp1976]

Could you explain exactly how step frequencies work? You had mentioned earlier about how you needed more than the 400kHz that the UC products offer. I am not sure what that means.

I don't need 400khz per say.
Pulse rate info calculations see posts #106 & #107.

I'm only running open steppers for my axis atm. The drivers themselves are only rated up to 200khz. Open loop look to generally be around that.
I've set all my switches / steps per rev / unit etc to a calculation of around 100khz.
Plenty for what I need.

[joeavaerage]

Hi,

6 digital out for fault to controller. I suppose I would get away with only using some of them

Six alarms...what the hell for? You mean that all those people that use Clearpath servos with one....yes....one alarm output are no good?

I have one alarm condition for my servos. It combines over voltage, over current, over temp and following error. I did contemplate having two alarms per servo,
one alarm that requires an immediate Estop like over voltage, over current and over temp and one that would cause a FeedHold, namely following error. In the event
the only alarm I've ever had is following error......so I feel no real need for further alarms. Who cares about SDRY?....isn't that the exact opposite of ENABLE?
My SmoothStepper reports that the servo is stopped, if that signal is requested by Mach.....so there is no need for ZSPD either.

The encoder is fed back to the drive...you don't need or even want feedback to the controller. The manufacturers servo drive will always do a better job than a general
purpose controller no matter how flash the controller is.

For example I have a 1.8kW Allen Bradley servo and drive that I use as a spindle motor. As part of the 'servo mathematical model' stored in the drive is a step-wise
linear approximation of the magnetic saturation properties of the servo. Have you ever heard of a controller like LinuxCNC apply a non-linear feed-forward term like that?

The current trend in servo design is for distributed motion control over a comms bus like CANopen, Ethercat or Profibus. They all work similarly in that the trajectory planner
issues to each slave servo drive a destination position and a time to get there....and the servo drive does it all from there. If you have a move that involves three axes
then the three separate servo drives each handling their own assignment arrive at the destination in coordinated fashion. Do you suppose that manufacturers would bother if
was necessary for encoder feedback to the controller? NO, in fact they are going the other way, each servo drive is responsible for its own motion control.

Where you get the notion that you need dozens of inputs/outputs to run AC servos is just plain wrong. You CAN if you wish use more inputs/outputs but why?

Craig

[joeavaerage]

Hi,

That's A+A-, B+B-, Z+Z-, CZ. So that's 4 inputs to my controlllers high speed optocouplers.
So I don't need to use this feedback then?. What about the CZ for positioning?. (spindle use).

No, you don't need the encoder fed back to the UC300......whats the point anyway, the UC300 is not a feedback controller and is precluded from being a feedback controller
by virtue of UCCNC/Windows not being realtime.

To use that servo as a spindle set it up as a Caxis...then you can coordinate it with other axes, linear and rotary if you want. If you issue a Gcode block like:
g0 c270
then the servo motor, ie your spindle, will rotate 270 degrees with the acceleration/deceleration profile according to your axis motor settings.
The servo will arrive at its destination.....or if it fails, and that would be detected as a 'following error', ie its lagging behind the exact trajectory by greater than
your 'following error window'. Such a following error would cause an alarm. Provided you have tuned your servo and do not demand more acceleration than it can produce
then it will not alarm 'following error' for weeks at a time, if ever.

Craig

[rodw]

I don't need 400khz per say.
Pulse rate info calculations see posts #106 & #107.

I'm only running open steppers for my axis atm. The drivers themselves are only rated up to 200khz. Open loop look to generally be around that.
I've set all my switches / steps per rev / unit etc to a calculation of around 100khz.
Plenty for what I need.

In my testing with a Mesa Card 100 kHz is a practical limit unless you have high quality gear. There are certain timings that have to be respected in the step pulse stream and the sum of these timings determines maximum frequency. If you push these to the limit, you can get overruns that impact reliability. Personally, I am very comfortable with open loop steppers if well engineered. I've been running 25x microstepping for years.

Hi,


The encoder is fed back to the drive...you don't need or even want feedback to the controller. The manufacturers servo drive will always do a better job than a general
purpose controller no matter how flash the controller is.

