Power for USB external 2.5" harddisk

[cil8]

Hi all!

I'm seriously thinking to buy a R-Pi board. Really a great idea... and a great price.

Just one question: I really like the idea of using a cell phone USB battery charger for this board. Using a power supply like the iPad battery charger (that gives 10W), do you think R-Pi board will be able to power a USB 2.5" harddisk through its USB port? Or it is necessary use a dedicated power source for the harddisk?

Thanks.

[RaTTuS]

if the hd draws more than 500ma then it will have to be powered itself

you only need 5w on the power supply

[cil8]

a 5W power supply will be able to power R-Pi board and a max. 500ma hardisk together?

[liamfraser280]

As far as I know, the Raspberry Pi's usb ports have a self-resetting fuse that only allows them to provide 100mA whereas your harddisk will need at least 500mA. Therefore you'll either need some kind of external power supply or powered usb hub.

I personally have invested in a powered usb hub because I want to try more than one device on the Pi. It's a 7 port dlink that I picked up for about 17 quid. The power supply is 3 amps which I am going to split into micro usb and then the DC barrel type connector to go into the hub. Easily enough power there to give 1000mA to the Pi (Model B at full load is supposed to use around 700mA) and 2000mA left over for 4 other power hungry devices using 500mA each.

I can also connect devices to the hub without having to fiddle with the small Raspberry Pi. Reduces chance of breaking anything even though the board is supposed to be fairly robust. Also, the fact that it is a hub won't affect performance because there is only the bandwidth of one USB port for the whole device: 480Mbits/s. The second usb port and ethernet adapter on the model B come from a usb hub chip and a usb to ethernet adapter chip.

Hope this helps.

Cheers,

Liam.

[abishur]

Liam Fraser said:


As far as I know, the Raspberry Pi's usb ports have a self-resetting fuse that only allows them to provide 100mA whereas your harddisk will need at least 500mA. Therefore you'll either need some kind of external power supply or powered usb hub.


I believe that is incorrect.  The self resetting fuse is where the power first comes into the r-pi and is a 1A fuse.  It's been a long time since it was discussed, but I believe there is no such limitation on the usb ports (it's just that USB 2.0 specs require a *minimum* of 100 mA)

At the same time, I agree with your overall assessment, you will need some kind of external power supply to power an external Hard Drive

[liz]

Ten points to Abishur!

[mobeyduck]

There probably are other fuses closer to the usb but those are of the kind

"OMG! it smokes! what now?!?"

electric components work on smoke, when it comes out its broken.

On-topic

The R-Pi probably wont give enough juice out over usb to power a hdd

But if you got one that doesnt need all that much power, like some 1,8" or ssd's.

a quick check with a ocz vertex 3 and a calculator it would use about 250ma so that should be safe, but ssd's dont perform as good as they can using on usb. and 1,8"hdd cost to much if available at all a ssd costs the same.

btw i dont think there is any trim over usb of if it is supported on an arm platform

[walney]

mobeyduck said:


But if you got one that doesnt need all that much power, like some 1,8" or ssd's…


I have  a Toshiba 1.8". According to the spec sheet it uses 500mA max. I'll try it out as soon as I can.

EDIT: they are still cheaper than SSD - you can get them from Ebuyer at between £70 - £95ish depending on capacity (120 - 250GB)

[jojopi]

liz said:


Ten points to Abishur!


There is as Liam says a polyswitch on each downstream USB port.  On the beta boards they are clearly 140mA (hold current) devices.  Are you saying that these are to be increased or removed on the production boards?

Otherwise, it would be good if admins checked their facts before "correcting" people, especially en masse.  Readers will be likely to assume that what you say is authoritative.

[cil8]

Liam Fraser said:

I personally have invested in a powered usb hub because I want to try more than one device on the Pi. It's a 7 port dlin...
An USB hub to power everything is really a good idea. And I get more free USB ports.

Ok... I'll buy one R-Pi

But the 500mA limit is still and interesting topic!

[abishur]

jojopi said:


liz said:


Ten points to Abishur!


There is as Liam says a polyswitch on each downstream USB port.  On the beta boards they are clearly 140mA (hold current) devices.  Are you saying that these are to be increased or removed on the production boards?

Otherwise, it would be good if admins checked their facts before "correcting" people, especially en masse.  Readers will be likely to assume that what you say is authoritative.



Well I did preface my statement by clearly stating that this is what I remembered off the top of my head, and left it there until one of the people actually affiliate with the RPF could comment.  So I did some more researching in order to give an authoritative response as requested.  Using the x-ray provided to us of the final board:



You can see that yellow is the power.  The power comes from the LAN9512 chip, stops by a coupe resistors and capacitors, then heads straight to the USB ports without touching the fuses.  However, if you look closely it would appear that the red lines (which represents data) might go through the fuses, meaning that they're protecting the board from a sudden surge of power over the data lines or someone plugging in a home made cable where they got some wires crossed.

