Desktop mill for optical parts

[CarbideBob]

Weird that the OP asks for help here but seems to knowing it all already and does not care to listen.
Bob

 

[mneuro]

I thought I had replied but my message seems to have disappeared, sorry.
I only measured the machine servo resolution, not any kind feature on the part. My application requires good angular direction of the surface normals which means I need a good local precision but don't require a great global precision. If the tool length expands of a few microns from the start of the machining and the end, it doesn't matter as it won't impact the local precision. Climate control is therefore not important.
If you will, I only care about the surface finish and not about the precision.

If surface finish is all that really matters, maybe a good surface grinder or lapping equipment would be more appropriate.

We do work in the .002-.003mm tolerance range. We have very expensive Yasda and Hermle machines. We also grind and lap. Temp control the shop to ±2 degrees F, control our inspection department to ±1 degree F, and we are still operating at the limits of what is possible with our shop setup.

I think suspect that your inability to really inspect your parts is leading you to believe you need more than you actually do.

 

[Mtndew]

Weird that the OP asks for help here but seems to knowing it all already and does not care to listen.
Bob

I was going to mention the same thing.

 

[carbonbl]

Thanks, making a test part could be interesting, but seeing that the spindle is 3.7 kW (5hp) makes me think this is a machine meant for rough work and steel machining. There will likely be a lot of vibration and big runout. A Nakanishi spindle has a power of 300W but is optimized for precision which I think is more adapted for what I'm looking for.

Others have alluded to this, but I'm going to state outright that you are going to have to discard your assumption that power (and/or mass) have a strict negative correlation with accuracy. In fact, for many use cases the opposite is actually true. In the CNC space, the exercise of plotting the relationship between the four variables of accuracy, spindle power, mass and machine cost is extremely complex and dependent on other additional factors. All the questions being asked in this thread are trying to tease out those other factors so folks can give you an informed recommendation.

I don't have any specific advice to you but I'm interested to see where this thread goes. The problem of the "high accuracy low cost desktop mill" is one that many machinists and tool builders have chased for years, with vary degrees of technical and/or market success. Welcome to PM!

 

[Halcohead]

Also out of your price (and size) range, but Fanuc also makes a robodrill with 0.1um programmable resolution and spindles in the 20-40krpm range. I've seen diamond milled optically reflective demo parts off this machine at expos, but of course I didn't measure them. You might contact a Fanuc representative to see whether they can sell you an entry level robodrill with an extra decimal place in the control.

We have a 1.5m swing Okuma VTL at my company with 0.1um minimum programmable increments (lol). And when you check with an indicator it even (sometimes) moves in these increments! But it's about 50x and 100x your respective price and weight budgets. Also the way the VTL does (and does not) follow commanded increments is illustrative of how these small increments can be very challenging: it has heavy axes running on box ways. On intermediate speed dynamic movements it can actually position in submicron increments, however it struggles with small motions due to stick slip; you can see servo overshoot and jerky motion due to trying to overcome static friction to move a 6000kg axis in tiny increments.

I assume the toolpath control the OP is referring to is planning toolpaths that are all one tool and the tool spirals or zigzags across the part so that each pass of the tool is close in time to its neighbors, that way thermal drift or axis scaling issues don't cause discontinuities. I would still expect to see big problems at axis reversal points if there are any mechanical issues (such as backlash or stiffness issues in ballscrews, classic example is reversal lines on interpolated circles). It's still not clear to me what the additional decimal place provides if the machine is an order of magnitude less repeatable than the current programmable resolution.

 

[DanielG]

It sounds like you have a fairly unique application. You are relatively unconcerned about global accuracy, but want a really good surface finish. If the Roland is almost there, but the spindle is the problem, would the most economical option be to replace the Roland's spindle with a high-speed NSK unit. They have some that direct clamp onto 6mm tools that have a great runout. It eliminates the runout of the toolholder entirely; there isn't one.

I think it would also be worth looking at the Haas CM-1. It's a pretty small machine at a decent price that might do what you need. It would probably at least be worth having the distributor do a test piece for you.

 

[mhajicek]

I think a Matsuura linear motor machine would do the job pretty well, but then you're looking at a full sized machining center and $350k or so.

 

[woodsrider845]

Somewhere in here is a joke about needing a supercar, but already winning your race with the plastic car on a merry go round. I just don't care enough to cut and paste pictures.

 

[precisionmetal]

Optical,

Have you tried bolting your existing Roland to a 100-200kg lead plate?

