Manufacturing a thin but long multi tiered 50-pin cartridge/cable connector

[cheater]

Hi all,
I'm looking to build (for a personal project) a system that requires a sort of "cartridge connector" or rather a gumstick sized, quickly patchable, hot plug, resilient up to 100k mating cycles, 50-pin, pcb-carrier, low frequency (DC..200 kHz), normalizable connector with a sturdy shell that will carry the cable. The idea is to have a relatively thick pcb the size of a stick of chewing gum, and have it mate with a port on the front panel of a piece of equipment. I might also want to add a run of fiber optic cable with associated mating hardware on the connector.

Let me go through the requirements a little first:

1. gumstick sized - this is going to go on a control panel where the user will be patching and unplugging a lot of connections. The control panel is very busy and front panel real estate is at a premium. The size shouldn't be much larger than a quarter inch jack, so something like a stick of chewing gum (see attached image) would be perfectly fine. I wouldn't like something huge like a D or Centronics-style or automotive multipin or military circular multipin Lemo connector. Those are just huge because they arrange the pins next to each other in the surface of the front panel they attach to. So the logical conclusion is to instead arrange the contacts depth-wise.
2. quickly patchable. one issue with all existing high pin count connectors is that they are very slow to patch. eg with milspec and lemos you have to aim the connector (it's not self-aligning) and then pray that you put it on at the right angle, and then screw it down. That's waaaay too much work for this use case which requires being able to add a patch within seconds.
3. hot plug. this will carry low analog voltages and should be hot pluggable. I don't see any issues.
4. resilient up to 100k mating cycles - on some ports, this will get plugged in and out dozens of times a day. I want it to last 20 years.
5. 50-pin. The pins will carry DC..200kHz signals at up to 50Vpp, but I could settle for DC...50kHz at a much lower voltage if there are unsurmountable issues, but I wouldn't necessarily really want to.
6. pcb carrier. It seems like the only way to really do this is to have a pcb mate another pcb using spring contacts, but I'm open to other suggestions.
7. low frequency - this will only carry audio
8. normalizable - the audio connections should by default be routed to another place rather out of (or in from) the connector. Just like a normalized audio patch bay. This can be done by having a mechanical feeler and switching using transistors, but it would be nice if the connector did it on its own in some way.
9. sturdy shell that will carry a cable - a 50 pin cable is pretty heavy and a bit of a boa constrictor, but if you have a heavy connector it should be able to carry the weight of up to 2-3 meters of such a cable.
10. fiber optic connection - I'd like this to ideally also patch a slow digital signal (~1-10 MB/s). As the signal runs together with sensitive audio, I can't have it run in electrical form, which would create very unwanted EMI in the analog wiring. It would mostly be used for carrying stuff like DC signal values to be put out via DAC, simple control values that don't change a lot, or digital configuration.

My main issue is that I don't know of any connectors that would fulfill this combination of needs. I've been looking for a bunch of years on and off and always come up empty handed. Maybe someone has a good idea of a connector that would work like this? But otherwise, I am looking at manufacturing this myself. I have a low initial need - in the tens to hundreds for the first set, later maybe single thousands over the next ten years.

I don't really know how to do a few things:

A. how to create the mating interface? If I have two PCBs above each other, I want them to connect using an elastic pin of some sort. Pogo pins feel like they'll have a failure rate that's too high. In addition, the two PCBs will most likely be sliding past each other. So I'm probably looking at something like leaf springs, like eg mobile phone battery connectors (see attached image). Is that a good idea? What are some better ideas to make this? Should I try to get the spring contacts themselves, without the plastic carrier package? Where would I get something like this? I've only seen them in the SMD package. Do they come in through-hole? that would likely make them more sturdy.

B. The connector shells for both the cable-bound plug and the front-panel socket will have to be sturdy. I made a simple design that includes locator pins and makes the connector slide in most of the way, and then in the final bit of travel it creates force to mate the two electrical contact surfaces. I'm not sure whether the rails should perhaps be on the front panel connector, since it requires thicker metal. I think mechanical retention of the connector could be accomplished by a spring loaded bearing ball in the underside of the connector. I guess either way the design means they'll have to be CNC milled, unless there's a better way of making them. I've included a crude mechanical drawing in the attachments - what would the best way be to create something like this? Is there a better design?

C. How to measure the reliability of the connectors? Or of the cable itself? I would like them to meet a certain amount of cycles. Obviously there's going to have to be a lot of testing, but other than that, how can I make an initial set of calculations to make sure I'm not doing something that'll obviously fail after ten uses?

D. What does everyone think of the design? It's pretty ambitious but I hope that with enough time and work I can figure out how to make something like this reliably and to good quality. Are there other designs you would suggest?

