Polyfuse F3 / RasPi power circuitry flawed


67 posts   Page 3 of 3   1, 2, 3
by mahjongg » Sat Dec 22, 2012 10:33 pm
The cost is fixed, as in cast in concrete fixed! And a perfectly suitable PSU with a mini usb connector shouldn't be more expensive than an unsuitable one, its just horses for courses.

The only way the PI would ever get a different power supply solution would be if it didn't have any influence on its price, (unless it would actually lower the price, that is). That is extremely unlikely, but stranger things have happened.

remember, the RPF wasn't even raising the price of the PI when they launched the 512MB model, that means they were unwilling to not wait for the price of the 512MB chip to drop, but instead to increase the PI's price with the few dollar that would enable them to launch the 512MB model a bit earlier. A commercial enterprise would perhaps thought it would be worth it, to increase sales, but that isn't how the RPF thinks.

The $35, price has been fixed from the beginning, and they will not increase it to make life a bit easier for someone who can afford that $10 extra for a PI that has a marginally more convenient power supply solution.
User avatar
Forum Moderator
Forum Moderator
Posts: 4962
Joined: Sun Mar 11, 2012 12:19 am
by gritz » Sat Dec 22, 2012 11:37 pm
Yes, hindsight is a wonderful thing, I suppose. In an ideal world a 5V 2A (for example) USB charger would be able to achieve both of these parameters at the same time.

Everybody in this thread has made valid points and it would be difficult / churlish to pick holes in anyone's point of view. Some random thought of mine:

1) Packaging the Pi with dedicated power supply that worked reliably and without fuss would maybe have been most desirable methodology. However, ensuring the supply of 500,000-odd psu (with whatever connector) to the four corners of the world would be a task for any small company.

2) As has been said, a charger is not a power supply! Selection of the USB charger as a psu was a pragmatic decision and has highlighted the problems associated with blindly trusting specifications and making assumptions in general. Valuable lessons for all budding engineers, if nothing else! More extensive testing may have promoted a different outcome though. DC-DC converters on the Pi would have provided the option for a range of psu and allowed separate supply lines for core and USB. Cheap is good, but cheaper is not always better - the Law of Diminishing Returns works both ways.

3) I've never found barrel connectors to be prone to intermittency, but unfortunately there are more sizes of barrel connector than one can shake a stick at - and polarity / voltage is similarly arbitrary. Here in my studio I have: 2.1mm 9V centre -ve connectors for floor fx (which is a sort-of standard), a 2.1mm 9V centre +ve for one of my synths and one of my DSP FX boxes takes a 2.1mm 9V AC supply connector. They all have to be clearly labelled! Every other piece of gear has a different supply (and a different connector), but at least those power supplies are provided by the oem and are of good quality.

4) It's all academic anyway. The Pi has a particular physical architecture and changing it now would "fragment" the platform. It's tricky...

5) If any good has come out of this rather scratchy affair it's that perhaps consumers will be better able to appreciate the "worth" and "value" of hardware items in future. A $35 computer (even one with a few wrinkles) is an absolute bargain - consider also a $500 x86 computer which has capabilities that our parents' generation could only dream about.

So don't be too surprised if that $3 USB charger you bought from some chancer on ebay burns your house down. :D
Last edited by gritz on Sat Dec 22, 2012 11:45 pm, edited 3 times in total.
Posts: 449
Joined: Sat Jan 28, 2012 2:33 am
by doveman » Sat Dec 22, 2012 11:38 pm
tufty wrote:Even the iPad charger, which is considered to be pretty much a safe bet, starts dropping its output voltage as more current is drawn; 4.4V at 2.3A, and it's pretty much a straight line - even at 1A, which is easily achievable with a micro-wifi adaptor plugged in, you're looking at ~4.8V or so. Woops. Indeed, only 3 of the chargers tested show decent voltage regulation, but that's seemingly at the expense of a whole shedload of spikes, noise and ripple. None of them are anywhere near "clean", though.


Well even with my UBEC I find the V drops as more current is drawn. For example, with nothing connected to the Pi (not even a SD card) I measure 5.22v across TP1-2. With a Sandisk 8GB Class 10 SD, 5.01v. I measured 5.12v with an old no-name, no-class 2GB SD but that might have been because it wasn't set up properly and so couldn't boot.

