Model B schematics

Now that there are Raspberry Pi boards in the wild, we thought it would be a good time to share our schematics with the world. In addition to giving you an idea of how the device works internally, these should also provide the information you need to build add-on boards which attach to the GPIO expansion connector and (in due course) the display and camera connectors.

Note: developers of expansion boards should not rely on the connection of expansion connector pins 9, 14, 20 and 25 to GND, pin 4 to 5V or pin 17 to 3V3 and should instead treat these pins as DNC (do not connect). They may be used to expose additional GPIO lines on a subsequent revision of the board.


Callum Burns avatar

Time for someone to go and make it on a breadboard…. :P *

To all those who worked on it are far smarter than I will ever be!

* Ofc that would be impossible. xD

wombat avatar

Don’t you mean pastry board? and would puff or shortcrust be better?

Cloudane avatar

Hmm careful with this, at least don’t encase it – it might end up defined as a pasty, and subject to 20% VAT!

wombat avatar

That depends how hot it runs, if it only heats up to just above room tempture then it will be VAT free!

Then again I usually have ice cream on pies. Wonders if anyone has developed a ice cream cooling system for RPi’s?

Cloudane avatar

Not on a meat and tatie one I hope.

You could probably have an ice cream sandwich on one eventually

FatBuzz avatar

Happy Geek…

mechos avatar

Best. Project. Ever.

John Doe avatar

Any special reason to not use the WP signal in the SD card?

liz avatar

A write-protected SD card doesn’t make sense on this device; it’s got your root filesystem on it.

John Doe avatar

Makes sense!

Thanks for the quick reply

oninoshiko avatar

I’m not completely sure it does. being able to write the system and then know that it can’t be changed might be useful to some (mostly embedded) types. Although I do understand the point of view.

Ofcourse, my real complaint is with the SD spec, that that is not enforced on the card.

DennisMe avatar

No way, this just reminds me more of the good old floppy disk!

Ville avatar

Wow, that actually explains why the lock “switch” on the SD card is not electrically connected anywhere – it’s actually the SD slot that detects the position of the lock tab.
I took one broken SD card apart a couple of days ago and was trying to figure out how the lock “switch” worked.

zardoz99 avatar

I also disagree. To have a bootable SD card that cannot be corrupted is very useful when you are transferring startup processing to either a network device or a hard disk over USB. With a little imagination, I suspect it would be possible to turn the Raspberry Pi into a thin client for use in education.

mahjongg avatar

Do you then also think that normal linux would work fine on a hard-disk that is write protected (if such a thing was possible)?
Any system, in the course of running normally has to have constant write access to its persistant storage medium.

jmarcelino avatar

It’s perfectly possible, Linux boots just fine from a write protected disk, you can then mount additional storage – local (e.g. USB drive) or remote (e.g. NAS) for anything you want to write.

This way configurations, kernel, boot area etc are protected against accidental damage.

Since the Pi MUST load its kernel from the SD card and doesn’t offer netboot, this would be a nice thing to have.

Ville avatar

It might not matter that much that the WP signal from SD slot is not connected. If that was connected, it would probably go to some GPIO pin and it would be responsibility of software to honour the “write protection”.
So if you would write buggy software that wrote to the card even if you didn’t want to, it might write to card even with the WP signal connected.
It’s pretty easy to make a distro that mounts all partitions on card in readonly mode.

Michael avatar

On the other hand, being able to determine whether the SD card is present or not (GPIO47) doesn’t seem to be that useful on the face of it ;)

mahjongg avatar

Well yes, I knew it was possible to boot a version of Linux from a read only medium. Its obvious, as Live CD/DVD’s do it all the time. But I was assuming that these versions were specially modified, just like in older days there were special versions of DOS that could run from a ROM. But you say that Linux boots from an -unmodified- version of linux on a read only medium?! If so I learned something new and unexpected.

Abhi avatar

IMHO, Read only file system should be fine with any ‘normal’ linux distro. Atleast it used to be so (during 2.2 – 2.4 days) Linux does NOT necessarily need write access to root FS during boot (offcourse it is good to have it so that status files like mtab etc can get updated). You can try to boot any unmodified linux distro with ro option at boot time. It will give you few errors/warning for not being able to write to some files, but will boot up (to prompt atleast, can’t say about GUI) Otherwise UnionFS / AUFS is always an option.

anonymous avatar

As long as /var, /tmp and /home are writable, it works fine. I use that configuration on my laptop every day.

Henri Tuhola avatar

Do you have plans for solving opengl drivers for non-linux operating systems?

liz avatar

Yes. ;)

oninoshiko avatar


Lars T. Hansen avatar

Oh, this is so great Liz. :)

Lars T. Hansen avatar

Text should had been:
Oh, this is great, Liz.