For example I have a 1.8kW Allen Bradley servo and drive that I use as a spindle motor. As part of the 'servo mathematical model' stored in the drive is a step-wise
linear approximation of the magnetic saturation properties of the servo. Have you ever heard of a controller like LinuxCNC apply a non-linear feed-forward term like that?
Craig

In some situations you DO WANT feedback back to the controller as well as the drive. Its not uncommon with Linuxcnc to have 2 feedback loops one on the drive and one from a linear scale. Dual feedback loops on both velocity and position enhance accuracy.

The thing with Linuxcnc is you can do anything. If you had a model to calculate and apply variable feedback, it could do it. But the PID routines out of the box probably would not ever require it.

With Linuxcnc you can be spoilt with the number of inputs available. A safe machine should have seperate min limit, max limits and home switch. Do that on a gantry and you've consumed 12 inputs which allows pinpointing exactly where the fault is which you can't do on an ESS. Throw in another 4 inputs for drive faults and you are half way there. In fact I have managed to consume all 32 available inputs on my machine. But Plasma machines need a few more inputs...

[joeavaerage]

Hi,
if I want dual loop I would use Delta A2 servos which offer that. I agree it would be nice to have a realtime computing platform but I can achieve the same result.

exactly where the fault is which you can't do on an ESS

Incorrect, all my limits and homes are on individual inputs as the ESS gives me a total 31 inputs and 20 outputs.

Craig

[dazp1976]

Incorrect, all my limits and homes are on individual inputs as the ESS gives me a total 31 inputs and 20 outputs.
Craig

One main reason I bought the UC300eth. So I could seperate all the switches. Has 36 out, 49 in + 2 in+out of analogue. Plenty.
Other reasons were: ethernet and the options of Mach3/4 or UC to run it.

It was close between ESS and UC300 but fancied trying other than Mach for a change.

[Sray]

I have lead on a free Dell Studio 9000 computer. Definitely more computer than is needed but it is free. I am sure the extra power/speed/features isn't a bad thing. I hope!

I just ordered the Mesa 7i76e.

I am in the process of ordering the motors/drivers and power supplies. I want to make sure I get the correct Power Supplies.

For the motors/drivers I am looking at this one:
Mini size 1500W Switching Power Supply AC-DC smps output 60V dc power supply input 110vac

I can get it in different voltages but I am thinking of going with the 1500W, 60V, 25A one. With my motors being rated at 5.66A will this PS be a good fit?

For the Mesa board:
I have read it can be powered using both 24v and/or 5v. Do I just need one 24V PS? Will this run everything that board can run or will I require multiple PS's?
I also read "There is also the option of using separate field and FPGA logic power if you want isolation between the FPGA section and the field I/O section". More lingo I need to wrap my head around. LOL.

Questions:

1. Will one PS run both?
2. Do I need to isolate the power between the two?
3. What is FPGA logic and field I/O?
4. I am seeing different types of PS's (Regulated Transformer Switching Power Supply / Switching Power Supply for Light Transformers / Switching Power Supply Source Transformer AC DC SMPS / AC To DC Led Driver For Industry Led Light). Which one do I need or will any of these work?

Thanks ALL!

[dazp1976]

I have lead on a free Dell Studio 9000 computer. Definitely more computer than is needed but it is free. I am sure the extra power/speed/features isn't a bad thing. I hope
I just ordered the Mesa 7i76e.
I am in the process of ordering the motors/drivers and power supplies. I want to make sure I get the correct Power Supplies.
For the motors/drivers I am looking at this one:
Mini size 1500W Switching Power Supply AC-DC smps output 60V dc power supply input 110vac
I can get it in different voltages but I am thinking of going with the 1500W, 60V, 25A one. With my motors being rated at 5.66A will this PS be a good fit?
For the Mesa board:
I have read it can be powered using both 24v and/or 5v. Do I just need one 24V PS? Will this run everything that board can run or will I require multiple PS's?
I also read "There is also the option of using separate field and FPGA logic power if you want isolation between the FPGA section and the field I/O section". More lingo I need to wrap my head around. LOL.
Questions:

1. Will one PS run both?
2. Do I need to isolate the power between the two?
3. What is FPGA logic and field I/O?
4. I am seeing different types of PS's (Regulated Transformer Switching Power Supply / Switching Power Supply for Light Transformers / Switching Power Supply Source Transformer AC DC SMPS / AC To DC Led Driver For Industry Led Light). Which one do I need or will any of these work?

Thanks ALL!

Power supply amperage is a little overkill but it defo won't struggle running 4 of those 5.66A motors that's for sure.

