Shapeways print length ultimatum (yeah you told me so Bryan :-))

[nkalanaga]

I see what you're saying, and understand the reasoning.  I've never heard the problem mentioned, but that doesn't mean it can't happen.

[Lemosteam]

If a two piece roof were done at shapeways the best way is using a center gap and good, dense & easily sanded filler IMHO.  This is feasible, I have already done similar on a different unreleased project.

I will investigate.

Added material for the tongue and groove would also add material cost to the part.

Cutting a roof to a precise length, plan view and side view square would be near impossible and would end up having filler anyway.  It would be best to have a known gap, with perfect edges from the printer, and a reasonable gap to fill in the center. Cutting a center piece would be made worse when having to cut TWO ends.

On the Harriman, a 0.005" strip could be used in the center to match the width of the rest of the raised panel seams already on the roofs and cover the gap at the same time.

To all, the phenomenon is in fact random from a printing process repeatability standpoint.  One day the prints can be perfect, the next 1mm short. Material shrinkage is not really a significant factor, and heat typically expands materials instead of shrinking them.  The New York office also uses external printing facilities to make your prints.

The problem is, as Bryan mentions they will not use Process Control methodologies to maintain their calibrations. A length measurement, twice daily on each machine could be used to predict that half-day's production and when the system encounters a bounding box with one axis longer than 100mm, it could automatically scale the part in only that axis making the process transparent.  This is what automotive suppliers do the ensure the parts' accuracy is maintained over time.  But they have no interest in such things.

Bryan's use of an etch is great, but would drive the cost of the roof through the ceiling if the roof surface were etched and would produce two seams that would not ordinarilay be there on a roof, unlike his tanker.

[towl1996]

@Lemosteam, are the P70 roof's available?

[Mark W]

To all, the phenomenon is in fact random from a printing process repeatability standpoint.  One day the prints can be perfect, the next 1mm short. Material shrinkage is not really a significant factor, and heat typically expands materials instead of shrinking them.  The New York office also uses external printing facilities to make your prints.

The problem is, as Bryan mentions they will not use Process Control methodologies to maintain their calibrations. A length measurement, twice daily on each machine could be used to predict that half-day's production and when the system encounters a bounding box with one axis longer than 100mm, it could automatically scale the part in only that axis making the process transparent.  This is what automotive suppliers do the ensure the parts' accuracy is maintained over time.  But they have no interest in such things.

I'm trying to think this through and if I understand this correctly, to me it makes more sense that it is a shrinkage issue, not a calibration issue. I'm basing this on how much material is printed for the body vs. how much material is in the roof, and assuming they're printed as the same "model" on the same tray at the same time, so virtually all other variables have to be identical (nozzle temp, material feed, ect).  Calibration would affect consistency across multiple trays, not consistency of multiple parts in a single tray.

As the body part is printed, all that extra material it uses helps hold it's initial dimensions against shrinkage, where as the roof part doesn't have that advantage, therefore the roof shrinks more.




 

[Lemosteam]

After two years of discussion with them on this topic, I can tell you it's not. It is purely calibration they tell me, and they are unwilling to account for it.

The reason why it's not noticeable on smaller parts, is because they have perfected the calibration,as they told me, to under 100mm, hence the "guideline" and no guarantee of accuracy for 100mm+ parts.

I have seen parts that don't really matter fit wise affected and different within the same order, note that not all parts for the same order are printed at the same facility sometimes. I can tell by the packaging practices.

I heat and cool my parts all the time in and out of my ultrasonic cleaner without a demonstrable effect.

@towl1996 , All roofs are still available for now, at your own risk,  until I find a better solution. I will be happy to provide cad model measurements to anyone that needs a reprint.  Please remember that with each complaint they will knock my printability rating, not that I care for now.

[bbussey]

It’s definitely a calibration issue and not a shrinkage issue.

That's one reason I went old-school on the heavyweight diner roof, making a master and having castings made. I mainly used FXD on those models for detail parts and the interiors. If the interior ends up being off by a millimeter, it’s no big deal.

I don’t mind mixing mediums. Whatever is best to build the model.

[thomasjmdavis]

That's one reason I went old-school on the heavyweight diner roof, making a master and having castings made.

Is that a new ESM product? 

More roof options are always welcome.

[bbussey]

Is that a new ESM product? 

More roof options are always welcome.

No, for personal use. I kitbashed a non A/C roof from four MTL parlor roofs and had it cast because I didn’t want to go through that process more than once.

