Aluminum expansion uniformity -- design advice for avoiding warping? 4

[ttuibk]

Hello,

I am designing a mechanical system which uses an aluminum rod as a temperature dependent length tuning element. With the exception of machinability and long-term stability I do not care about the mechanical properties. The main problem is that I need the end faces of the rod to stay parallel during temperature changes on the order of 30 deg C or so near room temperature. Really parallel. Like on the order of 0.1 arc second. The other design option is a piezoelectric stack, but for large displacements (50um or so) I am not sure things will stay parallel enough, and I do not like the idea of the length being susceptible to fast electronic noise.

Am I totally crazy here? Or am I worrying about this too much. I am going to try it anyway, but what can I do to increase the chances of this working? Some questions:

1. Alloy choice. I have been reading about "artificial aging" and "annealing" for various aluminum alloys. Is this relevant? I chose aluminum mainly for its high expansion coefficient.
2. Shape choice. I chose a rod shape only to save space and weight. Will a rod with a larger diameter maintain parallelism better? The current diameter is 12mm. (150mm long). The cross section is irrelevant but there needs to be a 6mm dia. clearance down the middle, and as mentioned previously the length must be 150mm.
3. Machining. Would it be better to start from some kind of extruded piece and then anneal it? And then machine the ends flat at the very end? Or will machining the entire thing give a similar (thermal expansion uniformity) result. Or should I have everything machined and then anneal it?

Thanks for reading!

 

[arunmrao]

Why would you like to use Al rod, why not consider Invar for such applications.

"Even,if you are a minority of one, truth is the truth."

Mahatma Gandhi.

 

[ttuibk]

Hi arunmrao!

Yes as I mentioned, the other option is choose a stable material like invar or glass and then tune the total length with a piezo stack. But then that just pushes the "maintaining parallelism while expanding" problem onto the piezo stack. Interferometers with small piezo displacements (a few um) work quite well, but 50um is a different story.

For the temperature tuning option, I think invar would require a huge temperature change to increase the length by 50um. The thermal expansion coefficient for invar is about an order of magnitude lower than for aluminum (22-24 um/m/deg C).

 

[MintJulep]

Do you really need this level of parallel?

Could you for example make the ends hemispherical? Or pointy? Or a knife edge?

In my experience things that "will work perfectly, if we can just hold the tolerances tighter" never work.

 

[ttuibk]

Hi MintJulep!

In general I agree that simply forcing tolerances tighter is a sign of laziness or lack of creativity. I'm not trying to achieve a high parallelism between the alumimum faces though. They are just to hold the interferometer mirrors. The standard technique to make them parallel is with glue and laser beams.

I'm trying to keep those mirrors from becoming unaligned once I change the temperature. Standard optomechanical techniques like phase plates and slipstick steering mirror mounts are either way way more complicated, or will not work for my optical requirements.

I have a very weird tuning requirement (tune to a specific length, and then leave it there) and the temperature change technique would be quite elegant and simple if it works. This might be one of those things where it's easier to just try it, than figure it out though. I thought I'd check with the metallurgy wizards first ;-)

 

[hendersdc]

You could use three rods holding the mirror on a triangular frame (one at each point) instead of one rod to remove the need for face parallelism. Clamp the triangular frame to the OD of the rods and when the temp changes all three rods will change length by the same amount and you won't care if an individual face is parallel or not.

 

[ttuibk]

Interesting! I suppose that a larger rod spacing makes any small differences in expansion between the rods less important.

 

[MagBen]

You may use Hi-expansion 72 alloy with CTE up to 28ppm/C, for a replacement of 50 microns, and a temperature change of 30C, you only needs a 60 mm long rod. I assume the shorter the rod, the easier to retain parallelism. Otherwise, not understand why you need a high expansion material?

Also, to increase the thermal stability, people often do a stress relief anneal, especially after all cold works are done.

 

[ttuibk]

Hi MagBen!

I did not know about those alloys. Thanks that is very helpful. You are correct that a shorter rod would be easier but there are optical requirements which make longer rods require less length tuning. I chose 150mm as a compromise.

