Fine pitch, high speed connector alignment issues (ex. RPi CM4 Headers).....

[Smokey]

I'm looking at putting together a carrier board for a Raspberry Pi CM4 module.  (yes I know you can't actually source them anywhere right now)
https://datasheets.raspberrypi.com/cm4/cm4-datasheet.pdf

These things use 100 position 0.4mm pitch SMT headers, and the recommended mates have no alignment pins.
https://www.digikey.com/en/products/detail/hirose-electric-co-ltd/DF40HC-3-0-100DS-0-4V-58/4282912

I typically much prefer to have physical alignment pins on connectors like this, but I don't see any for this case that have them and are compatible.

Assuming I make the PCB footprint per the manufacturer specifications, and my assembly house use the right reflow profile, how well should I expect these types of fine pitch connectors to self align during reflow?  Anyone have any experience with the reject rate for something like this?

[eugene]

I've used similar connectors for board-to-board connection without issue, including multiple connectors on a single pair of boards.

Precise location on the PCB depends on careful footprint design. During reflow, when the solder paste is liquefied, surface tension locates the components on the footprint. If the land patterns are much larger than the component pins, then there will be range of possible final positions that you might end up with. On the other hand, if the PCB pads are only slight larger than component pins, then the component ends up positioned surprisingly well. The good news is that higher pin count actually works better.

As can be imagined, you can expect on the location tolerance along the length of the connector to be much less the pin pitch. These have 0.4mm pitch, so a connector that's shifted even as much as 0.1mm in that direction would surprise me. (I'm not an authority, this is just my experience.) The other direction, side to side, depends on the pad design too. You need a little extra pad past the end of the pin, but I would make sure that is doesn't extend more than necessary under the part. About 0.5x the width of the pin seems to work for me, but contact your board house with questions.

[Smokey]

Cool.  Thanks for the reply.
I have a few concerns here.  The first one is that I don't see an alignment tolerance spec for those connectors.  I would have expected to see that on the data sheet.  Like how far off it can be before the connector binds and won't mate correctly.  With that many pins (and correct footprint), I get that it probably actually aligns better in the XY sense.  But I'm also concerned about rotation (not sitting flat).  I could see that being an issue as well.

I know I'm being overly paranoid about this, and a legitimate company like Hirose wouldn't make a product that was so sensitive that small pad design errors would cause a huge rejection rate.  But I've usually dealt with this by just picking parts with alignment pins and not worried about it.

[sam512bb]

Cool.  Thanks for the reply.
I have a few concerns here.  The first one is that I don't see an alignment tolerance spec for those connectors.  I would have expected to see that on the data sheet.  Like how far off it can be before the connector binds and won't mate correctly.  With that many pins (and correct footprint), I get that it probably actually aligns better in the XY sense.  But I'm also concerned about rotation (not sitting flat).  I could see that being an issue as well.

I know I'm being overly paranoid about this, and a legitimate company like Hirose wouldn't make a product that was so sensitive that small pad design errors would cause a huge rejection rate.  But I've usually dealt with this by just picking parts with alignment pins and not worried about it.

Good day,

I ran into this same issue and I posted about this on the RPi discussion board.  What I got was "don't" worry about it... which is a total joke.  Firstly, the connectors have an overhang over the pins and so reworking is a lot of work and problematic even with the proper equipment and skill.  the RPi moderators played this down, but you can see for yourself.  Secondly, the connector manufacturer, Hirose, specially stated that using two or more connectors in the arrangement is not recommended and their own specs stated this quite clearly.  Below is taken from Hirose's DF40 series Guidline document - ETAD-H1015-00, Page 5:

1.1.4 When using two sets connectors
When using two sets of connectors on one board receptacle and plug, the following mounting
accuracy is required.

(A diagram is given where X +/- 0 and Y +/- 0.01mm)

As mentioned above, very tight tolerances are required.
It is impossible to satisfy this tolerance in consideration of the board and mounting accuracy.
Therefore, please avoid the use of more than one set of connector receptacle and plug on single
board. When using two sets or more of connectors, please use the divided board for the one side.

(another diagram is given where the PCB only has one mated connection and the second is for a FPC connection) 

You will note the "X" tolerance of +/- 0 and the "Y" of +/- 0.01mm... which I would say is very challenging to meet consistently.

