[amasters]

First of all, I'm sorry if this is posted in the wrong place.

I've been lurking here for a while on CNCzone and trying to soak up as much as possible, but now I need advice.

I run/own a small 2 man manufacturing company and now have a somewhat urgent need (within 2-3 months) for a "specialized" machine to accurately slice a round bar stock of crystal into slices or wafers. The thickness of the slices is important. I am trying to keep them within +- 0.0003" of a target thickness.

I have attached a basic sketch of what I think the machine is to look like. I have specifically omitted the coolant drainage/collection and splash issues for now. Please look at the .PDF as it makes things clearer (I hope).

I envision a basic machine consisting of 2 axis'. An accurate/precision axis to step the bar along to create the slice thickness (8 inches of travel min.) and a far less accurate axis (3 inches of travel min.) to push the crystal bar into a position stationary diamond blade and then retract it when the cut is done. Cutting will be water cooled. The concept is very similar to a tile/brick wet saw where you push the object into the blade.

Each axis will be driven by either a stepper or servo and controlled probably by Gecko's and then Mach2 or other comparable software.

I am a Mech. Eng. and have a good bit of experience with designing building fixtures, jigs, parts, etc. I have previous experience with the blades/motors/etc and sawing of this material. I have minimal experience with stepper/servo motors and controls.

My plan is to try to purchase (new or used, ebay etc) either a complete 2 axis assembly or each axis seperately and join them. To fabricate the bar stock holder and mount it to the uppermost stage. To build the motor/blade mount myself and mount it and the two stages to a rigid base plate. To purchase a preassembled Gecko/controller/PS box. To use as much preassembled stuff as possible to save time and compensate for my lack of experience in this area.

I am doing this because machines designed for crystal wafering are typically $20K to $30K used!, and old, and oversized for what I need, etc, etc.

Does my plan sound reasonable?? I am expecting to spend $2k to $4K to do this.

I would be willing to let someone else construct it for me, if they had the experience and the price was reasonable.

Any and all comments/questions are welcome.

Thanks for your time,

Andy M.
(Ga. Tech, BSME '91)

[cbass]

Welcome to the boards Andy,

I don't have much experience building myself but what you have described (and drawn) seems very straight forward to me. Heck, you won't need much in the way of software either.

You'll probably want to spec out some ground ballscrew for your X axis if those are the tollerances you wish to hold. The whole blade setup will be just as crucial. What size blade are you going to use? Can you really hold it steady to .0003" while rotating it X rpm? This is the question that jumps out at me most of all. On the other hand, like you said, a lot of this has already been engineered, so I guess it is possible.

BTW, are you making silicon chips? I've done research on the silicon industry and the term "ingot" is frequently used...

Carlo

[DieGuy]

Well your price is probably a bit shy for a machine that must hold tenths. I suspect your therory is A-OK, it sounds like you know what you need. As long as you can determine what the kerf size is on the blade you should be able to attempt to slice your wafers,

Are the cutting loads fairly low? You will definetly want to use precision slides and ball screws which will drive up the price. 0.0003 sure is a tight tolerance to hold on a saw type operation, the alignment will be crucial, as will the work piece holding. is parrallel more important than size control?

To make things more rigid, you might consider a single axis for moving the work and another for moving the motor with the blade. That would allow a wider base across the motor axis and would be eaiser to get rigid. Both axis could then be fixed to a single plane, ie base. Possibly the use of a say 18 X 24 or 24 X 36 granite surface plate would make an ideal base with lots of mass. In 0.0002 flatness they are quite reasonable.

[CJL5585]

Andy,
Welcome to the CNCZone.

I think that building the machine is possible. However, I am concerned that the tolerance of 0.0003 could possibly be a problem.

I think that a servo motor with encoder feedback is required on the X axis. The encoder would be required to get the accuracy that is desired.

As was mentioned in the previous posts, a very high tolerance ballscrew or leadscrew with almost zero backlash would be required.

You might also want to consider having a brake on the X axis servo motor to hold the position once the correct position is reached. Any movement at all during the cutting stage could really mess up the wafer surface.

Any flexing of the cutting blade could also affect the tolerance.

I don't think the Y axis is critical, so you could use either a stepper or servo, but personally, I would use the same components on both axis so that spare parts would be minimal, and troubleshooting would be simpler due to using same components.

There is a complete wiring diagram for the Gecko 320 servo system on this site that I uploaded awhile back. It includes power supply, Gecko's, servos, encoders, and breakout board. It is listed under the Gecko Thread. I don't have the exact link at this time.

You will need a qualified electronics/electrical person to perform the wiring for this system.

Good luck on your project, and keep us posted with photos if possible.

Jerry

[Bear]

Hey Andy
check out "Low Cost CNC Retrofits" for ready to go units,sounds like what
youre hunting,Im satisifed pretty much ith the results I got.
Bear

[lerman]

I disagree with some of the previous postings. A precision acme thread can be as precise or more precise than a ballscrew. Since you are only moving in a single direction, backlash is unimportant. Since traverse speed is low, forces are low. Use steppers for both the X and Y axis.

Ken

[HuFlungDung]

I don't see posted here what shape the stock is, but I wonder if a lathe type operation (with a live tool running your diamond blade) would be more successful than a cutoff saw type of thing?

For one thing, wheel loading will be much reduced if you rotate the stock.

