how do you design RF cans?

[CopperCone]

I noticed if you take apart a high quality RF cans you have little tunnels basically, but I have seen on UHF stuff (say to 3GHz) they just whack a circuit board in a can that does not really have complicated internal geometry, it is just a well sealing aluminum box.

I have seen some designs where you basically have the signal path, and then the can fits over it with like a 1cm lotus , so you basically have little 'caves' or tunnels separating everything out, so its almost like some kind of coaxial shield. The top and bottom walls of the can are mated through the non edge areas of the circuit boards with heavy (isolated?) via stiching.. or are they nonisolated?

Also, how important are curves on the internal geometry?

Lets say I wanted to make this, but using steel sheet and welding, for the sake of not needing to use big milling machines (hassle to own), just welded nicely and maybe ground down a bit (not to that extent) with a die grinder and sanded. I guess you can really fill the shit out of the welds to make your own bevels with a die grinder, but is it really necessary?

http://www.gquipment.com/30-large_default/rf-enclosure-mini-aluminium.jpg

So I would not have the interior bevels.

Is this stuff done in a simulator? Or are there pretty good rules of thumb with frequency?

I just can't see myself owning something that can mill something like a 12x12 inch RF can.

I figure you can get a decent flat seal by puttin the thing in a fucking wood planer so long you set it for very tiny cuts, then finish it off lapping on some sand paper ontop of a big granite measuring block... otherwise you still need a damn mill to flycut it

and if its very uneven you can use bluing and an aluminum file to get it within a few thousanths (fairly easy so long you use a proper aluminum file), so it does not try to fly out of the planer and kill you, to account from distortion due to welding

much more fun to build then milling too. unless you have a CNC then you can just chill while you wait but I probobly wont be able to afford that kind of shit for a decade

I assume you can get a good enough surface finish by going over it with a few different grits of sand paper (maybe to 1000) and then polishing it with some rogue

then you can anodize it so it looks cool and laser ingrave or etch it and all sorts of stuff with the price of some kind of mill, plus you have welding equipment and developed shop skills for all your antenna, optics mounts and shit

[David Hess]

Larger cavity volumes have a lower waveguide cutoff frequency allowing more coupling between circuits.  Sometimes an absorber is added rather than messing with the shield geometry.

[Dubbie]

Good luck getting any of the precision you are talking about with hand tools.

You're better off getting an old manual mill.

[bd139]

Not even worth it for one offs. Look at S53MV for construction ideas. Several GHz are fine with mediocre enclosures: http://lea.hamradio.si/~s53mv/spectana/sa.html

The milled parts are ridiculously expensive. They cost more than the boards that went in them when I did assembly on them.

[Yansi]

Unfortunately I can not agree completely.

I have just  asked a few (CNC) local machine shops, if they would mill for me a small custom RF enclosure. The prices quoted were not exactly cheap, but affordable.  Also, the price goes down the more of them you order. Most of the price takes the setup (CNC programming).

At five pieces of small custom 60x40x12mm I was quoted $68.  As I've said: Not cheap, but affordable. 

So if you take time do design yourself some universal sizes you will use often, then I guess that may be affordable way to have professional looking enclosures.

Mind you, I have access to some decent machinery and metal stock. I could mill some of these myself, not a big deal. But if you count the time it will take you to make one (just cutting stock and squaring it on the mill), you will find that you already spent the amount of time equal to the price the CNC guys offered for 5 pieces.

Note I have nothing against the "mediocre enclosures". I use them often for prototypes. A sheet of tinned steel cost just pennies.  But when I get to the point making something nice and permanent, I think, that milled aluminium cases are just affordable.

[CopperCone]

Tollerances matter with these things? They should be accurate to 0.3mm or so if dilligent i think

Btw for making flats you can make really flat shit by hand using a scraper or file and a granite block and blueing.

[CopperCone]

Also doesent waveguide have a low and high frequency cutoff for the efficent mode of propogation, so it acts likea bandpass filter with impedance?

