[peteeng]

Hi Gecko - If you want to reflex the motor to the ballscrew that's fine but use a big belt! If you can use the ballscrew direct that's better. Peter

[joeavaerage]

Hi,
no need to get too carried away, OP is using 400W servo after all, say 1.27Nm (S1).

I used a 15mm wide by 5mm pitch belt on my fifth axis servo, and that is 750w or 2.4Nm(S1), and I suppose there must be some lost motion but its smaller than I can measure.

Craig

[Ansh_automation163]

Hi Craig - The current going to the brake coil is DC so is inductance a real problem? Peter

Hlo sir
im application engineer, working on Syntec & Fanuc controllers with servo motor & drives, If you need any help i can do dear.

Regards
Pankaj Rana ([email protected])

[GeckoSub]

Hi Gecko - If you want to reflex the motor to the ballscrew that's fine but use a big belt! If you can use the ballscrew direct that's better. Peter

Yeah, actually when I was still designing my own build, I was insisting on direct drive for simplicity and alignment ease - and that's also what this machine has, albeit with steppers. The reason I am considering belt now is strangely enough a bit of the same reasoning... Thing is, the ballscrew and its mounts are nicely aligned now so converting to belt may be a cheaper, easier and cheaper than a new mount.
But also, the main "issue" is that my machine has some non-standard ballscrew and motor mounts on it. They are 30mm in center height whereas everything else I can find to replace it is 25mm.
(I need new motor brackets to fit the slightly larger servos vs. the old stepper. I can't use a spacer plate on the old mounts as the gap between the ballscrew and servo is already pretty long.)

So far, I changed just the motor mount on the Y and now use a servo there. But I have a 5mm riser plate under that new mount to match the 30mm center height of the old one. But that plate is still just 3D printed and not stiff enough and seems not to be precisely the right height as there's there's some slight "binding" at the end of the travel. So, will be making one in aluminum shortly. If I nail the height and the ballscrew runs as smooth as it should, then I think I will just do the same for the other two axis and buy new direct motor mounts for those.

The other reason for a belt on the Z was just one of aesthetics. The Z servo I have with its brake is way taller than the stepper that's on it now. So, I will have to cut a hole in the top of the cover. Not a biggie, though.

[GeckoSub]

A Few More Bits & Bobs

I lost a few balls when I pulled the table off, so I 3D printed some dummy rails to help keep them in place. I have just ordered a used rail block which should have the same preload/same size balls and my micrometer is incoming so I can double check:


Oh, I am not always that stupid. Or rather when I am, I sometimes manage to mitigate my foolishness, haha. E.g. I replaced the missing balls on the table with some from the Z-assembly as I can take the Z on and off myself but can't take the table off alone. So, the new balls will go into the Z when I get them and it will be easier this way.

Also, this little baby arrived:


The tool setter was like most other parts here second hand and again at a price I couldn't resist. Let's hope it's not banged up inside. Looks good from the outside.

Also, I mentioned earlier that the machine had rolled ballscrews. I made that statement because on the sellers pictures it looked like that. I think the tell tale is that extra groove deep in the middle of the groove for the balls (I could be wrong). But that was just a line of fat, dark grease and I'm now pretty sure they are ground. Also, they say C5 on them and are proper Japanese made NSKs.

Oh, it turns out the pics didn't stick when I made this post.

Here's the pic of the dummy "ball keepers". But yes, once I get the replacement balls - and grease fittings for the X and Z - hopefully I wont need to take the blocks off again for a long while. This pic was taken when I was putting balls into the Y blocks. The table is really chunky but it doesn't look like the mating surfaces for the blocks are precision ground/scraped - I think just milled but hopefully good enough for my use.


