Anyone using MasterCAM in Optics manufacturing?

[Wes09]

Hi there, I was wondering if anyone is using Mastercam to generate toolpaths for optics grinding. I'm thinking spheres, aspheres and plano surfaces. 

The reason for this question is a machine manufacturer has things like aspheres locked behind a paywall in their conversational control software, and requires their own brand of CAM software to do freeforms. I have a significant amount of 5 axis programing experience with Mastercam, so I know I can generate toolpaths to create the optical profiles using Mastercam toolpaths. But maybe Mastercam cannot generate a precise enough toolpath for optics and so these manufacturers feel the need to create their own CAM software?

Curious what you guys think. Iv been a machinist/programmer for 15 years, but just moved into a position at what is effectively an optics startup to run a 5 axis CNC grinder and eventually CNC polishing. The folks here are very experienced, with extensive knowledge optics, but all of it from a non CNC background. Very interesting work and optics is mind blowing levels of precision/accuracy.

[Aaron Eberhard]

I have a very shallow understanding of the optics grinding world, but my understanding is that the biggest thing about the grinding software is that it controls dwells at certain points to remove the material, as the grinding amount is a result of both time spent in contact and pressure applied by the grinding surface.   Controlling that via a feedrate-based solution is going to be sub-par.   

But again, I don't pretend to know much about it!

I'm not sure what machines you're trying to program for, but I know that OptiPro Systems (https://www.optipro.com) is a Mastercam Reseller who might be able to help you out.

[JParis]

Optics, ugh....

It is such a specialized process.

I've milled surfaces, it wasn't for the lack of trying but they could never approach our diamond turning machines.

Think Red Dot type optics....

If they have their own proprietary abilities,  make them prove it. If they can,  as far as I'm concerned it would be a no brainer to use their stuff.

[Kyle F]

11 hours ago, JParis said:

If they have their own proprietary abilities,  make them prove it. If they can,  as far as I'm concerned it would be a no brainer to use their stuff.

My thoughts exactly... yeah maybe you can get it done in mastercam but if they have an option supplied by the machine tool builder and it's specifically designed for doing the type of work that y'all specialize in??? I'd probably talk my employer into biting the proverbial bullet on the add-ons. Honestly it should have been purchased with the machine by the sounds of it.

[Wes09]

14 hours ago, JParis said:

Optics, ugh....

It is such a specialized process.

I've milled surfaces, it wasn't for the lack of trying but they could never approach our diamond turning machines.

Think Red Dot type optics....

If they have their own proprietary abilities,  make them prove it. If they can,  as far as I'm concerned it would be a no brainer to use their stuff.

it would probably end up as me proving that mastercam can achieve the level of optical precision on a few test parts (or failing misserably!). Think of the machine I am running like its a 5 axis mill. table is x,y,c while the head is z,b. All of the tooling is diamond impregnated. I think spheres and plano could be done in mastercam. Sphere and plano are only one axis in motion (Z) and C spinning at whatever #rpm. Asphere and freeforms are where things seem like they will get tricky.

 

 

3 hours ago, Kyle F said:

My thoughts exactly... yeah maybe you can get it done in mastercam but if they have an option supplied by the machine tool builder and it's specifically designed for doing the type of work that y'all specialize in??? I'd probably talk my employer into biting the proverbial bullet on the add-ons. Honestly it should have been purchased with the machine by the sounds of it.

You are right, it should have been purchased at the same time. Unfortunately I joined this team after the machine was purchased (brand new and setup, but not running yet). The software is a $40k bullet, so I just want to do my due diligence before I throw that at them.

[Kyle F]

21 minutes ago, Wes09 said:

You are right, it should have been purchased at the same time. Unfortunately I joined this team after the machine was purchased (brand new and setup, but not running yet). The software is a $40k bullet, so I just want to do my due diligence before I throw that at them.

Very understandable

[Aaron Eberhard]

46 minutes ago, Wes09 said:

it would probably end up as me proving that mastercam can achieve the level of optical precision on a few test parts (or failing misserably!). Think of the machine I am running like its a 5 axis mill. table is x,y,c while the head is z,b. All of the tooling is diamond impregnated. I think spheres and plano could be done in mastercam. Sphere and plano are only one axis in motion (Z) and C spinning at whatever #rpm. Asphere and freeforms are where things seem like they will get tricky.

 

 

You are right, it should have been purchased at the same time. Unfortunately I joined this team after the machine was purchased (brand new and setup, but not running yet). The software is a $40k bullet, so I just want to do my due diligence before I throw that at them.

Sounds like it's time for the machine tool dealer and the Mastercam reseller to step in and and demo to me Let's see which $40k option actually cuts parts for ya!

[JParis]

43 minutes ago, Wes09 said:

it would probably end up as me proving that mastercam can achieve the level of optical precision on a few test parts (or failing misserably!). Think of the machine I am running like its a 5 axis mill. table is x,y,c while the head is z,b. All of the tooling is diamond impregnated. I think spheres and plano could be done in mastercam. Sphere and plano are only one axis in motion (Z) and C spinning at whatever #rpm. Asphere and freeforms are where things seem like they will get tricky.

There's a reason diamond turing machines cut in the millionths'...