Well more of a high pass filter, then some kind of notchyness/moding past its design frequency. I am actually looking ats parameter wideband testsof waveguide to see whats going on

I thought it was not really critical so long its sealed properlysince it does not really follow the pcb traces like a stripline or coax would, so what difference does it make if the wall developed a .5 degree bend due to weld stresses or a bevel is not exact?

The most important part is the top and bottom plates being parallel right.. But the waveguide behavior is hard for me to understand.

I know that in classical waveguide you needa certain geometry for the right TE modes. But since the edges in most cans are beveled dont you get some kinda half circular have rectangular shit that has complicated propagation modes?  When i cut apart real waveguide it is really square inside, it looks extruded or cast square

I also renember thatin my microwave antenna for 20ghz it uses some kinda oval waveguide which kinda fits but you really need a transition to a square endlauncher

I need to look at lance or wheelers? Book it had some information i think about te modes of some random geometries but thats proper waveguide, this seems likea bootleg waveguide with a really bootleg launcher (pcb).. But you want it not to couple so youdont lose signal power so like you gottathink backwords or something

So given that the traces are allover the place, and youdont have a real injector on the pcb does a slightly angled wall matter? Do i really need todeal with paying someone to run a robot orto get a three car garage to do this? I dont like paying any kind of contractors or tradesmen really...

Also, ineed to look for waveguide tollerances for manufactured waveguide i am curious now

And i am trying to get my money from my android phone back so thats why these daytime posts are a bit shoddy, enjoying the outdoors and people watching etc.. 400 dollars 3 years ago and i was hardly using it..

[Wolfgang]

Same experience:

- Milled enclosures are perfect, but only when properly plated and when the lids are screwed on with a LOT of screws.
- Up to a few GHz, tinplate boxes work with a little higher loss, but good enough for a lot of applications (yes, the 0.05dB NF amplifier is NOT one of them).
- Make the enclosures as small and flat as you can to avoid resonances, or at least to shift them up to a very high frequency where your circuitry has no gain left.
- Use chambers and absorptive material to dampen out resonances.

Much success !

[coppercone2]

What do you mean by properly plated?

Also what do you think about RF gaskets like BeCu fingers or foam rather then super tight mechanical design using many screws? Especially if they are glued onto a oxide clean surface using conductive epoxy rather then the shitty adhesive they are supplied with. Can you get away from having to use a flycutter and stuff on a home made box this way so it closes properly? I can file flat a alumium box real nice but it won't ave the quality of a flycutter.

I assume you mean that a slot antenna forms between the screws so you need alot of them to break up the little loop that forms right? How do finely spaced RF fingers compare to a typical machine screw heavy enclosure, if proper tension is achieved? How about multiple rows of fingers, with overlapping (so there is no slots), or fingers in combination with compressible foam? Can generalizations be made to the gap geometry to get similar performance from tight machining and creative solutions like fingers?

https://www.compeng.com.au/rf-shielded-enclosures-accessories/

And what if you just solder the can shut and rely on a special heating device to open it, using low temp high conductivity solder? Easy enough to make a custom fitted hot wire heater for a specific can. Or conductive epoxy seals?

I am trying to find ways around owning large, heavy, dirty and space hungry machine tools that I have little use for or interest in. And the tooling costs are absurd. And I don't want to learn all the settings, feed rates, etc.

 It is difficult for me to compare though, because I would need the milled enclosure to compare my home made ideas to.

Also, what are chambers?

[David Hess]

Also what do you think about RF gaskets like BeCu fingers or foam rather then super tight mechanical design using many screws? Especially if they are glued onto a oxide clean surface using conductive epoxy rather then the shitty adhesive they are supplied with. Can you get away from having to use a flycutter and stuff on a home made box this way so it closes properly? I can file flat a alumium box real nice but it won't ave the quality of a flycutter.

