[joeavaerage]

Hi,

I guess if this was running flood coolant, which it wont be, that would help. But I wonder what people do in these cases.

I thought you wanted to cut aluminum and at reasonable speed? And you intend to do that without coolant? I'm sure you've seen Piotr's machine in action and he has so much coolant
that you can no longer see the part being made.....why do you suppose he did that?

Craig

[GeckoSub]

Hi Gecko - Dowelling granite to alum is probably not the go. For parts that are not to come apart they can be bolted and epoxied (if you wax the surfaces they can be separated later). If they are to come apart use a wicking loctite like 290. This can be applied post tightening and works really well on lots of materials. Peter

Got it.
No dowels on the alu to granite but dowels between the alu parts. Thanks.
I don't think I will epoxy anything right off the bat, perhaps down the line. I have some nice laminating epoxy here as I do fool around with the odd carbon fiber from time to time.

[GeckoSub]

Hi,



I thought you wanted to cut aluminum and at reasonable speed? And you intend to do that without coolant? I'm sure you've seen Piotr's machine in action and he has so much coolant
that you can no longer see the part being made.....why do you suppose he did that?

Craig

Yes, I have seen the videos and it's quite possible I will go that way later on. I already have notes on his pump and its capacity, haha. But I have a tendency to overthink and while designing ahead is smart, I also don't want this project to become so big, it starts scaring me.

So, the thinking now is:
Start with the machine on a bench or stand, unenclosed except for end plates on the frame. Use MQL. In this phase, I wont be producing much. I will be learning the ropes, getting to know CAM, CNC software and generally getting to know the machine and making prototypes of the gadgets I have in mind (will post on that a bit later). So, I wont be cutting at max capability at all. Speed is not important to me in this phase and judging from a lot of other MQL machines, I think I will be more than happy with that at first.

Later on, once the product is proven I may need to make a bunch more on this machine or have a factory make it for me on industrial machines. At that point, I will know the machine better and it might be capable enough as a small batch production machine where speed starts to matter. At that point, an enclosure and flood coolant would possibly make sense. Also, if it turns out that a bit of experiments have shown I can cut titanium, then that will help drive the decision to go to flood coolant. Either way, I don't see that happening in the first 6-9 months of the machine cutting its first chips.

It's also very possible that I will need to get into home anodising and some laser cutting/engraving, so there are 2-3 other projects connected to this one. For now, I try to keep it as simple as I can and not fry my brain or scare myself with taking too much on.

As for the statement on experimenting with flood coolant and titanium, maybe it wont work, but I have a simple setup in mind: I have made allowance for screwing in acrylic end plates to the base. And with an o-ring channel cut into them or a gasket, I am sure I can make the whole base work as a waterproof tub. So, the simple experiement would be some low pressure coolant and a drain hole in the acrylic end plate to circulate the coolant. Now, it would probably still be flung over the sides despite them being relatively high. But might work for judging how much flood coolant will help. At the very least, the "tub" would be great for cutting carbon fiber parts submerged.

[GeckoSub]

What I am Setting Out To Make
Like a lot of you, I have a critical mind and often pick up a product and wonder if it could be improved upon.
Since I freedive this has manifested itself in wanting to make a better nose clip than what's currently out there. Nose clips are used for more advanced freediving where you dive without a mask since you don't want to have to use valuable air to equalize the airspace in a mask. With a clip on your nose, you also don't have to pinch with your fingers, meaning you can leave your arms down the side in a more streamlined position.

Anyways, I have been making them in 3D printed resin so far but they end up quite beefy in order to be strong enough not to break. So, with alu I can slim them down a whole lot and almost make them unbreakable.

Here are a few recent ones:



These are about 50mm on the max side so not big at all.

There is a toggle joint where the pads connect to the arms which is a feature that will take some clever machining and assembly solutions.
Yes, I need marine grade alu. Yes, I need anodizing. Yes, fixtures and work holding is not going be straight forward and yes, a 4th axis will likely be the first project once I have a bit more experience running the machine and doing CAM.
Suffice to say, I will have loads to learn, but I think I am up to the task.

I have a few other things I would love to make but for now, the nose clips are the main goal.

To tag a coolant note onto this. I think/hope MQL and airblast is more than adequate for these small parts for a long while.
Craig asked me why I think Fox uses flood. I think Fox is way more in production mode than I will be in the first 6-9 months and I speculate his parts have more need for it given that he does a fair bit of deeper slotting in enclosures where getting the chips out of the part is important. His parts are bigger, meaning a whole lot more chips to get in the way of the cut. Also, I think he said it really helped with the surface finish and lastly I think he wanted to do it for the satisfaction on doing it on a DIY machine.
In contrast, my parts are small, "open" and prototypes where I don't really care about being able to mirror myself in them.
Yes, it will be a tad messy, I will have to brush off or blast chips away, but I am ok with that for now. Part of the learning curve

[GeckoSub]

Hi,
no, optical ones are not the correct choice. Firstly they are optical, and when you get working there'll be chips and coolant all over the place an optical is bad in those circumstances.
Second issue is hysteresis. See if you can find a specification about how much an optical switch has to back off before you get reliable switch deactivation. The same applies with
Hall switches and proximity switches, you cant really tell when the switch is going to deactivate as the magnetic/ferrous target withdraws.

