USB Port Current Boost (solved)

[Lob0426]

I believe the spec for that micro USB connector is 1.8amps for all four pins. But people have put up to 2 amps through the micro USB. The usual is 1amp for it as connected right now.

Gert has stated that the traces were designed for 500ma. There probably is some safety margin there as well.

[rustybikes]

If I read this correctly, either by shorting out the polyfuses or adding a jumper (Pi-pass pictured by Larry_Adlard), I could run my old harddisk-based (80GB) iPod directly here. I've already connected it (iPod) through a hub and was able to lay down a partition table and the ext4 filesystem (the iPod's firmware was not happy about that) on it and mount it just like any other harddisk. Now I'm getting this idea that I could just plug it in directly (possibly boot from it). Am I getting too excited about this possibility? It entices me 'cos the iPod is about the same size as an RPi, and I have a funny image in my head of the two duct-taped together...

So, what do you guys think? Would that work? I really don't know how much power the iPod draws, but I'm sure it's more than a stock RPi could handle.

BTW, I also have no problems with hot-plugging things into USB. If it's a device that I've never tried before, I'll tail /var/log/messages to see how it gets recognised. I've had a couple that it had no clue about (no drivers), but it always seems to work and has never crashed my RPi. Now, I'm pretty certain that if I plug my iPod in as described above, I can get that to happen. Something would happen, I'm sure.

[dom]

@rustybikes
I've got a 2" hard disk extracted from a broken iPod 5G. I bought a USB to IDE caddy off eBay. After shorting the polyfuses(*), I was able to use the hard disk connected directly without a powered hub.
You need a decent power supply. Hot plugging the disk sometimes reboots the Pi, but otherwise it seemed fine

After that I tried a 2.75" laptop hard disk, though a USB to IDE caddy. This I couldn't use without using the Y shaped USB lead and giving it an extra USB power connection.

(*) I've got F1/F2 shorted. F3 has been replaced with a 1.5A fuse.
Possibly with F3 shorted the laptop disk would have worked. Possibly my board would have melted...

[rustybikes]

@dom - Well, that seals it. I'm gonna have to break out the soldering iron when I get home. I like the look of the Pi-pass (fewer connections to mess up), and that it'd be really easy to reverse the process (I have no SMD skills) if I decided to later. So, I think that's what I'll do. I'll post news of my success/disaster later..

FWIW, I found that BestBuy (of all places) has a good, cheap little 2.1A USB Wall-Wart that powers my RPi with no issue at all. I think it'll handle the additional load presented by the iPod.

[Larry_Adlard]

See also Evolving R-Pi. view topic; http://www.raspberrypi.org/phpBB3/viewt ... =63&t=9192

Having started with the Pi-pass, I decided to go for a more radical level of surgery.

I tried the using the video through RCA connector on an old CRT TV because I agreed with the principle of using an old TV. I think what everyone overlooked was that the BBC/Sinclair (I never used either) used a large character set and had only around 40 columns and 15/20 rows of characters. Fine for a BASIC program but I can't see kids used to HD TV putting up with the appalling quality of 80 columns and in 'X' only seeing a quarter of the screen in very low quality. I bought a HDMI cable and repaired a fairly new LCD TV. Having seen that I don't think composite video has any future, so the connector is pointless.

In the first distribution the alsa driver isn't active so there is no sound. It's not a priority for me and the HDMI will handle the sound eventually so the audio connector is redundant.

IFAIK the P2 connector is for test purposes and has no end user purpose. It just gets in the way.

I was wanting to wire into USB2 and the double decker connector prevents this. You can't safely remove this in one piece without risking damage to the board so it got dismantled. Forget about trying to salvage it - it's a write off. I replaced the lower one with a single connector although I'm not sure why, I don't expect to use it.

The unpowered hub got a similar rearrangement so that it fits (more or less) where the discarded components were.

Now the power supply feeds the USB hub and the R-Pi. The Pi-pass has now changed direction. It was originally intended to supply the USB from the Micro-USB connector. Now the Micro-USB is bypassed and the R-Pi is fed through it's fuses and capacitors through the Pi-pass.

A consequence of this is the USB ports and GPIO are accessible from the front and the HDMI, Power Input, and Ethernet Connector are at the back.

It isn't pretty, (lash-ups rarely are) but it points the way forward. I can't imagine that the board won't be redesigned at some stage. I know the headline 'Credit Card size' sounds nice, but slavishly following the exact format at the expense of ergononics seems to be taking things a bit too far. Even a compact keyboard is a foot wide. Alternatively, take all the connectors-off, make it Credit Card thickness and have an edge connector to all the interfaces.

