More on the beta boards

As promised, here’s some more info on the red wire visible in Dom’s video from last week.

When we first got the boards back from the factory, the core power supply for the SoC wasn’t working; this 1.2V supply drives all of the core logic in the device, and is generated directly from the 5V input via an integrated switch-mode power supply. A bit more investigation revealed that the 19.2MHz system clock wasn’t running, and after a bit of digging around we found that this was because its power supply balls weren’t connected to the system 1.8V supply. The red wire is there to bring in power to these balls.

The mistake turns out to have been in the original schematics for the board, rather than in the translation to the PCB. At several points in the schematics we have structures that look a bit like this:

Here, the SDC_VDD rail is generated by an on-chip LDO, decoupled by three external capacitors, and then pushed back into four balls. The offending bit of the schematic looks like this:

At first glance, this looks like the same structure, but in fact all eight pins are inputs, and the SDRAM_1V8 rail, which should be tied to the system 1.8V supply, ends up floating. Fortunately a small change to the PCB layout fixes this problem; the other beta boards have been fixed by carefully removing an area of solder mask and applying a blob of solder.

Thanks to Paul Grant at Broadcom for his help in tracking this one down.

119 comments

Avatar

Good spot, glad it’s just a minor hiccup!

Avatar

Is the issue relatively easy to fix for the Raspberry Pi team? Also, were there other flaws found in the PCB design or in the conversion to a physical board?

Myself, I’m still very amazed at the “5 layer” pcb design! Perhaps it’s a shame you didn’t get it down to 3, but it’s still good technology

Avatar

It’s so easy you wouldn’t believe it – all that we had to do was scratch a path of under a mm between two tracks on the top of the board and add a tiny dot of solder. The schematics have already been changed, so the problem won’t appear in any future PCBs, and we actually had pretty good fun doing the fix the other day! I’ve got some video of said fix in action which still needs editing (the family keeps shouting at me every time I open the computer, it being Christmas and all) which should go up soon.

This was, I’m very pleased to say, the only hardware bug on the board.

Avatar

Seems like the first beta-board’s value is increased already – by the manual work that needed to be done :)

Avatar

Glad to hear the good news, but I’m under the school of thought that “finding all the bugs” is always an unknown!

Avatar

D’oh!… Could this be a variation of the old Tech support question… “Have you checked if is it plugged in and powered?” lol!!

Glad it’s a simple fix!

Avatar

Actually, this looks like the “run the damn DRC check on the schematics before manufacturing” variation…

Avatar

Not really. If you screw up the schematics, then you build the wrong circuit. The thing will check perfectly (as it did); the mistake will only be obvious in physical testing (as it was).

Avatar

Interesting. It must mean that a bunch of “power input” type pins unconnected to anything else but each other is not set to fire at least a warning in your DRC matrix. Peculiar choice, but hey…

Avatar

Yes, Liz, some DRC’s are able to check for this kind of problems. Things like “two outputs tied together”, and “two inputs tied together”.

On the other hand…. it is all about “types”. So the power inputs can be marked as “power inputs” (duh!). But what do you mark a capacitor pin”?

Where does the power come from? On a LDO, you’ll find a “power output” pin…. But on a switching power supply, it ends up being the diode that “provides” the power…..

I’m starting to agree that it is difficult to have the DRC catch this…

Avatar

More power to your balls.

eben

:)

Avatar

Thanks for sharing the info Eben. The HW guys here that have been through things like this love to know the details! Isn’t it fun bringing up a new board like this! Very exciting. I assume all the board go through ATE or will for the 10,000 anyway. Do all the boards get system tested as well? At least for a few seconds? Just wondering. Best Regards BTW for a Happy New Year for you and Liz and your family.

eben

There’s not an enormous amount of scope for automated testing at the system level (most wires just go from an SoC to the edge of the board). We’ve budgeted 2-3 minutes of manual functional test for each board to check various interfaces are basically intact.

Avatar

“Here, the SDC_VDD rail is generated by an on-chip LDO, decoupled by three external capacitors, and then pushed back into four balls. The offending bit of the schematic looks like this:”

….

….