And the little baby I referred to is a Metrol tool setter


I actually tried testing it the other day by using a multimeter looking for continuity when pressing the plunger. No dice. I hope it's because it actually needs to be wired to 24V to even work. I was just thinking it was a simple mechanical switch - it only has two wires.
I'll need to go consult the manual, just too much other stuff to deal with these days

[peteeng]

Hi Gecko - I shudder when I see unkept balls! I've had a couple of times the balls have got out and in a workshop they seem to scatter really fast on the floor.... Madly counting and looking for the right amount to put back in. The first time could not find them all and had to run the car with one less. The keeper looks good & reduces my anxiety Peter breath in , wait , breath out ahhhhhh

[GeckoSub]

Hi Gecko - I shudder when I see unkept balls! I've had a couple of times the balls have got out and in a workshop they seem to scatter really fast on the floor.... Madly counting and looking for the right amount to put back in. The first time could not find them all and had to run the car with one less. The keeper looks good & reduces my anxiety Peter breath in , wait , breath out ahhhhhh

Yeah, those 3D printed ball keepers were nice - when I put the cars back on the rails, they just fell out of the way, no drama, no stress.
But this reminds me, I need to order another used car so I can cannibalize its balls. If it's the same preload, hopefully they are the same size, otherwise I have to track them down from the "replacement ball shop". There's a shop in the US but there should be a Chinese equivalent.
Anyhow, I am getting ahead of myself.

But I do need to order a granite square. I've measured and corrected what I can without one and yeah, not too impressed. The machine definitely needs to be trammed from scratch. E.g. I am 0.2mm out of parallel from one side of the table to the other (along the X). Can't wait to get into that rabbithole some more... haha.

As an aside, not sure I "admitted" to this, but when I was taking the first cuts, I had an issue where the zero on the part was kinda off. I mean, I would zero to where I had set it in the CAM and the CAM matched the CAD, but then when I hit cycle start it would do the tool path just behind the part. So, I had to cheat by setting the zero at the front of the part instead of on the rear as in the CAM so the tool path would actually be on the path.

I thought maybe it was a post processor issue until I decided it was more likely rookie me who was to blame. So, once I dared looking inwards I quickly realized I had my Y movement inverted in the controller.... I fixed that with a few button presses and a reboot and the tool paths now line up as they should.

Now, I just have to figure out how to delay the start of the first cut until the spindle has actually ramped up to full speed!

[joeavaerage]

Hi,

Now, I just have to figure out how to delay the start of the first cut until the spindle has actually ramped up to full speed!

Should not be that difficult.

In Mach4 you set the max and min spindle speeds and the spin-up and spin-down times. Thus when I call for the spindle to start and then cut a part, it waits for the spin-up time (10secs) before it starts
the cutting moves.

You could alternately program (in Gcode) a delay, for example:
M3 S10000
G4 P5
G1 X...Y.......this would start the spindle and spin up to 10000rpm but delay 5 seconds before moving at the first G1 move.

Craig

[GeckoSub]

Hi,



Should not be that difficult.

In Mach4 you set the max and min spindle speeds and the spin-up and spin-down times. Thus when I call for the spindle to start and then cut a part, it waits for the spin-up time (10secs) before it starts
the cutting moves.

You could alternately program (in Gcode) a delay, for example:
M3 S10000
G4 P5
G1 X...Y.......this would start the spindle and spin up to 10000rpm but delay 5 seconds before moving at the first G1 move.

Craig

That sounds like a nice feature, that Mach thing. But I am on a Chinese standalone controller which is based on Fanuc logic as far as I understand. Perhaps, once I make the switch to LinuxCNC it'll have something similar. I will also check again if my controller has a delay option, but I feel like I would have seen it by now as I have been poking around the settings a lot lately.

G04 seems to be the way, it also came up in my searches yesterday but good to hear it from you, too.
For now, since I wont have a ton of tool changes and it's probably a good thing to start actually familiarizing myself with G-code, I will insert it manually in an editor.

But the goal would be to have the post processor do this automatically after each tool change and/or M03 (I may be talking out of my butt here, it's still early days).