Now while I admit that the second half of that, the part about the data lines, is my personal speculation.  The first half is an authoritative statement.  You can clearly and unmistakeably see that the fuses do not interfere with the power being sent to the usb ports.

We admins don't say things authoritatively without checking our facts first, please don't go around accusing us of doing so and encouraging the spread of misinformation.  As with all things where the hardware of the r-pi is concerned, I am not where the buck stops.  Gert or Jamesh are where the buck stops and I never mind them stepping in and correcting anything I have to say.

[drgeoff]

It would be good to have an authoritative statement on this.

My interpretation of the "X-ray" is not in line with Abishur's.  I don't see any power tracks going to the USB sockets.  I see two yellow tracks going to the middle two pins of one USB socket and two red tracks going to the middle two pins of the other USB socket.  Surely those are the data lines.  The yellow tracks have breaks in them – I assume links that cater for the difference between the A and B versions.  (A version has one USB socket fed from the SoC and the B version has two sockets fed from the 9512.)

[rurwin]

I have no idea, but IIRC, Gert said it was a six-layer board, because 4 were not enough. I might guess that the extra two layers are under-utilized. So we should see at least four layers in the picture.

But I see only two layers in the picture: Red and Yellow. Blue and Magenta are silk-screen. If Red or Yellow represented multiple layers we would see similar-coloured tracks crossing each other. We don't see that.

Generally in a multi-layer board, two internal layers are reserved for power, and they are almost 100% copper, with holes around non-connecting vias. In pictures like this you generally turn off the power-planes, because otherwise they obscure everything else.

The conclusion would be, if you can"t see anything going to the fuse, power probably does.

But it is also possible that the connections are on the two layers that we cannot see.

[abishur]

I'd also enjoy Gert or Jamesh throwing down an official statement (we've had some discussions in the past about F1 and F2), if it helps I do know for a fact that yellow is power and that it's not a break in the line but a connection to two resistors and two capacitors.  I'm assuming red is data, but it's entirely possible that red is power on a different layer.

[Burngate]

Having a look at that gerber, and comparing it to the PDF data sheet for the 9512, it looks to me that the right-hand edge of the 9512 on the gerber is pins 1 – 16 counting up.

On the PDF, pins 1 – 4 are the USB data pins, and 14 & 16 are PRTCTL (power control.)

So pins 1 & 2 are traced in red to the centre pins of one USB, and pins 3 & 4 in yellow to the other USB (via a couple of Rs? which may have been removed? Can't work that out!)

Meanwhile 14 & 16 go to / come from a couple of diodes (D10 & D11) behind the fuses.

Outputs from the fuses and power to the USBs are not on the gerber, and not seen on the board JPGs, so will be on inner layers.

EDIT: Abishur both knows more and types faster than me!

[abishur]

really?  Interesting.  If pins 3&4 really are data then that would mean not all the yellow is power (which contradicts what I had heard previously, but even official statements can have mistakes from time to time I suppose)

[drgeoff]

I'm wondering if Abishur has the USB sockets and the LAN socket transposed in his mind.

There are definitely breaks in the yellow tracks leading to the centre two pins of one USB socket.

Power may be yellow but not all yellow is power.

For sure the Gerber is not showing all the layers.

[abishur]

drgeoff said:


I'm wondering if Abishur has the USB sockets and the LAN socket transposed in his mind.

There are definitely breaks in the yellow tracks leading to the centre two pins of one USB socket.

Power may be yellow but not all yellow is power.

For sure the Gerber is not showing all the layers.



Nope, I don't have them backwards (this time at least ) I'm talking about the USB ports, which lie directly under f1 and f2 and are kinda a pale yellow color on the centerish of the right side of the board.  The break I believe you're referring to are the four pale yellow squares outlined in pink.  This is the resistor/capacitors I keep referring to and not a true break (i.e. a disconnect with nothing bridging them) in the line.  Perhaps a different definition of the word break?

[drgeoff]

Abishur said:


drgeoff said:


I'm wondering if Abishur has the USB sockets and the LAN socket transposed in his mind.

There are definitely breaks in the yellow tracks leading to the centre two pins of one USB socket.

Power may be yellow but not all yellow is power.

For sure the Gerber is not showing all the layers.


Nope, I don't have them backwards (this time at least ) I'm talking about the USB ports, which lie directly under f1 and f2 and are kinda a pale yellow color on the centerish of the right side of the board.  The break I believe you're referring to are the four pale yellow squares outlined in pink.  This is the resistor/capacitors I keep referring to and not a true break (i.e. a disconnect with nothing bridging them) in the line.  Perhaps a different definition of the word break?


