twist an fr4 PCB through 90 degrees

[BMK]

I have a product design that includes 2 PCBs soldered together at 90 degrees. Currently the boards are jigged and then inductively soldered together at 6 points, with solder wire being fed through 6 separate mandrels. They've been doing it that way for 10 years and have been living with about 87% first time yield, and about 500 PPM of latent defects caused by poor soldering that pass testing but fail in the field.
This is an ideal case for a flex circuit, except that board rigidity is required in some areas, hybrid flex/rigid costs too much.
Is it possible to twist a narrow strip of FR4 pcb through 90 degrees? Say 20mm long and 6mm wide? Has anyone done this in a real product?

[PeteH]

Can't do some kind of board to board connector between the two? Or ribbon/cable harness...

Its hard to have good rigid areas in one area then enough flexibility to bend 90deg in 2cm. Is this a multi layer pcb?

Have done rigid bends before but you cannot put vias in areas that bend, the dielectric thickness should also be very thin....

[fivefish]

Are there other supporting L-brackets holding the board? Or is it just held by the 6 solder points?

I'm also using right angle boards, using a right-angle header soldered to the pads. But the board is also held securely in place by right-angle brackets screwed in place. So the solder points aren't carrying the stress.

[Ian.M]

Redesigning it so each track has two well separated joints in parallel should get the initial defect rate down to a couple of percent and improve long-term reliability.

[babysitter]

Hi,

mechanical enginners and technicians are the right people to tell you how to shape the PCB to get it thru your channel.
Also you might save some money by making it instead of a "real" flex-rigid into a flex-only and stick it to stiffeners yourself.

[IconicPCB]

How long is the right angle corner and how many nets go via the right angle joint?

Ever consider castellated corner joint?

[jeroen74]

Maybe a semi-flex PCB is an option?

http://www.eurocircuits.com/blog/192-SEMI-FLEX-pool

[tszaboo]

Or take a look at Würth Wirelaid. I have no idea about pricing, and only seen a demo, but I guess it is working, because the german guy said so.
87% is pretty bad, IDK how they lived with that... 

[nctnico]

Is it possible to twist a narrow strip of FR4 pcb through 90 degrees? Say 20mm long and 6mm wide? Has anyone done this in a real product?

Yes. I have seen this in an IBM PSU which I took apart about 25 years ago. I think the process is expensive though because I think the PCB needs to be bend somehow during production. A much better choice is to use a connector and a mechanical re-enforcement. Putting mechanical stress on a solder joint is never a good idea and should ALWAYS be avoided.

[X10an]

Found this old thread while searching for this possibility. Find it kind of suprising this subject is not more discussed and out in the open. Or is my 15 minutes of googeling not enough?

Now I am experimenting on a print for this box:
https://www.ebay.com/sch/85x58x33+Clear+Box
One PCB on top behind the glass with lots of LED`s, one PCB along the side for microcontroller, LED-drivers and ICSP/FTDI, and one PCB at bottom for accelerometer and radiomodule. The bottom PCB is cropped to keep the PCB under the 100mm pricelimit at Asian fab`s. The purpose of this circuit is of no importance in this discussion, but the fact is that I have now ordered the first prototype to find out if this actually is feasible or not. My first order is for 0.6mm FR4, and contains no traces on the exposed places. But this will still probably function as a proof of concept.

[ebastler]

But this will still probably function as a proof of concept.

Hmm, I'm not convinced, even if you manage to bend this arrangement without breaking it. How do you plan to attach this inside the enclosure? What stops the side-panel PCB from sticking out a strange angle or flapping in the breeze?

And what's wrong with a little stack of two boards, held together by one or two pin headers? (Assuming you can't fit the whole thing onto the top and bottom side of a single PCB, which would seem much preferable in many respects?)

[X10an]

Hmm, I'm not convinced, even if you manage to bend this arrangement without breaking it. How do you plan to attach this inside the enclosure? What stops the side-panel PCB from sticking out a strange angle or flapping in the breeze?

And what's wrong with a little stack of two boards, held together by one or two pin headers?

I absolutely love feedback and insight from other folks, this is exactly why I discuss this in an open forum. So thank you for replying.

This is an experiment which, if it succeeds, will pave the way for another project of mine which involves a stack of four boards and close to 50 contactpoints between each board and possibly in larger quantity. Contactpoints equals potential failurepoints and extra expences. The posibility to set up and manufacture only one board will also keep down price for both PCB and PCBA. All things regarding to development/repairs/calibration and revisions will also be much simpler without having to remove soldered headers for dismantling. Maby not a valid point in a finished product with non servicable parts, but worth mentioning anyhow.