In my search, I also found a post on the Fusion forum saying it should be possible to have the post do a 'dwell' where the dwell time is a function of the spindle speed. So, you could set it to e.g. 250ms per 1000rpm - which would dwell or hold the cutting for 6 seconds after the spindle starts for a 24K rpm spindle speed.
But when I tried inserting the proposed snippet of code into the post to make this happen, the post broke and wouldn't even show in the library.

So, yeah, G04 inserted manually for now

[GeckoSub]

OMG - First Post Processor Edit (It Works, kinda)!
It worked, I am kinda giddy, haha! So giddy I started a post with "OMG" which I swore I never would.

Granted, I actually don't know what I am doing which is not the best approach but I found snippets here and there of advice and the post processor now does a dwell relative to the RPMs of the spindle. This is actually the same code I found proposed yesterday where it broke the post processor, but for my post it had to go somewhere else and once I found that location, it worked.



It's set to dwell for 0.25 secs per 1000 rpm so it would dwell 6 seconds for a commanded 24K of spindle speed.

And it works, here's the resulting G-code of a simple 2D facing op:


14300 rpm, dwell of 0.25sec per 1000 revs which is 3.575s or 3575 as the G-code shows (which should vibe with my "Fanuc" controller where a P is in milliseconds).

And despite the fancy coloring of the editor, I know nothing about programming. I just followed an NYC CNC tutorial on what editor to use and how to import the Fusion preset and this is how it ends up looking.

One caveat though, it also adds the dwell to subsequent tool paths which uses the same tool and spindle speed and thus never M03'ed or M05'ed. That's the problem when you "code" by C+P without understanding what you're doing...
I would speculate the dwell gets triggered not by M03 as intended but by starting a new tool path/op but not smart enough to spot the difference in the code, yet.

[ajcnc]

If you have the space, move the table, not the gantry. Machine will be simpler, and for the same weight, much stiffer and capable of deeper cuts.

[GeckoSub]

If you have the space, move the table, not the gantry. Machine will be simpler, and for the same weight, much stiffer and capable of deeper cuts.

Heya AJ,
I went down a different path - bought a used machine in pretty much the same size as I set out to design and that one actually does have a fixed gantry.

[GeckoSub]

Servo and Stepper Mismatch - Tuning /Setting Ideas?

The machine is up and running but I am having some issues with circles ending up oval. It's not backlash, I have checked. But I am still about 0.2mm out of round on a 30mm circle. Same with squares orientated diagonally.

My lingo and knowledge may be flawed but I think it's an issue of the accel/decel of the steppers and servo not matching each other (I have a servo on the Y, steppers on X and Z). Maybe something do to with smoothing in the servo drive, too.

I have run auto tune on the Y servo but still a bit out of round. Also, I have disable S-curve on the servo drive which I think helped.

My theory is that I somehow have to dumb down the servo drive but any ideas on this? Any settings any of you would consider changing to somehow better match the stepper and controller?

@Joeaverage since you are running servos on all both X and Y did you ever have any issues with circles being oval? Or are they spot on?
I am asking since if I can't solve this with settings then I can either dumb down to a stepper on the Y or hurry up upgrading the remaining steppers to servos. But if I am going to have an issue tuning them, too then that's "scary".

BTW, Delta AsdaSoft tells me the inertia ratio on the Y is 4.0, so that should be within what the motor can handle. I am not having much luck getting the scope function to work to look for follower errors, though but will keep trying

[joeavaerage]

Hi,

My lingo and knowledge may be flawed but I think it's an issue of the accel/decel of the steppers and servo not matching each other

No, that is incorrect. The trajectory planner plans the cutting path and that includes different axes accelerating at different rates. The trajectory planner stipulates a set of point that the controlled point must achieve in
small, commonly 1ms, time slices. It does this knowing full well what the individual axes are capable of.

The only time you'd see an error is if you declared to the trajectory planner, ie your CNC software that the axis is capable of 0.2g accel when in fact the servo is tuned to only 0.15g
Provided the servos are genuinely capable of the acceleration that the machine controller believes thy are capable of then you should have no problem.