Quick look at the board tells me you run 5v to the step/dir signals near the front end of board and 24v to the field for the rest at the back
These 2 look like they can be isolated from each other and should be.
Read manual very carefully.

[rodw]

Good choice,
This video I did will answer a lot of questions about the 7i76e.


The power supply you have picked should be fine for your motors. Some people use toroid transformers for motors. Most drivers need to rectify this to DC but some accept AC direct. Hence I have some custom made 60 volt AC toroids for a direct connection to AC compatible drivers.

You don't need a 5v supply. There is a link on the board to power the 5v section from 24 volt field power. Something like a Meanwell MDR-60-24. I'm using a MDR-40-24 on my next build

There are 2 sections re power. the 5v logic section also used for stepgens, encoders etc. and the 24 volt field power section which is used for inputs and outputs. You only need to add a 5V power supply if you have a lot of daughter cards connected to the 25 pin expansion connectors.

[Sray]

The power supply you have picked should be fine for your motors.

You don't need a 5v supply. There is a link on the board to power the 5v section from 24 volt field power. Something like a Meanwell MDR-60-24. I'm using a MDR-40-24 on my next build

There are 2 sections re power. the 5v logic section also used for stepgens, encoders etc. and the 24 volt field power section which is used for inputs and outputs. You only need to add a 5V power supply if you have a lot of daughter cards connected to the 25 pin expansion connectors.

Thanks for the info. Will look into the Meanwell's.

Power supply amperage is a little overkill but it defo won't struggle running 4 of those 5.66A motors that's for sure.

What amperage would be best? If I supply less amps don't I lose torque? I guess I was thinking I needed to meet the max amps for the motors in order to get the most torque out of them.

Step frequencies would be the number of pulses needed to turn the motor the number of rotations for the desired speed.

Example:

200 step / rotation motor
1/8 micro step
1000 rpm max speed

For each full rotation of the motor, the controller needs to send the drive 200 * 8 = 1600 pulses.

To reach 1000 rpm, you would need to send 1,600,000 pulses.

Since there are 60 seconds per minute, 1,600,000 / 60 = 26,666 pulses per second or 26.66 kHz step frequency.

Thanks for this explanation. Very helpful!

Hi,
provided you don't go overboard with microstepping most hobby machines will run quite comfortably at pulse rates well under 100kHz.

Servos are different, one they tend to run much faster, 3000 rpm and more, and two, with high resolution encoders.

For example my Delta B2 servos have encoders that produce 160,000 counts per rev. Lets say I wanted them to run at 5000rpm, their abs max, then the pulse rate required:
160000 x 5000 /60=13.33Mhz!!! Very few controllers could mange such a pulse rate, and then there's the problem of trying to signal the drives with what amounts to
radio frequency signals.

All servos have 'electronic gearing' which allows you to manipulate the 'effective' encoder resolution up to the built in encoders max resolution. In my case I chose to 'set' the electronic gearing
to 5000 pulses per revolution. My servos are direct coupled to 5mm pitch ballscrews which results in a linear resolution of 0.001mm per pulse or 1um, a convenient number
and very adequate for my work.

The required pulse rate is:
5000 x 5000 / 60=416.66 kHz.

This is still a fast pulse rate, but well within the capability of my SmoothStepper (4Mhz) and would be easy with your Mesa board (10MHz) but a little too fast for the UC400 (400kHz).
The signaling rate (differential) to my Delta drives is 500kHz, and (single ended) 200kHz. So with differential signaling they run fine at 5000rpm.

The Mesa boards are very fast, second only, to my knowledge, to Gallil (22Mhz). If you have to ask what a Gallil is you can't afford one!

Craig

Thanks for this. Although I needed to go through the numbers a few times it helps me understand how it works.

[rodw]

What amperage would be best? If I supply less amps don't I lose torque? I guess I was thinking I needed to meet the max amps for the motors in order to get the most torque out of them.

Its generally accepted that the total motor amperage can be about 140% of the power supply amperage so you only need about 16 amps. I think I first read that on the Gecko site.

Its not well understood that the motors only need 100% power when working at capacity. That does not happen often and rarely with all motors working together.
On my Plasma table I have no cutting forces so the drives are only working on acceleration. So I cut the amps by 50% when they are at a constant velocity when required power is < 15% of the maximum (motor dependent).. I've been meaning to try 30%.of full amps. This lets them cool so I can work them harder when needed. Not many drives support this.