Sorry, for my bad english

mahjongg avatar

Great! RiscOS with openGL, that must be a winning combination. Now only these pesky USB drivers.

Oh, and what about hardware accelerated 2D drawing and blitting, will that only work with X-windows (if it will work at all that is), or could that also be OS independent?
I think that 2D hardware acceleration is even more important than OpenGL support, it will make the R-PI much faster, that and hardware (ARM) floating point support.

Darren avatar

Yay, thanks for that.

Fang avatar

Hey Eben, which software did you use to do the design and make the schematics? I am a first year compsci from York, we use Xilinx ISE. I think it would be nice if you could post some stuff on the design process. I think it would be helpful for us youngsters to understand the world better.

liz avatar

First of all, it wasn’t Eben – Pete Lomas is our hardware guy! He uses Mentor Graphics Expedition. Design process stuff is threaded through all of the posts we’ve written here – it’s worth reading what we’ve put on this blog in sequence.

covex avatar

Mentor is really not the tool you would get as a youngster. Just use some free tool like Kicad or Eagle, there is no need to start with Mentor. Important is to understand what you are doing and why, the tool may only help you.

liz avatar

Oh, sure – hearty agree. But he did ask what we’d used!

Abhi avatar

I doubt Pete Lomas would have used it as ‘youngster’ :)

jbeale avatar

I really appreciate seeing the schematics; thank you!

John Beetem avatar

Many thanks from me too!

RichardH avatar

It takes me back to 1979 and the schematics I got for my TRS-80 (cost me my life savings 8-) .. it was cool (for a 13 year old) to trace the tracks on the PCB against the schematics and see how the components all interacted. I still have it all and it all still works (though it would not pass CE tests these days – you could hear the chips working by placing a de-tuned AM radio near it .. great for debugging and performance profiling!).

Definitely think the schematics should be used as part of a school lesson.

mahjongg avatar

Well, the TRS-80 model 1 never did pass CE test then either, it was the bane of the radio amateurs who had to all kinds of things to try to use the thing, while being able to do radio. Even to the point that they had to wrap it in aluminum (aluminium) foil .
Only the later Model III was able to pass EMC tests, with its heavy internal shielding and ferrite rings for every cable .

GrahamC avatar

It’s amazing that it all fits on a credit card sized board. I count 70+ power rail decoupling capacitors alone – but then each of those is a tiny component.

liz avatar

There’s a hell of a lot of decoupling in there – it’s a very robust design. (Any idiot can make a board cheap by skimping on decoupling, but it takes some finesse to make a cheap and reliable board.)

RMW5 avatar

I think you have overestimated the capabilities of the average idiot.

tom avatar

BTW is it Open Hardware?

liz avatar

Not yet; but we do plan to release the PCB designs in the future.

mahjongg avatar

The “raft of chinese R-PI clones” would be nice as another source of cheap linux boxes, but they would miss one very important benefit, the enthusiastic support of a large group of supporters!
Without that they are not really R-PI “wanna-bees”, but who knows what will happen, they might get a large group of (Chinese) followers of their own. The R-PI is an idea that can change history!

Darren avatar

My guess is that in order to get the agreement with RS/Farnell to sell the Pi at a fixed price with margins much lower than they are used to that they insisted on an exclusive deal, so we will have to wait until the restrictions lapse in that deal before we see the full design.

liz avatar

Not the case at all, actually – both RS and element14 have been absolutely aware of our intention to open-source the hardware, and support it. There’s still plenty of business for them and us in selling an official Raspberry Pi blessed by the Foundation.

The reason we’re not releasing the full design at the moment is that not all the components in it are available to individuals (or even small/medium volume businesses). We’re working to fix this; it needs doing before we release the designs, though, because we don’t want to sour relations with our parts suppliers by creating a situation where a load of angry hackers are beating their doors down for something they don’t provide.

Darren avatar

Would be nice to see some more hacker friendly suppliers added to the suppliers list. Sparkfun and Adafruit for example spring to mind. As much as I understand the reasons behind the foundations decision I think I’m not alone in saying I’m not a fan of RS & Farnell, they have very little experience or interest in the average small scale enthusiast.

Would be good to see some of the outfits that are active in the open hardware movement benefiting from the success of the Pi, rather than feeding dinosaurs like RS and Farnell who’s business models are based on supplying business customers.

covex avatar

I do not agree, I order from Farnell and RS in low volumes and there are no issues. Thru Europe their delivery is sometimes surprisingly fast.