I mean "break" in the sense that without some components being soldered to those pads the pins 3 and 4 of the 9512 chip do not connect to the USB socket.

There are four pads.  Any usable component (except an aerial) has at least two connections so those four pads can take at most two components.  Thus one resistor and one capacitor, or two resistors or two capacitors but not resistors and capacitors.  And anyway why have those components (whatever they are) on the data lines to one USB socket but not on the other one?  I'm still guessing they are option fit zero ohm resistors dependent on whether the board is a model A or model B.

Hopefully Gert will be along soon and put us all out of our joint misery by telling us if there are fuses on the USB 5 volt outputs and, if yes, their rating on the production boards.

[jojopi]

drgeoff said:

There are four pads.
To be pedantic, if you are looking at the Gerber screenshot there are actually eight pads -- four on the top layer of the board and four on the bottom in the same place.  With 0R links populated on the top the front USB data lines are connected to the 9512 (Model B); with links on the bottom instead to the 2835 (Model A).  As you realised.

However, if you compare the photos of the beta boards, the top pads are gone and the short spur to the 9512 is permanently connected even on the Model A.  So there are no components at all on either the "red" or "yellow" USB data lines on the production Model B.  This must save a quarter of a cent.

[drgeoff]

jojopi said:


However, if you compare the photos of the beta boards, the top pads are gone and the short spur to the 9512 is permanently connected even on the Model A.  So there are no components at all on either the "red" or "yellow" USB data lines on the production Model B.  This must save a quarter of a cent.


That makes sense. I was wondering:

1. Where the complementary Model A 0 ohm resistors were hiding.

2.  Why the ones on the 9512 output were needed given that the 9512 won't be present.

All this doesn't definitively answer the actual question of what current can be pulled from the USB ports.

[abishur]

drgeoff said:


I mean "break" in the sense that without some components being soldered to those pads the pins 3 and 4 of the 9512 chip do not connect to the USB socket.

There are four pads.  Any usable component (except an aerial) has at least two connections so those four pads can take at most two components.  Thus one resistor and one capacitor, or two resistors or two capacitors but not resistors and capacitors.


Ah, then we were defining it differently   The x-ray does show four pads, but I was cheating and looking at the populated boards and saw that there are 2 resistors on top (pale yellow) and two capcitors on bottom (pink outline).

I sent Gert a PM and he was saying he didn't pick them but can confirm that they are a 140 mA fuse (which jojopi already mentioned).  Now it doesn't make sense for a 140 mA fuse to be protecting some LEDs nor does it make sense for there to two of them on the audio jack so he's 90% confident they are indeed protecting the usb ports.

So I'm a big enough man to admit when he was wrong, despite every bit of information I've received to date and the fact that I did check my information before giving an authoritative statement (None of us Admins just spot of random information thank you very much, though we are still wrong from time to time), it would indeed appear that the USB ports are protected by a fuse that limits them to a max current of 140 mA (not 100 mA).  Not that it matters, but they are also protected by a diode to prevent current from coming back down the ports.

My apologies for the misinformation, it's quite contrary to what we've all seen on the forums, but it's straight from the proverbial horses mount so I'm certainly going to take it to heart... of course I'm also probably going to remove them first chance I get .

[abishur]

As a side note I do want to make sure to specifically apologize to jojopi.  I found your comment about Admins needed to check their facts to be very rude, especially since we actually had checked our facts first, they turned out to be wrong but we did check first.  Ultimately I shot off my mouth and now I'm humbly eating my crow and tip my hat to you being correct on this issue.

[Vindicator]

So the next question is, Does the model A have a poly fuse on it's single USB port and what is the size of that fuse on the port ?

Not that matters either as long as you use a Powered USB hub LOL.

In the case that you use a passive hub it would matter as to the consumption of the hub +keyboard+mouse+Ethernet device = (greater)than the poly fuse.

Could be annoying your keyboard and mouse shutting on and off seemingly at their own will.

Also now that the model A has more memory that should mean it will consume more than the stated 300ma (maybe 400ma now) not sure.

[patrickhwood]

Vindicator said:

...
Also now that the model A has more memory that should mean it will consume more than the stated 300ma (maybe 400ma now) not sure.



I would bet not: most DRAM current draw comes from the pin drivers and row activity.  Also, parts of memory that aren't being addressed have very low current draw, at least for CMOS.  Of course, a lot depends on the configuration of the 2Gb devices vs. the 1Gb device.  If the rows are twice as wide, current draw for read and write operations will be something like 30% higher.