20mm long and 6mm wide for a twist of 90 degrees, as the OP mentions, is probably not at all feasible for a comercial product. But with multiple elements (and lots of U-turns) this could be done if the space permits.

I have already made some promising experimenting and testing on 1.6mm FR4, and my conclution is that a length/width-ratio of minimum 40:1 (200mm long and 5mm wide) will twist to 180 degrees without problems, and of course 20:1 for a 90 deg twist (no signs of cracking after "endless" manual twisting). To accomplish this ratio the solution can be the use of multiple "elements" In my attached PCB I have a lenght of each of the four elements of 48mm and a width of 3.6mm (total L/W-ratio>53:1 for a 180 deg twist). This will hold +/- 10 traces on each side (apart from components placed directly on the elements). I would recommend an extra 50% substrate in the turns  (only 36% on this PCB) to resist bending and emphasize twist. Cracking and damage to the traces is definitively most at risk in the "inner turns", so it is wise to route the traces as far out as possible. I have the exact same lengths on all the elements som my theory is that the "sidepanel" will be "centered by design" and the opposite forces cancels eachother out. I am actually very confident this will be the case, that the middle board vill be exactly centered in all directions with twisting of the top and bottom board. It is not much space for this to move either way regardless of this. The top and bottom will be held together with two soldered "legs" between them which I have "forgot" to provide holes for on this PCB. Both the enclosure and the PCB got som screw mounting holdes. I really need the space for batteries and a chargingdevice, so I will avoid using the void in between for standoffs, but the enclosure in itself will probably also hold things tight together with the constant "pressure" from the twisting elements. Although probably not much pressure left from 3.6mm wide, 0.6mm FR4.

On a sidenote I will also mention that the LED`s mounted on top of the twisted element will be manually soldered on at correct viewing-angles _after_ the PCB is twisted and fixed into its final shape (because solder does not twist well). But this is solely related to my specific design, and not a matter of this discussion.

[X10an]

Today I received my test-PCB. The result is all positive. This seems to be an exellent solution to this challenge. The center PCB centers exactly like it is supposed to and I can twist the PCB back and forth until my arms fall off without the substrate taking any obvious damage. I will now continue to route my PCB with this twist, with only minor modifications.

This particular PCB is for a remote controlled towlight, where the LED`s are supposed to shine through the lid, the microcontroller is mounted on the middle PCB and the bottom PCB is for mounting of radioreceiver, wireless charger and some other stuff. Batteries will fill the void in between.

[3roomlab]

fascinating, PCB ORIGAMI !

I wonder what widths should the twist bridge be if the PCB is 1.6mm thick. I am thinking it will crack for thicker PCB?

[asmi]

Everything I know about material stress is screaming to me that this is a disaster waiting to happen

[Mattylad]

I have seen this done on a very large board, a similar arrangement where ISTR cameras are attached to little sub sections on the end of the bent bit, when bent up they are then held in place in a framework, as the board never moves so as to not flex it again it seems to work well.

[X10an]

fascinating, PCB ORIGAMI !

I wonder what widths should the twist bridge be if the PCB is 1.6mm thick. I am thinking it will crack for thicker PCB?

I really appreciate your feedback :-)

I actually have good reason to believe this would work well also with 1.6mm PCB. But at this stage I really need the extra mm`s of space-gain in the box, so I will not order an 1.6mm board for test only at the moment. I will report back if/when I do.

[X10an]

Everything I know about material stress is screaming to me that this is a disaster waiting to happen

This testboard is actually "reasonally" happy with a twist of 180 degrees from the mainboard to the middle-board only (i.e. twice the twist in the pictures). I will twist it 90 degrees only in that space. Another important factor is of course that this will be a "one-time-and-stay-firm" bend only.
This testboard is really getting a fair amount of , but no signs of weakness so far...

You may have to consume a really big pile of popcorn before your expected disaster strikes :-)

[X10an]

Here I have some pics of the board, first twisted 180 degrees in one of the segments, and second with full 180 degrees twist in only one of the lengths... The board is a little stressed, but still no obvious signs of damage.