[lashedup]

Consistency in dimensions with repeatability is largely going to be dependent on a LOT of variables, many of which I assume Shapeways tries to control best they can:

1. Resin. Even within the same manufacturer/supplier, these seem to vary a bit from batch to batch. They are also very sensitive to being mixed properly before using and settling issues during a print session. Imagine mixing a resin properly, then it sits in a machine for the typical 2-6 hours or more it takes to print something in 3D. Settling occurs which is part of the reason Formlabs has a sweep mechanism in their printer to trying and minimize the resin mix from settling. Why the big deal?:

2. The printed models density can change if the resin mix gets thin, too warm, too cold, etc., etc. The ability of the laser/projector/led screen to cure the resin is affected by a large number of variables. Focus, mirror refraction (in the case of most laser systems), brightness, color or pigment in the resin being darker or lighter can affect how the light passes through to cure it and on and on. Higher exposure times (length of time the current layer is exposed to solidify the resin) typically mean a denser layer (additional light bleed from even a slightly higher exposure time can mean loss of detail and denser print layers). Lower exposure times typically result in thinner layers and mushy layers that can get pulled like taffy when the build plate moves.

3. In laser based systems, a mirror or series of mirrors are used to direct the beam. This causes some refraction and depending on the angle of reflection, items in the center of the build plate are often rendered more accurately than items on the edges of the build plate. Constant focus calibrations are also usually needed, but most machines handle this automatically. The mirrors and mechanisms that move them also need to be in consistent alignment.

4. DLP projector systems also have fringe focus issues that vary by the quality of the lens being used (identical to issues with photography lenses).

5. Some newer printers are using what is essentially an LED cell phone screen that is flush against the resin vat and eliminates the focus variability but introduces potential issues with pixel density and brightness consistency across the whole screen.

Typically the X and Y axis are fixed with the Z axis moving during the print. If all the above variables can be totally controlled (they all can't completely) then it comes down to the orientation of the model and the type and number of supports used. Usually as the model is being printed, it is stuck to the base plate, but often needs to be "peeled" from the build plate as each layer is made and it moves to print the next layer. The number and size of supports used has a big affect on the forces acting on the model being printed. Given that Shapeways orientation isn't consistent, the size and number of supports is always different each time you order a print.  This action introduces stress to the model and resin that is partially cured (remember you have to fully cure these models *after* printing to make sure they are completely cured).  The longer the model dimensions, the more opportunities there are for the model to go out of wack dimensionally. If a passenger car is printed straight up and down vertically on the build plate, there are more opportunities for the model to end up slightly longer or slightly shorter depending on which variables are acting on it. If we print the model at an angle, we can better control the dimensional issues by shortening the height. Printing the passenger car flat on the build plate is another option, but the layering affect will be more obvious and you also introduce potential suction issues (like a glass upside-down under water) that can ruin or distort the print as the build plate tries to move up and fight that suction issue.

Full curing also introduces new forces on the model as resin in various pockets, areas and densities hardens and changes dimensionally. Even if it is only .001 shrinkage or expansion depending on the resin, if that is multiplied across a longer distance, it can cause dimensional changes.

In the end the only way to ensure repeatable results in dimensionally correct prints is to run tests for the specific model being printed using a specific and consistent resin. The make, model and technology of the machine needs to be consistent too. Shapeways has probably done their best to try and control as many of these variables as they can, but they can't control them all - especially when they are ganging up multiple jobs on a single build plate and probably want to minimize resin waste and machine time (properly hollowed out models, fewer supports, lower build heights, etc., etc.).

- jamie

[Lemosteam]

@lashedup,

While I do not deny all of these occur, this is no different than any other manufacturing process and there are methods to maintain process control.  IMHO they cannot be used as excuses for not operating within their own STATED and documented tolerances of which the spec allows for tolerance growth beyond 100mm (which I designed my model to accommodate for).

In fact, every part assembled within everything you use on a daily basis, whether printed, cast, machined, molded, stamped, forged, die cast, plastic, metal, glass, or fabric, etc. are affected by all of the things you mention (and more, such as ambient temperature), and they are defined and controlled by tolerances on a drawing for fits and finishes that must fall within these tolerance ranges to allow the part to function in its designed environs.  Heck there are even color charts for painted and printed material and rust colors (parts shipped without paint) that must fall in a range.

Every production supplier worth their salt knows they must employ process control methods and tracking mechanisms (SPC) to ensure that their manufacturing processes are not out of control, to prove to the customer regularly that the parts are still within specification, at the risk of losing a contract no less.

The thousands of suppliers to my company do this for the millions of parts annually produced in a high volume sales environment.

I disagree that every model would need to be measured, only one surrogate model 100mm long per shift or material change would be necessary to measure, track and monitor shifts in accuracy so adjustments could be made on the fly to manage the print quality.  I even sent them a CAD model with various length features that could be used for such a process.

They simply do not care that I care, that my designs are not printed to their spec and their lawyers have told me to pound sand, end of story, unfortunately.