Michael avatar

Both Farnell and RS also have excellent enthusiast communities of their own – do head over to and to check out their forums, blogs, design resources, articles and videos.

JennyPeters avatar

Darren – Not sure why you think we’re dinosaurs.In Europe for example 75% of our business is transacted via the web, we have a community for engineers and designers not unlike this one with hundreds of thousands of members where Pete Lomas was a regular blogger before raspberry Pi. Our specialisation is high service….ie low volume for next day delivery . We hold a huge inventory to enable us to do this and have 48 websites globally in local languages/currencies. Yes, we are used to dealing with business but have a large hobbyist following too and with companies liek CPC in the Group a great source for some of the accesories we don’t sell.Raspberry Pi has been different for 3 primary reasons – the huge surge of interest on day 1 the like of which we’ve never seen, the initial supply problems around replacement parts and compliance testing and the most significant the fact that whoever distributes RPi needs to have the financial backing and resources to source components and set up manufacturing for hundreds of thousands of boards at low margins. We are excited by what the RPF are doing and want to be a part of it so i’m sorry you’re disappointed with our support so far but in this early phase, with the manufacturing complexity i think many of those you mention would also have struggled. If there is a particular service issue you’d like me to address, please let me know. Thanks

JamesH avatar

Nice one Jenny! Let’s hope that damps some of the more ardent commentors!

peter green avatar

“not all the components in it are available to individuals”
Looks like it’s the broadcom SOC and the PoP memory that are the sticking points. Am I correct?

jbeale avatar

Given the aforementioned low margins, it will be interesting to see whether, and how quickly, any clones appear.

liz avatar

I (personally, not on behalf of the Foundation) think that’s going to be a *very* interesting thing to watch.

Nick Scott avatar

Couldn’t agree more with that comment, already there is the Allwiner A10 CPU module that is being worked on (different target market), and no doubt there will be a few platforms released that target the same market. Ignoring the price factor, more options is often (not always though) good, I can think of quite a few applications where just a CPU module would be more useful (including in a few courses that I have done at uni). But at the price points involved, its becoming less important.

Darren avatar

Hopefully rather than a raft of Chinese clones shaving a couple of $ off the price we will see people supporting the foundation by buying the official Pi. The core design could see its way into custom designs and almost become a standard ‘Pi compatible’ rather than PC compatible. But that is rather in the hands of component manufacturers, if the barrier is set to high for obtaining the components then it will significantly restrict the potential and hold the door open for an alternative platform.

Overdose avatar

Thanks Eben – nice to see this.

Phil Olynyk avatar

I haven’t found the electrical specs (max/min voltages, etc) for the GPIO pins anywhere. Would you point us to them please? Are they3.3v tolerant, or very low voltage :) like 1.8?

Abishur avatar

They are 3v3

gregd99 avatar

that answers my question about why SDA/SCL are pulled up to 3.3V. I have only ever used IIC/TWI with 5V devices…. I guess there are 3v3 variants available.

p4trykx avatar

I’m not sure but probably you can use this converter

Frank Buss avatar

Yes, there are many 3.3 V I2C devices, like the PCA9555, which I’ve used in my IO-Expander, designed to work with the Raspberry Pi:

mahjongg avatar

Most 5V I2C devices will accept “high” levels as low as 2.5V fine, and because their drivers are open collectors (or drains) it doesn’t matter on what core voltage they run, as long as the pullups go to the highest voltage that all connected devices can handle.
Only very low voltage devices (1.8 Volt) will have trouble with 5V devices on the same bus.

Techmeology avatar

As Abishur already pointed out, the GPIOs are 3.3 V. You should be aware, however, that they aren’t protected against over voltage. Much more information about the GPIOs here:

Thanks, everyone for all your hard work!

lb avatar

What’s much more interesting is how much current the GPIOs can source and sink. Any data on this?

mahjongg avatar

That is -sinking- 5mA, sourcing current might be much less (<1mA).
If you want to drive a LED (in an opto-coupler) drive it by sinking current, also if you want to switch a transistor open (NPN transistor by driving current into its base) it wouldn't harm to have (in addition to a 10K base resistor) a 2K2 pullup to 3K3 on the GPIO pin, just so that the GPIO output doesn't have to struggle too much to inject current into the base of the transistor.

Harrkev avatar

Just a small comment…. On sheet 3, area C9, should that not be a (B) instead of (A) under the “IC3?”

I used to do a lot of schematics before I switched to ASIC design…

Burngate avatar

Also D10 close by …

Michael avatar

I’ve created an errata at to track these and any others that get identified.

gregeric avatar

More trivial errata on sheet 3:
R22,23,24,25 & C28 all missing (B) designation.

mahjongg avatar

and R28 too

AC avatar

Sheet 4 D6: DSI0 should be DSI1

Jim Manley avatar

Just. Simply. Amazing.

