Doomed to use Powered Hubs

[Lob0426]

So now we have a little more information than we had yesterday. The RasPi \"may\" handle 1amp in so there may be more than 100ma per port after the devices uses what it needs. If this is true then it is good news. The limitation may be what the fuse on the input power is. A lot of this information has just been hanging in limbo for over two months now. It would help if some of these new specs would be released. Some of the specs are just not going to be available until the production run is made.

@kme the max output spec per USB 2.0 port is 500ma. There is a stepping up of supply as the demand increases. On the A model the USB power output is probably controlled within the BCM2835. On the B model it is controlled by the LAN9512.

The Alpha boards are using up to 2.25 watts. With the changes that have been made I think we will see something up near 2watts overall. Again we just do not know. The question is when will we get more info on the new design.

[carlosfm]

This and a USB cable would do:

http://www.ebay.co.uk/itm/1pcs-TGreen-5 ... 415ecf5846

[toadwarrior]

Just get your OS set up and then do everything via SSH then you don\'t need a keyboard or mouse. I believe that\'s what I\'ll end up doing.

[Johannes]

USB power is only limited by the fuse for the grand total current. Gert van-Loo explained in the \"Power supply news\" thread that \"the 5V coming in goes straight to the USB.\" The maximum allowed total current (which we don\'t know yet) is the only limit that the Raspberry can enforce via the fuse, and that limit is not under software-control.

USB devices are required to negotiate for power: When a device is plugged in, it is allowed to draw 100mA. Then there\'s some communication between the host and the device during which the device transmits a configuration block listing the desired current allowance (up to 500mA for USB2.0). Then the host can allow the device to draw that current or deny it. Only after the host has allowed more than 100mA may the device actually draw more than 100mA. The important thing to note is that this is just a software protocol. A host which passively passes current from the power supply to the USB devices won\'t even know whether the device stays within its allowance. Additional hardware would be needed to measure the current and cut a device off. The simplest implementation would be a self-resetting fuse per USB port. If the \"input\" fuse doesn\'t leave room for dangerous currents, then there\'s no point in adding the cost of two more fuses, so that leaves the Raspberry Pi with no control over the individual USB device power consumption.

http://www.raspberrypi.org/?p=260 has this bit of information: \"Model B owners using networking and high-current USB peripherals will require a supply which can source 700mA\"
For this to be safe, there has to be a fuse with a trip current close to 700mA on the board. Otherwise someone could connect a non-compliant high-power USB device which ignores the 100mA limitation, exceed the total 700mA without tripping the fuse and overheat and damage the power supply. Strictly speaking that wouldn\'t be the Foundation\'s fault as it can only happen with non-compliant USB devices, but there are so many non-compliant devices that ignoring this hazard potential in the hands of children seems a little reckless.

[carlosfm]

[quote]Quote from Johannes on October 31, 2011, 23:32
USB power is only limited by the fuse for the grand total current. Gert van-Loo explained in the \"Power supply news\" thread that \"the 5V coming in goes straight to the USB.\" The maximum allowed total current (which we don\'t know yet) is the only limit that the Raspberry can enforce via the fuse, and that limit is not under software-control.[/quote]

Something doesn\'t add up.
So the power to the USB ports is direct from the PSU?
We need a power switch on the board, so that we can turn the Pi off, along with all the connected USB devices.
I don\'t like to be disconnecting and connecting (power USB) cables or PSUs.
I\'ll have to put a power switch and so the mini USB power plug is no longer practical. Unless the input power trace on the PCB is visible (top or bottom layer) and can be cut.

[abishur]

[quote]Quote from carlosfm on October 31, 2011, 23:47
Something doesn\'t add up.
So the power to the USB ports is direct from the PSU?
We need a power switch on the board, so that we can turn the Pi off, along with all the connected USB devices.
I don\'t like to be disconnecting and connecting (power USB) cables or PSUs.
I\'ll have to put a power switch and so the mini USB power plug is no longer practical. Unless the input power trace on the PCB is visible (top or bottom layer) and can be cut.[/quote]

When the r-pi has power it\'s on, there is no off switch on the r-pi. That means that the only way for your usb to be getting power is if the r-pi is getting power and therefore on.