I use roller plunger mechanical switches. You see the the switch is mounted on slots and allows the switch to be slid back and forth a little. The roller is activated by a shallow ramp
that is attached to the axis. It too is slot mounted, but especially because of the shallow ramp angle the adjustment is quite fine. Another important feature is that if the axis over travels for whatever reason
the switch will not be wiped out or damaged as the ramp alone passes by without ever impacting the switch nor does the ramp cause the switch to overtravel and get damaged.



Yes, that is about it, adjust the switches to be as square as you can and use software for the last bit. Not sure how UCCNC handles that but I do know how Mach4 and the Ethernet SmoothStepper handle
it.

Mach4 has two programmable features to do with Homing, namely Offsets and Back-off. They are applied to each home switch. So with a gantry you'd have a master motor with its home switch, and a slave
motor with its own home switch. Both motors are home simultaneously but each will have (in addition to whatever slight mounting discrepancy between master and slave switches) but will have individual Offset and/or
Back-off applied for that last few um.

These switches are about $20USD depending on where you buy from and with care can get repeatability of 0.01mm, I regularly get 0.02mm. The Metrol switches you have proposed may result in better absolute accuracy,
but really who cares....what you want is repeatability and that can be had for less money.

Craig

Thanks so much for explaining the back-off feature and the pics of your switch and mount. Will put it on the list off what to check up on in relation to UCCNC before I buy into the system. I guess if it can be made to home to the "on" latching and not the "off" that widens my options. Will try to lock it down fairly fast though so I can incorporate the switch and mount into the design.
Already on the list: checking hysteris for switch options.

[pippin88]

If you want serious accuracy switches are not great.

Home to linear encoder index, or servo encoder index.

[GeckoSub]

If you want serious accuracy switches are not great.

Home to linear encoder index, or servo encoder index.

Gotcha, and I was hoping to do that but sadly UCCNC doesn't offer this as far as I am aware.

That said, I have been thinking a bit more. I think I am being overly paranoid about the power cuts and being able to resume an operation with super precision. As such, the squaring of the gantry is the most important and again, I think I am aiming for too much.
E.g. some really quick math shows that if I have a +/- 0.02mm switch on each end of a 500mm wide gantry, the most I can be out is 0.04mm which is 0.004 degrees and that's worst case.
For now, I will check hysterisis on the Metrols and if there's a kick ass deal on them to be had, I may still go for them.

[GeckoSub]

Hi,



I thought you wanted to cut aluminum and at reasonable speed? And you intend to do that without coolant? I'm sure you've seen Piotr's machine in action and he has so much coolant
that you can no longer see the part being made.....why do you suppose he did that?

Craig

Please don't take this as me trying to pick a fight, I am not
Not sure how stiff Andreas' FS3MG epoxy granite mill in the video below is compared to my "real granite" machine with thinner structures but his MMR with MQL is impressive to me and I am not sure he is pushing to the max yet given it's still fairly new. His machine has 25mm rails, 20mm ballscrews and steppers as far as I know but with a 5.5kw spindle(?). But anyhow, he moves alu fast enough for me. Datron, too comes to mind and nope, I am not saying I will end up with anything worthy of comparison. Just saying, there are machines out there that use MQL which impresses me and leads me to believe I can live without flood coolant for now.

https://youtu.be/ICIFecRdEQQ?si=LhC__VfewWEtrLt5&t=85

[peteeng]

They use high velocity air blower with mist lubricant. You can see this in the video. Plus he has his chip loads dial perfectly for the tool. His Fz=0.1mm which is perfect for al. Peter

[GeckoSub]

They use high velocity air blower with mist lubricant. You can see this in the video. Plus he has his chip loads dial perfectly for the tool. His Fz=0.1mm which is perfect for al. Peter

Ah, I gather the high pressure/velocity air does spray chips fairly far and his machine is definitely enclosed. Either way, for now, I will stick to my gut feeling that an enclose and/or will have to come a bit later. Too much other stuff to sort out now.

BTW. I really like when people put their feeds and speeds on their videos. I know machine stiffness has a lot to say but it's nice seeing what a dialed in tool should look like.