_This_ is how we can really start explaining what the Pi is, what it does, and most importantly, why and how it does it.


JQ avatar

i have to ask what is the raspberry pi part number at RS online pls

Michael avatar

Just go to their site and search “raspberry pi”. This will take you to the expression of intrerest landing page. Fill in the form and submit and the system will place you into the queue.

When RS are ready to fulfil a delivery, they’ll take the people from the top of the queue and send them a link to a secure payment page where you can enter your payment details.

Jonathan Roberts avatar

Any chance of a properly dimensioned layout file for the completed PCB? Designing an addon board would be sooo much easier if one were available..

Ville avatar

This has exact dimensions for everything (250dpi)

Jonathan Roberts avatar

I have seen that but a .dxf or similar would make things much much easier.

Burngate avatar

Just want to say Thankyou!

roli avatar

This is great. But how about the display connector? If I understand this properly we need an additional driver before we can use it. I would realy like to connect a simple touchscreen (like that from an iphone) to it and use it instead of an real display on HDMI.

JamesH avatar

Correct. The DSI connectors is direct to the GPU, and there needs to be a driver on there. The Foundation intends to produce a display and driver in the future. Not yet though.

Mark B avatar

Will be on my to buy list when you get one

John U avatar

Any way to use the DSI pins as GPIO to expand IO capability? Seems the one thing it’s lacking is a few more IO pins brought out – even if only to small pads or an unpopulated connector/header footprint.

Gert avatar

Short: No!
Long: No way!
Very long: No way at all, whatsoever!

Michael avatar

The CSI connector on the other hand…

Jack avatar

That looks truly terrifying.

Rob avatar

Loveelly… thanks very mooch

betatier avatar

Hello …. Thanks for the schematics.

I heard about power supply about HDMI ? There is a 5V available, but is there also 700mA ? Can anybody tell me, thanks.

Greets betatier

JamesH avatar

You cannot power the board over HDMI I am afraid.

Ellis Birt avatar

It is interesting to see how the analogue audio out has been implemented – PWM into a LPF. This could be another place for folks wanting to speed control motors to look. Would require soldering onto the end of a surface mount passive component though!

Gert avatar

And disable the sound driver. You would not want you motor to waltz to the music!

david avatar

That might look pretty cool though.

hamjudo avatar

As a demo, I once drove a 4 wire stepper motor with a portable CD player, hooked up to a regular audio amplifier. I generated appropriately phased sine waves on my Linux box, and burned them to a CD. I also drove the stepper from a USB speaker plugged directly into the laptop.

It makes a great demo. Speakers are low impedance coils, as are the windings in a stepper motor. Electrically they are very similar, so there is no reason you can’t drive a stepper motor from a stereo amplifier, as long as the impedance of the motor is close enough to the impedance of the speakers that the amp was designed for.

The phase between the channels determines the direction. The frequency determines the speed. Amplitude determines the power.

The students can tweak parameters in the program that generates the sine wave, and see exactly how the stepper behaves. Back then, I used C. Now I would use python.

There were series and parallel capacitors in the output circuit of the CD player, just as there are in the R-Pi. The series capacitor prevents DC coupling, which means there is a minimum frequency/speed. The parallel capacitor is part of a low pass filter, which means there is an upper frequency limit. The upper limit is well beyond the maximum speed of the stepper motor, so it can be ignored. The series capacitor will block DC, which means the motor can’t be commanded to hold a position.

You will need an audio amp to do the demo with any commonly available stepper motors. The audio output on the R-Pi is only a few milliwatts, enough for headphones, or driving an amp.

mahjongg avatar

Its probably not exactly (normal) PWM, that would be difficult to filter, it may be similar to the “1-bit D-A converter” Philips introduced for their CD players.

Tulio Adriano C Muniz avatar

The project Raspberry Pi is wonderful.
I was wondering when they would release the Broadcom BCM2835 full datasheet.

JamesH avatar

Very unlikely that the full datasheet (including GPU spec) will be released to the general public. It’s not much use to the public anyway.

Tulio Adriano C Muniz avatar

Thanks for the clarification!
I assume that only the datasheet released by Raspberry Pi Foundation is enough for experienced programmers to write or porting their operating systems for the Raspberry Pi board.

Frank Buss avatar

Looks interesting, but on which page of the schematic is the RAM?