PM

 

[Optical]

Optical,

Have you tried bolting your existing Roland to a 100-200kg lead plate?

PM

What would be the purpose? The machine is running so slowly (<200mm/min) that the movement of the axis apply only very small external force.

It sounds like you have a fairly unique application. You are relatively unconcerned about global accuracy, but want a really good surface finish. If the Roland is almost there, but the spindle is the problem, would the most economical option be to replace the Roland's spindle with a high-speed NSK unit. They have some that direct clamp onto 6mm tools that have a great runout. It eliminates the runout of the toolholder entirely; there isn't one.

I think it would also be worth looking at the Haas CM-1. It's a pretty small machine at a decent price that might do what you need. It would probably at least be worth having the distributor do a test piece for you.

I though about replacing the spindle but that's probably not enough. The positioning accuracy is not good enough.

Also out of your price (and size) range, but Fanuc also makes a robodrill with 0.1um programmable resolution and spindles in the 20-40krpm range. I've seen diamond milled optically reflective demo parts off this machine at expos, but of course I didn't measure them. You might contact a Fanuc representative to see whether they can sell you an entry level robodrill with an extra decimal place in the control.

We have a 1.5m swing Okuma VTL at my company with 0.1um minimum programmable increments (lol). And when you check with an indicator it even (sometimes) moves in these increments! But it's about 50x and 100x your respective price and weight budgets. Also the way the VTL does (and does not) follow commanded increments is illustrative of how these small increments can be very challenging: it has heavy axes running on box ways. On intermediate speed dynamic movements it can actually position in submicron increments, however it struggles with small motions due to stick slip; you can see servo overshoot and jerky motion due to trying to overcome static friction to move a 6000kg axis in tiny increments.

I assume the toolpath control the OP is referring to is planning toolpaths that are all one tool and the tool spirals or zigzags across the part so that each pass of the tool is close in time to its neighbors, that way thermal drift or axis scaling issues don't cause discontinuities. I would still expect to see big problems at axis reversal points if there are any mechanical issues (such as backlash or stiffness issues in ballscrews, classic example is reversal lines on interpolated circles). It's still not clear to me what the additional decimal place provides if the machine is an order of magnitude less repeatable than the current programmable resolution.

Backlash and low stiffness is probably a source of problems on the Roland. The extra decimal is indeed probably not essential at this point, definitely not with this machine.

I'll try to rephrase my initial question: do you think that a machine built for small precision work like the Minitech GX (granite base, THK linear rails, precision ball screws, Nakanishi spindle) can get a surface finish close to mirror (on brass for example)?
And a side question: their GX model has a granite base, do you think it provides a benefit if you only take very light cuts?

 

[mhajicek]

I'll try to rephrase my initial question: do you think that a machine built for small precision work like the Minitech GX (granite base, THK linear rails, precision ball screws, Nakanishi spindle) can get a surface finish close to mirror (on brass for example)?

I'd be leery of buying a machine from a company that can't maintain a website. But like I suggested before, have the company do a test part for you. That's the only way you'll find out for sure if their machine is good enough for you. You should have several companies do it, so you can choose what's right for you.

 

[implmex]

Hi again Optical:
You wrote:
"I'll try to rephrase my initial question: do you think that a machine built for small precision work like the Minitech GX (granite base, THK linear rails, precision ball screws, Nakanishi spindle) can get a surface finish close to mirror (on brass for example)?"


I have two machines with "granite" (polymer concrete) bases , a Defiance VTX 1 benchtop 3 axis CNC mill, and an SNK Prodigy CNC lathe.
Neither is the holy grail for machining precision by any stretch of the imagination...both are usable machines but the granite does not confer some special magic.
Of the two, the Prodigy is the better machine...it's smooth and rigid and will drift a couple of tenths in a day due to thermal effects, and it's predictable.

The Defiance is a piece of crap if your goal is accurate machining...the problem is the fact that it weighs almost nothing (500 lbs), it has a cheap-ass spindle in an aluminum housing, and the ballscrews are just mounted in thrust bearings with no thermal comp and no support other than at the motor end, so they whip at high speeds.
It also has a Bridgeport DX32 control on it which is a source of much pain (it cannot decelerate properly at the end of a commanded motion so it is jerky...a major DX32 flaw that was never fixed and cost Bridgeport millions in lawsuits).
It's construction looks almost exactly like the Minitech GX but with an enclosure, and it has brand name rails and ballscrews too.