As a quick test I had 10 test boards assembled using mega $ professional PnP machines and found that 8 out of 10 were placed ok... and 2 had shifted (no longer parallel) such that mating with the CM4 could not be done or was forced... and so long term contact reliability concerned me.  Although the CM4 has great specs and a great price, I am very concerned about the long term contact reliability, potential board yield issues that would be extremely difficult to rework, and connector and CM4 availability risk.

Cheers,

Sam

[langwadt]

maybe it is possible to make a "carrier" of some sort, say a PCB to they can be mounted accurately as one unit?

[sam512bb]

maybe it is possible to make a "carrier" of some sort, say a PCB to they can be mounted accurately as one unit?

Good day,

I tried this and this is where I noted that even a simple carrier board had connector to connector tolerance issues.  As for the footprints... I used the same ones used on the RPi base booard design.  Perhaps modifying and tweaking the footprints would help, but given that Hirose states that it is not recommended and that their tolerance specs are challenging, I think that one has to expect some level of placement issues. 

Having worked in the heavy Industrial industry a lot of attention is focused on the mechanical aspects of daughterboards.  A failed field unit can translate to a lot of $ and so these companies tend to be quite particular.  In fact their manufacturing process dictates that absolutely no assemblies can have any level of rework.  That said, perhaps I am also being particular as well, but to me it is very important to provide products that are highly reliable and so assembly issues or connector fatigue can really impact the final product, but also my reputation and potentially future work.

However, YMMV.

Cheers,

Sam

[Smokey]

Ha!  That is pretty much the stuff I was worried about.  Scary!

That's awesome Hirose recommends against using the connectors like this in pairs. 

I had concerns about the Raspberry Pi guys from the beginning when I saw they did the "official" CM4 carrier design in a version of KiCad that was at the time not stable.  It didn't even use any of the new experimental features.  That's not cool.

I had that connector mate "carrier board" idea, but that's not really viable for production.  You would probably bend the pins hand installing the mates on to the carrier and it would fit worse than just normal reflow offsets. 

[sam512bb]

Good day,

Indeed, the Hirose statement and x/y tolerance requirements for dual connector use is very important... well... at least to me.  I mentioned this to the RPi moderators and it was played down big time:

a. Lots of people are using these connectors with "no issues" (ahem) and so it is just me I guess...despite the warning provided by the connector manufacturer.

b. The Hirose statement, etc came out after they designed in and used the connectors and the arrangement

c. It is not a big deal to hand place this components... yeah right...  I have been soldering PCBs since I was eight and so almost 50 years and have some higher end gear and experience and it is not a trivial task.  Can it be done?  Sure, but if I am building a number of devices, spending a lot of time hand soldering ... or having poor yields on assembled boards is an issue for me.

Now, to be fair... Rpi never designed their products for use in industries outside of Educational use and so their internal experience in other markets is probably limited.  However, the CM4 and other modules are marketed towards applications and industries outside of Educational use and so one would think that simply talking to a few people in the industries intended may have been prudent?  Secondly, we all make mistakes and so why double down and defend a poor design decision and simply offer a newer rev that uses connectors that use an alignment pin?  If anything provide a heads up to those that the current design can have assembly issues and so plan accordingly.

Given the attitude and demeanor of the Rpi support moderators, I am wondering what other design issues potentially lurk.  Indeed, in general the Rpi's, have been quite reliable once you deal with the SDCard issue, etc  and so based upon this empirical data the Rpi stuff could be perfectly fine.  However, if the application is sensitive to failure/issues, then one really should test the crap out of the unit and/or look for another solution.  On m side I am going with the latter, as the risk of connector contact fatigue (do to a slight misalignment) and low assembly yields is too much and so I using and designing a different solution that is better suited to me needs.  My cost will be certainly more... but I am in full control of the design and have a single board that addresses all of my I/O needs, along with a form factor that is better suited towards my applications.

Cheers,

Sam

[langwadt]

maybe it is possible to make a "carrier" of some sort, say a PCB to they can be mounted accurately as one unit?

Good day,

I tried this and this is where I noted that even a simple carrier board had connector to connector tolerance issues.  As for the footprints... I used the same ones used on the RPi base booard design.  Perhaps modifying and tweaking the footprints would help, but given that Hirose states that it is not recommended and that their tolerance specs are challenging, I think that one has to expect some level of placement issues. 

yeh, thinking about it, no matter how accurate you mount the connectors on your board, there is no guarantee on how accurately the connectors on the CM4 is mounted