There is another kind of gasket material which is springy and has sharp "fingers" on both sides to puncture any insulating surface like a star washer.  I have also seen copper braid used.

I assume you mean that a slot antenna forms between the screws so you need alot of them to break up the little loop that forms right? How do finely spaced RF fingers compare to a typical machine screw heavy enclosure, if proper tension is achieved?

Exactly and it happens with pretty much any material.  Springy materials like steel may have the fingers built in.  Anything which forms an insulating oxide surface like aluminum is especially a problem.

And what if you just solder the can shut and rely on a special heating device to open it, using low temp high conductivity solder? Easy enough to make a custom fitted hot wire heater for a specific can. Or conductive epoxy seals?

I have seen this with tin plated steel shields and hermetically sealed enclosures.

[Ian.M]

For rectangular one-offs, if you don't have a box & pan brake with small enough fingers to form then out of tin plated steel, just use double sided coper clad FR4.   Solder the inside seams, and copper foil tape the outside seams.   Either make up L section fingered strips to seat it on from springy brass shim stock and solder them down, or solder it directly to the PCB, or solder on brass angle flanges and bolt it down on a tinned copper braid gasket.

[coppercone2]

How do all these methods compare to the industrial RF standard of using a milled machine box that is surface ground and sealed with screws every 1.5cm or so? Does anyone have any plots that show attenuation?

[mzzj]

If you are just making enclosures and rf  "boxes" they don't need be accurate, just tight seams and lid with enough many screws. Corner rounding radius or wall paralleism doesn't matter.

If you start making parts where the "box" is not just a box (like cavity filters) then things get more "interesting"   

[coppercone2]

Yes, I am interested in things that enclose PCB's, when I took apart some waveguide stuff it looks like you needed a deep shaper machine and a lathe to make the square sides and disks, not to mention various plating technology to make  the silver plate on the aluminum waveguide. This was just a cavity filter thing.

It does make me wonder if you can make waveguide out of welded stuff so long you have a very precise X-Y table on a very stable arbor press to use as a shaper, you can technically shave down the sides to parallelism, if you have a special tool that has scrapes on all 4 sides. It would be a lot of plunge cutting, I don't think a regular arbor press would be suited though. If you used thicker copper and solder instead of weld or braze it might not deform too much, so you might just need to shave a few hairs off, if you solder with spaces in place to ensure a consistent gap geometry, with minimum spacing between the parts (0.5thousanths maybe), since mechanical strength is irrelevant. It would be a nice machine if it can work because the travel would need to be like what, 3 inches?, and it would fit the same form as a drill press, so it would not occupy much floor space. It would only be hard if you need to make a big ass wave guide for like <2GHz. The tool would need to be long for for shaving copper or aluminum I doubt it would need to be very stiff? I imaging strait waveguide parts up to like 1.5 feet could be made that way, but hell if I know how to smooth out the interior of bends and stuff, maybe you could live with some irregularity there 

The stuff-in-waveguide looks like you need a serious machine and chemistry shop. I do have a feeling from reading that so long you are under like 15GHz you can get away with significant butchery of waveguide geometry with little consequence 

microwave101.com
"During the brazing and hardening process the walls of the waveguide can become deformed from their ideal geometry. At frequencies Ku-band and lower this tends to not be much of a problem; a few mils of error represent less than 1% error and do not cause an appreciable effect on the insertion and phase of the signal at these frequencies. However, when reaching higher frequencies such a Ka, V, and definitely W, this becomes a problem worth thinking about when considering dip-brazing."

Taking some liberties with his numbers, I figure a rule of thumb might be 0.005 inches bulk geometric change = 1% error, since a few probably means 3 and 3 is less then 1 so we can make 1 = 5. Engineering math. Might be as high as 0.01 inches.

