USB Port Current Boost (solved)

[gjs]

If I need more than 100mA from one of the onboard USB ports, is there any reason I cannot connect the +5V GPIO pin directly to the +5V pin of the USB port? Or will this end in a puff of smoke?

[Gert van Loo]

Easiest solution is to remove the USB fuses and short them out, which is in fact the same thing your are proposing.
On second though: just short them out. Removing the fuses is difficult unless you happen to have two soldering irons.
I'll see if I can change a board and show a picture.

[gjs]

A picture would be great, thank you.

Why is it neccesary to short out the fuses? If the GPIO pin is connected after them (on the port side) they should not blow, no?

[Burngate]

If you connect the GPIO pin to the ports, then you are shorting them out!
Someone somewhere (I can't find it for the moment) posted a pic of wiring directly from the power i/p to the USB output. That means you can take upwards of 500mA out of the USB without risking blowing the input fuse. Of course it also means the USB ports have no protection ...

[Ed Raket]

On this here blog it is clearly shown with pictures how to "bridge the polyfuses";
http://theiopage.blogspot.nl/

Maybe this is what Gert was referring to?

[Larry_Adlard]

On this here blog it is clearly shown with pictures how to "bridge the polyfuses";
http://theiopage.blogspot.nl/

Maybe this is what Gert was referring to?

That's very helpful. I had already decided to go direct from power input to USB, but this is a lot easier. This is more protection for the power supply than for the devices on the USB. The Rpi can't police the USB and each device is expected to protect itself.

I think we are coming to a general consensus that the Rpi needs a more powerful supply than 700ma, probably 1.5A to 2A. I want to put a WD hard drive on the USB and it says it needs 550ma plus a small amount for USB to SATA interface. I'm not intending to turn the Rpi into a desktop - I use the WD as a backup device and although there isn't anything serious on the Rpi right now I'm hoping there will be a great deal of work to protect later.

[dom]

On this here blog it is clearly shown with pictures how to "bridge the polyfuses";
http://theiopage.blogspot.nl/

Maybe this is what Gert was referring to?

That looks a bit more complicated than necessary. How about this?
https://dl.dropbox.com/u/3669512/IMG_1359.JPG

[hippy]

I think we are coming to a general consensus that the Rpi needs a more powerful supply than 700ma, probably 1.5A to 2A.

Remembering though that the R-Pi has a 700mA power inlet fuse. Though it won't blow at 700mA if intending to draw a substantial amount through the micro-USB socket it may become necessary to short or replace that fuse. The easier solution is to provide power in via the GPIO power pins with appropriate current protection external to the R-Pi

The current design, with USB ports notionally limited to 100mA each, has a logic and rationale to it but, as it doesn't support higher current USB devices, it is awkward where that's desired with a single PSU. Plus there's the USB polyfuse resistance issue which can cause problems at lower currents with some devices.

[clickykbd]

Plus there's the USB polyfuse resistance issue which can cause problems at lower currents with some devices.

Are these issues often manifested by kernel panics (seemingly at random) when doing i/o to the usb device in question? I've got a USB DAC that is doing that to me right now and haven't narrowed it down yet. It has it's own external power, so wondering if it's possibly the lower-current version of the problem?

[Larry_Adlard]

I think we are coming to a general consensus that the R-Pi needs a more powerful supply than 700ma, probably 1.5A to 2A.

Remembering though that the R-Pi has a 700mA power inlet fuse. Though it won't blow at 700mA if intending to draw a substantial amount through the micro-USB socket it may become necessary to short or replace that fuse. The easier solution is to provide power in via the GPIO power pins with appropriate current protection external to the R-Pi

The current design, with USB ports notionally limited to 100mA each, has a logic and rationale to it but, as it doesn't support higher current USB devices, it is awkward where that's desired with a single PSU. Plus there's the USB polyfuse resistance issue which can cause problems at lower currents with some devices.

I think you have just convinced me to return to plan 'A' - wire power input direct to USB. I'm not happy going via GPIO because I intend using it later and I don't want it cluttered up with kludges for some other problem. Ideally, I don't want to mess with the R-Pi too much. The plan is to use it rather than mod it. The polyfuses seem to be a problem generally but not on my R-Pi (AFAIK).

I sympathise with the Foundation. A machine that was originally designed to get kids programming has become a device expected to be all things to all men. The recurrent themes are power, power connector, connectors in general and i/o performance using an SD card rather than USB. As a programming platform it is already adequate without the numerous improvements in the pipeline. I suspect that some will be tempted to take the work of Raspberry and fork a more capable Strawberry.

[bredman]

On this here blog it is clearly shown with pictures how to "bridge the polyfuses";
http://theiopage.blogspot.nl/

Maybe this is what Gert was referring to?

That looks a bit more complicated than necessary. How about this?
https://dl.dropbox.com/u/3669512/IMG_1359.JPG

Still too complex.