Also, I have disable S-curve on the servo drive which I think helped.

Don't use it, unless your motion control supports it. Let's say that the motion control expects an axis to accelerate at 0.15g but because you have some BS 'S curve' tuning going on in the servo
then the servo will never faithfully follow the trajectory that has been planned for it. If you wish to use S curve acceleration, then the trajectory must be composed that way.

@Joeaverage since you are running servos on all both X and Y did you ever have any issues with circles being oval? Or are they spot on?

Never had an issue. I do not drill holes, I use an endmill and circularly interpolate a hole. Means you do not have to stop and change drills all the time.....so I do dozens or even hundreds of circularly
interpolated holes daily. I expect them to be round to better than 10um, in fact I get often better than 5um. Having said that I have 750W servos on both X and Y axes, and both axes are heavy,
in excess of 115kg. The percentage difference between the two effective axis weights is small, ie you might as well say they are the same. Also while at rated (S1) torque each axis is capable of 0.27g,
but because I find that greater than I need, and more than my nerves will stand I have detuned all axes to 0.15g. Thus, I know that all of my axes a very VERY capable of matching the 'easy'
trajectory that Mach4 serves up.

BTW, Delta AsdaSoft tells me the inertia ratio on the Y is 4.0, so that should be within what the motor can handle.

Why are you relying on the software to measure and interpret that for you? The software is looking at the dynamics and estimating the inertia ratio, it is not measuring it.
You should be measuring it yourself and supplying that as corroborated data TO the software, not the other way around. Have you not calculated the inertia ratio?
My machine the X and Y axes are both 8.75:1 with only modest variation due to the weight of the vice and the workpiece, and that is what I calculated, not guessed or let the software
guess for me.

Second issue is that the scope is USB, and is fairly slow. If you are going to use it and the other tuning aids, I suggest you use the best and most powerful PC you have. My machine control PC
is a little dual core Atom and is too underpowered to do a decent job of Autotune or using the software scope. If I wish to do that then I roll up my main development PC, still only an i5, but
then it works fine. My liittle Atom PC is fine for programming the drive, but just not for high data throughput.

Craig

[peteeng]

Hi Gecko - I'm not familiar with your system but as a general rule here's my view; 1) check that the axes are calibrated correctly ie that they move 10mm if commanded to move 10mm 2) In the configuration there will be a tolerance setting check that it is appropriate for what you are trying to do (eg simple woodwork 0.1mm high tolerance metal work 0.001mm etc). The controller is driving the tool down a road the tolerance width. If the controller can cut a corner it will, just like you do in a car, you take advantage of the wider road so you can go through the corner faster. The S-curve settings and accel settings will not directly create ovals. The trajectory planner will simply slow down the machine so it goes around the path within the tolerance limits. I expect that's the issue here unless your axis calibration is incorrect. There will be linear error settings and radius tolerances to consider....Peter

[GeckoSub]

Hi Gecko - I'm not familiar with your system but as a general rule here's my view; 1) check that the axes are calibrated correctly ie that they move 10mm if commanded to move 10mm 2) In the configuration there will be a tolerance setting check that it is appropriate for what you are trying to do (eg simple woodwork 0.1mm high tolerance metal work 0.001mm etc). The controller is driving the tool down a road the tolerance width. If the controller can cut a corner it will, just like you do in a car, you take advantage of the wider road so you can go through the corner faster. The S-curve settings and accel settings will not directly create ovals. The trajectory planner will simply slow down the machine so it goes around the path within the tolerance limits. I expect that's the issue here unless your axis calibration is incorrect. There will be linear error settings and radius tolerances to consider....Peter

Hi Peter,
Will start with you first

1). Axes Calibration
The axes are good. I checked them early on as the Z was actually set wrong when the machine arrived (the ballscrew on the Z was assumed to have 5mm pitch when in fact it is a 4mm). Also, I had to figure out how to set the gearing on the servo for the Y. I ended up setting it so it takes the same number of pulses as the steppers do. All in all, the axes are good now. Also, proven by the fact that the machine cuts squares to dimension as long as they are parallel with the X/Y.
The reason I thought it was some kinda of mismatch between the X stepper and Y servo "control part" is because the issue only turns up on diagonal squares and circles.