[peteeng]

Hi Gecko - Yes his chips are blissful to watch. OMMMM Peter

[peteeng]

Hi gecko - To put 70mm of granite into steel perspective. 70mm of granite has the same bending stiffness as 49mm of steel. But the granite is damper and it weighs 210kg/m2 and the steel weighs 387kg/m2 so the granite is ahead if you can work it. Peter

[GeckoSub]

Hi gecko - To put 70mm of granite into steel perspective. 70mm of granite has the same bending stiffness as 49mm of steel. But the granite is damper and it weighs 210kg/m2 and the steel weighs 387kg/m2 so the granite is ahead if you can work it. Peter

Thanks so much Pete! Much appreciated. I am pretty much set on granite. I sent the drawings to one factory a few weeks ago when the frame was at 50mm wall thickness and the initial quote wasn't that scary which is why I am moving ahead with this. I will go with 70-75mm and see what happens to the quote and also find one or two more factories to have some comparisons.

Here's the type of stuff they do:


If this goes well, it might be viable for others so don't mind sharing costs and such once we are bit further in the process. That said, what will make this work - if the quality is there - is that I also have access to really affordable shipping from China to the Philippines.

[edit] Oh, looking a bit closer at those pics it looks like at least one of the machines has a milled lip to but the rail up against. If they can do that precisely, I reckon that might be worth having. And the last one, there are some much smaller inserts on either side of the bigger one. If the bigger ones are for rail screws, the smaller ones might be for those little clamps or rollers that serves to align rails, too?

That said, I have seen a few hobby builds getting really good results in terms of aligning rails without either the lip nor the alignment rollers/clamps. I don't know how needed they are or if they also serve to make sure the rails don't move after the fact?

[peteeng]

Hi Gecko - I understand why your going down this path. I nearly did it a couple of years ago as the chinese granite quote was quite good. But I decided that I wanted to do it myself so have been working along the home casting route. Granite is minimum 70GPa but my grout is 50GPa. But I can make up for that in casting thicker with honeycomb style backs. Lighter too. Keep at it your getting there. I've just ordered the aluminium for my next build so I'm excited!! Tomorrow I order the plywood so its all go down under... Peter

[joeavaerage]

Hi,

f you want serious accuracy switches are not great.

Home to linear encoder index, or servo encoder index.

BS. As I've posted I get 0.02mm repeatability with mechanical switches, and with care I could probably get more. What is wrong with 0.02mm repeatability?

Mach4 has Index Homing....but I can't be bothered with it. Is that last 0.02mm going to make any real difference in the parts I make?.....no....or at least 99% of the time. Save your money/time/energy for those things that make a difference,
trying to extract sub micron homing is not a productive use of resources. Going from ball bearing cars to roller bearing cars, that is a useful step. Going from C7 to C5, that is a useful step. Going from manual tool change to ATC is a useful step.
Going from plain old flood cooling to a higher pressure filtered cooling is a useful step. Making a fourth and/or fifth axis is a useful step. Upping your CAD/CAM to to simultaneous five axis CAM to take advantage of your fifth axis is a useful step.

Will put it on the list off what to check up on in relation to UCCNC before I buy into the system. I guess if it can be made to home to the "on" latching and not the "off" that widens my options.

Tough s******t, UCCNC and Mach both home at switch de-activation.....end of story, and as I've already said its immaterial, you can use UCCNC to good effect without Index Homing. Yes, you could do better but it won't show in any
of the parts that you make....so its only about 'bragging rights'....nothing genuinely useful.

Craig

[joeavaerage]

Hi pippin88,
how do you home your machine?

Do you have sensors of some description and then home to linear scale index or what?

How about some photos of what hardware you use and what software settings you make for homing? In your best estimation, or measurement, what is the accuracy and repeatability?
What sort of accuracy and/or repeatability do you need to successfully make parts?

Craig

[pippin88]

Hi pippin88,
how do you home your machine?

Do you have sensors of some description and then home to linear scale index or what?

How about some photos of what hardware you use and what software settings you make for homing? In your best estimation, or measurement, what is the accuracy and repeatability?
What sort of accuracy and/or repeatability do you need to successfully make parts?

Craig

Craig

I use LinuxCNC - the docs explain the different homing options.

My router uses proximity switches. It's a wood router, tolerances are loose, 0.1mm doesn't matter
From memory I have it set to home onto switch at fast pace (so I don't have to wait a long time), back off switch, then at very slow feed advance onto switch until switch triggered (and that is the home signal).

For my new mill build, in very slow progress, I have Delta A3 servos with absolute encoders, plus linear encoders. I won't need to home regularly. I may just use hard stops for initial homing.
A3 have inbuilt homing functions, explained in detail in the manual.

I agree with your point above that super precise home switches are not the best way to spend money and better accuracy improvements can be had elsewhere

My point was that if you want very good accuracy, expensive switches are not the best way. Using a high resolution encoder is.