PS. For those who don't believe plugging in a USB device can cause a reset - keep trying. I can plug in some devices and no problem (8Gb TDK), others always cause a hard reset but if they are present at boot then they work fine (4Gb TDK). I agree with manis404 - a larger capacitor is required. This hub has one built in. If necessary there is space to replace it with a bigger value.

[alexeames]

Flippin' 'eck. You'll be telling us you're gonna overvolt it next. That'll void your warranty y'know.

[Lob0426]

@rustybikes
I've got a 2" hard disk extracted from a broken iPod 5G. I bought a USB to IDE caddy off eBay. After shorting the polyfuses(*), I was able to use the hard disk connected directly without a powered hub.
You need a decent power supply. Hot plugging the disk sometimes reboots the Pi, but otherwise it seemed fine

After that I tried a 2.75" laptop hard disk, though a USB to IDE caddy. This I couldn't use without using the Y shaped USB lead and giving it an extra USB power connection.

(*) I've got F1/F2 shorted. F3 has been replaced with a 1.5A fuse.
Possibly with F3 shorted the laptop disk would have worked. Possibly my board would have melted...

The 2.5" HDD's generally need about 4 watts to run reliably, but they surge above this. It might be interesting to replace the 47uf capacitor getween F1 and F2 with a 220uf and see if the drive works then. Another option is the Pi-pass wire. They put a wire from the input side of the .7 polyfuse down to the USB ports under the board. It bypasses all of the polyfuses. I believe it needs a capacitor in there someplace to give the best performance. Leaves the SoC with the .7 polyfuse as protection, but no protection at all on the USB. I would cut the power connections to the board completely and let the bypass power it only. No voltage drops through the polyfuses. Too much draw and you smoke the USB connector.

[rustybikes]

Failure..

I got the Pi-Pass wired up (that was easy), but then the lower USB port fails to respond to anything. I tried 3 different keyboards before it dawned on my - "what about the top port?" - that made me feel like an idiot, but it did work. Deciding that a single-USB port wasn't gonna cut it, I removed the Pi-Pass jumper. The lower USB port still doesn't work.

FWIW, the iPod refused to spin up on either port.

I fear I've damaged my RPi. I also wonder if one of the polyfuses tripped. My plan now is to halt it, and leave it completely unpowered over night. Do you guys have any other ideas what I could do here?

[mahjongg]

Have a good nights sleep, then check everything you have done, maybe you made a stupid error (it happens to the best of us).

Also it will give the polyfuses some time to recover, as well as you.

[Lob0426]

If it does not work for you tomorrow try bypassing the polyfuse with an alligator clip onto both sides of the polyfuse. You should also look to see if you ended up with a tiny solder bridge under the board. Since it is the lower connector you may have made some excess solder run under the USB connecter as a bridge also.

[Larry_Adlard]

Failure..

I got the Pi-Pass wired up (that was easy), but then the lower USB port fails to respond to anything. I tried 3 different keyboards before it dawned on my - "what about the top port?" - that made me feel like an idiot, but it did work. Deciding that a single-USB port wasn't gonna cut it, I removed the Pi-Pass jumper. The lower USB port still doesn't work.

FWIW, the iPod refused to spin up on either port.

I fear I've damaged my RPi. I also wonder if one of the polyfuses tripped. My plan now is to halt it, and leave it completely unpowered over night. Do you guys have any other ideas what I could do here?

The USB ports are protected in two ways.

Firstly there are two diodes D10 & D11 (Fairchild BAS16) which protect the LAN9512 chip from a reverse voltage through the USB ports. The manufacturers data states that it can take up to 85 volts reverse and there is nothing which can produce such a voltage. More importantly, the maximum forward current is only 200mA (average).

We do know that some USB devices can cause a reboot, although the exact cause is unknown. It is possible that these exceed the current limit that the diode can provide. As far as I can tell, the hierarchy is that the diodes protect the chip and the fuses protect the diode. Unfortunately, fuses take time to blow and may not be fast enough to protect the diodes.

The point of the Pi-pass was to bypass the whole machine leaving the fuses and diodes intact. Despite this, there are still times when plugging in a particular device causes a reset because the capacitance on the board can't bridge the gap.

The brutal truth is that the USB ports do not meet the USB specification and everytime you plug something in there is a danger of a transient power draw damaging the diodes or something else. As detailed in my previous post this can be a small thumb drive. It's a lottery. Apart from the most basic keyboard and mouse I am coming to the conclusion that it isn't really safe to plug anything into the USB without a complex level of modifications. You will get away with it 95% of the time but the real answer is to bring the USB up to specification which as they say, "is way beyond my pay grade." I have been lucky so far, but the whole power system is so marginal I could have a problem any moment. It relies entirely on using a very stable lab bench power supply.