Not knowing your application, I certainly won't say it can't be done but I do know in our application, regardless of how good the finish was, it wasn't close enough.

[Aaron Eberhard]

 

5 minutes ago, JParis said:

There's a reason diamond turing machines cut in the millionths'...

Not knowing your application, I certainly won't say it can't be done but I do know in our application, regardless of how good the finish was, it wasn't close enough.

I know that for a ruby grinding application someone contacted me for, the blanks each cost somewhere north of $60,000 and there was no equivalent substitute material to practice on...    No quote.

[Aaron Eberhard]

If you do go down this path, you'll need to play with the File > Config > Tolerances page, the default system tolerance is only .00005", which you'll want to tighten up. 

On the toolpath side, if you're playing with a Unified then you'll need to not only adjust the toolpath tolerance, but you'll also really need to to tighten up the curve tolerance (Cut Parameters > Advanced Options).  There's some other things you can play with to, depending on the tool contact point, etc.

I'd recommend going with a Flowline if you can, though, as that's the most direct/least "converted" option.   You'll be back in old-school land for tolerances, so make sure you tighten them up.  That probably won't work out well if your tool shape is odd, though, so you'll be back to Unified to get proper comp points. 

Also, make sure that you're working with your reseller to get 3d tool compensation output.  All of the Multiaxis toolpaths contain not only the tool tip and center line, but also the contact point of the surface and surface normal for each vector.   Make sure your control is taking advantage of that.

[Aaron Eberhard]

Oh, had another thought while eating lunch...  

I'm not sure what the max tolerance is for the post, but I know I've had some issues with rounding/fuzzy math before.  I'd imagine it's worse in this case.  @Colin Gilchrist - Anything you can add here?

Can you run that machine in Vector mode instead of traditional positioning mode?   That way you're not getting the rounding errors compounded by MP conversion. 

[Colin Gilchrist]

I would recommend forcing the tolerance in Mastercam to 8 decimal places for Inch. You always want the "input" data to be more accurate and granular than the output data. Keep in mind > more precise tolerances require more mathematical calculations and time to process.

What is the 'least increment input' for the machine? Will you be running in Inch or Metric? Often, with Japanese machines (and most machines in general), you'll be able to hold tighter tolerances in Metric mode.

Keep in mind > your NC Code is an approximation of a path to follow. Translation of complex curved shapes into point-to-point, or line/arc motion, is always an approximation of a complex shape.

Is your machine's control capable of NURBS Interpolation (G06)? This would be the best and most accurate form of curve control at the machine, but good-luck supporting that output in Mastercam. The "NCI Data Format" (Mastercam's Raw Toolpath) is not capable of feeding "NURBS Curve Data" into the Post. Every Toolpath will linearize the input curve, and the output will be point-to-point motion. Trying to take those point-to-point moves and then derive an accurate NURBS curve output from that input, is a recipe for failure. There may be some advanced tools in MP where you could literally capture the starting NCI Block where the curve starts, and then scan forward through the blocks to determine the end of the curve, but there is a whole lot of math involved in just capturing the start/end blocks of the curve, let alone using that input to generate accurate NURBS output.

Could you do this with Point-to-Point, using Mastercam? Sure. Will it be accurate enough to give you what you need? Only one way to find out!

There are other CAM software packages with a deeper bench of engineering tools that may better support what you are trying to accomplish specifically for machining optics.

The "rounding issue" that Aaron is referring to, has to do with how Floating Point numbers are stored in your computer (and how the windows kernel reads and writes values. For example, if your input is "4.5", that number is likely stored as a very close approximation: 4.499999999999 (or similar). It is through Rounding, that we are seeing nice, round, numbers.

There is a Command Variable in the Mastercam Post: 'round_opt$', which "changes the internal windows library which fetches the data, and performs the rounding.

The 'round_opt$' variable is only needed in cases where you notice "bouncing" of the final digit of a rounded value.

(For example, a Z value that bounces from Z-.6605 to -.6604, where only that 4th decimal place digit is bouncing between 4/5, or 5/6.) This could be a result of an input value of 0.660455, for example.)

The round_opt$ variable can fix these instabilities in floating point conversions to rounded values.

'round_opt$' only has a couple of valid values, I recommend trying them all! NOTE: you should only initialize this variable near the top of the Post, and only one time. (Set the value, Post your code, examine the output, change the initialization value, save the Post again, and re-test the output. Do this for each of the following values.)

round_opt$ : 1

round_opt$ : 10

round_opt$ : 11

round_opt$ : 21

 

 

 

 

[Chally72]

On 12/20/2023 at 1:57 PM, Aaron Eberhard said:

Oh, had another thought while eating lunch...  

I'm not sure what the max tolerance is for the post, but I know I've had some issues with rounding/fuzzy math before.  I'd imagine it's worse in this case.  @Colin Gilchrist - Anything you can add here?

Can you run that machine in Vector mode instead of traditional positioning mode?   That way you're not getting the rounding errors compounded by MP conversion. 

^this

2024 scales 5ax path vectors to a unit vector before inserting into the NCI to avoid some of these rounding errors throughout the process, and vector mode will further reduce the possibility of fuzz/jitter.