Yes

Avatar

[…] jQuery("#errors*").hide(); window.location= data.themeInternalUrl; } }); } http://www.raspberrypi.org (via @Raspberry_Pi) – Today, 2:08 […]

Avatar

You know the schematics are great and all… NOW GO AWAY YOU ARE ON VACATION, go out, catch some air, have fun :P

Avatar

What NO WAY!!! they can’t do that until I have one :P

Avatar

Hmmm I remember a layout guy forgetting to put a ground into the PCB, now THAT was a hairy board with lots of blue wires :)

Avatar

Wow, you were lucky to have been able to fix the problem without re-spin. It could easily have been under the SoC in which case you would have been SOL,

I remember the first run of a board with a BGA where we forgot to connect to a ball: it had to be pulled up instead of left floating. Fortunately the ball was on the edge of the BGA so it was possible to solder a red wire to it by pressing the wire against the ball and heating up the wire with a soldering iron. That was in the “good old days” of 50 mil ball pitch.

Hey, as long as you’re doing a PCB re-spin anyway, I’m going to repeat a minor request I made earlier. The traces connecting to power input S1 have nice vias going through the board. Those would be a great place to connect for an alternate 5V power input. Can you clear the solder mask from around those vias so the holes aren’t filled with solder mask? I’d also recommend moving the trace connected to S1 pin 1 away from S1 so there’s no chance of shorting to S1 pins 2-4.

Avatar

No. Sorry.

Avatar

> Wow, you were lucky to have been able to fix the problem without re-spin. It could easily have been under the SoC in which case you would have been SOL,

As I read it, the decoupling caps were present and connected to all the right SoC balls. So the worst case hot fix was a wire from the LDO to a cap.

The neat later solution appears to bridge the via at the south of C87 to a large adjacent copper area. That is a little fortuitous.

Avatar

I never trust schematics. When I was doing PC board design, I always produced a textual netlist with nets sorted by net name and pin within the net. Then I painstakingly checked the printed netlist against the schematics to make sure everything that looked connected on the schematic really was, especially making sure every power and ground was hooked up to the correct signal and not a pseudonym.

I usually found at least one problem that way. Plus, my sorting program resulted in a canonical netlist so you could tell the differences between board revisions using diff.

Avatar

Amazingly enough, we do that too! (As Eben notes, this slipped through because the structure looks very similar to another structure.) We’ve folders upon folders of printed materials which were checked by more the one person against the schematics; this is just part of the long design process. That thoroughness is probably why there was only one hardware bug in the whole of this very complicated board.

Avatar

You did that on purpose to aggrandise the revison 0 boards for auction. I am not fooled.

Avatar

Yup – if you look really carefully at the tiny dot of solder, you can see an microscopic, upward-pointing middle finger stamped into the centre of it. ;)

Avatar

Thats one cold solider joint.

Avatar

I would like to add my thanks to Paul Grant at Broadcom for his help in tracking this one down. Great Support at this time of year and a very good spot, the kind of error you can spend ages not seeing?

Avatar

Since you are going to have to redo the PCB. Any way to move the USB connector back flush with the edge of the PCB?

Avatar

That would require a much larger change to the board, and won’t be included.

Avatar

You should explain your article a bit more, several abbrevations are unknown to me and the conclusion(s) should be pointed out more precisely for the layman.

In German I would say: “I only understand railstation.”

Btw: What about CE conformity and other “cerificates” mandatory nowadays …

Avatar

Those certificates aren’t mandatory for dev boards, which is what this is. (Your Arduino doesn’t have those certificates either.) When we sell a cased version to schools, that will have to be certified.

On your other point, I think Eben is basically incapable of expressing things like this in layman’s terms; that’s what I’m here for. In short, the power didn’t work because two tracks in the PCB (printed circuit board) which were meant to be joined together were not joined together. Now they are, and everything works again.

Avatar

We know that dev boards will be available soon in January, but when can we expect cased versions? Month, 2 months, half year later ?

Avatar

No work at all has been done on casing (bare board more important), so some time yet.

Avatar

Thats Understandable :) maybe some contest on case design? I belive that in R-PI fans community are some talented 3-D designers which can create some nice case projects (notning fancy but maybe little bit more than standard cuboid with few I/O holes). Logo competition gives great results!

Avatar

http://arduino.cc/en/Main/ArduinoBoardUno Looks like a CE mark to me

Avatar

Yup – when you start to sell something as a reference board rather than a dev board, you have to register. Mea culpa: that Arduino has a CE mark.

Avatar

I like seeing updates like these……. really nice…. best wishes for the success of the RBPI in users hands.

Avatar

Be careful – Apple steal your ideas. :) >_<
Awesome.
– When can I buy it from your Shop?

Avatar

Should be Jan 2012

Avatar

About eben
general “troublemaker”? is this a exemple of tipical english humor?