The presence of the granite in no way mitigates all of the other things that are deficient with it; yet it was originally marketed as a super accurate electrode making machine for sinker EDM.
It was hopeless for that and Defiance went out of business after ditching the Bridgeport control for a better one, and re-marketing it as an educational machine but never really gaining traction with it.
It was about like your Roland but for about double the price.
Defiance was about marketing...they built a relatively cheap toy and spun a great volume of bullshit around it.

The point of this long winded rant is that the granite is pointless if other things are no good...it's a nice marketing buzzword, but is actually pretty irrelevant.
The control is super important, the quality of the components is super important and the engineering is super important.
The first two cost money, and the third requires long experience.
That's why a Hermle is so expensive.

Cheers

Marcus
Implant Mechanix • Design & Innovation > HOME
Vancouver Wire EDM -- Wire EDM Machining

Another thing...I just had a closer look at the Minitech website...this thing runs on steppers.
It's a big step down even from the Roland which at least has AC servos

 

[CarbideBob]

...
I'll try to rephrase my initial question: do you think that a machine built for small precision work like the Minitech GX (granite base, THK linear rails, precision ball screws, Nakanishi spindle) can get a surface finish close to mirror (on brass for example)?
And a side question: their GX model has a granite base, do you think it provides a benefit if you only take very light cuts?

I can get a mirror finish where you can read the printing on the the shop's overhead lights reflected to the part on a well worn, 0.100 backlash 1950 B-port.
Anything that says "granite base"speaks to me like late night infomercials.
This has only been done well by the biggest and even then not very accepted. At one time I most certainly thought this would be just so good particularly in the small machine and small grinder world.
I know this the rage or Holy Grail on some sites or users but it is very much not all that. It seems like a good idea but the real world is different than what seems a good idea.
For start ask what class of THK rail and go to THK's info for how good that rail is. Not sure what you consider a precision ball screw as this tossed out freely, your world and mine perhaps different.
For sure I'd be a least into roller rail guides and not balls and even that so many problems.

I do not know what you making, volumes scrap allowance or how measured.
Mirror finish fence posts are a bit different than other parts that go inside a engine or worse yet optical.
One of these need finish shiny, the other may need accurate finish.
How does one check a profile at 1/10 or even 1/4 wave? The inspection stuff so way past the price of machines you link too so that makes me wonder a bit.

The tooling and cutting conditions you use may influence "shiny and mirror" just a tad.
Bob

 

[TheBigLebowski]

Also out of your price (and size) range, but Fanuc also makes a robodrill with 0.1um programmable resolution and spindles in the 20-40krpm range. I've seen diamond milled optically reflective demo parts off this machine at expos, but of course I didn't measure them. You might contact a Fanuc representative to see whether they can sell you an entry level robodrill with an extra decimal place in the control.

We have a 1.5m swing Okuma VTL at my company with 0.1um minimum programmable increments (lol). And when you check with an indicator it even (sometimes) moves in these increments! But it's about 50x and 100x your respective price and weight budgets. Also the way the VTL does (and does not) follow commanded increments is illustrative of how these small increments can be very challenging: it has heavy axes running on box ways. On intermediate speed dynamic movements it can actually position in submicron increments, however it struggles with small motions due to stick slip; you can see servo overshoot and jerky motion due to trying to overcome static friction to move a 6000kg axis in tiny increments.

I assume the toolpath control the OP is referring to is planning toolpaths that are all one tool and the tool spirals or zigzags across the part so that each pass of the tool is close in time to its neighbors, that way thermal drift or axis scaling issues don't cause discontinuities. I would still expect to see big problems at axis reversal points if there are any mechanical issues (such as backlash or stiffness issues in ballscrews, classic example is reversal lines on interpolated circles). It's still not clear to me what the additional decimal place provides if the machine is an order of magnitude less repeatable than the current programmable resolution.

Agree with this. The robodrill eco model with a 10k spindle comes in at under 50k. I am not 100% sure but I would think being fanuc hardware the ballscrews are preloaded, which should really help with surface finish. The new ones also have lots of toolpath optimization/smoothing options to help with improving surface finish. Dollars to Donuts it would be very hard to beat.

 

[CarbideBob]

....
That's why a Hermle is so expensive.
Marcus

I want a band new true five axis Hermle with the all the options for $30,000 or less.
Why is there no such machine? I do not need big.
Bob

 

[Mike1974]

I know it was mentioned before, but there is a member using a small sub micron accuracy (?) machine here.. But definitely not under 30k!!

 

[hanermo]

It would not be very hard, technically, for someone who does these types of things, to make a really high-resolution version of the machines the OP mentioned.
The Mira is actually a pretty good one.
Adding about 1000 kg in mass and 30k$ in stuff plus 30k$ in "skill" you c/would get what you want in about 6-8 months.