[Dirk Jan Blikkendaal]

Hi John (and others),

Following this thread for some time now I would like to through in a bit of my experiences.
I use Shapeways frequently (the Dutch factory, and they are helpful but not perfect either) but also a guy in Spain (3dTum) who has the same Projet machine. And I also do resin casting myself.
To start with the last: resin castings do vary in length also (random)..... and I take care to really mix equal amounts etc but room/work temp does make the difference.
The 3D printing (FXD mostly) is basically consistent in length but longer prints are certainly a bit shorter than the CAD dimensions quoted. The guy in Spain printed about 10 RSD-15 shells (different batches) and they all were about 0,75 shorter than the advertised  12.405 mm. Although the etched siderailings are a bit too long now they can be adjusted.......... I appreciate your eye for perfection and I certainly agree that Shapeways promised better adherence to tolerances but I frankly think they are doing a (reasonably) good job. Fixing the print orientation (Shapeways has that option as we all know) yourself in unfortunately the most expensive version also helps consistency is my experience.
We all 'know very well that where temperature is involved expension or shrinkage is a factor.....
I guess (not sure) that the removal of the support material (by heating) and the subsequent cooling of the print also is a factor. Furthermore: I assembled one of my boxcars lately (resin cast shell) with an etched roofwalk (glued with Zap-a-Gap ACC). After some time (big outside temp and humidity change, which affected my trainroom as well) the roofwalk had a little bulge...... Could easily be corrected but it shows what temp will do.

My conclusion: I would suggest to simply add 1% to STL drawings for prints longer than 10 - 12 cm. I realize its not perfect but it solves a problem.

All the best, Dirk

 


 

[peteski]

My conclusion: I would suggest to simply add 1% to STL drawings for prints longer than 10 - 12 cm. I realize its not perfect but it solves a problem.

All the best, Dirk

I think that John's point is missed again here.
As I read it, he has added length to his prints after some extensive test prints.  But the problem is that the error in printing is not constant. It changes. So even his prints designed longer than desired will  can still be shorter or longer than what he considers usable.  That's why he is unhappy.

[C855B]

I think that John's pint is missed again here. ...

Given all this, I could use a pint myself.

[peteski]

Given all this, I could use a pint myself.

LOL!  Fixed.  Fingers too fast for the kbd, missing letters. . . 

[lashedup]

@lashedup,

While I do not deny all of these occur, this is no different than any other manufacturing process and there are methods to maintain process control.  IMHO they cannot be used as excuses for not operating within their own STATED and documented tolerances of which the spec allows for tolerance growth beyond 100mm (which I designed my model to accommodate for).

In fact, every part assembled within everything you use on a daily basis, whether printed, cast, machined, molded, stamped, forged, die cast, plastic, metal, glass, or fabric, etc. are affected by all of the things you mention (and more, such as ambient temperature), and they are defined and controlled by tolerances on a drawing for fits and finishes that must fall within these tolerance ranges to allow the part to function in its designed environs.  Heck there are even color charts for painted and printed material and rust colors (parts shipped without paint) that must fall in a range.

Every production supplier worth their salt knows they must employ process control methods and tracking mechanisms (SPC) to ensure that their manufacturing processes are not out of control, to prove to the customer regularly that the parts are still within specification, at the risk of losing a contract no less.

The thousands of suppliers to my company do this for the millions of parts annually produced in a high volume sales environment.

I disagree that every model would need to be measured, only one surrogate model 100mm long per shift or material change would be necessary to measure, track and monitor shifts in accuracy so adjustments could be made on the fly to manage the print quality.  I even sent them a CAD model with various length features that could be used for such a process.

They simply do not care that I care, that my designs are not printed to their spec and their lawyers have told me to pound sand, end of story, unfortunately.

I guess a couple things to consider (if you will)...

- Every example you give above for manufacturing type and process is a very mature method of manufacturing that has been around, in some cases for thousands of years. We tend to learn a lot over time in how to leverage the technology and refine the process involved. 3D printing is new technology that is evolving and developing quickly. You can't logically hold a new immature and developing technology to the same standard as other manufacturing methods that are very mature and been around for literally hundreds of years in a lot of cases. That said, Shapeways shouldn't publish a specification they can't meet and that point I completely agree with you on. They are just shooting themselves in the foot.

- More specifically to 3D printing. I was merely pointing out why things don't turn out the way we hope. Shapeways business model itself makes it very difficult to get what you are looking for. Every object placed on a build plate (through an automated process I would imagine) is a different orientation based on Shapeways trying to get as many objects on the the build plate as close together as reasonably possible and with the minimum build height they can get away with to lower the print times and move on to the next job. This means your item orientation is likely different every time and thus subject to variations every time. If you were making D&D game pieces, you probably wouldn't care if your 3D printed object is off by a little bit. But when we need a Clerestory roof in N scale to press fit into place, things are a lot harder with the current state of the technology and the way a company like Shapeways operates on a mass production level. There are just too many variables at work here and the longer the object is vertically on the build plate, the more opportunities there are for it to become dimensionally challenged.

Again though, I understand your frustration as Shapeways shouldn't advertise something they can't deliver.