I have some avionics waveguide for around the Ku band, I wish I had a CMM to see what was considered acceptable for its manufacture. I doubt a micrometer will give me terribly useful information but I will give it a whirl. 5 thousandths is not that crazy accurate. Low melting point soft solders might make this really achievable. 1% of a high power signal is like 1dB. For 10GHz thats in the same ballpark as RG141, which is like 0.6dB/Foot @ 10GHz, wheras commercial WR90 waveguide is expected to have something like 0.03dB/foot. I wonder what he means exactly. If he means error from the ideal target specification, then we are golden, since 1% of 0.03dB is basically nothing. Who gives a shit if its 0.03dB +- 1% ?? Its still a dead short compared to anything else you can do. I doubt anyone here can even measure better then 0.05dB @ 10GHz anyway, unless they work for Lecroy or Agilent.

I think I have seen hybrid cavity filter/PCB enclosure thing inside of a spectrum analyzer, but I never took apart the cavity filter bit, It's the part where you use screws to adjust little 'pistons' to tune the thing right?

[David Hess]

For rectangular one-offs, if you don't have a box & pan brake with small enough fingers to form then out of tin plated steel, just use double sided coper clad FR4.   Solder the inside seams, and copper foil tape the outside seams.

I made my first FR4 boxes that way but they much stronger if the outside edges overlap so they can be soldered as well.  This can also physically protect anything protruding through the FR4 like piston trimmers and feedthroughs as shown below.

[coppercone2]

All I can complain about is that it needed some lacquer after soldering, like a coating of canned polyurathane spray, shellac, etc. And rubber feet.

[bd139]

Lacquer just slows it down. Doesn’t solve the problem. Been there, found that out. Tinning the parts is probably better.

Edit: or just shrug. Function over form

[coppercone2]

how about top-coat clear-coat spray paint that uses acrylates? Lol, or just paint.

I wonder if you can solder together waveguide out of FR4 thats decent if you coat the inside of the waveguide to prevent solder wicking, to minimize the multimode rounding that would occur from the solder wick. Like expose your PCB to have solder resist everywhere, then just wash it out with a solvent after soldering. You can glue it down first carefully with cyanoacrylate to provide the mechanical strength wich would ordinarily come from the large solder bevel when making a PCB box. Since less solder is used, there should be less deformative stresses, since the solder is being used for an electrical connection only. If you mate them well with glue, and have resists, you might just be able to lay down a bead of 0.2mm solder to make the joint at the seam. You would just need to very carefully glue it, using a microapplicator, and align everything correct. I wonder how much error that would result in, and the minimum amount of solder you can use to make a seamless electrical connection with minimum bevel to maintain the right angle as much as possible. I would angle the thing so the angle is pointing directly dowards, the thin solder wire is pressed into the fluxed bevel, with minimum copper exposure, and heat it with hot air or maybe pulsed laser?

If 0.2mm of solder is pressed into a right angle, the bevel should be like 0.005 inches radius if the idea works.

Maybe you can do even smaller if you figure out how to properly apply the fine grade solder paste. The most aggressive flux can be used since it would be cleaned afterwards.

[bd139]

Paint works: https://aa7ee.wordpress.com/2013/10/19/the-vk3ye-micro-40-dsb-transceiver/

Here's one I did recently, a simple return loss bridge. Didn't need to be terribly RF tight as it's HF only so the lid is just screwed on and it's single sided board so all the skanky copper stays on the inside:



Edit: guts:

[David Hess]

I used to clean and spray them in the past but the spray coating always rubs off and I think that makes it look worse than if it was just plain tarnished copper.  That particular one is a helical resonator (1) so silver plating it would have some performance benefit (2) but it is largely more trouble than it is worth and it would still tarnish.  Electroplating nickel on copper would be my first choice because it is easy to do well but there are other chemical surface finishes which could be used.  Or send it off before final assembly to a professional to plate it.

I have an idea for the next helical resonator that I build which will include variable coupling.  That one I might try silver plating because it will have finger wiping surfaces.

(1) It is the second helical resonator I ever built (3) and it shows its age.  It works fantastic but overcoupling resulted in about 0.05dB of loss and a wide passband which has its advantages in some applications.