[asmi]

I guess I wasn't clear enough. It's not the board I'm worried about - it's traces. The ones on the outside of the bend will have to stretch quite considerably, simple back-of-the-envelope guesstimate (assuming it bends around center, if it's around corner than it will be even bigger) is 2 * pi * (h/2) / 4 ~ 1.26 mm, and we're talking about 0.035 mm thick traces (this is for 1 Oz copper). I know that this is a very crude simplification, but it demonstrates the point I'm trying to make.

[X10an]

asmi: You have of course good and valid points in your argumentation. Copper can become very brittle after two to three (larger) bending-operations and will benefit from a heatingcycle before continued bending. Because this is not possible to do in this application the traces WILL break after continued bending given enogh bending cycles at large angles. This is of course the no.1 reason it is so important to twist and not to bend. I really have no impression that the copper actually will stretch much through the "straits". The stretching is limited to the 180 deg. turns. And all the traces will twist to the exact same degree regardless of the placement on the "straights".

To me, it is obvious (by feeling and experience only) the single most exposed stresspoint is the trace in the "inner" 180 deg. turn, not the outer turn as I think you imply. To remedy this I will drag the traces as far to the "outside" of the 180 deg. turns as possible (leaving the extreme inside traceless), in addition to that, this is the exact reason I have made the fr4-width in the turns about 50% wider than the rest of the straights to stabilise this exposed part of the tracks. I can also make the turns even wider if experimentation should find it needed.

When I do receive the finished board with traces, I will extreme-bend the board with the tracks serial connected to my scope to observe what kind of abuse this will actually tolerate before first sign of glitch. If it holds up 100 times (normal twist) it will be fine for my finished product (most of them I will only twist once). But I will continue the testing at extreme angles until failure. Remember this is ment to be a test to see if this is possible and acceptable, the first two to three runs will not be mounted in human pacemakers...

[asmi]

I don't quite understand your desire to play games with these weird design techniques.
I'd design this in 3 boards connected by RA headers. It would be simple and reliable. No need to try your luck in finding out if board/traces will break or not, and if that result will be consistent over several batches of PCBs.

[Ian.M]

FR4 is notorious for its tendency to creep under high mechanical stress.  I would be concerned that any small defect at the edge of the torsion strip could trigger gradual progressive delamination with the stress concentrating in the as yet unaffected portion of the strip width, until it has insufficient strength remaining, leading to sudden deformation with a high probability of track damage.

There is no good way to do accelerated ageing tests for this so PCBs under extreme torsion should be avoided in all except short-life disposable products where you can actually test their behaviour over at least twice the required product life time.

[AndyC_772]

I'd bet £1 that the SMD components that are actually mounted on the twisty bit will suffer cracks before the PCB itself fails. Ceramic parts like these have no ability to accommodate bending; they rigidly maintain their shape up to a point, then fracture. Same goes for the solder joints attaching them to the board.

If you really must bend and twist the PCB like this, then at least don't multiply the risk of failure by putting components in the stressed areas.

[X10an]

I'd bet £1 that the SMD components that are actually mounted on the twisty bit will suffer cracks before the PCB itself fails.

On a sidenote I will also mention that the LED`s mounted on top of the twisted element will be manually soldered on at correct viewing-angles _after_ the PCB is twisted and fixed into its final shape (because solder does not twist well). But this is solely related to my specific design, and not a matter of this discussion.

-------------------------------------------------------------------------------------------------------

FR4 is notorious for its tendency to creep under high mechanical stress

This is exactly the reason I go a long way to make the mechanical stress as low as possible under this conditions. My point is that with these long twisted sections the mechanical stress is actually relatively low in this application.

Angeled headers will of course give much of the same results apart for the space requirements and the issue of dismanteling in case of servicing. Then I will have to use connectors with extra component cost, solderpoints and failurepoints. The cost for larger series is also a fair bit lower with only one side of the PCB to be populated under PCBA, with no manual soldering.

[donotdespisethesnake]

I understand the desire to eliminate inter-board connectors, they are a pain aren't they? However, I predict that with this particular solution, thermal cycles, vibration, etc will cause a lot of field failures. Quite open to failures caused by assembly as well. Perhaps the worst case is you waste a lot of time and have to redesign using conventional techniques. At least I would have a "safe" design prototyped, and ready to go...

There are some manufacturers who make "bendable" FR4 (semi-flex), might be worth reviewing their design guide.

[babysitter]

If you reduce the bended part of the PCB to single layer, consider milling the boards: mill a "trench". where it is more easy to bend due to the thinner substrate. However, do your homework.