2). Tolerance Setting
The controller has a 'Contour Accuracy' setting of 0.020mm. That's how it was set when I received the machine. It sounds low enough to me that it's not what causes 0.20mm errors but I honestly wouldn't know. But I can try to set it lower.

I should also add that it seems the issue is less severe when I slow the machine down a whole lot. So, I CAM it with a fair bit of radial stock to leave so I am sure the finishing passes actually have something to cut and correct any ovality. Then I slow down the finishing passes, and repeat, and I think it helps (lost track a bit on my tests...). We are talking final cuts at 700mm/min in alu where the roughing would be more than three times faster (and I think the roughing is still not being pushed hard). I am not sure what this says, though...

[peteeng]

Hi Gecko - Is the tolerance +/-0.02mm or path deviation 0.02mm? Since the "error" changes with speed I expect it to be the tolerance settings. if the tolerance is +/-0.02mm then on the straights the "road" is 0.02mm each side of the tool and the controller takes it down the correct path. Around a rad due to the velocity limits or accel limits it will cut the corner 0.020mm to smooth the path. There is also functions within the path planner that simplifies the code eg if two straight segments are co-linear to the tolerance the controller will combine them to be one segment. This does not show up in your Gcode it's an internal thing with the planner. Same can happen with rads. Investigate your tolerances further or tighten it up and see what happens.. Peter

[GeckoSub]

Hi,



No, that is incorrect. The trajectory planner plans the cutting path and that includes different axes accelerating at different rates. The trajectory planner stipulates a set of point that the controlled point must achieve in
small, commonly 1ms, time slices. It does this knowing full well what the individual axes are capable of.

The only time you'd see an error is if you declared to the trajectory planner, ie your CNC software that the axis is capable of 0.2g accel when in fact the servo is tuned to only 0.15g
Provided the servos are genuinely capable of the acceleration that the machine controller believes thy are capable of then you should have no problem.



Don't use it, unless your motion control supports it. Let's say that the motion control expects an axis to accelerate at 0.15g but because you have some BS 'S curve' tuning going on in the servo
then the servo will never faithfully follow the trajectory that has been planned for it. If you wish to use S curve acceleration, then the trajectory must be composed that way.



Never had an issue. I do not drill holes, I use an endmill and circularly interpolate a hole. Means you do not have to stop and change drills all the time.....so I do dozens or even hundreds of circularly
interpolated holes daily. I expect them to be round to better than 10um, in fact I get often better than 5um. Having said that I have 750W servos on both X and Y axes, and both axes are heavy,
in excess of 115kg. The percentage difference between the two effective axis weights is small, ie you might as well say they are the same. Also while at rated (S1) torque each axis is capable of 0.27g,
but because I find that greater than I need, and more than my nerves will stand I have detuned all axes to 0.15g. Thus, I know that all of my axes a very VERY capable of matching the 'easy'
trajectory that Mach4 serves up.



Why are you relying on the software to measure and interpret that for you? The software is looking at the dynamics and estimating the inertia ratio, it is not measuring it.
You should be measuring it yourself and supplying that as corroborated data TO the software, not the other way around. Have you not calculated the inertia ratio?
My machine the X and Y axes are both 8.75:1 with only modest variation due to the weight of the vice and the workpiece, and that is what I calculated, not guessed or let the software
guess for me.

Second issue is that the scope is USB, and is fairly slow. If you are going to use it and the other tuning aids, I suggest you use the best and most powerful PC you have. My machine control PC
is a little dual core Atom and is too underpowered to do a decent job of Autotune or using the software scope. If I wish to do that then I roll up my main development PC, still only an i5, but
then it works fine. My liittle Atom PC is fine for programming the drive, but just not for high data throughput.

