Newbie CNC Router Rebuild Seeking New Controller Advice

[dfbittner]

Peter,
In my day job I actually develop very high speed and high performance (ie very tightly tuned, not auto tuned) mechatronic solutions with industrial servo systems - as a mechanical engineer and partnering with the controls engineer. I appreciate exactly where you are going with measuring the compliance and performance of the mechanical drive train and "designing the overall system" based on that knowledge and the end goal - speed of the system as you pointed out in your example. This reminds me of a very common statement uttered in the industrial servo control world when a customer asks if the end effector will be at its desired position. The typical 'auto tuning' system engineer will respond "it's a servo and it will be there!". The reality is the servo itself will NEVER be at it's command position and of course then the end effector won't either - based on the presence of load inertia and drive train compliance. Thus the real question is not if but how far will the servo motor and end effector be out of position at any point in time? A well designed and tuned overall mechatronic system will minimize the position error and perform acceptably, but a poorly designed, well, not so much... Anyway I can fully appreciate how optimizing literally everything does enable higher speeds while maintaining acceptable position control. This is a great example of exploiting the seems between disciplines and as a result improving the system performance - a common fruitful ground for innovation. ie what is an S-Curve really, really? Is the motion profile continuous through acceleration? Or is it discontinuous? How about jerk? Is it infinite, finite, continuous? etc...

I can also appreciate the opportunity for adding arguably functionally limitless axes in the future with the Duet3D system. Also that there is much more development funding flowing into the additive vs subtractive manufacturing space now and for the foreseeable future. While I'm still pretty convinced I'll be very happy with the Ethernet SmoothStepper, I'm intrigued by this other option to include alternate software types and availabilities. I'm not really interested in migrating this system to a printer, so that's not a consideration.
Thanks,
DB

[peteeng]

Hi DB - nice summary. I see many opportunities with the Duet or BTT type solutions that are not available in the CNC tech and cost is parity or cheaper because they sell probably 1000's more printer boards then CNC components and the CNC components are disjointed (SW, firmware, machine controllers, motion controllers all separate and different developers going in different directions) not Integrated like the printer boards are. As you say the additive space is the fast moving area. I have only been poking into this for a few weeks, so a long way to go. I had decided to use a particular product for my next machine but it was unavailable due to the microprocessor shortage so it pushed me down a path I hadn't gone before. But the Duet path is attractive and that's where I'm going. I know the learning curve will hurt but when the throbbing stops I'll be in a happier place!

In industrial machines hybrids are becoming common, additive subtractive complementing each other. I think the Maker market may go the same way. So DB keep chewing on it and have fun along the way. Looking back at your #1 list I think Duet fits better than M4... I'm about to download Marlin and have a play...Peter

[joeavaerage]

Hi DB,
given your experience with industrial servos then you will be undoubtedly familiar with with notch filters within the servo drive tuning loop. These are great for taming low level vibrations
especially when the inertia ratio gets up to 20:1 or so. The one thing that they don't, or cant do, is predict the phase of the vibration, and yet if you are going to apply a corrective action
to the control input (over and above a simple envelope correction) you need phase information. Then and then only can you apply an 'active correction'. I would suggest that high performance
high acceleration machines, say pick and place machines, or maybe highspeed industrial robots would benefit from the increased complexity required to provide phase feedback. Such machines
have high acceleration in relation to their compliance.

Conversely CNC machines are very stiff, and have to be to accommodate cutting forces, and are very stiff in relation to their acceleration, and the benefit of the 'active vibration damping' is very much less.
Certainly in regards to the extra feedback necessary is well into the 'diminishing returns' category.

Any of the existing CNC software solutions including Mach, UCCNC, PlanetCNC, Acorn or even GRBL will provide perfectly adequate control to a CNC machine such that the limitations to what can
be achieved is NOT the control solution but your imagination and physical limits of the machine.

Craig

[ardenum2]

.
is this what you guys are talking about? :

[peteeng]

Hi Ard - Yes that is a simple demonstration of jerk control. It's great to have so my beer does not spill when I rest it on the spindle rack. Not sure how tech you are but "jerk or Jounce" is the differential of acceleration. ie its the rate of change of acceleration. Jerk creates vibrations as accelerations create inertial forces on masses. If there is no acceleration then there are no inertial forces. Vibrations can and do create very high accelerations which creates chatter and machine wobbles that we don't want. Trapezoidal velocity profiles have infinite accel at each change of velocity so we have very big unwanted inertial forces with this profile. S-curve velocity profiles smooth out the edges so velocity change is smoother and jerk is reduced. S-curve controllers do not know anything about the machine so the machines vibration character is not included, it's an open loop solution. Input shaping (passive damping) measures the machine natural vibrations, identifies the vibration modes then uses electronic damping at those modes to modify the accel or velocity profile to minimize jerk. The rexroth system is active ie the servos feedback vel, accel and jerk info in real time to the drive and it modifies the motion so its smooth as it goes. I have a jerk thread if you do a search. Peter

a note - I have found that marlin has an s-curve option so that's useful.
[2.0.x, Question] S_CURVE_ACCELERATION and JUNCTION_DEVIATION seem to have no effect? · Issue #11672 · MarlinFirmware/Marlin · GitHub

[joeavaerage]

Hi,

Trapezoidal velocity profiles have infinite accel at each change of velocity so we have very big unwanted inertial forces with this profile.