Avatar

Wow… One error. On a PCB as complex as this. Not bad. ;-)

Also, I hope this little computer gets on the market Soon. I can’t Wait playing with it! Will the GertBoard be released at the same time?

Avatar

The Gertboard will follow a few weeks later. It was always one generation behind the Raspi boards. I also do not want to spend the money for a 5 day PCB turn-around time. (Don’t forget it is not paid for by the Raspberry-Pi foundation, it is all paid for privately)

Avatar

Are you sure you have that the right way around?
I would have thought that problems would arise if balls were connected to a psu, not the other way around……

Avatar

Other HW guys will know this is largely normal for hardware design… :-/

3 Spins at a schematic then layout is my average. But 5 spins has been my worst case thus far (damn Instrinsic Safety)….

Hopefully I’ll have a RP for my birthday (end of January) ;-)

Avatar

I’m looking forward to buying one of these little goodies..

Is the USB 2.0 or 3.0… I’m guessing 2.0?

All the best folks, Merry Christmas and a happy new year to you all..

Avatar

2.0

Avatar

Is there a distributor set up for the United States ?

Avatar

No disties anywhere yet. For the moment all will be sold from the website. As time goes, we may go for distributors, but it won’t be a good way to make money – you’ll just save on batch shipping.

Avatar

Hi, I’M from America Central (Guatemala, Costa Rica). will you sell here too?

Avatar

Worldwide.

Avatar

I hope you’re ready for massive numbers of chargebacks and fraudulent credit cards coming in from some countries then. Rather you than me, unfortunately. :-(

Avatar

I’m guessing the error wasn’t present on the Alpha boards because it is related to the power supply which was changed after Alpha? It probably also helped track down the bug because you knew it wasn’t present on the Alphas.

Did you experience any interesting bugs on the Alpha?

Avatar

Is it possible to buy one of these prototypes ? I am able to modify the hardware based on the problem description . Could these defective boards be made available to enthusiastic hackes with no warranty etc.
Thanks

Avatar

No, afraid not. You will be able to buy form the website early next year.

Avatar

early next year? no release date?
when I made my statement I was insulted by you guys – especially by liz.
I am still willing to help.
my suggestion: shop will openJuly 1, 2012
price will be 40 and 60 USD.

Avatar

Oh, shut it.

If you cant say anything nice, why are you even here?

Avatar

Well, I don’t like your suggestion. If things go well, we should be selling in January, at the original prices – $25 and $35.

Sorry if you felt insulted, but to be honest, we are pretty polite here, so if you were insulted, you must have started it!

Avatar

ASDA selling LED HD ready freeview TV’s for £50 ! Single HDMI input. Got mine already, can’t wait for the Pi to plug onto it !! Great news on the progress. It’s going to be a Happy New Year !

Avatar

[…] week, the test models had in fact encountered a problem. Specifically, the power supply balls weren’t connected correctly on the motherboard and the […]

Avatar

Is there any indication at all to release? January?

Avatar

From the outside, it looks like there are still a few issues left before they can get the factory churning out assembled boards. They have a strict no pre-orders policy, due to a rare condition called “common sense”.

They have fixed the board design, but I’m not sure they have tested it. Back before boards were designed on the computer, every change, no matter how minor, had to be built and tested to make sure that there were no stray cat hairs on the master image. In these modern times, every change, no matter how minor, has to be tested, to make sure a cat didn’t walk across the keyboard, and change an unrelated part of the design. These tests also detect a whole class of human errors.

There was also a problem with volume ordering of a suitable SD card socket.

Once they have the right parts in hand, and everything is retested to make sure it all really works together, they will have to program the pick and place robot to assemble the parts it can assemble. Then they need to train the humans to assemble the remaining parts. Then they need to train the humans to test the final assembly. (There will be much sadness, and delay, if testing detects any systematic problems.)

Then they have to negotiate a schedule with the factory to make the boards.

Then they will sell the first 10k boards. We will each attempt to buy a board. A few of us will succeed and report that it was a generally wonderful experience.

They will sell many more batches, and by February or March, supply will catch up with demand. Adventuresome teachers will be able to get enough boards together, so they can use them to teach classes.

Then they can tweak the board to put in mounting holes, additional GPIO, line up the connectors, and address any additional issues that have come to light.

Over the same period, the community will be spewing out software, tutorials, howtos, and other such support stuff, so the boards will be ready for a much larger crop of teachers to use to teach students.