Ignoring that tid-bit, however, what he\'s actually describing is no different from a regular computer. You have the main PSU and from that PSU voltage independently runs to the processor and the peripherals. It\'s not as if the power for your USB ports on your desktop or laptop run through your processor

[novelty2145]

As there seems to be a LOT of confusion about power to the usb ports ... has it been confirmed by the devs that the RPi won\'t power a 2.5\'\' HD or is this just speculation?

I have to say that this would kill the device for me. Having to use a powered hub on a device that is connected to the wall outlet is ridiculous.

[Lob0426]

The problem is what is that self-resetting fuse set at. Was the 700ma quoted as that is the average USB PSU output or is that the fuse? What is the projected draw of the RasPi with the new power setup? What is the maximum amperage of a Micro USB connector? The LAN9512 chip is capable of controlling the output current, did they wire around this? I doubt that they did. The 100ma that was being figured on before came from that quote of 700ma, is that in error?

A lot of unanswered questions here!

By the way the wiki still states it is A DC jack(J1), states it is 6 to 20v and that the target consumption is 1watt. Might be time to update the wiki! Might want to mention that it has been changed to a Compliant USB PSU with a Micro USB charge plug. It might also be good to state whether J11,J12 and J13 are going to mounted, dismounted or must be provided by the customer.

[abishur]

Indeed there has been. Best way to explain it is with a handy dandy photo!

[img]https://lh3.googleusercontent.com/-2osm ... 20Flow.png[/img]

So in the above example let\'s say your PSU has the following specs
Output: 5V DC, 1A

So from the PSU it goes straight into the R-pi via a DC power jack using the micro usb form factor.

Once it hits the r-pi it goes through 3 protections features (I make no claims that this is the order it goes through internally, I\'m just listing them):

1) Polarity Protection Diode - Keeps the voltage traveling from PSU to R-pi and not the other way around. While damage to the PSU from the r-pi wouldn\'t technically be the RPF\'s (raspberry pi foundation) fault, they\'re really focused on their mission here and are looking out for the end user

2) Voltage Clamp - Basically a surge protector, it makes sure the voltage coming in isn\'t greater than the r-pi can handle

3) Self-Resetting Fuse - Protects from too much current going into device. Technically, I think the purpose for this is more of a feature to protect your PSU than the r-pi. In theory, under normal circumstances, you shouldn\'t receive more current than absolutely needed. This fuse then is to protect the PSU in the event that the attached devices request too much current. Of course, the r-pi team can\'t know the capabilities of each of our PSUs, so they\'re choosing a limit they think is reasonable and safe.

Okay, so once it leaves the protection features it splits, 5V going to the Soc (where the internal switching regulator steps it down to 1v2 and, I believe, the 3v3 used by the GPIOs) and to the USB devices.

I realize this is a long, roundabout way to answer your question novelty, but it\'s important that we clear the air about this confusion.

So let\'s assume that the self-resetting fuse is 1A as well and we can use every last mA the PSU has to offer. For the sake of argument and just because this is an illustration, let\'s also assume that this is a perfect system and we loose nothing through the entire process So the power comes in and we immediately use 350 mA to power the r-pi itself, from the moment it turns on it\'s working at 100% power draw and stays that way the whole time. Let\'s also assume that the HDMI port is using it\'s full 50 mA as well (which for the record, I should have drawn one more line coming off the 5V \"rail\" right after the protection features going to the HDMI port meaning that the 5V \"Rail\" goes three places (HDMI, USB, and SoC) rather than the two places I show in my photo).

This will leave 600 mA available for the USB ports to use. Maybe enough to power some 2.5 drives, but it will have a slow spin-up. Of course all of this is just an example! If you use a 2A PSU then obviously you\'d have more mA available by the time you got to the USB ports provided that the fuse allows up to 2A!!! That\'s really the million dollar question for me at this point. What are the specs of that fuse??? Of course for the more intrepid, you could always try replacing the fuse

[img]https://lh3.googleusercontent.com/-39z5 ... images.jpg[/img]

[Thunder__X]

All I know is that according to windows, I\'m going to need 500ma to power this nano usb dongle and 200ma for a keyboard and mouse. Does anyone know how to check usb power draw on Debian?

[kme]

[quote]If you use a 2A PSU then obviously you\'d have more mA available by the time you got to the USB ports provided that the fuse allows up to 2A!!! That\'s really the million dollar question for me at this point. What are the specs of that fuse??? [/quote]Nice analysis. But the bump up to six layers PCB needs a reason and typically extra layers are used for power or grounding. We just have to wait and see.

