Design flaw? Polyfuses for USB current limiting

[W. H. Heydt]

...20-20 eyes!

That leaves me out....

[SirLagz]

Same haha...
Guess I'll need to peg it down to solder it or something...

[W. H. Heydt]

Just another data point....

I received two Pis today, ordered (Allied) on 5 July. Both have polyfuses, not zero ohm resistors.

There appear to still be boards with polyfuses working their way through the supply chain...

If it still an issue in 7 weeks, I'll let you know.

[Kestrel]

As a first post...

I received my Pi earlier this week, and upon seeing this thread and examining the PCB, it appears that the green polyfuses are gone and have been replaced by two black 'jumpers' with '0000' scribed on the top of each. This doesn't match with the single large '0' shown earlier, but I believe them to be the same. No access to a voltmeter (my tools are a few hundred miles away!), so I'm unable to verify a voltage drop through the USB.

I actually headed out to purchase a keyboard and mouse for the system last night, but the local shop only really carried a couple of options like the Logitech MK360 wireless mouse & keyboard set. This keyboard is famous for having sticky keys, though it seems like my revised board may actually not suffer any of the issues? I may purchase one a little later in the day to test it out.

[sorvad]

When I first got my Pi I spent a long time trying to get it to work correctly with a powered hub and wifi dongle and it seemed really unpredictable, eventually I gave up thinking I was being a bit thick and left the Pi on a shelf for a while, but still frustrated I couldn't get it working as I needed it. It seems I was a possible victim of this issue. I want to run my Pi using 2A powered hub which also powers the Pi, which is the best mod I need to do to my board (and pictures would really help here) to eliminate these USB issues ?

[mahjongg]

The best solution would be to do what the new Rev2 board does, and to completely short out the USB polyfuses. Like so (picture is from the "USB current boost (solved)" thread):


With this patch you can backfeed power from the hub into the PI without any issues, so no more loopback power cable needed to feed the PI from a hub port.

[Ed Raket]

The best solution would be to do what the new Rev2 board does, and to completely short out the USB polyfuses. Like so (picture is from the "USB current boost (solved)" thread):


With this patch you can backfeed power from the hub into the PI without any issues, so no more loopback power cable needed to feed the PI from a hub port.

This mod Probably couses hotplug problems (with the Edimax wifi adapter (rtl8188cus) for one...) becouse of the voltagedrop (powersurge) on the PCB when inserting the adapter. Could you test this with whatever USB stuff you got? (for me 1-Ohm resistors are still making all USB stuff hotplug )

Allso read;
http://www.raspberrypi.org/phpBB3/viewt ... &start=200

[mahjongg]

Ed Raket (Seems to be a Dutch name, as "Raket" is Rocket in Dutch) may be right about the hotplugging problem, but for backfeeding the PI from a hub 1.0 Ohm it is still much more than the 0.18 Ohm of the input polyfuse, so it could cause a significant drop for the PI itself, how much of a bother the voltage drop (about 0.4 Volt) will be for the PI remains to be seen. But note that the rev2 board seems to have no resistance at all anymore, (unless they hid tiny 1E resistors on the board somewhere, not seen the schematics yet) and no easy way to patch the board to get it!
The motto seems to be that any script relaying on plugging in any USB device should do so with the PI turned off! Then after a reboot continue where left off.
I know that isn't an ideal situation, but its a compromise the RPF seems to think we can live with.

That doesn't mean you have to do the same, so what you could do is to patch just one of the two polyfuses with 1 Ohm, and leave the other a wire. This way you can see what effect having 1E while backfeeding the PI has. F1 is used for the top USB connector.

[Ed Raket]

Ed Raket (Seems to be a Dutch name, as "Raket" is Rocket in Dutch) may be right about the hotplugging problem, but for backfeeding the PI from a hub 1.0 Ohm it is still much more than the 0.18 Ohm of the input polyfuse, so it could cause a significant drop for the PI itself, how much of a bother the voltage drop (about 0.4 Volt) will be for the PI remains to be seen. But note that the rev2 board seems to have no resistance at all anymore, (unless they hid tiny 1E resistors on the board somewhere, not seen the schematics yet) and no easy way to patch the board to get it!
The motto seems to be that any script relaying on plugging in any USB device should do so with the PI turned off! Then after a reboot continue where left off.
I know that isn't an ideal situation, but its a compromise the RPF seems to think we can live with.

That doesn't mean you have to do the same, so what you could do is to patch just one of the two polyfuses with 1 Ohm, and leave the other a wire. This way you can see what effect having 1E while backfeeding the PI has. F1 is used for the top USB connector.

First of all "Raket" is verry Dutch indeed, well spotted!
Secondly i never thought about "backfeeding" the Pi, i can imagine the 1-Ohm / 0.4V loss could be a problem then... Indeed, MrEngman could adjust his script with an extra reboot for the new or unmodded Pi's and your "single port mod" is allso a good idea
Im allso verry currious how they adjusted the new boards... maybe the RPF have included an other puzzel for us to solve?

That is not a bug..., it is a feature!