Gert avatar

There is no RAM. It is all a scam to hide that the processor does not require memory and not give us a Nobel price for it at the same time.
Now seriously: the Memory is PoP (Package on Package) and the BCM2835 symbol does not have the top PADS (Where the memory lands on) brought out in the symbol used there. So you can add a memory to the schematics but you have nothing to connect it to. It is up to the company/designer policy to add it in or not. Here it is not in the schematic. You must NOT forget it to your BOM though and also add it to the pick-an-place robot things-to-do.

Ville avatar

So if I understood correctly they come as two separate parts and are put on top of each other by the pick&place machine.
For what I’ve understood normally there’s solder paste put on the board with help of stencil even for BGA components. So how does it work when a bga component itself acts as a board for another BGA? Does it have paste on top? Are they both soldered on one go? Does one have to program the pick and place machine to put the RAM chip on 1mm or so higher plane than usually?

eben avatar

The key step is getting flux to the joints between the top and bottom packages. For low-volume PoP you can apply a flux paste manually to the top of the bottom package, but for higher volumes you need an SMT line which has the appropriate attachment for dipping the top package gently in a tray of flux immediately before mounting. There some helpful further reading in this Texas Instruments file:

using the BeagleBoard as an example.

mahjongg avatar

In principle BGA balls do not need solder paste, but often its applied nevertheless, not for the solder, but for the flux. On POP I can imagine they just put a layer of flux on top of the package before soldering the RAM on top.

Jon Allen avatar

Looks like a decent job of cost efficient hardware design. I would put in more series elements in the path of external lines for ESD and EMC issues.

WASD avatar

Hey I recognize some of those symbols…

arcol avatar


What I would like to ask is how many grounds are there? I see two kinds, one for usb shield and other for signal ground.

The real question is, if I want to design a metal case to the raspberrypi, where can I contact it with the pcb? Is it safe to contact it with GNDs? Or should I only contact with the shields? (usb, hdmi metal case which is PE with circle on the schematic).

And a copyright question, if I intend to sell locally (or even on the web) the metal case, can I put a raspberry logo onto it? (a sticker or anything else). Or should I ask through legal way (ie. hire a lawyer, etc. Which I can afford obviously).


arcol avatar

> Which I can afford obviously
Which I CAN’T afford obviously

Sorry for the typo.

Matthew avatar
DeliciousRaspberryCake avatar

Aw it’s such a pity like half of all the BCM2835 pins are not connected to anything. I see at least 54 GPIO connectors in these schematics and they’re used for various other stuff. Or am I entirely mistaken and they are all connected?

What’s up with all the GND and input-voltage connectors, though? 32 ground connectors on the bcm2835, 12 individual 3v3 connectors on the lan chip? Is this a case of excess pins that need to be given a function?

IbanezJem avatar

I’m guessing, and please correct me if I’m wrong!
Due to the construction of the chip and interconnection to the outside world, the minute size of the wires from the silicon bit to the solder pads will limit the current available to pass through these paths. And to join all these within the chip will add to size and cost. An alternative is to leave them as separate circuits and use a ground layer, or +3V3 on the PCB to make all the interconnections.

jbeale avatar

On BGA (ball grid array) chips, it is common to use very many separate connections for high-current nets like power and ground, because each individual solder-ball is tiny, and has some limits to current carrying capacity. If you tried to run the whole chip with just one power and one ground, the little contacts would probably melt.

jbeale avatar

…and apart from current carrying capacity, there is inductance. When you have fast signals you want to minimize inductance to reduce “ground bounce” so it helps to have many grounds spread out all over the chip. For the most critical lines you’ll see signals driven as differential pairs. USB does this, with a “Data+” and “Data-” line, in between power and ground.

IbanezJem avatar

It’s been a while since my electronics education, but what JBEALE said! :-)
If I’m understanding these drawings correctly, I’m counting 17ish unused GPIO from the BCM and 5 from the LAN chip. There may be routing issues with these that made them impractical to pass through to use. The lines out with a diamond are used on the same page, the lines out with a block arrow are used on another page.

eben avatar

Indeed. We’re a little constrained in terms of the number of balls we can escape cleanly given our board technology level. We may on a subsequent revision escape a few more and replace some of the redundant grounds on the GPIO connector with actual signals.

Thinks: better edit the post to discourage people from relying on the presence of all those grounds.

tank avatar

Are the unused GPIO balls brought to the bottom layer where we could perhaps hard wire some of them to a connector ?

eben avatar

Wow. Rock and roll! We could have a look I suppose and see if there are some unrouted ones which don’t have caps too nearby on the underside.

Gert avatar

It is a matter of fan-out. Take a grid 18×18 of circles. You can draw 1 line (wire) between the circles. You may place 1 via in the centre of the circles but then you can’t route any more wire there. You can route wires on other layers but again you can only route 1 wire between 2 adjacent vias. Now bring out all the connections to the outside. Impossible unless you start using buried and blind vias. (Those are vias which do not block wires on other layers) But then you find that your PCB costs suddenly doubles!