[Lob0426]

@abishur that is indeed the million dollar question, what is the rating of the self resetting fuse? Tried a search on the Micro USB max current. Appears to max at 1.8 amps(30v). I think it really is meant for about 1 amp. Not real clear in that link.

http://www.molex.com/webdocs/datasheets ... ECTORS.pdf

[tufty]

[quote]Quote from abishur on November 1, 2011, 04:41
1) Polarity Protection Diode - Keeps the voltage traveling from PSU to R-pi and not the other way around.[/quote]
More to save you from connecting +5V to GND and vice versa, surely?

[quote]Quote from abishur on November 1, 2011, 04:41
3) Self-Resetting Fuse - Protects from too much current going into device. Technically, I think the purpose for this is more of a feature to protect your PSU than the r-pi.[/quote]
A fuse /always/ protects the supply (and, in the case of your house, the wiring).

In normal circumstances, a device will only draw as much power as it needs. If the device is broken (or broken by design), it may draw as much current as it can. Consider, for example, a device with a broken USB cable, shorting between 5V and GND. A normal USB port will notice that the \"device\" instantly draws more than the 100mA it\'s allowed to, and shut it down. An \"uncontrolled\" USB port (if that\'s what we are to believe the Pi has) won\'t even know, and will happily source as much current as it can until magic smoke is released somewhere else. Hence the fuse.

Page 18/19 of the LAN9512 datasheet http://www.smsc.com/media/Downloads_Pub ... s/9512.pdf implies that you can, at least, detect overcurrent situations.

[obarthelemy]

\"still 5V\" is always much better than \"choppy 5V\", right ?

[jzu]

I beg to differ.

Providing still 5V is always better.

Accepting 6 to 20V would have been nice for money-tight hackers in developing countries. This should be alleviated somewhat in the next years when 5V-micro-usb wall warts become ubiquitous - this is not the case right now, however.

[Burngate]

If I were designing the circuit (which I\'m not - there\'re cleverer people out there) the first thing the power would hit is the fuse. Then the reverse biased diode - connect the power the wrong way, it\'s shorted by the diode, blows the fuse. In parallel with that is the over-voltage protection which could be seen as a thyristor-zener combination - the voltage goes above the combination, the zener fires the thyristor which shorts the power and blows the fuse. There should be some capacitance on the power rails after the fuse for various reasons, so the thyristor must be able to dump the stored energy and take more current than the fuse rating.
According to my reading of the device spec, the USB chip can\'t do any current negotiation - it has one pin per port, which can be used either as an over-current sense input or as an on/off switch output (or both). So Ras-Pi can\'t negotiate 100mA / 200ma/ 500mA, either it\'s on or it\'s off. Also, Ras-Pi has only one poly-fuse, not one per port. This means to me that the power rails go straight from the fuse to the USB socket(s). Therefore all downstream protection / current limiting is done by that one fuse. If that fuse allows the current your devices want your\'re ok, otherwise not.
The only other area of interest is what else does the 5v feed. We know it feeds linear regulators for 3v3, 2v5 & 1v2, and we assume the over-voltage protection will limit the supply to below whatever those regulators can handle, but if it is limited to, say 6v5, but your USB devices don\'t like anything over 5v25, then you\'ll have blue smoke if you power it from a 6v battery. Bring the OVPT down to 5v2 and your ok - the fuse blows.

[abishur]

[quote]Quote from tufty on November 1, 2011, 07:46
[quote]Quote from abishur on November 1, 2011, 04:41
1) Polarity Protection Diode - Keeps the voltage traveling from PSU to R-pi and not the other way around.[/quote]
More to save you from connecting +5V to GND and vice versa, surely?
[/quote]

Not everyone here is technically versed, I\'m going for a basic explanation to help people understand what\'s going on in the power process and the basic function of a diode is one way electricity

[quote][
3) Self-Resetting Fuse - Protects from too much current going into device. Technically, I think the purpose for this is more of a feature to protect your PSU than the r-pi.[/quote]
A fuse /always/ protects the supply (and, in the case of your house, the wiring).
/[quote]

Well... not technically. A fuse is a two way street thing. It\'s purpose it to protect both sides of the equation, it protects the PSU from over current draw and it protects the device from over-current push (such as in a surge situation). My statement was actually referring to their specific stated reason for putting it on.