A typical measured resistance of the fuses on my board is 1 ohm or less. Provided you only use one power source, the difference between the Pi-pass voltage and the other side of the fuse is a small fraction of 1 volt. The Pi-pass may encourage people to be more ambitious about what they can plug into the USB ports. We have learned in the last couple of days that other aspects of the circuitry (capacitance) isn't adequate to deal with sudden demand.

I've done all the mods I intend to do. From now on the machine is restricted to programming activity only. I had hoped to backup though the USB but from now on the card will be removed and the backup done on another machine.

The only advice I can give is to measure the resistance of the fuses and check if the diodes D10/D11 are still functional. The USB data lines are directly connected and there is no protection.

[Burngate]

Not quite right.
D10 & D11 are connected to inputs on the 9512 (actually, they can be configured as outputs as well, but that isn't done on the Pi). Since those will be high impedence, not much current will flow through the diodes, even if >85v is applied to them - though enough volts would destroy the 9512!
So what those diodes achieve is to provide a 0 to the 9512 if (a) F1 or F2 has blown; and (b) whatever is connected to the USB port is trying to draw current.

As for the value of C32 is concerned, it's in parallel with C2 right on the input but after F3. So we've actually got ~267uF across the 5v supply.

If it's any interest, F1 and F2 measure ~3R on one board, ~4R5 on another, and a wopping 7R on the third.

[kta]

Is it a good idea to do the 2-wire pi-pass mod and remove F1 and F2? This way you loose usb hotplug capability but can supply more power without stressing the pcb traces.

[Larry_Adlard]

Not quite right.
D10 & D11 are connected to inputs on the 9512 (actually, they can be configured as outputs as well, but that isn't done on the Pi). Since those will be high impedence, not much current will flow through the diodes, even if >85v is applied to them - though enough volts would destroy the 9512!
So what those diodes achieve is to provide a 0 to the 9512 if (a) F1 or F2 has blown; and (b) whatever is connected to the USB port is trying to draw current.

As for the value of C32 is concerned, it's in parallel with C2 right on the input but after F3. So we've actually got ~267uF across the 5v supply.

If it's any interest, F1 and F2 measure ~3R on one board, ~4R5 on another, and a wopping 7R on the third.

That's odd. D11/D12 are shown as blocking diodes of my version of the circuit diagram (RPI00021.sbk). Maybe that's wrong (out of date). It makes more sense to have them supplying the 9512 chip.

[jbuehl]

The 9512 datasheet http://www.alldatasheet.com/datasheet-p ... N9512.html shows the diodes going the same way as the RPi schematic http://www.raspberrypi.org/wp-content/u ... s-R1.0.pdf, i.e. the arrows point away from the device.

[Lob0426]

When I looked up the LAN9512 it had several different schematics for powering the USB. Some were for active control of the power and some were passive as the RasPi uses, polyfuses.

[Burngate]

Looking at Raspberry-Pi-Schematics-R1.0.pdf sheet 3, LAN9512 pins 14 (PRTCTRL2) ana 16 (PRTCTRL3) are attached to the anodes of D11 and D10 respectively. and are sensing the state of the power pins on the ports. They are NOT supplies to the 9512. On a different circuit they can be used to control a switch for the port power, but that isn't how they're used here.
If for some reason the voltage on the positive power pin of the port drops low enough, a zero is detected on the appropriate PRTCTRL, and the system can respond accordingly. See LAN9512/LAN9512i - USB 2.0 Hub and 10/100 Ethernet Controller Datasheet - SMSC

[mahjongg]

AFAIK the diodes are simply used to detect a blown fuse, they have nothing to do with any "hot swapping mechanism".

[Larry_Adlard]

Looking at Raspberry-Pi-Schematics-R1.0.pdf sheet 3, LAN9512 pins 14 (PRTCTRL2) ana 16 (PRTCTRL3) are attached to the anodes of D11 and D10 respectively. and are sensing the state of the power pins on the ports. They are NOT supplies to the 9512. On a different circuit they can be used to control a switch for the port power, but that isn't how they're used here.
If for some reason the voltage on the positive power pin of the port drops low enough, a zero is detected on the appropriate PRTCTRL, and the system can respond accordingly. See LAN9512/LAN9512i - USB 2.0 Hub and 10/100 Ethernet Controller Datasheet - SMSC

Thanks. Following an earlier post I read " LAN9512/LAN9512i - USB 2.0 Hub and 10/100 Ethernet Controller Datasheet - SMSC" and understand how it works now.
The point I was making much earlier is that the USB voltage can't damage this port because the anode voltage on the diode is lower than the normal USB voltage so although it isn't designed for the purpose of a blocking diode, that is a side effect of it's operation. This is the first time I've investigated USB operation. Perhaps you know how it detects the presence or absence of a USB device. If you have the time, a quick description of how the USB operates might be useful for all concerned.

rustybikes has a machine down, I'm trying to figure out what might have gone wrong.