There is no way you will find one anywhere that will give you consistent sub-micron repeatability and overall accuracy for 100k$, even used.

It´s not too hard to do sub-micron mechanical resolution if you can afford 5000$ components, multiple, per single axis.
Say 15k$ per axis, for a refit, ballpark.

My clients when I was sales manager of Haas Spain had grinders that made 0.7 micron overall accuracy parts.
5 axis.

They used to make solid carbide stamping dies -- accurate to 0.7 microns -- that they then used to bash tuna cans from sheets, 400 per minute.
True story.
Accurate to 0.7 microns - overall envelope.
17 km from me.
500k€ each for the grinders.

For your machine parts:
Single-axis sub-assys with 0.1 micron real world mechanical resolution and 0.1 micron repeatability - might cost around 20-30k each, depending.
For 100-200 mm travel, of less than 50 kgf or 500N max push force in mechanical work (trending sub-5 kgf in these types of work).
Half the cost is "skill", not hw.

You are welcome to source hw yourself and have a go.
I did.

To others who posted:
1 micron mechanical repeatability on a local level (say less than 100 mm distance) is not hard to do - for a single axis only.
1 micron local accuracy is not hard to do if you can use limited spindle power, like the Nakanishi stuff mentioned.

I tried NSK with the Japanese rep in germany (Emo), and they are amazingly good, and yes nearly noiseless at 20.000 rpm with less than 1 micron TIR.
A slight whine at 30k rpm. Much less noise than a hairdryer, less than a laser printer.

To OP:
1 micron mechanical resolution / axis is fine, but
1. over 3 axis the error is 8x, or == 8 microns volumetrically, and,
2. at a mount-to-tooltip distance of minimum 100 mm,
you get into harder territory.

The only way to get really accurate 1 micron tooltip accuracy or tir is to have about 2x pivot-tooltip or 200 mm distances from the pivot point of the spindle to the spindle mount rear support.
This means the spindle gets taller, the spindle mount gets somewhat larger, everything gets heavier by 8x, and preferably heavier by 3x times 8x - empirical data and my own experience.


None of the steps are all that hard, themselves, intrinsically.
But they cost a few grand each in parts.
Some parts won´t work - too sad.
And you need many parts, about 20-30.
And you need some integration between many of them.
And none of it is really standardised and some really wont work they way it should.

All above is based on using high end import parts and not name brand front office deliveries.
Aka not IKO or SKF or THK or Fanuc or Siemens parts from their local rep, but their second level actual manufacturers exports - that one manages oneself and one already mostly knows what to do and how - as and OEM.
But still using the high-end stuff like the P2 ballscrews and best mounts.

Example of complex issues:
*edited out - boring I think, maybe *

- - -

If I thought there was a business case to sell 0.2 micron machine tool linear slides, I would make some.
But to be viable, it needs about 20-30 pre-orders, and I doubt such demand exists at the value level- with no brand.

The kinetic tracking mounts and other similar endeavours buy their stuff from the established vendors at 15x the price.
And top end research institues and science labs.
And this is not all wrong, as these are somewhat documented systems, and well supported, even if the support has an egregious cost, typically.
The buyer knows that they bought a solution that is good for their career.

E. non-commercial.

I know at least 2 ways to make 0.2 micron resolution/accuracy small milling machines with no expensive tools.
Both involve skill and large amounts of hand work.
3 ways actually.

Google first ruling engine and see how 2 guys built a sub 0.1 micron accurate mechanical engraving machine.
Using a 100$ base South bend lathe to make the parts, they first made into a 1 micron lathe.
Took 17 years of course, but ..

Google offhand compensation for leadscrews, known since 1900 or earlier.
25x more accurate in mechanical resolution than the base leadscrew.
In Moore and SIP jig grinder times, == 1960, I believe these were sometimes used, perhaps to make the leadscrews accurate to 1 micron.
Not hard to do.
I can do it.
Anyone here on PM can do it.

3rd way is hydrostatic bearings using oil or air.
Actually dirt cheap and very forgiving.
Car internal combustion engines, about 2 billion, use them.
Crankshafts are almost always running on hydrostatic bearings - they float on oil and have zero contact.
They would run for thousands of years if the oil pressure was maintained and no contaminants were added.