[MrEngman]

Ed Raket (Seems to be a Dutch name, as "Raket" is Rocket in Dutch) may be right about the hotplugging problem, but for backfeeding the PI from a hub 1.0 Ohm it is still much more than the 0.18 Ohm of the input polyfuse, so it could cause a significant drop for the PI itself, how much of a bother the voltage drop (about 0.4 Volt) will be for the PI remains to be seen. But note that the rev2 board seems to have no resistance at all anymore, (unless they hid tiny 1E resistors on the board somewhere, not seen the schematics yet) and no easy way to patch the board to get it!
The motto seems to be that any script relaying on plugging in any USB device should do so with the PI turned off! Then after a reboot continue where left off.
I know that isn't an ideal situation, but its a compromise the RPF seems to think we can live with.

That doesn't mean you have to do the same, so what you could do is to patch just one of the two polyfuses with 1 Ohm, and leave the other a wire. This way you can see what effect having 1E while backfeeding the PI has. F1 is used for the top USB connector.

First of all "Raket" is verry Dutch indeed, well spotted!
Secondly i never thought about "backfeeding" the Pi, i can imagine the 1-Ohm / 0.4V loss could be a problem then... Indeed, MrEngman could adjust his script with an extra reboot for the new or unmodded Pi's and your "single port mod" is allso a good idea
Im allso verry currious how they adjusted the new boards... maybe the RPF have included an other puzzel for us to solve?

That is not a bug..., it is a feature!

I've had enough problems with the script, and there are still a few outstanding, never mind not having to work out how to restart half way through and then continue on after a restart. Although I have been looking at it. Rather than a reboot, a shutdown and then power off, plug in the USB and then restart is probably more practical. Gives people more time to find the wifi adapter and plug it in.

However, If someone can point me in the right direction as to how to run the script after a restart it would be an interesting exercise I suppose.

As it so happens I ordered a new Pi from Farnell on Thursday and it arrived this morning - Rev2 made in the UK. So I can test it with the new hardware.

Just noticed one thing that's not going to help power stability is the main fuse being reduced to 700mA. As far as I'm concerned not a good idea at all.

MrEngman

[hippy]

Just noticed one thing that's not going to help power stability is the main fuse being reduced to 700mA. As far as I'm concerned not a good idea at all.

My understanding is that all R-Pi have been fitted with 700mA main fuses so nothing has changed there; it was just that the schematics showed 1.1A

[mahjongg]

It seems nobody really understands these parts.
The change to Rev2 did not affect F3!
It is and has been a fuse that trips at 1.5 A (when in room temperature, as a polyfuse is temperature sensitive).

The fuse that is used is a Raychem Tyco MINISMDC075F-2


It has an IH of 0.75A (750mA) and an IT of 1.5A (IT is normally double IH)

IH is the "holding current", the guaranteed safe current through the device.
But IT is the more important "tripping current", and its 1.5 Amperes (at room temperature).

Normall "glass tube" fuses would only mention the tripping current, (which is more or less independent of the surrounding temperature) so an 1.5A T fuse would be a slow fuse with a tripping current of 1.5A.

Also note that F3 has a "minimum" (read "normal") resistance" of 0.1 Ohm, which can rise to a maximum of 0.45 Ohm after tripping and recovering (several times).

[MrEngman]

It seems nobody really understands these parts.
The change to Rev2 did not affect F3!
It is and has been a fuse that trips at 1.5 A (when in room temperature, as a polyfuse is temperature sensitive).

The fuse that is used is a Raychem Tyco MINISMDC075F-2


It has an IH of 0.75A (750mA) and an IT of 1.5A (IT is normally double IH)

IH is the "holding current", the guaranteed safe current through the device.
But IT is the more important "tripping current", and its 1.5 Amperes (at room temperature).

Normall "glass tube" fuses would only mention the tripping current, (which is more or less independent of the surrounding temperature) so an 1.5A T fuse would be a slow fuse with a tripping current of 1.5A.

Also note that F3 has a "minimum" (read "normal") resistance" of 0.1 Ohm, which can rise to a maximum of 0.45 Ohm after tripping and recovering (several times).

Well unfortunately I do understand these devices and have used them often during my career as a hardware designer. The hold/trip currents don't bother me. But the resistance is another thing altogether.

[mahjongg]

Agreed, but the point I tried to make that calling this a "750mA" fuse, or even claims that it has been reduced to a "700mA" fuse, are unfounded.
F3 has always been what it was, and has not been a big factor in the problems, most report it to be about 0.18E.
The almost doubling of the "off-factory" 0.1E is explainable because the act of soldering it to the board has had had a negative effect on it.

At, say 0.18E, it would mean that at a normal current drop of 600mA (an average 400mA for the SoC and 9512, and twice 100mA for two not really low power USB devices) a 0.2E resistance will cause a 0.12V drop, which the PI can take easily (as its mostly powered by the 3V3 low drop regulator which doesn't mind a 0.12V drop at its input), and most of the USB devices, including non low current ones would also not mind. 0.12 volt is a big difference from the 0.6 Volt each original USB polyfuse would induce for a 100mA device.