Super precise homing only helpful if doing a critical job over multiple sessions. Doing a part in one machine session (machine on whole time) -> precise homing makes no difference

What tolerance do I need? None, I'm a hobbyist, so nothing matters except my enjoyment. I'm doing a build with aim of ultra precision because I like the idea. I spend my money on this stuff, not fancy cars. But it is not necessary at all. Then I'm using the machine in a non-temperature controlled environment, so it's all out the window anyway.

OP - if you are wanting very accurate, you need to consider temperature. Alu and granite have different temperature expansion. Steel and granite much closer.

[joeavaerage]

Hi,
yesterday evening I set a PCB job running on my machine before I left for home for the evening.

This morning I came in and just for the hell of it I re-Homed my machine, flipped the board to do the bottom side. I use little pins to flip the board....so the top and bottom
line up nearly perfectly....but I am relying on the home position being the same as the day before, ie the repeatability of my home switches.

As you can see the holes, which are drilled from the bottom side and any misalignment or variance in the home position will result in the drill holes not being centered in the top traces,.....
and in fact they are not perfectly centered, I would estimate 0.02 to 0.03mm from perfectly centered. This is still a highly useable PCB. So while mechanical home switches are not the last word in
accuracy they can and are used to advantage here for a perfectly acceptable result.

I use a minimum trace width of 0.2mm and a minimum isolation between features of 0.2mm. In order to get useful boards 100% reliably I need to get about 0.02mm repeatability, ie about
ten times the size of the smallest feature. That's what I get with mechanical switches....and that's good enough.

Craig

[joeavaerage]

Hi,

I use LinuxCNC - the docs explain the different homing options.

I used LinuxCNC briefly about 8 1/2 years ago but did not encounter that particular option. The version I installed homed on switch de-activation as I had been used to in Mach3.

From memory I have it set to home onto switch at fast pace (so I don't have to wait a long time), back off switch, then at very slow feed advance onto switch until switch triggered (and that is the home signal).

Mach4 has similar programming/ homing tuning features. The travel of my mill is 375mm on all three axes, so if I set 20% of g0 speed (15m/min in current tune) it still only takes 8 seconds to traverse the full
length of the axis in the unlikely event it needs to travel that far. I don't sweat it, I can wait 8 seconds if that's what it takes! The back-off speed is more critical for repeatability and that is set to 5% of g0.

I agree with your point above that super precise home switches are not the best way to spend money and better accuracy improvements can be had elsewhere

Yes, that is very much my point, use your resources to best effect, and quite frankly good mechanical switches are adequate and thereby save funds for that new spindle or whatever.

My point was that if you want very good accuracy, expensive switches are not the best way. Using a high resolution encoder is.

I'm happy enough with the X,Y,Z home switches that I have, but I rather suspect that I'll want or need better with the two rotary axes. If I wish to still use the machine as three axis then the
fourth and fifth axes will have to be set to zero degrees with high precision or otherwise my X,Y,Z will be off by an excessive amount. I wont really know until I try. My ESS is down to about five
inputs to spare, so I'm going to have to be a little conservative about home and limit inputs. I have some ideas.

In particular I will have the trunnion axis (A axis) limits direct wired to the CCW and CW limit inputs of the servo. To have the trunnion rotate beyond +-90 degrees would be a spectacular fault!
I will use probably an optical sensor for limits and homes. I can totally enclose these so should be OK for contamination. I will probably have a small microprocessor to assist with homing and thus relieve
the need for too many inputs to the ESS/Mach.

A fourth and fifth axis is a 'project that keeps on giving'. If I'd known at the outset just how much time and money and energy a trunnion five axis would be I would have considered declaring myself insane
and be voluntarily committed!

Craig

[GeckoSub]

A Bit More Info on the Metrol Switches
These are the 5um repeatability Japanese small switches (they also have 3um and 1um versions).
Not sure why I haven't given up on them, yet... Oh, I know, it's likely because I can get them for 20-30 bucks used which given that they are USD 60-70 sounds legit. Anyhow, I went looking for some info on hysteresis and Metrol claims there is none. As in zero. How they made that work, I don't know. But here's a quote from a page of theirs:

METROL's contact-type sensors have no hysteresis.
In other words, the movement position (on) and the return position (off) are equal. The lack of hysteresis enables detection of fine motion of the workpiece.
[the on and off annotations are mine, but I think that's what they mean]

There's also a tech guide here where they call it movement differential and state that their switches have none (screenshot from that PDF):




As you have pointed out, the other solutions are fine too and if the job doesn't get interrupted none of this really matters.

Anyhow... I will go with the Metrols if the ones at 20-30 bucks aren't banged up. Should I be worried I can only get NO versions of them secondhand? Let's say I have a cable failure and they don't sense - can I set up the Delta B2 servos to stall out and not try to wreck the machine and themselves?