Further possible delays to the Pi

So the first batch hasn't left the factory yet.

I guess I don't understand how we are at this point.  My understanding was that the manufacturer was done with the first 10k on 20 Feb.  At that point the team from the Foundation were on site to test the boards.  If you read the RPi blog link in the first post, the author said that without the right RJ-45 port, there is no ethernet.  What kind of test was done that didn't find that the ethernet was not working?  I've seen the Newark date on their website move to 27 July, likely due to the rework on the original batch and the delays sourcing proper parts and making the next batches.

I assume that at one point the team of the Foundation, RS and Farnell will determine how many they need to fill current demand (paid orders) and to stock shelves for future orders.  At that point they can make an order from China and get delivery dates.  I'm happy with the effort that all parties are putting in, but at the same time I am severely disappointed that I likely won't get the RPi I ordered until 6 months later.  I still say that the launch should have been delayed until all of the original 10k units were on the tested and on the shelves at RS and Farnell.  This situation doesn't bode well when no one is sure that parts exist to make more RPi's.

I believe it had left the factory, having been QA tested, been received at Raspberry Towers, and it was at final QA testing, whuich also included ethernet connectivity, by the Foundation themselves that the problem was identified, so the whole batch has been returned from whence it came...

A great shame, but better that than sending out Raspberry Pi units that don't work ...

I'm somewhat worried by the QA testing in factory..... that does not assure the quality of the product to a satisfactory level.... hummm

Tanks for the up date Steve , I'm only replying coz it's puts things in contexts . Patients ,don't want a half cooked pi . .

Actually the substituted jacks would work, though only reliably on short cable runs. More importantly they would be considered illegal in many jurisdictions as there would now be no electrical isolation between the Pi side and the LAN side as well as no impedance matching which is required for long cable runs for signal integrity.

Non magnetic (no transformers) jacks are fine to use, but then the equipment normally has the transformer as a separate component - look inside an old modem and you'll see the transformer. Because the Pi was designed to be small that requires the network jack to have the transformer built into the jack.

How they found out they haven't siad, but it could have been they tested  signal reflection on a long network cable or they were getting unreliable signal integrity or they simply measured the resistance between pin pairs and found it too low. Some nice X-ray pictures are on the R-pi site.

The pinout of an Ethernet jack with magnetics may be different from a non-magnetic jack.  If you look at the RasPi PC boards (there are links at http://elinux.org/Rpi_Hardware) it looks like the Enet differential pairs go to connector pins 1+3 and 4+6, which is one of the possible pinouts for the PCB side of a magnetic jack.  For a non-magnetic jack the pairs would be 1+2 and 3+6, the same as an Ethernet cable.  Given this, it would be obvious that Ethernet wasn't working.  The RasPi designers are lucky that they didn't use a magnetic jack with pairs on 1+2 and 3+6, as they may not have noticed that the magnetics were missing.

It is quite normal for the manufacturer to perform only an electrical test. It is the responsibility of the integrator (in this case the RPi Foundation) to perform the protocol tests. There is nothing unusual in the acceptance tests of the RPi.

The problem was compounded by the fact that the manufacturer seems to have used the correct jack for the trial production run, which did pass the protocol tests. The manufacturer then used the incorrect jack for the real production run. This is a rare event, but this is why there are so many tests to ensure that faulty devices do not shipped to customers.

In my experience it is usual for the manufacturer to perform complete end of line testing which should be designed to ensure 100% functionality of the part. Any reasonable test would show up the difference between a jack with magnetics and one without.

I doubt that the RPi Foundation going to 100% test the boards when they arrive in the UK so one is left wondering if they have any proper test plan at all.

Michael Kellett

Eben debunked the claims that 10000 boards ever left China in a recent interview: http://www.theinquirer.net/inquirer/news/2157021/raspberry-pi-pcs-served-april

Only a "very small number" of "sample devices" had left China, he said.  They were

shipped "to check build quality, components, and so on."

So no need to wonder how 10000 boards could get returned to China so quickly,

or why the factory might have built 10000 boards before the ethernet jack was tested

on any of them, or how long it would take someone to desolder 10000 through-hole

ethernet jacks.