When I measured 4.93v across TP1-2 at 95% CPU, at the fuse before the GPIO I measured 5.05v. That was with an 8GB SD, hub (powered directly from the UBEC) with 8GB USB, RF remote receiver and dual DVB-T tuner (although this isn't installed/working yet so probably isn't drawing anything much). I'd tidied/shortened some of my wiring by this point as well, so that might have been helping. Prior to that I tested with a 16v/4A PSU going into the UBEC and measured 4.87v across TP1-2 with all the peripherals, 4.89v with the hub data cable disconnected and 4.93v with the hub power disconnected.

I think 4.93v is fine, just saying I do see a drop. Nothing like the 4.44v I got with the 5v PSU I extended the lead on though (or the 4.68v I got with the unmodified identical 5v PSU).
Posts: 163
Joined: Wed Dec 07, 2011 11:52 am
by mahjongg » Sat Dec 22, 2012 11:48 pm
Obviously no device specifies "exactly 5 V", they always use a range of acceptable voltages, with the PI that is normally the industry standard 5%. Not that the PI itself really needs it, but there could be devices connected to the PI that do. As long as the PI gets 4.75V (5V minus 5%) it should be OK.

As for voltage drops occurring, well, as no practical ways have been found to incorporate superconductors in PCB's yes, obviously you will get voltage drops an any circuitry, they are actually unavoidable. :P

That was a joke, but actually any polyfuse by its very nature also has to have a non negligible resistance, otherwise it couldn't heat up when a short circuit current ran through them, so yes, there will be a slight voltage drop occurring, but normally is should be passable. Perhaps it would be better to lower the resistance of the polyfuse even if it meant less effective protection, but the real problem seems to be polyfuses that behave abnormal, and have too much resistance to start which, and by too much I don't mean in the upper margins of their rated values, but more like they have tripped up, and have resistances ten times the normal maximum. That shouldn't normally happen, but it does, and is a point of attention for the RPF i'm sure.
User avatar
Forum Moderator
Forum Moderator
Posts: 4962
Joined: Sun Mar 11, 2012 12:19 am
by Dilligaf » Sun Dec 23, 2012 12:07 am
You call me a troll?? (Maybe it wasn't aimed at me) and then post a link to a 5.25 volt power supply which is what I was stating, that the PI needs better than the 5.0V +-5% standard. I have no problems with the micro usb connector, I do have problems with the assumption of any old power supply will work, even a supply that meets standards may not work.
Posts: 283
Joined: Wed May 23, 2012 6:48 pm
by doveman » Sun Dec 23, 2012 12:46 am
mahjongg wrote:Obviously no device specifies "exactly 5 V", they always use a range of acceptable voltages, with the PI that is normally the industry standard 5%. Not that the PI itself really needs it, but there could be devices connected to the PI that do. As long as the PI gets 4.75V (5V minus 5%) it should be OK.

As for voltage drops occurring, well, as no practical ways have been found to incorporate superconductors in PCB's yes, obviously you will get voltage drops an any circuitry, they are actually unavoidable. :P

That was a joke, but actually any polyfuse by its very nature also has to have a non negligible resistance, otherwise it couldn't heat up when a short circuit current ran through them, so yes, there will be a slight voltage drop occurring, but normally is should be passable. Perhaps it would be better to lower the resistance of the polyfuse even if it meant less effective protection, but the real problem seems to be polyfuses that behave abnormal, and have too much resistance to start which, and by too much I don't mean in the upper margins of their rated values, but more like they have tripped up, and have resistances ten times the normal maximum. That shouldn't normally happen, but it does, and is a point of attention for the RPF i'm sure.


I'm bypassing the polyfuse by going in the GPIO though. I'm using an external 2A fuse but there's a drop on the Pi (4.93v) compared to the output side of the fuse (5.05v). I haven't done extensive testing but that was what I measured on the last test with the CPU at 95% and I measured greater differences before when I was using the 5v PSU.
Posts: 163
Joined: Wed Dec 07, 2011 11:52 am
by Max » Sun Dec 23, 2012 2:19 am
mahjongg wrote:The $35, price has been fixed from the beginning, and they will not increase it to make life a bit easier for someone who can afford that $10 extra for a PI that has a marginally more convenient power supply solution.


