18 responses to “Fixturing and Constraints”

  1. Jacob

    Just on a hunch – by any chance, did you study under prof. Alex Slocum? (He’s my favourite reference when it comes to these matters).

    1. Miss Outlier

      I did indeed – and I still have his precision machine design book on my reference bookshelf, too!

  2. Stanley Ma

    Do you know anything about wafer steppers and what how they get 0.25 micron accuracy in alignment?

    1. Miss Outlier

      Photolitographic mask alignment is usually done with incredibly complex
      high performance positioners. The biggest difference is that kinematic couplings are passive – they will go back to the same position every time, whereas wafer steppers require active feedback control. Wafer steppers generally can move on the scale of resolution <1nm, by using laser interferometry for position sensing and comparing the mask alignment features with features etched into the wafer in order to overlap the mask with the wafer.

      1. GEARS

        The whole wafer stepper system is indeed pretty complex. Just to give you some more details on it, the wafer stage is generally floated magnetically to eliminate unwanted vibrations into the stage. And yes, laser interferometers are generally used for position feedback but they’re not referenced to the stage itself, but rather the wafer on the stage. They generally set some nominal, known position on the wafer to zero and then go from there.

        Because they need to control all 6 DoFs, they also need to zero all six DoFs but they’ve got some nifty ways to reference that off the wafer.

        It’s only the lowest levels of the wafer that have the 30~nm position uncertainty. After the first few layers (say 3-4), the tolerance stackup and imaging uncertainty means the positioning doesn’t need to be as robust because they’re fitting a larger picture over a smaller one.

        In general though, it’s not the stage position that limits the alignment, but rather the imaging uncertainty. Because the illuminated mask will be blurry at the edges (around 20-ish nm uncertainty), that’s going to set your lower limit for wafer positioning. The stage feedback system is probably good to 10x better over short time scales (1-5 minutes). If you need better than that, sponsor my research and I’ll make you something better 😀

        1. Fluxor

          I think this whole topic deserves a separate post. 🙂

  3. Henrik Sandaker Palm

    Thanks for a great post. It was nice reading some mechanics stuff here, and the content is well explained and easily understandable. I like how engineerblogs is getting more and more specific on engineering content, like technical stuff. You and Paul, keep it coming.

    1. Miss Outlier

      Thanks! I appreciate it, and will try to do more of the same. 🙂

  4. agammy

    I have been seeing more attention being paid to fixtures used in the design process (and particularly for manufacturing) because it can be more complicated than people expect. Especially if you’re making a product that requires a high level of accuracy, you need to use a little bit more sophisticated means to make sure you have a reliable process. I don’t know if this is interesting to this audience, but I wrote a piece myself about steps you can take to document your fixtures, in case something happens to them – http://blog.arenasolutions.com/documenting-fixtures-and-jigs/ – I hope it’s useful to someone else who reads this blog.
    Anyways, great post – and I will second the comment that I like these technical articles, so few of them out there!

  5. GEARS

    if you use the 3 sphere and V-groove combination, you’re thermally centric (last figure, left picture). If you have thermal issues, then you’ll still have a center point that doesn’t expand (at least to first order). In the other case, the flat-groove-cup, you choose the point (that’s the cup) where you want zero thermal expansion. So depending on your design, you can pick your thermal datum.

    Miss Outlier, did you do any repeatability experiments to see how long the wear in period was? Generally, those surfaces need to be either hardened or strain hardened via coupling and uncoupling due to Hertzian contact forces. If not, your alignment might change a few um’s over time as it wears in.

    1. Miss Outlier

      Very true about the symmetry comment – I think that’s probably one of the reasons the 3 sphere and V-groove is more common.

      I haven’t built the new machine yet, so I haven’t yet worked through the wear-in period. 🙂 But I do plan to do that – the wear-in period does indeed vary depending on the material, and possibly also the preload applied. Usually I’ll put three cap probes on one of the plates, and measure the gap between the two plates. As I take data while cycling through putting the plates together and taking them apart, I can get nice exponentially decaying curves converging to the worn-in position, and I just use judgement as to when I’ve reached enough cycles.

  6. Bentech

    do you have any rules to follow in order to determine the angle of each groove in the case of 3 V grooves on one side and 3 balls and the other side, for the case of non equilateral triangle?
    Thanks in advance for your help!

  7. Dan Merrick

    Being a civil/structural engineer I am unfamiliar with this but I find it very interesting and applicable to some of the work that I have done. I sometimes consult with a company that makes architectural concrete panels for buildings – the outer shell that we see. Downtown San Francisco, LA, Sacramento, etc. have many examples of their work. These panels need to be constrained in all 6 degrees of freedom with what equates to your high precision – 1/4 inch for a 25 foot panel. The connections also need to allow for building movement due to wind, earthquake, etc. We always try to design these panels so that we constrain the 6 degrees of freedom but no more. The indeterminacy that results from more constraints can generate enormous forces in the structure when things move and result in failure. Unfortunately, we usually work in a rectilinear space and the tripod solutions aren’t really applicable. We end up using XYZ “pinned” connections and links. Links, like adjustable connecting rods, make alignment in the field easy.

  8. Dr Richard Skipper

    Miss Outlier,

    I am very impressed with your approach to the challenge and it is possible that your work is very timely to assist with a problem I have and would like to describe the approach you have used to engineers in the UK. Is it possible that you would share the drawings and specifications you used with your suppliers so we could replicate your work here?

    Many thanks

    Richard Skipper

  9. Dyna

    Good job … thank you for writing this article .. it’s very give my inspiration

  10. Iain McClatchie

    Jet aircraft fly up where the temperature is -40 C or less, but their pressurized interiors are at around 20 C. So the fuselage is at a higher temperature, and a much higher pressure, than the wings. On smaller jets, connecting these two structures is done with a set of links that kinematically constrain the two relative to one another, and allow for differential expansion. This is why all those Cessna jets look pregnant. The wing box passes completely underneath the fuselage.

  11. Craig

    I like the detail of your artical as well.
    So many designers do not understand the proper use of constraints in my organization.
    I love the fact you addessed overconstaint as well.
    Thanks For sharing.