Volumetrically averages down errors, for the cost of a 100 $ pump.
0.1 micron accuracies in air bearings (diamond cut optics, lenses) use this.
And several tens of tons for lifting stuff sometimes uses air bearings.
And it lasts forever, 70-year old stuff running on wooden spindles in major powerplants still run on hydrostatic bearings using water.
There is a roman mill 2000 years old still running.
Wood bearings.

 

[mhajicek]

Edmund Optics® utilizes a wide range of CNC machining equipment including ultrasonic machining and high speed cutting (HSC) technology from DMG MORI® for the machining of complex optical component geometries. CNC machines offer the highest amount of precision and flexibility when it comes to fabricating components beyond basic shapes. Our true 5-axis precision machine centers have the capability to machine lightweight structures and high aspect ratio components with specialty bevels, precision holes and inserts, and mounting interfaces. In addition to optical glass, we can machine a wide range of demanding materials including ceramics, corundum, tungsten carbide, and even composites. Some high-tech materials, such as Tungsten Carbide and composites, cannot be machined using conventional machining and can only be shaped using ultrasonic machining.

CNC Machining | Edmund Optics

Dimensional Accuracy Linear: 2-8µm
Parallelism: 2µm
Concentricity: 7µm

 

[hanermo]

To add to the thread.
It is very very easy to test any machine.

Ask the manufacturer rep to mill 2 20 x 20 mm 316L studs next to each other, from a single block,
20 mm deep maybe.
Maybe 30 mm separation between the two towers.

Goal being 20.000 mm +/- ? as near as possible, to a 20 mm depth.
Using any tools of their choice, any cost, as slow or fast, just as accurate as possible.
Probing to confirm stuff, whatever, no limits, just make it as accurate as possible.

And for the tower 2, some steps of 1 micron x 3 mm wide, 2 um, 3 um, across one edge say x, and 2, 4, 6 um across the other edges say y.
Just say 5 mm deep from the top.
Will they do this ?

They need to plane the block first so it´s as square as possible.

Then any naybo can use a std 1 micron electronic mic to measure the sizes and consistencies across about 16+ planes and datums.

I suspect most manufacturers would not do this.
Most manufacturers used to hate public tests or trials.

Gene Haas has single-handedly changed this.
Including Youtube video of how they make VF series mills.
He is right - and his sales and brand has grown as has his social media guru potential.
Gene Haas has made more extra (incremental) sales and money from openly sharing stuff than others have from hiding everything.

Obviously Gene Haas is right.
He and many others I know (IT, other stuff), all mostly agree on the same thing.
Let them come and copy, as much as they can, from the product ...
It just is NOT that easy.

More copiers of Haas made more advertisements for others and for us.

The copiers make the garden grow .. and the original front runners take most of the profit.
Whatever the copiers get - they are welcome to it.
Good for them !
They are growing the market - not hurting the market.


When I was at Haas or a,b,c ...
edited as boring ...

 

[hanermo]

Em ... ?
I´m sure it was somewhat true at the time the press release was authorised.

Does this somehow mean that a large part of their production is done on such machines to such accuracies.
xxx - Maybe. Edited.

Edmun does 118M$ with 400 employees across 11 companies,
Annual Report on Edmund Optics's Revenue, Growth, SWOT Analysis & Competitor Intelligence - IncFact

Which is actually a fantastic result.
EO makes good money, about 300k$/yr in revenue per employer with probably 50% or up of gross margin.
Optics and similar typically have 80%+ operating margins, and EO did not seem to have major debt due to adquisitions or other silly opertions.
It´s still afaics a mostly family held company where most of the top posts go to edmunds surnames.
Good for them.
Afaik - first 1 min look.

Edmunds is exactly what companies should be doing .. and what the government(s) should be supporting.
Edmunds is high-tech, local-workers, highly-paid expert workers leveraged via automation, making high value-add products, for great profits.
And also global offices and representation and sales (light).

Without any financial data from disclosures etc. I have no actual info.
But I bet EO pays about 1.6 - 2x the typical median salary for staff positions, aka below management, with probably a 1200-1500$/month package of medical benefits/unemployment insurance/other benefits on top of that in the US.
It´s probably one of the places everyone wants to get into, and almost no-one ever leaves.


Fwiw..
Most new companies and jobs in Finland (Scandinavia) are like that.
Great productivity, everyone works very hard, everyone gets paid well, plenty of bennies, and almost no buro-crap.
Most people can work flex-time, and remote-work, and as long as spirit and results keep up, no-one wants to poke into details.

CNC Machining | Edmund Optics
Dimensional Accuracy Linear: 2-8µm
Parallelism: 2µm
Concentricity: 7µm