That’s how I think it will probably go, but I am most certainly not a spokesbeing for the R-Pi foundation.

Avatar

Thanks :)

Avatar

Ohhhhhhhhh…. that’s what it is!

Avatar

Could you made USB SATA converter’s speed test for beta board?

Also it is interesting to see how it works with /swap.

Thank you.

Avatar

Can wait to see the final boards :) The POE option would be awesome :)

Avatar

There isn’t a POE option, unless you DIY.

Avatar

Are there any POE to Std ethernet + 5V modules?
At the chip level I looked at some Maxim, TI & Lineaar Technology IEEE 802.3af standard PD front end chips but they all seem to need lots of other external components and some other chips!
Am I looking at the right things?

Avatar

Google for ‘PoE module’; the first page gives plenty of sources. After 5 minutes of extensive literature review(!), PD solutions seem quite simple so long as you are not going to try to redesign the r-pi mainboard to keep it all on a single board, and you can cope with the potential thermal issues.

Avatar

Yes, there are.
The magic phrase is “PoE Splitter”.

5VDC versions are available, a webcam a colleague has uses one.
You’ll probably need to solder on the USB plug to power the Raspberry though, as barrel connectors look more common.
Might be a bit pricy though – the ones I found in a quick Google search were around $20 to $30.
Eg:
http://www.ipphone-warehouse.com/Cisco-Power-Over-Ethernet-Splitter-5V-POES5-p/poes5.htm

Avatar

PoE is power over ethernet

Avatar

[…] that problem was simply bound around a red wire and a discerning soldering job, and a group promises it won’t seem in any destiny units. Once […]

Avatar

[…] that problem was easily fixed via a red wire and a quick soldering job, and the team promises it won’t appear in any future units. Once […]

Avatar

[…] that problem was easily fixed via a red wire and a quick soldering job, and the team promises it won’t appear in any future units. Once […]

Avatar

[…] that problem was easily fixed via a red wire and a quick soldering job, and the team promises it won’t appear in any future units. Once […]

Avatar

[…] that problem was easily fixed via a red wire and a quick soldering job, and the team promises it won’t appear in any future units. Once […]

Avatar

[…] that problem was easily fixed via a red wire and a quick soldering job, and the team promises it won’t appear in any future units. Once […]

Avatar

[…] that problem was easily fixed via a red wire and a quick soldering job, and the team promises it won’t appear in any future units. Once fixed, […]

Avatar

[…] that problem was easily fixed via a red wire and a quick soldering job, and the team promises it won’t appear in any future units. Once fixed, […]

Avatar

When can I get my hands on one of these? I have a fairly tech illiterate roommate who loves Linux ( I know it’s bad) and I think this would be a perfect computer for her, since all she does is use Facebook and other things that are on the web.

Avatar

Should be ready to go Jan 2012.

Avatar

Just one error and it is easily fixed! And this on a 5-layer board! That is amazingly well done!

Great Job!

Avatar

[…] that problem was easily fixed via a red wire and a quick soldering job, and the team promises it won’t appear in any future units. Once fixed, […]

Avatar

[…] that problem was easily fixed via a red wire and a quick soldering job, and the team promises it won’t appear in any future units. Once fixed, […]

Avatar

[…] that problem was easily fixed via a red wire and a quick soldering job, and the team promises it won’t appear in any future units. Once fixed, […]

Avatar

You guys are super fine! Honesty is a rare trait these days. Having done some PCB design myself I know how it is. Thanks for the update.

Avatar

Hi Andy here jast to clarify wat is the situation in the cases? because i was going to use a dvd case style box to contain it as dust may make the ICB prone. so would like to know when or if a case is being released shortly after raspberry pi is released.
thank you in advance (even if the info is unhelpful) so anything on the matter is still worth knowing :)

Avatar

We don’t have a definite date for a case yet, but we are looking at a number of options at the moment and we should be selling them pretty soon after launch.

Avatar

[…] that problem was easily fixed via a red wire and a quick soldering job, and the team promises it won’t appear in any future units. Once fixed, […]

Avatar

[…] that problem was easily fixed via a red wire and a quick soldering job, and the team promises it won’t appear in any future units. Once […]

Avatar

[…] that problem was easily fixed via a red wire and a quick soldering job, and the team promises it won’t appear in any future units. Once fixed, […]

Avatar

At the date of launch, will people be able to buy it from a foreign country?
I mean: if I live in Brazil, can I buy a Raspberry Pi through the Internet, from your shop? Are you going to be able to ship it to my country?
Are you going to accept Paypal?

