HTSSOP dead bug cooling for high vacuum (Solved)

[mongo]

First of all this is going to be an obscure question but as so many experts look on here I will try.  I have a project where I need motion control under high vacuum, due to limitations on surplus connectors and the size limitations on my vacuum container I have to have a significant amount of electronics in the chamber.  I am stuck with this until I can produce a POC to justify the insane costs of modifying or building a purpose built chamber.  I have a 60L camber which was purchased as surplus from a biomedical company and thus it uses BioPharm Clamp fittings which are typically not built to the 10^-8 Torr level I am using so I am stuck for now.

I have been producing single layer boards by electroplating copper or kovar on drilled borosilicate glass, I then would mill a PEEK, copper, or alu water manifold for cooling.  But I would love to use the L6470 which is designed to cool through VIAs.  I have tried to use various metal rod and other solutions to get appropriate cooling but as my self plated PCBs trace strengths are not high and because I cannot find a thermal grease that is HV compatible I am planning on trying the dead bug method.  I think that by having the glass under the "top" of the package that I will be able to get enough pressure against the package to get fairly good cooling but bridging the gap to the pins with HV compatible solders like Sn95Ag5 seems to be problematic.  I also tried to mechanically couple two heat sinks but have run into problems with differing expansion coefficients.  I have dedicated cooling lines for my sputtering head which makes it easy to liquid cool but I was hoping someone had tips on a way to bodge this together.  Or if you know of a small prototype service which will produce traditional boards with known compatibility that would be great too, but I am quite sure that the solder mask or glue contaminated my diffusion pump oil, and I assume that this will be true of most modern rohs compliant boards.

[ConKbot]

At 10^-8 you're getting into the territory where everything matters. (As I'm sure you know)  for PCBs in vacuum, check out if any Rogers RF laminates are suitable(a lot of them are ceramic loaded teflon) , I have seen outgassing information about them online, but cant recall what sort of vacuum range they were doing. IIRC there was a recommendation for no solder mask, everything gold plated to prevent a bare copper surface from oxidizing, and holding moisture in the surface porosity. (Or perhaps just gas diffusing/adsorbing in? Cant remember)  If you had to prevent wicking, they were using just a little stripe of solder mask or silkscreen between the trace and the pad to prevent the solder from wicking down the trace.

For thermal grease, just use whatever high vacuum grease youre using anyway. Literally anything is better than a vacuum, so as long as the gap is filled with something in a thin layer, you've got a decent start. If your surface is particularly rough, maybe fill the grease with a finely powdered ceramic thats thermally conductive and plays nice with vacuum?

With Rogers material, you should be able to get a PCB fab to do double sided with vias for you, just work with them on your oddball requirements for solder mask. Or if single side works for you, drill a hole though the PCB and have a stud that sticks up though it to touch the thermal slug that way. If youre inside the range that Hysol 1C works without outgassing, you could use some of that to relieve some of the strain on the leads by gluing the chip corners to the PCB.

[lincoln]

hmm motors going to be under vacuum? I understand you need to get things running but this sounds like a "unique approach" to the problem of moving some thing in a chamber.   The applications that I have made mechanical parts for has also been very hot so putting motors or control electronics in the chamber would be a non starter. Depending on how much torq is required almost always there was some sort of magnetic coupling through a stainless steel housing.

[mongo]

ConKbot:

Thanks for the tip, I'll dig into Rogers RF's data sheets and yes everything is hard at this level, I never imagined I'd get to this point in the project so far.  There is a good reason I stopped at 10^-8 as things get even harder below 10^-9.

lincoln:


This is not a long term solution, it is purely a POC before investing 5-6 figures plus on the appropriate gear if the product ends up being viable.  That said I also would had not tried this at all if I hadn't acquired some stepper motors designed for spacecraft use at an auction a while ago.  They are an older version of these.

ftp://ftp.phytron.de/phytron-usa/stepper_motors/physpace-us.pdf

They will overheat using traditional stepper drivers which is why I am adding the complexity of a more modern driver. Luckily I do not need to bake at this prototype stage, and I never expect to have production ready anything from this system.  Even though I have them well shielded they will die at some point due to spurious deposition so hopefully they don't die before I am done.

