Right angle PCIe-4x connectors?
Bill Henning
Posts: 6,445
Anyone find any yet at reasonable prices? (qty.100+)
After scribbling on paper about the modules, it seems that right angle connectors would be better for many uses.
After scribbling on paper about the modules, it seems that right angle connectors would be better for many uses.
Comments
There may be benefit in us coming up with a small right angle adapter board, that also had a prop plug connector on there, and possibly some other compact features (eg test points or user leds or uSD socket, as well as breaking out to a 0.1" DIL pitch for breadboarding or IDC ribbon
Also I raised with Chip a little while back the effective prop plug routing at one end of the connector. It's "almost" easy - but perhaps we would be better rotating the pins in that block of 4 by one. The ground pin can still stay on the same side, but moved across by one. That way it would line up, and you could always just solder a 4 pin header against the PCIE pins to connect a prop plug, with no link wire required.
Sorry Jazzed not sure I follow. What bit was it you objected to previously (just getting back up to speed with prop threads now)?
Value is a complicated thing. Cost doesn't necessarily have to be reasonable (!)
That's great, and a 2 slot backplane can make a sandwich possible, but a single right-angle connector on a board would make a sandwich design almost half as tall as a 2 slot backplane. Maybe a dual-row right-angle header could be put on a right-angle breakout.
The other best option barring the availability of a cheap right-angle solution is to put a dual-row header on Chip's board. Chip had no comment on that.
That's not the whole truth really. There were two strong reasons: 1) dirt cheap, and 2) expected use.
Combined with #1, the #2 item was the final decision point causing us to go with the straight socket because the intent was to provide a Parallax-made module to be used for prototyping work and for easy/low-cost "replaceability" (which goes hand-in-hand with prototyping). It would be easier and cheaper to replace a blown Propeller 2 module on an expensive development board than it would be to replace the entire development board that had a Propeller 2 built-in.
Along with this, our intent was to make it as small as we could practically make it so that it wouldn't stick up unreasonably high, and with the small connector and upright orientation, it gives the development board designer/manufacturer more board real-estate to pack in their desired features. If it were a right-angle header, as much real-estate as the module's footprint would have to be left blank on the development board (plus a little more to allow for insertion/removal of the module from the socket). This seemed to be an impractical demand to place on every development board (plus the extra cost of the right-angle connector and the 12 fewer available connections) and as such, using the Propeller 2 Module for such boards didn't seem that it would provide any realistic advantage over building in the parts directly into the development board.
Also, it doesn't seem reasonable that a non-prototype product would want to use the Propeller 2 Module; but for cost (and other) reasons, a product meant for production use, taking extra board real-estate, will usually opt for the components themselves built right into the board.
The fact that a high-density, super low-cost, prevalent, and unlikely-to-be-discontinued-soon connector seems to not be available in both a vertical and right-angle form factor, with the same pin-count and pitch, made us opt for just one - and the one was the vertical style for the reasons above. Our desires were initially to find this in both form factors, but the only options that allowed for that seem to be 3x or higher in cost.
We've even exercised our Chinese contacts in search for this Holy Grail of a connector. The result was more of the same with a strong indication of limited and unsecured availability and no definite cost unless we talk many thousands in quantity.
That was one of the first options we looked into. So far we haven't found any that comes close to matching or beating the cost of the solution overall.
And the cost of this solution ultimately will have to be paid by the users and developers.
I haven't spoken with Chip recently about this (I'm hoping he responds to this thread with more recent information), but last I knew, we were adding castellations (sp?) around the other edges of the Propeller 2 Module board to allow a development board manufacturer to solder it right onto the board. This at least allows the module to be parallel with the board for those that want it that way, with the advantage of providing access to all the connections that the vertical connector affords, plus it saves a little development board real-estate if this 4-layer module is being soldered to a 2-layer development board.
If someone has researched and found a solid solution that solves all these issues (cost, form factor, availability, product life, etc.), please let us know. The current choice (64-pin vertical connector) delivers a complete connection solution for just $0.27 to $0.54 per module. We've spent countless hours considering, discussing, searching, losing hair, and toiling over this... but we will gladly put all that behind us if a better solution is found that pleases all.
Food for thought:
2x32 2mm deep 1.27mm x 1.27 mm SMT connectors are about $0.58-$0.65 qty.1000
They are also available as through hole.