E. What sort of costs could I be looking at?

F. What is the best way to strip and solder 50 wires to a pcb, if you have to do this a dozen times? A thousand times? Ten thousand times?

G. Are there other places I should ask about this, or is this the correct place to go?

Thanks for reading and looking forward to the replies.

[SMdude]

I don't know that you'll find what you want in the bubblegum packet size, but interesting idea.
I had a similar requirement(minus the mating cycles and optic fibre) and went for pciE 64pin connector. You'd have to design a robust cartridge port and cable holder to go with it. Heck, if you designed the housing you could also integrate an optic fiber connector.

As for how to strip and solder 50 wires to a pcb, either one at a time or get someone else to do it!! 

[Algoma]

Do the signals need to be realtime / low latency?  Perhaps there may be some opportunity to Mux the signal into a smaller groups of channels / or convert some of it to high speed ethernet protocol on a software side, and go with more standard cabling types.

[Algoma]

Highest pin-density in an existing small connector cables with off the shelf parts, and known reliability would be on mini-display port cables, at 19 pins each for 15meter cables connections.. Not sure if there is active chips in those cables.

https://glenwing.github.io/docs/mDP-1.0.pdf  - Specified Cable is listed as Bulk Cable at 30V rating.. looks pretty passive, and usable for Audio applications. See page 18 for pinouts of the cable.

[cheater]

Thanks. The signals need to be analog, at DC-200 kHz. No digitization is possible due to how the signals are being used.

[mikeselectricstuff]

F. What is the best way to strip and solder 50 wires to a pcb, if you have to do this a dozen times? A thousand times? Ten thousand times?

IDC connector

A 50 pin D connector or amphenol plug (as used for SCSI ) probably comes pretty close, and will certainly be way cheaper than anyting custom.
Trade off cost vs. mating cycles and replace it as needed.

[mikeselectricstuff]

For a spring-loaded solution, something like these
https://www.mill-max.com/products/new/high-current-small-scale-spring-loaded-pins
Rated at 1M cycles

[fcb]

Not sure I've come across a D connector with more than 10K cycles.  As Mike suggests, you could swap these out.

Mill-max and CODA Systems (UK) both make suitable ganged parts rated to at least 1,000,000 operations, you'd need a bunch of these, but job-done (6way seems to be highest pin count).

http://www.coda-systems.co.uk/catalog/Swift_Dock_Mini.html
https://www.mill-max.com/search/keywords/858

I've used the Swift Docks before, so no doubt they'd do what you want.

[mikeselectricstuff]

Will people really be mating these multiple times a day, 365 days a year?

[phil from seattle]

Will prople really be mating these multiple times a day, 365 days a year?

Actually, 13 times a day, every day for 20 years is still less than 100K cycles. Perhaps a tad over spec'd?

[mikeselectricstuff]

Will prople really be mating these multiple times a day, 365 days a year?

Actually, 13 times a day, every day for 20 years is still less than 100K cycles. Perhaps a tad over spec'd?

..not to mention will the product still be in use and not have been long superceded in 20 years?

[Cerebus]

F. What is the best way to strip and solder 50 wires to a pcb, if you have to do this a dozen times? A thousand times? Ten thousand times?

If you can use ribbon cables either bare cable soldered with a hot bar (there is tooling that can strip the entire width of a ribbon cable in one go), or via an IDC transition connector vis:

[cheater]

Has someone been deleting my posts? I replied to this thread and now my replies are gone...

Edit: Ah, yes, something about a backup restore in the news bar up top.

[cheater]

just an idea. the latch @ 3 could be a thumb screw or somekind of thumb size quick pull tab lock latchy thingy (or magnetic?). i think the latch thing needs to have a size to last 100k uses. edit : looking at latch PB switches, the contacts wear out and the metal residue smears across the contact points undoing the "open circuits" in the switch, i think sliding is not good for long term reliability.

Great idea, but maybe making the thing pivot towards the bottom (basically, turning it upside down) would be a better idea. This way, it would not only be held by friction fit, but gravity + the cable's weight would help as well.

[cheater]

F. What is the best way to strip and solder 50 wires to a pcb, if you have to do this a dozen times? A thousand times? Ten thousand times?

If you can use ribbon cables either bare cable soldered with a hot bar (there is tooling that can strip the entire width of a ribbon cable in one go), or via an IDC transition connector vis:

Can't use ribbon cables - the cables connect front panel connectors that are next to controls and displays that need to be accessible. There will be many cables and using ribbons would completely cover up the control surface.

[cheater]

Will prople really be mating these multiple times a day, 365 days a year?

Actually, 13 times a day, every day for 20 years is still less than 100K cycles. Perhaps a tad over spec'd?

..not to mention will the product still be in use and not have been long superceded in 20 years?