I admit it is a complete bodge and if this POC is successful I will be looking to contract the work out before attempting to deal with this for an actual product.  I just have some specific material needs and complications in this application that makes the cost of having this done professionally outside of the budget of an individual.  This complexity is also needed due to my limited power budget but yes, I have to at a pretty low duty cycle to protect everything inside but as long as I can keep the critical parts below 200C I am OK.  Luckily part of the reason for the complexity here is due to my limited power budget with residential electrical service which also limits the thermal considerations.  There is also the complexity of the target materials tolerance of super high temperatures so it is not difficult to make sure the electronics are first in the cooling loop.  With the exception of the stepper driver cooling so far things are working OK with just a few degrees of rise at the target's intake.

Even if I pay to have a more passthroughs installed in my current chamber I would still have many challenges to make it viable for long term use like the fact that I only have a liquid nitrogen cold trap and know the sound of a contaminated oil diffusion pump all too well. This is the main reason I am so paranoid about material selection.

Thank you for the warnings, they are quite valid. I have far less into the vacuum system than a cell phone costs and wish the same could be said for the argon and liquid nitrogen bill.

If you are still making parts for this type of system message me your contact info.  If I succeed in the POC yet fail to find a contract producer I'll try to send a few quotes your way.  I promise those specs will not be engineered by me.

[tszaboo]

The plastic package itself can absorb moisture, which will outgas when you decrease the pressure, due to water starting to boil. Then you have a crater in your IC, and they dont like that.

[ConKbot]

Vapor pressure of water is 1 atm at 100c (worst case for die temperature) vs 15atm for 200c when soldering. If someone can popcorn an ic in a vacuum without exposing it to soldering-like temperatures, I'd be impressed.

[peter.mitchell]

have you considered IMS/MCPCB? i'm not sure specifically what epoxies and the like are used to insulate the copper form the aluminium, but i think it would be worth looking into.

[tszaboo]

Vapor pressure of water is 1 atm at 100c (worst case for die temperature) vs 15atm for 200c when soldering. If someone can popcorn an ic in a vacuum without exposing it to soldering-like temperatures, I'd be impressed.

Well, IC packages are stored in controlled enviroments. Soldering pre-heating happens at 150 degrees, so boiling is at that temperature, where the vapor pressure is 4.7ATM, while the outside pressure is 1ATM. We know that stuff can go wrong with this setup, 3.7ATM difference.
The die will run hot, since conduction is not really possible. Say, the usual limit 125C is reached, where the vapor pressure is 2.2 ATM, difference is the same. 2.2/3.7 is 66%

[TerraHertz]

Wouldn't it be simpler to put all the electronics in its own small container inside the vacuum chamber?
Then only the exterior of that container (and its feedthroughs) need to be HV compatible, while PCBs and components can be garden variety. Sounds easier and cheaper to me.
It can contain air, or an oil, and conduct heat to outside the main chamber via your liquid cooling loop. Or a heat pipe.

Or, doesn't your chamber have some small ports? Can't you make a cheap feedthrough plate with as many connections as you need?

For bulk quantities of feedthroughs, see ebay 261672048845, 231658118550, 251041737116 etc.

[Kleinstein]

These SMD power chips usually use the pins to get rid of most of the heat. So dead bug style could be a problem if may pins stay free or use only thin wires. Usually a motor does not need that many wires - the electronics inside also needs power and control signals - so I doubt one will save very much on the number of feed through pins.

Outgasing from the electronics also get worse at it gets hot. So low power parts like a preamp for an STM are usually not a big problem, but at higher temperature even normal plastic cases might give off to much. This is especially a problem if you can't bake your system, as it will take very long to get contaminants out.