DeliciousRaspberryCake avatar

Do you perhaps know how this problem is solved in mainboards such as one with a LGA1155 CPU socket (which has 1155 protruding pins to make contact with the pads on the processor) ?

I understand adding layers to a circuit board increases its costs, though is it common for modern computer mainboards to have >16 layers for routing all the pad connectors away from the CPU?

One more question; why is BGA used if it is so difficult to route it on a PCB? Traditional chips with horizontal pins have easier access to pins. Is this a case of costs?

IbanezJem avatar

A DIP package with 1155 pins at 1/10inch pitch would be nearly 1.5 metres long!

IbanezJem avatar

How many Altoids tins is that? :-)

JamesH avatar

Remember this is a device usually used in mobile devices. Mobile devices are small, BGA chips are small.
Mainboards are bigger – more space to do stuff,. They also have more layers which as you say makes them more expensive.

mahjongg avatar

And they use hidden and buried via’s

Stefan Brüns avatar

Another possibility is “Pad in via” – it is a real space saver, as you dont block the are between the pads, and on the bottom layer you can place capacitors directly on the vias as well. It is also beneficial from a thermal standpoint.

Regarding the cost, at work we recently received a batch of PCBs with all vias plated, at no extra cost – if we had known before …

Gert avatar

Just a word of warning: Now that you have the schematic you will notice that the GPIO connector does NOT match with what is on the WIKI page. Do NOT update that wiki page. The pins which are marked as NC should not be used in order to be compatible with possible future designs. The plan is that if a new design comes along and if the layout permits it we will connect additional GPIO pins to those NC pins. (My first vote is for GPIO 19 and 20, That gives us the second PWM, second SPI and I2S)

Nial avatar

“Note: developers of expansion boards should not rely on the connection of expansion connector pins 9, 14, 20 and 25 to GND, pin 4 to 5V or pin 17 to 3V3 and should instead treat these pins as DNC (do not connect).”

Only leaving _1_ GND could be a mistake IMHO. I have seen problems before with parallel interfaces with insufficient GND connections between boards. I’d recomment _at least_ pin 6 and pin 20 being nailed down as ground connections.

(I can’t seem to reply to Eben’s post at the top hence the reply to Gert)


Michael Wilson avatar

I am really encouraged that a project like this has come to fruition to fill the knowledge gap about computing that we lost over the last few years. I was one of those who in the 80s helped turn the $10,000 business PC into a household item by building clones for friends out of Taiwan PC XT motherboards and 10Gb hard drives – what you can get on a USB stick today! xD
Today, a few of us are on an edcuational council for an IT tech school called Per Scholas in the Bronx. The mission is to break the cycle of poverty with IT.
Does anyone have details about or a guideline of a training plan that could be modeled with the Raspberry Pi? It would be a great addition to the program as Per Scholas also puts computers into the hands of families beneath the poverty line to break that divide and to give them access to the internet as well. All suggestions welcome! Thanks!

Thomas Doubty avatar

Who actually believes this fairytale. 10Gb hard drives on XT in the 80s….I’m not sure you understand computers if you think that makes sense.

Did it have Wifi and Bluray also?

Please stay away from sharp objects.

JamesH avatar

Thomas- please be polite or be banned. There was an obvious typo in the OP, should have read MB. Otherwise I see nothing that deserves your rude reply.

Thomas Doubty avatar

Maybe re-read the self indulgent Munchausen stretch about co-fathering the home computer. Building an XT clone during the ‘PC compatible’ era and fathering home computing are pretty different.

JamesH avatar

On the other hand, his post, whilst a touch self aggrandising, wasn’t rude.

Ray_GTI-R avatar

Is there a mechanical drawing of the board lying around anywhere/link on the Interweb?
Alternatively, can someone with a ruler and a board pass on the following dimensions:-
a) board length
b) board width
c) board thickness
d) CPU length & width
e) GPU length & width
TIA, Ray

Ray_GTI-R avatar

Self- reply …
So far I have the following data for the board (all in mm):-
84.81 x 55.92 or
85.0 x 56.17 or
85.7 x 57.1 or
85 x 56
(Average 85.1275 x 56.2975)
1.5 thick
1.6 thick
(Average 1.55)
I have also just seen which is a great help.
No data for the CPU & GPU yet … I’m only an amateur BTW.

Matt avatar

Out of curiosity, will this be eventually drawn in NEMA for the stubborn americans refuse to use IEC like the rest of the world? It would kind of make sense to do both since it is created as a training device right?