There are devices in this category being used productively since 60 years, and they constantly get restored to as-new condition because they are irreplaceable. They will probably continue to be used for another 100 years if not indefinitely.

[cheater]

Will prople really be mating these multiple times a day, 365 days a year?

Actually, 13 times a day, every day for 20 years is still less than 100K cycles. Perhaps a tad over spec'd?

Depending on the day it could be zero cycles or much more than 20 cycles. I don't know the distribution well enough to tell. However, I think 100k cycles is probably not hard to achieve given the right kind of mechanical design. I am more worried about the MTBF which will be independent of mating cycle count. Dozens of contacts per connector easily takes its toll due to basic probability theory. Which is also why you want to spec for high mating cycles.

[cheater]

For a spring-loaded solution, something like these
https://www.mill-max.com/products/new/high-current-small-scale-spring-loaded-pins
Rated at 1M cycles

It feels to me like a pogo pin is going to be much more failure prone than a leaf spring. And much more expensive: a pogo pin is at least three parts (two high precision turned parts plus a spring); a leaf spring is one low-precision piece (a piece of laser or jet cut foil, bent to shape). Would you agree that it's better to look for leaf springs?

My other problem is that these pogo pins are all through hole - I cringe at having to place tens of thousands pins before soldering; I imagine leaf springs should be more easily available in SMT, but let me know what you think.

Also, won't the coiled spring add weird inductance? We're only talking about 200 kHz max, so I don't know if that would even be relevant. However, those pins will be part of an analog feedback circuit, at least at times.

[mikeselectricstuff]

For a spring-loaded solution, something like these
https://www.mill-max.com/products/new/high-current-small-scale-spring-loaded-pins
Rated at 1M cycles

My other problem is that these pogo pins are all through hole - I cringe at having to place tens of thousands pins before soldering; I imagine leaf springs should be more easily available in SMT, but let me know what you think.

There are SMD versions of these, e.g. https://www.digikey.co.uk/product-detail/en/mill-max-manufacturing-corp/806-22-001-30-003191/ED10982DKR-ND/6645200

Also plenty of SMD leaf spring contacts https://www.digikey.co.uk/products/en/connectors-interconnects/rectangular-connectors-spring-loaded/408?k=spring%20connector


 

[fcb]

If you use leaf-springs you'll need to very good plating and minimize sliding if you are using a PCB as one contact).

50 contacts, 100,000 cycles - the likelyhood of a solder joint cracking is pretty high. I would certainly avoid using a PCB in your connector.

[cheater]

If you use leaf-springs you'll need to very good plating and minimize sliding if you are using a PCB as one contact).

50 contacts, 100,000 cycles - the likelyhood of a solder joint cracking is pretty high. I would certainly avoid using a PCB in your connector.

I'm not sure how you'd else build a connector like this. Bear in mind also every usb connector is a PCB.

[cheater]

For a spring-loaded solution, something like these
https://www.mill-max.com/products/new/high-current-small-scale-spring-loaded-pins
Rated at 1M cycles

My other problem is that these pogo pins are all through hole - I cringe at having to place tens of thousands pins before soldering; I imagine leaf springs should be more easily available in SMT, but let me know what you think.

There are SMD versions of these, e.g. https://www.digikey.co.uk/product-detail/en/mill-max-manufacturing-corp/806-22-001-30-003191/ED10982DKR-ND/6645200

Also plenty of SMD leaf spring contacts https://www.digikey.co.uk/products/en/connectors-interconnects/rectangular-connectors-spring-loaded/408?k=spring%20connector

Thanks a lot. Great links. Any reason I wouldn't go with something like this? https://www.digikey.co.uk/product-detail/en/avx-corporation/709155001610004/478-8741-2-ND/4490441

Any reason I should use the plastic-cased versions instead? Like EG this: https://www.digikey.co.uk/product-detail/en/molex/0788641001/WM11204TR-ND/5056769

That's a 40% increase in price.

[fcb]

If you use leaf-springs you'll need to very good plating and minimize sliding if you are using a PCB as one contact).

50 contacts, 100,000 cycles - the likelyhood of a solder joint cracking is pretty high. I would certainly avoid using a PCB in your connector.

I'm not sure how you'd else build a connector like this. Bear in mind also every usb connector is a PCB.

USB connectors are not made from PCB's, or put it another way very few USB connectors are made from PCB* - they have plated metal contacts with one side a spring. They also aren't rated for 100K operations.
*I have seen a few products and homebrew that use a PCB to form a Type A connector.

[fcb]

For a spring-loaded solution, something like these
https://www.mill-max.com/products/new/high-current-small-scale-spring-loaded-pins
Rated at 1M cycles

My other problem is that these pogo pins are all through hole - I cringe at having to place tens of thousands pins before soldering; I imagine leaf springs should be more easily available in SMT, but let me know what you think.