Craig

Craig! Thanks so much.
A lot to digest here.

Accurate Bores
First off, I am very happy to hear about the accuracy you are getting on the bores. I was also hoping to avoid too many tool changes (I am using ER collets) so boring and not drilling would be fine for me. This is important as it might speed up me switching the remaining steppers out for servos.

Inertia
Sadly, I wasn't smart enough to weigh the table when I had it off of the rails. So, I can't do the inertia calculations. But I don't think I will be as "lucky" as you in matching X and Y. If I had to guess, then maybe the moving mass on X would be 50-70% of the Y but that's really only a guess based off man handling the parts and allowing for an increase in mass with the bigger spindle and a few other things to be added on the X.

PC for Tuning
I did see you mention the need for a good PC somewhere else and it's a good point to make and not one I have seen others make at all, nor Delta themselves when I have wandered around the internet.
I bought an older mini Lenovo PC to run LinuxCNC on at some point and that's what I have been using. It's 2015 or so but does have an I7gen4 chip and 8gb of ram so hoping that would suffice. Again, I actually don't know. Maybe I can borrow a friend's newer, fancier laptop for a day to check.
But yes, I have been running the auto tuning on this.
It could also be that my AsdaSoft version is a bit wonky or my cloned data cable is. I will try some other AsdaSoft versions, too.

S-curve & Smoothing
I will double check that I have that off, but I am almost sure I do. Actually, I think I will go back and run the test cut with it on vs. off.
I did come across a poster here having ovality issues with some no-name Chinese servos and he said he found a smoothing setting in the drive software that he turned down which solved the issue. So, I will go look for something similar in the Deltas. On that note, just in case you can think of some settings off the top of your head that I should also try to kill, please do let me know

Trajectory Planner & Accel/Decel
Honestly, I don't know at all how the controller accel settings work in real life and zero help in the manual. Also, I am lost as to how many Gs an axis can run or where to start and how it corresponds to the numbers in the controller. The controller has three accel settings, I will list them here:
Rapid Acceleration:
10000mm/min2 (range: 0-600000)

Feed Rate Acceleration:
10000mm/min2 (range: 0-600000)

EMC Acceleration:
50000mm/min2 (range: 0-600000)
[I have no idea what EMC accel is and then manual doesn't mention it at all]

I'll attach a screenshot but the 'e.g. 1G 600000' statement confuses me. Do they mean 1G = 600000mm/min2? How did they get there?

For some reason, I feel like the accel settings are really low, but I don't know...

[joeavaerage]

Hi,

Sadly, I wasn't smart enough to weigh the table when I had it off of the rails. So, I can't do the inertia calculations. But I don't think I will be as "lucky" as you in matching X and Y. If I had to guess, then maybe the moving mass on X would be 50-70% of the Y but that's really only a guess based off man handling the parts and allowing for an increase in mass with the bigger spindle and a few other things to be added on the X.

Too fu***king bad, you'll have to guess the weight.

Have you read some of my posts on the momentum equation of an axis ......it is now time. My axes despite weighing something like 150kg each constitute only 7.5% of the total momentum for that axis with the ballscrew (32mm diameter, 750mm long)
being 80% and the armature of the servo being 12.5%. It is entirely probable that the inertia of your axis IS NOT dominated by the weight of the table but the ballscrew. This does most peoples head in, they cannot conceive that a large mass
travelling slowly (ie the table)is less important than a much lighter mass but spinning blindingly fast....but that's physics for you.

You need to do the momentum calculation...no guessing...no BS...measurement followed by careful and considered calculation. That will tell you what you can expect from your machine....and THOSE are the numbers you plug into
your controller. You can't guess....it just does not work that way.

Craig

[joeavaerage]

Hi,

I'll attach a screenshot but the 'e.g. 1G 600000' statement confuses me. Do they mean 1G = 600000mm/min2? How did they get there?

That is correct. 1g is near as damn it 10m/s²

That is 10,000mm/s² or 10,000 x 60 =600,000mm/min²

Craig