(2) It is easier to make the resonator slightly larger to make up for the loss of Q if this is an issue.  In a production environment, there is no reason not to silver plate it.

(3) The first one was constructed like bd139's example but that was how I decided that double sided FR4 copper should always be used with overlaps for fillets on both the inside and outside for strength.  Plus the extended sides allow the enclosure to sit flat when things like piston trimmers extend through the sides as shown.

[coppercone2]

Assuming I found a copy of a expensive HFSS EM simulator that fell off the back of a truck, how difficult would it be to bodge a model of a waveguide made of PCB so I can determine what effects the solder bevels would have on the propagation characteristics?

Is there a tutorial that shows you have to make the different shapes and set up the simulation? I never used 3d cad software before. I just wanna see a rectangle vs a rectangle with tiny solder bevels.

I am looking a piece of brass waveguide coupler right now under a magnifier, and there is not really a discernible bevel with a standard magnifying glass. You can see some kind of edge discoloration but it seems pretty flat when held up to a light source.  The edges on the exterior waveguide (which the signal goes through) look a bit galled towards the interior though, maybe from the mating process? It is around 10GHz.

Is there any EM programs that I can upload a solidworks model into? (there is alot of youtube videos on how to make solidworks models but I really don't wanna learn how to make the same shit in a different program)

I get the idea of how 3d cad programs work, but I feel that I am going to get clobbered by hard to find menu settings and stuff.

oh i might be in buisness, http://www.padtinc.com/blog/the-focus/importing-and-splitting-solid-models-for-ansys-hfss-18-0

[David Hess]

Actually if I electroplated the outside, then I would bead blast it or coarse sand it first to lower glare.  In that respect, tarnished copper is actually pretty good.

[coppercone2]

I am running an experiment to see how well car wax does at oxidation prevention on some old step stool I sanded last night

[bd139]

Notes here on copper tarnish and lacquer: https://aa7ee.wordpress.com/2015/10/21/how-do-lacquered-boards-stand-up-over-time/

[coppercone2]

I found this on microwave 101

"
Electronic discharge machining waveguide

There are two types of electronic discharge machining: wire EDM and plunge EDM. Both methods use high-voltage to melt away metal, rather than mechanically removing metal with cutting tools.

Wire EDM uses a very thin wire that travels from one spool to another, is energized with high voltage, and is placed in contact with the part being machined to make a linear cut by blasting away material. Picture a tiny, round bandsaw... The part being machined is typically held stationary, while the two spool feeds are moved to steer the saw. In the best wire EDM machines, the top and bottom spools can be moved independently. This allows more complex shapes, for example, a cone can easily be machined. The part being machined is suspended in an oil solution, which is used to carry away and suspend the machining "dust". The accuracy of modern wire EDM machines is one mil or even better. Wire diameter for wire EDM is typically just a few mils, and brass is the material of choice. A five-mil brass wire will allow you to cut a 2.5 mil inside radius on your waveguide part.

Plunge EDM uses carbon elements that are energized to high voltage, and are pushed into the part being machined to blast away unwanted material. The part being machined is suspended in an oil solution. The plunger can be a complex shape, and can be moved in X, Y, and Z axis, as well as rotated, enabling more complex shapes than wire EDM. Accuracy is better than one mil. Plunge EDM is far more expensive than wire EDM, because the plunger must be machined, and it has a short life. Plunge EDM is best used in tool and die manufacturing, or creating extremely accurate molds for injection-molding plastic parts. The expensive but highly accurate plunge-EDM part that is then used to create thousands more parts."

It seems that they manufacture waveguide using EDM, so unless they follow it up with some kind of shaper, the waveguide will tolerate 0.0025 radius, which maybe achievable with fine solder in a PCB board corner. You would just need to make the side flat and use right angle blocks while gluing it to make it perpendicular

I also think that using technologies like indium corporation nanofoil might be a good option for sealing your waveguide parts without a bevel or warping.