There are SMD versions of these, e.g. https://www.digikey.co.uk/product-detail/en/mill-max-manufacturing-corp/806-22-001-30-003191/ED10982DKR-ND/6645200

Also plenty of SMD leaf spring contacts https://www.digikey.co.uk/products/en/connectors-interconnects/rectangular-connectors-spring-loaded/408?k=spring%20connector

Thanks a lot. Great links. Any reason I wouldn't go with something like this? https://www.digikey.co.uk/product-detail/en/avx-corporation/709155001610004/478-8741-2-ND/4490441

Any reason I should use the plastic-cased versions instead? Like EG this: https://www.digikey.co.uk/product-detail/en/molex/0788641001/WM11204TR-ND/5056769

That's a 40% increase in price.

The single part is rated for 1000 cycles.  The ganged part just 30 cycles.  Granted, these will be conservative ratings and often at load, but pretty unlikely to get to 100K!

Start with a connection system that is at least rated >1% target life-cycle.  Say you used USB C connectors, you can easily find them rated to 10K cycles - and that gives you 24 connections each.

[cheater]

I thought about USB-C but it isn't great. 10 pins too few (I need *at the very least* 34), not enough penetration depth to support a heavy cable, and existing cables are not straight-thru analog DC coupled connections - they're active and have reduced lead counts. Finding cables would be a nightmare for users. Much better off making a special purpose cable here...

[fcb]

I thought about USB-C but it isn't great. 10 pins too few (I need *at the very least* 34), not enough penetration depth to support a heavy cable, and existing cables are not straight-thru analog DC coupled connections - they're active and have reduced lead counts. Finding cables would be a nightmare for users. Much better off making a special purpose cable here...

Requires abit of work, but make a guiding mechanism (block of nylon, tufnol, whatever with stainless pins on one side, bused holes on the other) - this has a pair of connectors securely mounted in or on it.

[cheater]

I thought about USB-C but it isn't great. 10 pins too few (I need *at the very least* 34), not enough penetration depth to support a heavy cable, and existing cables are not straight-thru analog DC coupled connections - they're active and have reduced lead counts. Finding cables would be a nightmare for users. Much better off making a special purpose cable here...

Requires abit of work, but make a guiding mechanism (block of nylon, tufnol, whatever with stainless pins on one side, bused holes on the other) - this has a pair of connectors securely mounted in or on it.

Having a problem figuring out what you mean. What are bused holes? Are you thinking of something similar to a d-sub connector?

[fcb]

*bushed - sorry

If you make a surround with locating features (such as a couple of guide pins) that keeps a pair of USB-C connectors aligned as you mate to another another pair.  Wurth have a collection of USB-C plugs and sockets that might be suitable.

Another option might be Hirose 3800 series.  Rated at 20,000 mating cycles.

[cheater]

*bushed - sorry

If you make a surround with locating features (such as a couple of guide pins) that keeps a pair of USB-C connectors aligned as you mate to another another pair.  Wurth have a collection of USB-C plugs and sockets that might be suitable.

Another option might be Hirose 3800 series.  Rated at 20,000 mating cycles.

Oh, sorry, I should've noticed you meant bushed.

I like where you're going with the usb-c idea, but the issue remains that I would have to make the cables myself, at which point they're entirely proprietary anyways. Hirose 3800 are flimsy and pretty terrible and nothing I would call a good experience. And again hard to find cables for them.

The issue with all the connectors being mentioned here is that they're two-dimensional: all the pins are in the plane of the front panel, and they get squeezed together. Instead I'm thinking of something that works more like a card, like a SIM card or SmartMedia card for example, where pins are all laid out in the 3rd dimension (not in the plane of the front panel). Think of a front panel that accepts a SM card. the pins aren't laid out flat in the plane of the front panel - in addition to being laid out width-wise (which is a direction that's in the plane of the front panel) they're laid out depth-wise (which is a direction perpendicular to the front panel).

The Amphenol C702 10M008 325 4 is rated for 0.5M mating cycles. And it's built very flimsily, but I'll easily trust that it can hold that. Nothing's stopping anyone from producing something like this but more rugged and with more pins... maybe that's the kind of design I should be going for.

Something like those belly-shaped leaf pins at the port side, and a pcb at the plug side. The pcb is surrounded by a piece of C-folded steel shroud, shaped kind of like USB-B. Basically like it's U-shaped, but additionally folded around the pcb to hold it. And the steel shroud could have a conical nose in order to self-orient. The C-shape would prevent it from getting warped even when abused. Question remains how to hold the pcb in it without hurting the PCB. Some sort of plastic layer between the PCB and the steel maybe. Something very stable that won't become brittle even after 50 years would be useful. The simple design would make it possible to re-manufacture this by third parties. The shroud could also double as cable strain relief (thick cable after all).