So I am afraid it would be more economic to buy the feed through parts and have the electronics outside, even if this means to buy or custom make an adapter from one type of flanges to a standard type. I know the prices for the feed troughs can vary a lot - so some search there can pay off.

[mongo]

Thanks for all the input,

To be clear this project started when I won the default bid on some gear on dovebid, originally I just wanted to make a sputterer to know how it worked but then I realized that a type of sensor I needed in a project years ago was possible. 

The chamber does have two small ports and one does have a commercial Type-C Subminiature passthrough and the other is used for power, cooling etc... Unfortunately due to the design of this chamber, which is a flip top, I would have to have an entire new top produced and the cost of getting a proper chamber made is not significantly higher and my pin count in available in the port size is small.

With the provided input I think I will just encase the entire electronics in a pressure vessel with a cooling loop ahead of my sputtering target.  I have to keep my target and surface temperatures at around 100C to get the chemical composition and grain structure I need so high heat isn't a huge issue in my application.  Due to the huge capacity of my container, the impermeability of the surfaces I am working on and the fact this doesn't need to support high volume production I can solve most of the moisture and other off-gassing issues by just ramping up the vacuum very slowly.  I do pre-bake all new items in an oven before putting them in the chamber and I also flood the container with argon with a "day tank" in-between use and this has been OK for this experiment so far but the chamber I have was also rated to some insane level of helium permeability so back filling an 5 gallon empty tank to 50PSI of argon will keep it above 1 ATM for a couple of weeks.

I am not sure why I didn't think of the enclosure solution except that I have a sputterer, so it is probably just a case of when you only have a hammer...  To be honest the hardest part was getting the gcode right to reliably drill glass without breaking it and if I hadn't run into the issue of having steppers overheating while holding I wouldn't even need modern parts.  But the powered off holding torque of my steppers is to low so I just need to use a stepper driver that will drop voltage on stop to prevent the windings from overheating. Unfortunately precludes most of the chips with heat sink or through hole construction. Plus I have become a L6470 fanboy after realizing how simple they are to deal with.

I am sorry if this entire post is a bit obtuse, but I had to solve some production problems around the consistency of sputtering 304 onto tubes for this project.  When I can I will share the information and it will make a lot more sense.  I am hesitant to share too much at the moment due to the weakened prior art requirements in the US patent system.  I am not going for a patent for this process but I need to produce a publishable, peer reviewable paper on my solution to prevent a very well known company in this space from filing their own.  They will still probably file within the 12 month limit but hopefully it will limit the cross licensing costs or defensibility of the solution for their competitors.

[TerraHertz]

Three cheers for patent blocking. That entire system has become a monopoly by big corporations, impeding general technological advance.

My own vacuum system project is remarkably similar in terms of needing to keep silent on the details at present, for exactly those reasons. Don't aim to create any patents, but if it works it's crucial to publish in a way that blocks any bastard corporations from trying to suppress it via the patent system.
Frustrating, isn't it?

Glad the idea was a help.

[kony]

...but if it works it's crucial to publish in a way that blocks any bastard corporations from trying to suppress it via the patent system. ...

And that is done best how? I have such issue to solve too, as it seems. 

[jeremy]

...but if it works it's crucial to publish in a way that blocks any bastard corporations from trying to suppress it via the patent system. ...

And that is done best how? I have such issue to solve too, as it seems. 

The absolute best way is to publish it in a peer reviewed journal. I've seen multiple patents thrown out due to a small paragraph in a paper.

Its complicated if you aren't associated with a university, so if that's not what you are looking for, maybe
Arxiv?

Even a nicely detailed blog post which shows up on a few tech sites is better than nothing.