This old blog post mentioned a more modest cost saving of $ 2 by choosing 5V micro-usb over 6-20V
Wonder what the price difference at the current production scale would be.
by tufty » Sun Dec 23, 2012 6:17 am
mahjongg wrote:
is almost certain to not, in fact, be suitable.

Why do you keep saying that! there might be a chance a random USB 1A power supply you already have laying around won't be suitable, granted, but "almost certain"?

I "keep saying that" because it's the case. The USB PSUs people have lying about are chargers, not regulated power supplies. The USB PSUs available in "the shops" are chargers, not regulated power supplies. Indeed, finding a USB PSU that's guaranteed to be "in spec" across it's rated range, as opposed to a charger, is hard. I'm not talking about lightweight chargers, here, but expensive "5V 2A" devices with one USB A port.

mahjongg wrote:Being forced to also add a regulator on board that can do at least 12V to 5V @1A

mahjongg wrote:Also being forced to add protection against reverse polarity (not a big deal but in practice would prevent direct 5V powering with a barrel connector)

We had these before. Yes, they cost a bit of money, but the only thing you really had to worry about WRT PSUs was that they were in *range* vis a vis voltage and could source enough current.
mahjongg wrote:being forced to add a very big electrolytic capacitance, to prevent large ripple on the input

DC input, not AC. Yes, you need *some* capacitance, but you're not taking over the AC-DC conversion. As it stands, you're delegating all of the above to the PSU, and just trusting that it "gets it right". If you feel a large capacitance is required to eliminate ripple from some hypothetical PSU putting out somewhere between 6V and 12V, it's also required *now*. Unless it's barrel-connector-induced ripple, of course.
mahjongg wrote:all this means, not being able to keep the dimensions of the PI as small as it is now, which also add costs and makes the PI less attractive.

The dimensions of the proto were, I believe, identical to the version shipping now.
mahjongg wrote:Barrel connectors are also less reliable in letting through power without interruption during vibrations.

Not an issue I've seen myself, but hey, anecdote is not evidence. Fit it with screw terminals, then.
mahjongg wrote:less power efficient (hotter)

Ummmm - nope. The original switchmode regulator on the Pi was massively more efficient than the linear stuff we now have. Yes, I understand *why* we have nastly inefficient linear stuff now, but you *cannot* make that claim without looking ridiculous.
Posts: 1364
Joined: Sun Sep 11, 2011 2:32 pm
by stormy1 » Sun Dec 23, 2012 6:27 am
If your in the US the Kaito PST-15U, also sold as IQ PST-15U is a rock solid 2a 5v supply.
Good supplies are hard to find, most are junk.
Posts: 60
Joined: Fri Jan 06, 2012 3:44 am
by jamesh » Sun Dec 23, 2012 10:58 am
I use my Kindle charger and it works fine.

My laptop has a barrel connector, I keep having to replace the socket on the motherboard.

Perhaps the Foundation/someone should produce a barrel to USB adapter...??? I've actually got one on my desk at work for a Nokia phone for when they moved from barrel to USB.
Soon to be unemployed software engineer currently specialising in camera drivers and frameworks, but can put mind to most embedded tasks. Got a job in N.Cambridge or surroundings? I'm interested!
Raspberry Pi Engineer & Forum Moderator
Raspberry Pi Engineer & Forum Moderator
Posts: 11576
Joined: Sat Jul 30, 2011 7:41 pm
by redhawk » Sun Dec 23, 2012 12:13 pm
My 10 year old Toshi*a laptop also has a barrel connector for a 19v@16.3A PSU and I've never damaged it yet but that's probably because I look after my equipment. :)
To honest I don't think there is really a suitable connector for the PI each design has it's strength and weaknesses the fact that you're tugging and inserting all the time the metal contacts would eventually lose their grip.
That's actually happened to one of my USB extender cables it has a wonky end so I have to make sure it cocked over in the right direction.