Mattylad avatar

Perhaps it can be seen as an opportunity for those same stubborn Americans to learn to read schematics that are not drawn in your own standard. I.E. Use the one that the rest of the world uses lol….

After all it is perfectly readable as it is.

Matt avatar

I can read both. And I actually prefer IEC. But I doubt that schools will feel like they have a reason to teach both IEC and NEMA

Ray_GTI-R avatar

… also noticed that the newly-released schematics say “USB Power Input” is 700mA min.
Emphasis on >>> min <<<.
What happens if someone plugs in a 5V PSU with, say, 5 amps? Or 699mA?? Or one that fluctuates between, say 600mA & 701mA depending on local mains power to the PSU???
Help with the previous (physical size) questions also welcome.

Gert avatar

AMPS are not like volts. AMPS says how much you CAN draw from a supply. It mean you can safely connect a 100A supply to a 0.01A device. (Compare it to liters/pints. If you need 1 liter/pint of water you don’t care that you have a lake available. As long as there is enough).
If a supply has not enough amps your Raspberry-Pi will work unreliable. We have seen that the first thing that starts failing is the Ethernet, next is the keyboard.

Abhi avatar

WoW!! What an example Gret, you’ve earned yourself (another) fan today. I’ll surely remember this example if I ever go to academics.
Coming to your question, Ray_GTI-R, as Gret mentioned, if you are using a PSU rated 5A (Off-course it has to be 5V, not much compromise here) device will function fine. However, if your supply is of lower current rating there are number of possibilities.Ranging from, may work fine if SW is not using dev fully & hence chip is consuming lesser power, to coffee smelling PSU, if it is very cheap quality unprotected one.

Ray_GTI-R avatar

“If a supply has not enough amps your Raspberry-Pi will work unreliable. We have seen that the first thing that starts failing is the Ethernet, next is the keyboard.”

Excellent point. Thanks Gert.
I guess what I was getting at is that if someone connected a 5A (5V) PSU then connected a peripheral that draws 5A what next? That is, can the Pi cope with a 5A (x 5V, or 25 W) throughput? And the associated thermal impact. Note I’m leaving aside the original 700mA vs 700mA MININUM in the newly-released schematics.
As for the CPU & GPU dimensions I guess I’ll have to wait for my own order to be completed. :-(

JamesH avatar

The max the Pi’s USB can output is 100mA each before they trip a fuse (that might be 150, therabouts anyway). So a 5A power supply should be fine!
What do you mean by CPU/GPU dimensions?

Flitzpiepe avatar

Just watch out for dimension of F3 – seems to be a fuse for 1.1 A – even if it is dying @ 1.1 A and 6V it is not possible to get 5A @ 5V over this fuse – Ptotal,max would be 6.6 Watts.

Ray_GTI-R avatar

The length & width of the two main chips, one is the CPU and the other is the GPU.

JamesH avatar

Odd. The GPU and CPU are in the same package, not separate chips. I presume you mean the SoC and USB/Ethernet chip?

Ray_GTI-R avatar

Oops – my mistake, yes … those two chips.

Ray_GTI-R avatar

“The length & width of the two main chips”
that is …
“the SoC and USB/Ethernet chip”
Liz, Eben, Anyone … got a ruler?

JamesH avatar

They are 12x12mm for the CPU/GPU and 8×8 for the ethernet.,

Ray_GTI-R avatar

Thanks JamesH for those dimensions. Most helpful.

Andrew Scheller avatar

“developers of expansion boards should not rely on the connection of expansion connector pins 6, 14 and 20 to ground”

According to pin 6 is the “official” ground pin?

Michael avatar

Good spot! Eben’s updated the post to clarify that it is pin 9 not pin 6 that is do not connect; and clarified a couple of other DNC pins that were’t listed in the original post (six pins in total).

Stefan Brüns avatar

Would it be possible to add the information about the “DNC” GPIO pins directly in the schematic, a simple text box would suffice.

Mark Dunsford avatar

I know that the upgraded boards will almost certainly be labelled Version 2 or something like that but in an educational environment that kind of detail can get overlooked when you have 200 devices and some of them behave differently to others. (or as the teaching staff will describe it – “don’t work”)

Rather than upgrade the model B might I suggest that these improvements be rolled out as a part of a new model C including other design revisions that come out of feedback from initial field use. With any luck there will be an extra connector or different layout to help those of us who are observationally challenged.

JamesH avatar

What upgraded boards?