If the module could support both 2x32 0.05"x0.05" and PCIe 4x, then the PCIe could be used for vertical orientation, and the 2x32 headers for horizontal. This would also result in shorter traces than would be needed for the cassellations on the edge of the module.
Bill
Thanks Bill.
I don't know where the source of these is (I searched a bit for them before writing this), so I'll have to assume that they are a male 2x32 SMT connector for $0.58 qty 1,000, and we'll also need a female connector of the same form to mount onto the Propeller 2 module, with the assumption that it's also $0.58 (or more because of the extra structure involved). So that's a $1.16 solution qty 1,000 versus a $0.27 solution qty 1,000.
The cheapest I could find back in Feb was around $2.90 for the pair of connectors.
If the cost is really that low, and the cost difference is not a big issue, we can certainly go that route if all/most development board producers like yourself are keen on a lay-flat module.
Any links to source(s) you could give us?
Has anyone else found other options to evaluate?
I bet they would give a much better price in 10k quantities.
Cluso99, Steve, tubular, other volume board makers... I did not post the URL (you probably already know it) so that end users reading the thread don't swamp register at the supplier and whine about not being able to get connectors in less than 1k unit lots (ie qty.10-100) at such prices. PM me if you need it.
I'd recommend female on the module (~$0.58@1k, probably a lot less at 10k) so your P2 module BOM cost would increase by $0.31 per module
BUT
board maker BOM costs would decrease by about $1.60/unit (compared to right angle connector)
Regards,
Bill
Looking at PCIe-4x connectors 64pin, they seem to be female straight @ 1.0mm pitch. They are made for a pcb with gold fingers edge connector?
So, does this mean the P2 PCB has the gold fingers as an edge connector? This would make the most sense since there would be no connector cost for the P2 module.
A tiny pcb with gold fingers and a straight PCIe-4x connector on it would be a quick and cheap solution for a right angle transition to anther pcb below.
The straight connectors are 31c-48c in qties of 2,100 or 3,000. I have not found cheaper ones but presume they are available.
This one is .32c in qty 3,000 #16137
I've been watching this connector space for years, and I think PCIE-4x is a great way to go. You have to keep the cost as low as possible on the P2 PCB to afford maximum flexibility in a market that is tightening up. There is nothing as low cost as gold fingers (and perhaps 20 degree chamfer) on the PCB, other than pads/holes/castellations like you seem to be considering anyway. Standard pitch castellations would be a bonus.
A few years back a manufacturer I know was having problems soldering those 0.05" pitch through hole connectors. They ultimately had to move away from those connectors due to yield issues. Whether the parallax selective solder equipment does that or not, you'd have to run lots of trials to find out. One benefit of 0.05" is you can connect it to high density idc ribbon (eg FCI Minitek 127 range), and the connectors in that range are really compact.
Regarding choice of connector, when you have that many poles you really need to align with something already in common computer use. So you have things like PCMCIA or CF connectors,
One other option I can think of is the 2x32 0.1" connector used in PC104 bus systems. It is available in self stacking or not, and is a "one part" solution like edge connectors are. One advantage if you went PC104 and followed the bus pinout, P2 has enough pins and speed to be able to talk to all those existing PC104 peripheral cards, including network adapters, communications cards, data acquisition etc. But the module would be double the length of the PCIE version being proposed.
PCIE-4x is the best option so far...
regards
Lachlan
It is possible (just) to alternate holes (castellations) on a 0.05" grid such that you have a 0.9mm/35 mil holes alternating with 0.7mm/28 mil holes. That allows either 0.1" or 0.05" (ie standard IDC ribbon pitch) to be used. You'd get around the right number of pins broken out, then.
If you have a smorgasboard handy there is a small castellated prop1 board on 0.05" pitch
We could do a small pcb with gold fingers and a straight connector (to be used as a RA transition connector), plus the straight connector for the baseboard way cheaper then $4 if done in volume. But for someone really requiring the RA connector it is available.
Therefore, the basic requirement of the cheapest P2 module has been met (no connector, just gold fingers) and the straight connector is also extremely cheap. IMHO this is the main requirement. And we have also solved the RA solution - may not be the cheapest, but it can be done easily.
As for castellations, I would suggest 1.0mm pitch. 1.0mm pitch male headers are available at almost reasonable pricing (but not female!). Anything larger would not bring out enough pins. Alternatively ditch the castellations altogether.
It would make more sense to me to have another P2 module without the SDRAM and just have castellations around the edge - ie no gold fingers. Alternately, have dual rows of 0.1" around the pcb.