[quote] Quote from Burngate
According to my reading of the device spec, the USB chip can\'t do any current negotiation - it has one pin per port, which can be used either as an over-current sense input or as an on/off switch output (or both). So Ras-Pi can\'t negotiate 100mA / 200ma/ 500mA, either it\'s on or it\'s off. Also, Ras-Pi has only one poly-fuse, not one per port. This means to me that the power rails go straight from the fuse to the USB socket(s). Therefore all downstream protection / current limiting is done by that one fuse. If that fuse allows the current your devices want your\'re ok, otherwise not.
[/quote]

Technically true. The usb chip doesn\'t do current negotiation. But, we\'ve been told that the software (USB Drivers) itself handles that such that it will still be limited to a max current of 500 mA per port. And yes, just as my picture depicts, the power goes straight from the fuse to the USB port.

[Burngate]

.... the software (USB Drivers) itself handles that such that it will still be limited to a max current of 500 mA per port. How? or are you saying the Pi tells the device it can only use 500mA so the device behaves like a good boy and only takes that? ... makes sense, I can understand that!

[novelty2145]

If a compliant USB charger should be able to source 1.8 Amps by specification, why would there be any reason to make the fuse blow at a lower current?

1.8A should be enough to allow at least one port to draw the maximum .5A while still powering the Pi itself and maybe another usb device at .1A

[Lob0426]

The LAN9512 data sheet states \"Full Power Management with individual or ganged
power control of each downstream port\".

The LAN9512 does not take in 5v at all. It only has input for 3.3v. (@abishur that is where your 3.3v goes to.) It provides \"port control\" pins that use external components to control the USB port output. The case here seems to be a ganged output control using a single poly fuse. I believe there is still power control on the USB ports. They did not change this as far as I know. The diagram for this is in the data sheet. The 5v does go to the USB ports but it goes through the poly fuse first. Of course this is only the case on the B model. The A model must have some sort of port control from the BCM2835.

http://www.smsc.com/media/Downloads_Pub ... s/9512.pdf

[USvER]

When you say MicroUSB you don\'t say just a MicroUSB plug... But also a USB spec.
So why we nead to double-check that?
And the question is How many USB devices do that?

I don\'t want to additionaly power up my USB hub...
Damn i dont want even additional hub!...
When i read about MicroUSB i said HALELUYA... But now it\'s just useless... =\\ so disapointed about this...
Why you use MicroUSB and don\'t use the features wich come with it? So... .... ... . . .. . . =\\

[kme]

[quote]When you say MicroUSB you don\'t say just a MicroUSB plug... But also a USB spec.[/quote]It\'s NEVER been microUSB at any time. It\'s been UCS. Period. That UCS just happen to use a microUSB form factor physical connector is irrelevant.

[USvER]

Hm... eben posted exactly this \"Power supply confirmed as 5V micro USB\"...
Ok... I can understand what you say, but can\'t understand why NOT???
Why not to use USB spec when we use microUSB plug? Why then MicroUSB plug... It will just confuse people... as i\'m realy confused.... =\\

[gbl08ma]

The advantage of using a microUSB plug is that it allows the Pi to be fed by a normal phone charger. It doesn\'t mean the Pi will implement the USB protocol on that micro port, though.

By the way, I just noticed it\'s microUSB and not miniUSB. That\'s bad for me, since I have a lot of miniUSB chargers and USB->miniUSB cables, but none that are microUSB. This means I\'ll have to buy some adapters...
Or I can hack the Pi by desoldering the microUSB connector and soldering a miniUSB one

[Lob0426]

The Micro USB in the case of the RasPi is just for a power plug. The mention of specs for it were as to what current is it rated to carry. This plug will carry no data at all.

@USvER you will not need a hub if you do not intend to use any high draw devices or too many devices.

The real purpose of this thread is to get the clarification, that we need on what we really can expect from the USB ports on the RasPi. can it run a USB HDD from its own ports or not? Can it be made to run higher draw devices off the ports? What is the rating of the fuse in the power circuit? Did they remove current control from the USB ports? We have picked up some new info in this thread already but there are still many unanswered questions.