BS. With a second order planner the acceleration will go from zero when at constant speed or stopped, to its commanded acceleration, say 0.2g (or at some lesser acceleration as determined
by the coordinated toolpath) when it must accelerate (decelerate) to a different constant speed. There is at no time 'infinite acceleration' The acceleration changes from zero to its maximum instantly
certainly....but that is third order....acceleration is second order.

Craig

[peteeng]

Hi Ard et al - Here's a snapshot of the rexroth damper. It has an accelerometer on the motion platform which feeds back to control. The controller uses a PID device to filter the signal then feeds a modified velocity signal to the motor. You select damping ratios and pass limits etc or these can be automatically generated probably. The diagram also shows it can be selected to sense position. If sensorless this would be using input shaping. Some sort of initial vibration test would be done to identify critical freqs then these would be used for the damping. Good find Ard. Peter

[dfbittner]

OK I have the ESS/Ethernet Smooth Stepper powered up, connected by my breakout board via ribbon cable, and communicating with my PC by ethernet. Seems I need some information on my breakout board, which I don't have. I am unable to clear an error that seems to be related to E-Stop. Is anyone familiar with this BOB shown in this photo, and if so any advice if a hard wired E-Stop is required and where to connect it? Thank You

[joeavaerage]

Hi,
I've never seen a BoB like that.

In the first instance go to the ESS plugin Pins Config tab and reverse the 'sense' of the Estop pin, if its Active Low currently swap it to Active High.

The most likely explanation is that the BoB and whatever is hooked to it is Active High, and therefore the ESS interprets that as an Estop condition.

Craig

[joeavaerage]

Hi,
Looking closely at the photo you posted it look like numbers printed on the board may give you a clue as to what is what.

Can you read the numbers clearly?. I can see a two pin connector labeled 'EStop' in the lower middle of the board for instance.


What are the numbers alongside the row of plugs in the center of the board? They are partially obscured in the photo, but you may well see them clearly
and make some sense of them. What are the connectors on the right-hand end of the board, we are seeing only the solder side of those plugs. Are they connectors to the motor drivers?

Craig

[dfbittner]

Craig,
I can provide some insight that might help, and answer some of your questions.
-The large white junctions/blocks on the left side are the drive output connections. Further to the left on this board, not shown in the picture, is DC power conditioning (downstream of an off board rectifier).
-The row of soldered connections on the right side of the board are to a parallel port plug.
-Yes I can make out all the numbers/letters on the board clearly with the right photo.
-The numbers alongside the black row of plugs in the center board are J14 on bottom/one side and R3 on top/other side.
-The markings that I would bet are board level markings are www.k2cnc.com on bottom right in black and DC0744 on top right in white
-The large chip has what looks like a TI logo, and the following markings:
—Row #1 immediately after logo, 65H14FK E4
—Row #2 SN74ACT244N.
I’ve also uploaded another photo here that’s more clearly legible.

I will experiment with the jumper labeled E-stop and check the ESS plug in pins configuration as you suggested. Great tips.
Thank You,
DB

[joeavaerage]

Hi,
I wouldn't worry about the IC, its just a buffer, and the empty socket looks like a resistor array.

What I really want to see are the numbers of the pin/sockets that you believe are the motor outputs.

In this circumstance what I would do is disconnect all wires and plugs. Using Mach4 and the ESS Pins config set Port1,Pin 1 high, and then find that output on the BoB with a multimeter or oscilloscope.
Then repeat for Port1 Pin2, and so on until I had identified all outputs 1 thru 9, 14, 16 and 17.

I would then use a 5V probe with say a 1kOhm current limit resistor (just to be safe) on inputs 10 thru 13, and 15 and observe the ESS diagnostics on the Mach4 screen until you have identified each and every input.

This procedure is slow and tedious but way less error prone. The problem of trying to identify a 'block' of pins is that its very confusing....and so while you think it
will be faster and easier it turns out to be slower and harder than the methodical 'one at a time' approach.

Craig

[dfbittner]

Thanks Craig. Makes perfect sense. I will go through that tedious process.
DB

[joeavaerage]

Hi,
may I suggest that if you assign say Motor0Enable, one of Machs output signals to each output pin of the ESS in succession, you should be able to toggle th corresponding
pin in the BoB by Enabling/Disabling Mach.

For example in the attached pic I have assigned Motor0Enable to Port1,Pin1. Now each time you Enable or Disable Mach the ESS pin will toggle. Thereafter it remains to find
the corresponding output on the BoB.

Once you've found that pin then reassign the same output (Motor 0 Enable) but now to Port1 Pin2 and repeat. Slow and tedious but doable.

Its entirely probable that the numbering on the board will start to make sense and therefore you can identify many of the outputs with high confidence without this tedious procedure,
but this is a good and secure way to make that start.

Craig