I use a tiny piece of aluminium foil and sticky tape on top of each polyfuse.

[Gert van Loo]

Below is a set of pictures which show how to short the fuses.
I do not remove them as that is very difficult (unless you have two soldering irons)
So I put a short straight over them.

Top to bottom:
1/ Fuses before changes.
2/ Fuses with solder blob added to one side.
3/ Fuses with wires soldered to one side.
4/ Fuses with wires soldered at both sides.
5/ Cut off the wire parts which were sticking out.

[thexman]

if you want to temporary short the Polly fuse for a trial or test before going the whole hog of soldering try a crocodile clip they open far enough and act as a short and don't require a soldering iron or damage to your pi should you ever need to return it for warranty.

This sort of thing. theres lost listed these are an example not a recommendation.http://www.ebay.co.uk/itm/Qty-10-Insula ... 1c20fcf538

[Ed Raket]

if you want to temporary short the Polly fuse for a trial or test before going the whole hog of soldering try a crocodile clip they open far enough and act as a short and don't require a soldering iron or damage to your pi should you ever need to return it for warranty.

This sort of thing. theres lost listed these are an example not a recommendation.http://www.ebay.co.uk/itm/Qty-10-Insula ... 1c20fcf538

Thats a good idea! allso usefull to see if it solves "your specific" problems

[Ed Raket]

On this here blog it is clearly shown with pictures how to "bridge the polyfuses";
http://theiopage.blogspot.nl/

Maybe this is what Gert was referring to?

That looks a bit more complicated than necessary. How about this?
https://dl.dropbox.com/u/3669512/IMG_1359.JPG

"More" complicated?.... you completely removed the original F1&F2...

[hippy]

Plus there's the USB polyfuse resistance issue which can cause problems at lower currents with some devices.

Are these issues often manifested by kernel panics (seemingly at random) when doing i/o to the usb device in question? I've got a USB DAC that is doing that to me right now and haven't narrowed it down yet. It has it's own external power, so wondering if it's possibly the lower-current version of the problem?

That sounds to me more like a software issue rather than hardware as it's got its own power source.

The polyfuse issue is that each USB fuse has a fairly high resistance which at a moderate current draw causes the voltage supplied to the USB device to drop. That voltage becomes too low for some ( but not all ) USB devices and they cease operating reliably.

The proposed 'fix' to short out the USB polyfuses resolves (1) their current limiting affect, and (2) the voltage drop they cause as current increases.

[Ed Raket]

You see? thats the magic of the Pi! I havent even got one yet, but i allready want to tinker with it....

[manis404]

Hi guys. I am updated the blog post on modding the fuses for the case of high current devices. I recommend that you take a look at the USB Spec whitepaper I linked in there.

http://theiopage.blogspot.com/2012/06/i ... -host.html

Basically, for high-power devices, you'll need to increase the capacitance between the Vbus and Gnd. The 47uF already present on the pi is way too low for plugging external hard disks directly as they will cause a significant voltage drop that cannot be countered by the polyfuse mod alone.

I found that there is some kind of flaw on the schematics...Take a look at how C32 is before F1 and F2(and not after). This basically means that the fuses will blow each time the bulk storage cap (C32) is really used to its fullest. The whitepaper also states that placing the bulk storage cap before the fuses is technically incorrect.

[Larry_Adlard]

Well here it is - The Pi-pass. It simply bypasses the R-Pi and all it's fuses.

The R-Pi is full of surprises. It appears that the USB isn't hot plugable. You plug in a small device and the machine does a hard reset.

[manis404]

Larry_Adlard, that was exactly what I was planning to do actually. Its way neater this way, nice mod. I didn't do it like that 'cause I had no spare wire around at the time.

[selsinork]

Well here it is - The Pi-pass. It simply bypasses the R-Pi and all it's fuses.

Hmm. only problem is that your wire-add is probably 10x thicker than the one in your micro-usb cable

[Larry_Adlard]

Hmm. only problem is that your wire-add is probably 10x thicker than the one in your micro-usb cable

It is 7 strands x 0.2mm. I'm not using a standard 'phone charger and it's actually thinner than the twin core figure of 8 micro-usb cable. It just looks fat alongside the components. A fat wire wouldn't be a problem as such - less resistance.

[selsinork]

It is 7 strands x 0.2mm.

Ok, so only 7x thicker than the usb cable

I went to 16/0.2 via the gpio header for power quite quickly as all of the micro usb cables I could find had shocking voltage drop at fairly low current.

Your nice wire-add = good, I'd just think it'll show up weakness elsewhere... Now if you've got a 500 watt 'phone charger' it's probably not going to be that

[Larry_Adlard]

Now if you've got a 500 watt 'phone charger' it's probably not going to be that

Nice one

[Ed Raket]

Now if you've got a 500 watt 'phone charger' it's probably not going to be that