Perhaps Adafruit could sell a PI PSU adapter one that accepts a wide range of input voltage from 3.7v li-po to 12v car batteries and dodgy unfiltered unregulated wall adapters.
All it would need is a step-up / step down buck converter, some smoothing capacitors, rf chokes and perhaps a low battery warning light.

Richard S.
User avatar
Posts: 3328
Joined: Sun Mar 04, 2012 2:13 pm
Location: ::1
by mahjongg » Sun Dec 23, 2012 2:44 pm
tufty wrote:I "keep saying that" because it's the case. The USB PSUs people have lying about are chargers, not regulated power supplies.

Then they shouldn't use those as I have pointed out multiple times, they should only use a PSU that isn't a charger, or is one of the GSMA’s Universal Charging Solution types, or similar. What I mean is the type of power suplies with an USB-A connector like this one:
Image
If you have one of these laying around you have a good chance they will work.
This was the " random USB 1A power supply you already have " I was talking about.
If you don't have one like these you can try whatever you have that has a micro USB (except this in the wiki that are mentioned as not working) but don't be surprised if they don't work!
We had these before. Yes, they cost a bit of money, but the only thing you really had to worry about WRT PSUs was that they were in *range* vis a vis voltage and could source enough current.
Its the "cost a bit of money" bit thats the problem, it's its not much money perhaps, but nevertheless enough so that the $35 price couldn't be kept.
DC input, not AC. Yes, you need *some* capacitance, but you're not taking over the AC-DC conversion. As it stands, you're delegating all of the above to the PSU, and just trusting that it "gets it right". If you feel a large capacitance is required to eliminate ripple from some hypothetical PSU putting out somewhere between 6V and 12V, it's also required *now*. Unless it's barrel-connector-induced ripple, of course.
regulated 5V supplies do not output ripple, thats why no more than 220uF is on board, and as we have seen that capacitor is sometimes even optional.
Yes, maybe you don't always need 4700uF but most circuits I have seen that use a barrel connector use it, not only to ensure less ripple, but also to smooth over interruptions caused by poor barrel connector conductivity and vibrations. But its a minor point I grant you that.
The dimensions of the proto were, I believe, identical to the version shipping now.

No the alpha board that had the switcher was quite a bit bigger, now I believe that such a switcher itself could be squeezed in about 1.5 square cm, but the barrel connector itself also takes more space!
Not an issue I've seen myself, but hey, anecdote is not evidence. Fit it with screw terminals, then.
Yes, there are some specialised barrel alike connectors with screw terminals, but the PSU you will have or buy won't have them.
Ummmm - nope. The original switchmode regulator on the Pi was massively more efficient than the linear stuff we now have. Yes, I understand *why* we have nastly inefficient linear stuff now, but you *cannot* make that claim without looking ridiculous.
You forgot that even with a barrel connector and switching regulator you would still need the other linear regulators, so all the heat coming off the additional switcher is 100% extra heat.

Also, all the time you are arguing that the current solution is so bad, you seem to forget that we now have a million units running fine. Yes, a consumer type of user would possibly be better off with a conventional solution, from a purely technical point of view, but the PI isn't a consumer product.

I grant you that a purely commercial product wouldn't have dared to use this solution, unless the PSU was included of course, but the PI isn't a commercial product, and some experimenting with finding a proper PSU shouldn't be an issue.

Again, if the RPF sees fit to change this solution to a more classical one, without raising the price, and dimensions of the PI (for a moment ignoring other factors, such as heat, and the impossibility of bypassing this switcher in special use cases) then perhaps they will. Its more likely though they will try to mitigate whatever problems there still are with the current solution. In fact I offered some solutions to them, and have been pleading with them to implement these. But its up to them to use my suggestions.
They do indeed read these forums carefully you know.
User avatar
Forum Moderator
Forum Moderator
Posts: 4962
Joined: Sun Mar 11, 2012 12:19 am
by doveman » Sun Dec 23, 2012 4:46 pm
mahjongg wrote:
tufty wrote:I "keep saying that" because it's the case. The USB PSUs people have lying about are chargers, not regulated power supplies.