I know that in Australia, you can (it's been a while for me though) file a provisional patent for about 80 bucks with the intention of never following through with the PCT. Among some of my colleagues this is known as "spoiling the earth", because it means that nobody can patent it later (or at least, your provisional will override theirs as prior art). This a tactic used by academics who want to share their creation and be free to work on it without worrying about patents.

[calexanian]

I did not read the entire thread but are mechanical feedthroughs not an option due to cost? That is the traditional way of doing this sort of thing.

[helge]

My vacuum experience is limited to buffer amplifiers that drove quantum dot structures in lHe cryostats where I guess outgassing was prevented by the environment temperature. 

Maybe instead of investing so much in trying to get bare PCBs clean it may be worth buying hermetic ceramic or metal packages like these and tuck all the stinky FR4 socks inside:
http://www.streamtek-electronics.net/pape/Photonics%20Package.htm
http://www.ametek-ecp.com/solutions/electronicpackaging/ceramicvsglasspackages
should also work with components what dissipate significant power and different metal lids may be available to increase the usable volume.

ceramic LCC packages are <$20 and if you're doing prototyping you could start out with a proper o-ring seal on top of such a package. Then your outgassing problem is reduced to pumping vs. leakage rates. Think of it as an investment to save the time you'd otherwise spend on workarounds to fix the thermal, reliability and possible EMI issues.

Thinking of it that way, could you try a box only open on one side facing towards the pump port? Perhaps you can redneck your way around it that way (only applies to setups with TMPs / IGPs directly attached to the recipient and large lambda at.... < 10^-6 mbar?).

[Howardlong]

I haven't seen much power are you dissipating, sorry if I missed that.

Does the package have a thermal pad?

The stuff I've been involved with for space applications is usually plain old FR4, but often we don't use solder mask or silk screen. We generally test in a thermal vacuum chamber extensively and there's a lot of thermal modelling which is dependent on anticipate in orbit parameters such as varying amounts of time in eclipse.

We also often conformally coat boards, and teflon coated wire is used for interconnects.

At the end of the day the thermal design of the board and system as a whole is key, and you are reliant on local heat conduction and radiation to dissipate that heat away, so methods dependent on convection or fluid conduction don't work... as I'm sure you know!

About ten years ago I was involved on a project where the designer of a solar shunt used a mosfet in its linear region to prevent battery overcharging. In the thermal vacuum chamber, there was no solar simulator, the use case for the shunt in vacuum was never tested. The satellite lasted about a day in orbit before the batteries completely depleted, as the mosfet's failure mode is generally to go short, essentially continually shorting the solar array.

I was involved in a retrospective peer review, it turned out that the designer had read the spec sheet, seen that the max junction temperature was 150 deg C, and designed his heatsink around that figure. Now while it worked in the presence of air to convect the heat away, in a vacuum, well, not so much. Designing for a junction temp of that kind of temperature is daft anyway, and thermal cycling over large temperature ranges will cause repeated physical stresses on the part from differences in coeficients of expansion leasing to early failure.

In general, introducing a pressurised module brings its own complications, but I've never been directly involved in a project that resorted to that, however it was common in Russian space craft to do exactly that for thermal and outgassing reasons. ISTR more recently a hard drive was sent into space in a hermitically sealed pressurised box I believe to stop any lubricants evaporating. Mechanical things in a vacuum is hard precisely because of the lubrication problem, but I am sure you've addressed this too in your motors.

So in short I would steer clear of exotic designs and keep it simple, but also pay close attention to any thermal modelling and thermal vac testing that you do, and that it reflects all the real use cases of your device.

But if you let us know a bit more about the amount of power you are trying to dissipate, maybe we can help further.

[mongo]

I have some parts on order and some lathe time but I have a direction forward.  I found an old picture from when I was getting my chamber out of storage that may make the story a bit more clear, although it is dirty in this old picture.  This is a research vessel from a bio pharmaceutical company which has a drain port on the bottom and a port for a mixer on the top.  As I am using the top port for the high voltage and gas inlets. I am running my digital and vacuum tree through a T junction on the bottom port.  I have ordered parts to "invert" the NW/KF style port on the bottom, and I will be milling a tube to fit in a coaxial fashion in to allow the largest passthrough possible.  It will pass through the entire top part of the tee and will then attach this to an off the shelf vessel inside to house the electronics.  I will attach my vacuum pump to the base of the tee.