The travel accuracy of the saw would be better if it only has to travel the radius distance to complete the cut. If the cuts are concave, the tool slide could be tweaked to obtain better flatness, or at least, you will be able to determine if the flatness fault is in the blade wandering or in the setup, if you find that no amount of tweaking will give perfect results every cut.

With a lathe type cutoff, you can also position the part for cutting several slices without rechucking the stock. This permits you to make proper offsets to allow for the blade thickness (and possible wear amount), between cuts, and permits you to get the maximum accuracy for the next cut by simply indexing the carriage screw. You could use any old screw, but use a precision linear scale to determine the position accuracy.

Hold the cutoff part with a suction cup in a live tailcenter.

[Geof]

You have this sentence in your first post:

"Does my plan sound reasonable?? I am expecting to spend $2k to $4K to do this."

You want +/- 0.0003 on a target thickness which requires two cuts separated by one advance of the material. Each cut has to be better than +/- 0.0001 and the advance has to be the same. This is a very tight tolerance; if you look up the specs for many CNC machining centers and lathes you will find the mid range machines quote accuracy of +/- 0.0002 and precision of +/- 0.0001.

For $2k to $4k no your plan is not reasonable.

[swarfmacdaddy]

Andy,
My first thought was cut the wafers without any regard to tolerance then lap them.I have seen rotary lapping machines used on ebay for the cost you are talking about. You could even build one as i am doing now. Google speefam or rotary lapping table. They can lap to within a half a wave length of a single lightband, which by the way is how the semiconductor industry does it. Also look up lapmaster model 15c for an idea of how they work. One could build the frame,table,slurry dispensing system,etc and buy the conditioning rings,and lap plate. Good luck!

[DennisCNC]

Here is a lapping machine on eBay small and in your price range:
Click

This is most likely the simpler way to go.

[ViperTX]

Well you're an ME...so you probably already know that the main problem is with how accurately your diamond blade is rotating.....so the spindle for the saw is the biggest problem....figure out how much tolerance different spindles can hold....once you know that then you know how much of your budget is in the spindle.

You could actually use a stop mechanism....you know where you feed the ingot until you meet the physical stop. I'm sure you get the idea.

[braidmeister]

Making solar cells...or gonna make some semi-conductors?

-Brady

[amasters]

Actually, I'm making thermoelectric (Peltier) devices. I don't make 10,000 a day like the Chinese manuf's, but make the custom and specialized stuff you can't just buy off the shelf. I make them from scratch; buy the raw materials, grow the semi-conductor crystal, slice it (hence this machine), coat them, dice them, and build the Thermoelectric coolers.

BTW, from previous replies, the lapping idea works for the odd slices here and there, but lapping every slice as a process turns out to be too labor intensive. The Bismuth-Telluride semi-conductor material is soooo soft that it's like lapping rock candy, dissappears before your eyes! You have to watch the lapping process and continuously stop, check thicknesses, and restart.

I would like to thank everyone who has given their advice and $.02 worth. I'm going ahead and have already purchased a stage off of ebay 2 days ago. Can't wait for it to arrive <grin>

Andy

[2muchstuff]

Don't forget..... pictures, pictures, pictures...

[Switcher]

I'm not sure about the tolerance/cost, Have you considered waterjet ?

I found this cool website, check out the photo pages.

http://www.waterjets.org


.

[mdreitzusa]

have you thought about using a dimond bandsaw.it would allow you to cut in both durrections.that way you could get two cuts in the same time that the round blade takes for one.heres a link to one i found that has a good price.all you would need to do is make the cross slide and clamp.
http://www.khue.com/dept/grind/gryband.htm

[amasters]

mdreitzusa,

Thanks for the info on the diamond bandsaw. It may or may not work for the slices, but I do need to cut alumina (Al2O3) ceramic from time to time. I already have a Disco 2H-6T diamond dicing saw (see pic), but you can't do cut and stop actions that are needed for notching etc. The Disco can cut through anything, steel, glass, ceramic etc., but it has to do it in a continuous path.

And if there is anyone that may need to have any small (it has to fit on a 6 x 6 " chuck no thicker than .13") parts diced or sawed, I would be willing to help out. I believe it's accurate out to 0.0002"

As soon as I get the first parts in, I'll post photos.

[NC Cams]

In an earlier post it was stated: "Since you are only moving in a single direction, backlash is unimportant".

When you are trying to hole +/- 0.0003" thickness tolerance, I'd contend that backlash in your driving member is critical.

Whenever you have clearance in the screw (ball or acme), the cutting tool can/will move. We had 0.0003 slop in our ball screw (traced to bearings and backlash) and would see up to 0.001 flats on a profile as we milled around it at direction change points.

We got it down to zero backlash with reploaded ball screw and super precision, high preload ball screw bearings. We can no hold size to within 0.0001". If you don't have a preloaded rigid drive system, tool can/will move when you cut due to cuttin loads/chatter, etc.

As we grind cams day in and day out and work in tenths regularly. We find that the closer we machine the part to near net shape to begin with (machining, rough grinding, etc), the easier and faster it is to make the final part exaclty the way we want it when we finish grind.

I dare say, the product that AMASTERS is making would respond in a similar fashion. Thus, I'd say, buy the best you can afford, this way, you won't be disappointed.