By "all cold works done" do you also mean machining?

 

[IRstuff]

You have not said anything that suggests that the length of 150 mm is related to the 50 um displacement, i.e., there is nothing to suggest that a single material has to do both.

TTFN (ta ta for now)
I can do absolutely anything. I'm an expert!

https://www.youtube.com/watch?v=BKorP55Aqvg

faq731-376 forum1529 Entire Forum list

http://www.eng-tips.com/forumlist.cfm

 

[ttuibk]

Hi IRstuff!

According to my optical simulations, if I reduce the interferometer length to 60mm then the required displacement range goes up to something like 1200um.

 

[MagBen]

Yes, cold works include machining, grinding, stamping etc. which can create stresses.

To assure optimal dimensional stability, Invar alloy is to heat 1000°F to relief the presence of cold work stresses. To promote temporal stability (when necessary), it is to age at 200°F for 24 to 48 hours.
For Hi-expansion 72, heat to 1450F for 1 hour, then furnace cool.
Not know much about Aluminum.

 

[IRstuff]

That's not what I was implying. You could have 60 mm of Alloy 72 and 90 mm of Invar.

TTFN (ta ta for now)
I can do absolutely anything. I'm an expert!

https://www.youtube.com/watch?v=BKorP55Aqvg

faq731-376 forum1529 Entire Forum list

http://www.eng-tips.com/forumlist.cfm

 

[ttuibk]

Thx Magben. That seems reassuring that the stresses can be annealed.

@IRstuff, I don't understand. How is this better? 150mm of annealed Alloy 72 with less temperature shift seems to do the same job without an additional mechanical joint? I'm not a mechanical engineer, but it seems like you would want to avoid more joints? The Alloy 72 faces are closer together, but then you have to increase the temperature 2.5 times as much as you would for a monolithic rod?

 

[Compositepro]

Aluminum alloys have a crystal grain structure that can adversely affect your goal. Pure aluminum would be your best choice (A1100).
To get uniform thermal expansion, good thermal conductivity is required to reduce temperature gradients.
A change in parallelism of the end faces implies bending of the rod. The greater your rod diameter to length ratio, the more stable that will be.

 

[ttuibk]

Hi Compositepro!

That makes a lot of sense about the thermal gradients. I am seeing 8 W/m/deg C for Alloy 72, and 235 W/m/deg C for Al. The current design has a cylindrical Al heat shield. I will try to servo the crap out of it at several points and try to keep the Peltier arrangement symmetric as well. Come to think of it, maybe it's even possible to align the interferometer with temperature gradients.

I will also think about the diameter.

I will try to find out: if pure Aluminum can be annealed? how well it can be machined.

 

[EdStainless]

Pure Al is soft and difficult to machine cleanly, you have to take your time.
And after you are done with everything take the assembly and temperature cycle it a bunch of times over a larger range than you intend to use. This will help remove any little irregularities.
I like the idea of using three rods with the ability to trim heating on each to fine tune position.

= = = = = = = = = = = = = = = = = = = =
P.E. Metallurgy, Plymouth Tube

 

[ttuibk]

Hi EdStainless!

Thanks for the info. I think maybe even Solidworks can do a thermal stress analysis.

http://help.solidworks.com/2012/Eng...lysis.htm?id=43e4391fc5ba49c1b6e4bffc0ab5983e

 

[Compositepro]

Using three more temperature controlled rods or bolts is a well established method of making high-force micrometer actuators. But if you want to maintain parallelism rather than to control it, then one large solid rod is desired rather than several smaller ones.

 

[ttuibk]

Hi Compositepro!

>Using three more temperature controlled rods or bolts is a well established method of making high-force micrometer actuators. But if you want to maintain parallelism rather than to control it, then one large solid rod is desired rather than several smaller ones.

I imagine that, in terms of maintaining end parallelism, it is going to be the case that the most temp-gradient sensitive position for a single rod is going to be the middle? and that transverse gradients are going to be more detrimental than longitudinal ones? Just thinking about how many temp servo points I really need on the heatshield, and how much longer it should be than the rod itself, since the ends have a hole where blackbody radiation can escape.