PS. You would think the spellchecker on this thing would recognise USB by now.

[rustybikes]

Failure..

I got the Pi-Pass wired up (that was easy), but then the lower USB port fails to respond to anything. I tried 3 different keyboards before it dawned on my - "what about the top port?" - that made me feel like an idiot, but it did work. Deciding that a single-USB port wasn't gonna cut it, I removed the Pi-Pass jumper. The lower USB port still doesn't work.

FWIW, the iPod refused to spin up on either port.

I fear I've damaged my RPi. I also wonder if one of the polyfuses tripped. My plan now is to halt it, and leave it completely unpowered over night. Do you guys have any other ideas what I could do here?

Confirmed. The lower USB port on my RPi refuses to acknowledge any devices connected to it. I metered the polyfuses, and as Larry mentions, they both measure at well under 1Ω. They fluttered a little between .01Ω and .08Ω, but my probes are a little big to make reliable ehough contact to get a solid reading on SMC devices.

So anyway, the polyfuses seem fine. I've triple-checked my work on the back of the board for cold-solder joints and the like, but find nothing. The Pi-Pass is simple enough that those should be rare anyway. Aside from completely removing the USB housing to check that I didn't get a bridge underneath, is there anything else I can check/do to light up the lower port again?

Please understand - I'm not about to go all negative about the platform at all. I did something to it that isn't endorsed by the Foundation, so I'm on my own. Also, this is not my only RPi, and I had intended to run it as a headless LAMP server (low volume), using a 32GB USB stick in the TOP port. So, it'll still see (a lot) of use, but knowing that there's a component on it that's broken and that I might be able to fix it... well, it's tweaking my OCD something fierce right now..

Any ideas?

[lrvick]

I too did the pi-pass mod.

I was then able to run higher powered devices like my USB-N10 off the top port, but as mentioned here by others, the bottom port became unresponsive. Also the USB-N10 must of been plugged in at boot for it to work. I can confirm that hot plugging a USB-N10 with the pipass also resets the board.

I then unsoldered the mod and found the bottom USB resumed responding normally.

I really need that second port, but I also need enough juice to power my USB-N10.

Seems the pipass is a lose/lose. You get one higher current port, that is unable to deal with hot plugging higher current devices, but you sacrifice the other port in the process.

Any other ideas?

[jbuehl]

I'm running a USB-N10 in the top port and a Logitech C250 webcam in the bottom port with the pi-pass. Maybe I'll try swapping them and see if it makes a difference. It doesn't make sense that the ports would behave differently unless there was solder bridging an adjacent pin or something.

Update: Yep, I swapped the devices and it still works.

[Lob0426]

The PiPass bridges the two USB ports power leads. There is no reason you could not just PiPass one port instead and have one high power port and one RasPi normal port. The reason the hot swap stops working is you have bypassed the 47uf Capacitor between the USB polyfuses, which is not large enough anyway. Putting a capacitor of the right type, and size, back into the circuit should fix the swap problem, I would think.

[jbuehl]

I don't think the pi-pass creates the hot swapping problem for a high current device like the USB-N10 because it just didn't work in the first place, at least for me. In my case, I couldn't get it to function until I did the pi-pass. I'm pretty sure it would reboot the RPi if you plugged it into an unmodified board, but I'm not going to undo it just to test that. I'm completely satisfied with the way mine is working now. Hot swapping the wifi device isn't something I need to do.

[Lob0426]

I don't think the pi-pass creates the hot swapping problem for a high current device like the USB-N10 because it just didn't work in the first place, at least for me. In my case, I couldn't get it to function until I did the pi-pass. I'm pretty sure it would reboot the RPi if you plugged it into an unmodified board, but I'm not going to undo it just to test that. I'm completely satisfied with the way mine is working now. Hot swapping the wifi device isn't something I need to do.

Hot swapping is not an issue for me either. I turn it off to swap stuff around. But some people may need to hot swap so we might as well investigate it.