Then they shouldn't use those as I have pointed out multiple times, they should only use a PSU that isn't a charger, or is one of the GSMA’s Universal Charging Solution types, or similar. What I mean is the type of power suplies with an USB-A connector like this one:
Image
If you have one of these laying around you have a good chance they will work.
This was the " random USB 1A power supply you already have " I was talking about.
If you don't have one like these you can try whatever you have that has a micro USB (except this in the wiki that are mentioned as not working) but don't be surprised if they don't work!


Probably generally true but I've got a Samsung adapter with a USB-A connector that's only 4.4v@400ma according to the label, so that probably wouldn't work. That came with my camera. It does say Travel Charger on the label and is a lot smaller than the one in that picture, so it's probably not the sort of thing you had in mind and anyone who knows what the Pi requires wouldn't try and use it but I'm just pointing out that having a USB-A connector doesn't mean it's suitable.
Posts: 163
Joined: Wed Dec 07, 2011 11:52 am
by mahjongg » Sun Dec 23, 2012 5:24 pm
Outputting only 4.4 Volt is an infringement on the USB standard. But you are right that exceptions to the rule will always occur, and no guarantee of suitability can thus be implied from having an USB type A port. But yours is the exception to the rule. In general having an USB type A port (and specs like 5V 1A) would give you a better chance than using a tiny GSM charger with a fixed cord.

Common EPS chargers (the new standard for chargers) are now in the field for about a year, if you use one of these they should be able to power the PI just fine (they do 850 mA at 5V +- 0.25V). They do have to have an USB type-A port too.

So look for a charger with an USB-A connector, and these specifications or better (more current, up-to 1800mA). It may carry the text "common EPS" or "universal charger" which may indicate they are real universal chargers. The chinese equivalents of these should also be suitable.

I haven't been able to find out yet if they also carry a specific logo, I'm working on it.
User avatar
Forum Moderator
Forum Moderator
Posts: 4962
Joined: Sun Mar 11, 2012 12:19 am
by rurwin » Sun Dec 23, 2012 6:02 pm
I think the comon EPS is standard EN 62684:2010, but I can't find an online version; everyone wants paying for it.

The USB Battery Charging Specification v1.2 appears to be insufficient, although I'd be surprised if a conforming supply failed to power a Pi except under exceptional conditions. It guarantees that a dedicated charger will supply 500mA at at least 4.75V, but allows the voltage to drop to as low as 2V at higher current. On the other hand a Charging Downstream Port (that's a laptop socket with a red tongue IIRC) will supply 1500mA at at least 4.75V.

Standard IEC 62684:2011, which is supposed to be the same as the EPS, only specifies 4.75V at the socket of the charger. It then allows a further 0.125V drop in the cable, more if you are drawing over 500mA. (A fixed cable is better, since then you have to get 4.75V at the micro-B plug.) It does however guarantee 4.75V up to the full rated current, and if that current is considerably higher than the RaspPi will draw it would be unlikely to only supply minimum voltage to the Pi.
User avatar
Forum Moderator
Forum Moderator
Posts: 2903
Joined: Mon Jan 09, 2012 3:16 pm
by W. H. Heydt » Sun Dec 23, 2012 10:21 pm
Dilligaf wrote:(Maybe it wasn't aimed at me) and then post a link to a 5.25 volt power supply which is what I was stating, that the PI needs better than the 5.0V +-5% standard.


I posted a link to the Adafruit 5.25v, 1A power supply. So far as i know, that impetus behind setting the out put to 5.25v is to compensate for the myriad of really poor quality cables, so as to insure that the Pi get power within spec.

I'm planning to order a fistful to Adafruit power supplies next month and "standardize" on them for my Pis.
Posts: 1374
Joined: Fri Mar 09, 2012 7:36 pm
Location: Vallejo, CA (US)
by mahjongg » Sun Dec 23, 2012 10:38 pm
Yes, and if I may point out, their supply is a 5.0V +-5% device!

I applaud their effort.
User avatar
Forum Moderator
Forum Moderator
Posts: 4962
Joined: Sun Mar 11, 2012 12:19 am