Sorry this is so dirty in this pic, but you can see the top port in the red box, surrounded by the mixer mounts.  For scale that is a 12 pound rounding hammer.




While I could have removed one of the viewing ports I need them at this "shade tree" point of the process.


I have good cooling plates which will fit into the "pressure vessel" which will really be at 1ATM.  I will simply mill a mount for my existing electrical passthrough and cooling lines into that vessel.

These cooling plates are overkill for the application but they came with my Haake immersion coolers and I will just be using some sparkfun controllers or another board I have on hand for the stepper controllers.  It will be trivial to thermally connect them and I think I'll just use kapton tape to do so.

Note that the reprap board is just set on one plate for scale.



These plates will just be in front of the same cooling loop that is cooling my sputtering target.  As you can see they are obviously anodized and thus not usable at HV.

After I thought of just creating a reverse NW/KF tube I did consider just creating a higher density passthrough myself or attaching a larger passthrough inside but the cost was actually cheaper to do it with an internal canister at 1ATM. Plus I will be down to a small number of tig welds and one lathe part that will not be commercially designed and tested for the job.  I do have too large of a chamber for this need so removing capacity from inside will also speed up my cycle times.

Thanks for all the input, I have learned a lot from your posts and they have lead me to even more information.

[Kleinstein]

Pumping time does not depend very much on the volume - it's more the surface that takes time. Less volume only saves you a little of argon (or whatever gas you use) for backfill. The port for the pump also needs to relatively large - for high vacuum pumps a reduction is diameter should be avoided. For that size I would expect at least 63 mm diameter, better more (e.g 100 mm).

There were reasonably priced feedthroughs with sub-D connectors, so you can get quite some lines through a single port. So even if you need a cone to go to a larger diameter this should be easier than having a separate pressurized (gas filled) box inside. Normally a 2 phase stepper motor needs 4 wires and might even work with 3. The stepper driver will need at least 3, more like 4 wires. So don't see much saving in having the motor driver inside, unless there are several motors.

[mongo]

I am 3 axis + sensors and yes I am using a commercial sub D and running out of pins is, this is the major attraction for using an MCU that locally uses SPI to talk to steppers and one wire to talk to temp sensors. It looks like I need to break out the math. but even my main port is suboptimal.  I am fine with this for right now, as it is a pure proof of concept, I will be having a proper chamber made if I can get within a factor of 10 from my end goal...which I am close to.  If I can improve just a couple more items I may also be able to just give up a view port.

[Howardlong]

Immediate problem that springs to mind is the use of electrolytic caps, forget it in vacuum. Tantalums like AVX TPS have been found to be OK, or ceramics.

[sarepairman2]

pour some water on it

[Howardlong]

pour some water on it

Lol!

[Smokey]

#1)  Stop making your own boards.  There is no justification for this if you are actually trying to accomplish something other than making your own boards.
#2)  Start using PCB material designed for thermal transfer such as Metal Core or IMS.  Think high wattage LED boards.  Yes it will be expensive, but it's the right tool for the job.
http://www.multi-circuit-boards.eu/en/products/printed-circuit-boards/special-production/metal-core-pcb.html

[mongo]

In this new plan the caps will not be under a vacuum.



Although these are triclamps the concept will be the same



If you look at the central strainer you will notice they just used a flipped trilamp.

Checking the bundles of my cable and cooling needs I can get away with 1/2" heavy wall so the ratio of the coaxial tube will be similar to the displacement on this 2" triclamp.



This is effectively less than 10% of the surface area of the circle so the impact on flow will be minimal in this application.