AFAIK, the Model A and B are it for the foreseeable future. No design changes (except for CE compliance). And those changes will not mean a change in behaviour. In fact, changes in behaviour are unlikely even were there to ba a model C – it’ll still run Linux and the same apps as any previous boards.

olaf van schoten avatar

James, Mark refers to the response of Gert about the NC pins on the PCB. Gert predicts these NC points could be used to get some more GPIO pins out in a later revision.
And people do now expect there will be a new revision that has these extra GPIO pins hence new version.

JamesH avatar

I’d advise against holding breath waiting for that one.

Mike Cook avatar

From the sounds of it others can open that PDF. I am on a Mac and it gives me an error saying the file is corrupt. This happens with the default Apple application as well as the Adobe reader 8.1.1. Has the file become corrupt or do I need a more up to date reader?

liz avatar

That’s odd – I’m using a Mac, and I’m fine just in the browser (I haven’t tried Acrobat reader, though). Anybody else having any problems?

Andrew Scheller avatar

I’ve just uploaded the results of the “schematics breakdown” I’ve been working on most of the weekend… :-)

Flitzpiepe avatar

Guys – if you give developers no chance to supply their stuff through expansion header there is more or less no way to do something useful with it … even high-current-LEDs might be a problem without +3V3 and +5V … you force everyone just wanting something normal like rs232-signaling (max3221 for instance) to use extra supply – wtf?

Flitzpiepe avatar

Sorry – just ignore this post – havent seen Gerts post and havent seen the actual pinning – my bad.

So far nice job folks – i am myself mainboard developer, so i know this is lots of work

matt avatar

I’m sad now seeing all those unused GIO pins :(

SteveOll avatar

Is it possible to mount the device on pillars or motherboard standoffs (I don’t see any holes for this purpose on the board) as I would like to mount the Raspberry PI inside some kind of enclosure for a Mechatronics Project at Uni.

JamesH avatar

Lots of thoughts on mounting in the forum. Search sugru for examples.

Aditya Mehendale avatar

What is the purpose of diodes D12 and D13? Won’t the low-side diodes actually clamp the swing of the audio-out-port to -0.7V (when actually it sould be doing -1.7V or lower due to the decoupling?)
Hats off for the good work though!

ukoda avatar

The company I work for currently produce Sigma Designs based media players used in vertical market applications. I have been looking at cheaper solutions, like the Amlogic or F10 solutions but being a fan of Linux and Open Source solutions the Raspberry Pi has appeal. I like the idea of building a version in the form factor I would typically use and I only see the PoP as being a manufacturing challenge. However there are two other considerations.

Firstly I would want to add a second SD card for media storage, independent of the Micro SD card with the OS on it, Has anyone given any thought this yet? I would be wanting an interface fast enough for direct HD playback so bit bashing would be out of the question. Using USB storage would be plan ‘B’, but SDHC cards are what we currently use.

Secondly I hear talk of it being Open Source hardware but I notice the schematic has a copyright to Norcoft Tecnologies Ltd. Just because our factory is in China it does not mean our company does not respect copyright. What is he legal status of derived works from the published schematic?

JamesH avatar

You need to talk to Broadcom about your use case since you would need to buy the SoC from them. Depending on a number of factors you then may be able to get tech support from them for your particular requirements. What I don’t know is which part of Brcm you nee to talk to or who the contact may be. I’ll try and find out.

ukoda avatar

Thanks for that. Our volumes would be in the range of 5K to 50K PA, which is sometimes too low to get support. I haven’t tried dealing with Broadcom so I’m not sure what they are like.

Ainsworth Kohler avatar


Greg Smith avatar

I suggest you put a ‘resources’ tab at the top of the site, and put the schematic on the resources page. And any other useful resources that have are presented in the news posts.

donde avatar

Gee, I wish there was small value resistor at the very beginning, so I could measure current. I = E / R

Richard avatar

Is there any chance of producing a searchable PDF where the net names etc are stored as text rather than an image? I’m not sure if Mentor can do it (I’m used to Altium at work) but it’s something I find very useful.

Bensilver avatar

I also would like a pdf “searchable”!

Adrian avatar

it would be fantastic if the schematics were in orcad /eagle/… format, more than pdf format, could be?

Abishur avatar

I’d come discuss this in the forum, people tend to look at only the most recent front page post ;-)

Dp avatar

What is the Microcontroller in the Pi board?

Rahul Rasal avatar

Hello Friend,

Can you suggest….

I try to CLFS from rsp.
but, i didnt ask about which type board.

Try to bootloader but, target specify ..
i dont know target..

Any one can suggest me about {target}_config..?


[${CLFS_TARGET}= arm-linux-musleabihf]

Rahul Rasal

James Hughes avatar

Please post questions like this on the forum, you won’t get a reply here.

Comments are closed