Ideally (obviously) no voltage drop from the PSU to the end-of-the-power-chain-device should occur, but we are still living in the real world.

But as I said, this has been what F3 was about all the time, no change from rev 1.

[sorvad]

Thanks for the picture mahjongg, I will implement this fix on my board and re-visit the wireless set up.

[Lob0426]

Sorry @mahjonng, but my first RasPi had a 700ma at F3. So some Rapi were manufactured with a slightly lower rated F3. RasPi #1 came from Farnell in June. All of the newer ones I have seen have 750ma at F3. This really has little or no effect on its performance. I replaced it with a 1.1A anyway.

[LilSnoop40]

hi, i have the rev 2 pi board. i don't want to use any other device plugged into the usb ports on it but my usb wifi dongle. http://www.amazon.com/gp/product/B005CL ... 05_s00_i00
i noticed that when this is plugged in and i am viewing certain addon's the raspberry pi reboots. i was to to get a usb hub or short out the polyfuse. if i short out the polyfuse would that fix the reboot issue when the wifi dongle is plugged in.

thanks!

[abishur]

hi, i have the rev 2 pi board. i don't want to use any other device plugged into the usb ports on it but my usb wifi dongle. http://www.amazon.com/gp/product/B005CL ... 05_s00_i00
i noticed that when this is plugged in and i am viewing certain addon's the raspberry pi reboots. i was to to get a usb hub or short out the polyfuse. if i short out the polyfuse would that fix the reboot issue when the wifi dongle is plugged in.

thanks!

The best way to find out the answer to that question is two-fold. First off how many mA does your wifi dongle use? While it shouldn't be a problem on the usb port side, if it take a high amount then maybe it's causing the entire current for the pi to exceed the 1A polyfuse under load conditions. (though that should cause the pi just to shut off, it should create a reboot situation)

Second test the power on the pi. If you could check the voltage at TP1 and TP2 we could see if was running a little low.

[LilSnoop40]

hmmm i am not really that good with a meter. what do i set it to, to check what your asking. i know i have to take the board out of the case to test the t1 and t2. please advise


thanks

[mahjongg]

http://elinux.org/R-Pi_Troubleshooting# ... r_problems

[pluggy]

The best solution would be to do what the new Rev2 board does, and to completely short out the USB polyfuses. Like so (picture is from the "USB current boost (solved)" thread):


With this patch you can backfeed power from the hub into the PI without any issues, so no more loopback power cable needed to feed the PI from a hub port.

Sheesh. with the flak I've had from mods on here, its a bit strong you're using my photo on my server without permission...........



Only kidding.....

[Rasoul]

I encountered what may be a problematic design flaw in the Raspberry Pi Model B.
In conclusion:
1. If you can't get your wifi dongle or other USB device to work with the Raspberry Pi, you might be able to fix it by bridging F1 and/or F2
2. Current limiting on the USB ports (its implementation or even having it at all) may need to be rethought.

First of all, I do agree with your analysis of the schematic. In addition to your conclusion, in the schematic it can be observed that the fuse F3 (miniSMD) is rated as 1.1A, which is also problematic. To make the power of USB ports meet the standard spec, I would replace F3 with a 2.5A fuse, and USB fuses (F1 and F2) with 630mA types. This simple, low-cost modification prevents one from buying a USB Hub (~30 USD) just to compensate for the design flaw.

[mahjongg]

@rasoul, regarding the F3 fuse I Agree, but the RPF people seem to think that, true or not, such a high rating would be no better than no protection at all! And they do have a point!
As for F1 and F2 I'm afraid you make the same mistake as the original designers, in that you seems to assume that a fuse rated higher than the highest expectant current would have no effect on the protected devices. Unfortunately that isn't true!

Even a 630mA type would still have an unacceptable "cold resistance", especially if you connect USB devices that are rated more than 100mA, and therefore generally use components rated 5V ± 5%. The resistance quickly would cause a greater than 0.25V drop.

Anyway, these fuse were already removed from the latest PI's months ago, in the revision 2 PI's.

I only wished they had replaced them with a substantial, but unharmful resistance. For example 0.1 Ohm. This would have helped against the hot-plug rush-in current problems the revision 2 PI's now have as a result of simply removing F1 & F2. Especially if they had used the money saved by replacing the fuses with resistors to use a larger 5V decoupling capacitor near the USB ports. But perhaps the money saved went to extra drill holes for ATM testing and for P5 (holes cost money too).

[FrTedCrilly]

At what temperature will the polyfuse go?
I left my Pi on a home router which was hot. And after testing with multimeter it appears to be blown

[mahjongg]

there is no "temperature" at which the fuse "will blow", but increasing ambient temperature will lower the current at which the thermal runaway could take place that ends in the condition we refer to as "blown".

That is why polyfuses have two specifications, not just one, like a normal fuse. One is the current that is guaranteed not to "blow" the fuse under whatever reasonable ambient temperature, and the other is the current that the fuse is guaranteed to blow with under whatever reasonable ambient temperature there is.

when thermal runaway is in progress the fuse will probably heat up shortly to several hundred degrees Celcius.