Thank you!

Avatar

Yes, and probably yes.

Avatar

You can test some, or all, of the purchase process by going to the shop at http://www.raspberrypi.com. To do a complete test, you would have to spend some money on a sticker. You can test most of it for free, by stopping before you give your payment details.

I got my stickers, so I know it works to the US. I used paypal. On the day I placed my order, the exchange rate was such that it cost me $5.03 including shipping.

It isn’t quite a complete test, as they are just stickers. They’re sent as a letter, rather than as a parcel. It would have cost quite a bit more to do the test, if they had shipped the stickers in boxes.

Avatar

[…] that problem was easily fixed via a red wire and a quick soldering job, and the team promises it won’t appear in any future units. Once […]

Avatar

Do you know a more concrete release date? Any chance that you will start selling at 1st January? :D

Avatar

Not as far as I know! We still have testing to do, and we wont release unless we know everything is OK.

Avatar

1st of January is a Sunday, and the 2nd is a national holiday in the UK so those are out – they couldn’t possibly ship anything then.

Avatar

[…] that problem was easily fixed via a red wire and a quick soldering job, and the team promises it won’t appear in any future units. Once fixed, […]

Avatar

[…] that problem was easily fixed via a red wire and a quick soldering job, and the team promises it won’t appear in any future units. Once fixed, […]

Avatar

The project open some ways to digital inclusion. Congratulations!

Avatar

Titanic work and wonderful project!
Hello from Ukraine!

Avatar

Just reading story about the treasure trove of the Apple Archive (big Mac fan, here…). Raspbery Pi is a whole new ballgame. Looks very exciting!

Avatar

This might be the first board with balls!

Avatar

[…] that problem was easily fixed via a red wire and a quick soldering job, and the team promises it won’t appear in any future units. Once […]

Avatar

Kudos to the people who checked the schematic and the board. No forgotten balls under the part, and no fatal mechanical interference between components, makes for a good prototype.

On my last project I forgot a ball under a 324 pin BGA. Something called ‘boot mode select’… Fortunately the paranoid* layout guy hadn’t culled the via for the pad, so a little scratching later and it lived.

Seeing your complex machine come up is such a rush.

*You can’t have him!

Avatar

I’m so excited for you guys. You’ve come such a long way and we’re so close to getting our grubby little paws on these things.

Just out of curiosity, do you think you will quickly sell out of them once they’re released or are you sending your orders straight to the factory? Because I know personally that I will buy at least 3 or 4 of your Model B’s.

Avatar

If I were Bender from Futurama – I’d have a huge e-Boner just by starting at this post!

Avatar

[…] that problem was easily fixed via a red wire and a quick soldering job, and the team promises it won’t appear in any future units. Once […]

Avatar

[…] ^ More on the beta boards (Raspberry Pi Blog) […]

Avatar

[…] that problem was easily fixed via a red wire and a quick soldering job, and the team promises it won’t appear in any future units. Once […]

Avatar

Hey this is practical and very educational stuff for those who want to follow your step and create/design their own board from scratch.

Amazing, the first article I read already make you guys project my favourite.

Need to read more!

Avatar

[…] of first-edition hardware. The pre-production systems have a tiny circuit track modification after an error on the original schematic (which will be fixed for the first production run) and a large SD card slot held on by Araldite. […]

Avatar

[…] of first-edition hardware. The pre-production systems have a tiny circuit track modification after an error on the original schematic (which will be fixed for the first production run) and a large SD card slot held on by Araldite. […]

Avatar

[…] been put in at schematic level now, and won’t be present in the production boards.) You can read here about why we needed to make a change. The area being worked on is around a square mm (so a bit […]

Avatar

[…] has been put in at schematic level now, and won’t be present in the production boards.) You can read here about why we needed to make a change. The area being worked on is around a square mm (so a bit […]

Avatar

[…] has been put in at schematic level now, and won’t be present in the production boards.) You can read here about why we needed to make a change. The area being worked on is around a square mm (so a bit […]

Avatar

I can’t wait for the first set of public release Pi’s come out! I’m gonna have to start pitching these to my school district because they spend A LOT of money on computers where it could be used to keep more teachers in jobs at the same time as filling up the computer labs.

Thank you for working on making such a small computer so affordable!

Avatar

… [Trackback]…

[…] Read More: raspberrypi.org/archives/470 […]…

Leave a Comment

Comments are closed