EEVblog Electronics Community Forum
Electronics => Projects, Designs, and Technical Stuff => Topic started by: XaviPacheco on October 05, 2018, 01:39:16 pm
-
My 5V power supply is losing about 0.05V in the wiring when I connect the load. The board that needs the 5V is 8 feet away from the power supply, and this board is low-voltage sensitive, so it needs strictly 5V. How can I overcome this? I have a LM2596 little adjustable module, which I'm not sure if it can help regulate the power.
-
Is it really sensitive to a 50 mV drop?
If so, you will probably encounter other issues IMO. Adding a step-up regulator? Well, ensuring that its output will never drop by 50 mV under all load conditions seems a bit optimistic?
AFAIK, the LM2596 is a step-down regulator so it does the exact opposite (lower the input voltage), so it could only be worse.
If you absolutely need a very stable and precise 5V, your best bet would probably to add a step-up regulator that provides a reasonable margin (eg: 5V PS -> 5.5V or 6V) and then followed by an LDO down to 5V. You'll lose a bit in efficiency but that's the price to pay.
-
As mentioned, for a critical precision like that, you have 2 choices - Supply at 5.5-6.0V and an LDO .. OR .. add a sense wire and correct the 5.0V from the load.
-
Another vote for increasing the raw DC flowing across the long wire run, and regulating locally. Lots of reasons why that is a FAR better solution.
-
We're talking 1% here. An application that's that sensitive is incorrectly designed IMHO. If you need that kind of precision, it should come from a local reference and not be dependent on supply voltage losses.
-
We're talking 1% here. An application that's that sensitive is incorrectly designed IMHO. If you need that kind of precision, it should come from a local reference and not be dependent on supply voltage losses.
Definitely. If you're working with 5V digital logic you probably have at least 10x that in specified tolerance and if you're working with analog, unless 5V is the very minimum your stuff needs, are all your elements so precisely configured and characterized that 1% on the rail is going to make any difference?
-
I too recommend local regulation inside the device if you need precise supply. Also be aware that in most cases a generic power supply will vary 5% to 10%. (Obviously a lab power supply behaves better than that.)
Maybe you should analyse your circuit to find out what the sensitive bits are? Some precision analog? Could you run these precise bits from a 4V regulator?
Also, if 1% supply voltage change is an issue, what happens to your design when the _temperature_ changes?
-
If this is just for running the load in your lab, then there are lab PSU's available that have sense inputs. So you basically connect the load to the power supply, and then run a pair of wires for sensing the voltage at the load. Such PSU'are usually high amp capable, because that's when voltage drop across a wire can get significant.
But if you need a solution that needs to be run standalone, even outside your lab, then the best solution would be to give the load a higher voltage and locally regulate it down to 5V.
-
If your LM2596 is one of those little premade modules you can find all over eBay/Aliexpress then the LM2596 is counterfeit. It does work but can't handle the advertised current and switch at a far lower frequency. The ripple coming out of that will be far worse than the 0.05V drop you're currently complaining about so don't use that to regulate a higher voltage down to 5V for your board.
-
And also don't use any standard voltage regulator... Many of them have only a +/- 100mV or even a +/- 200mV precision.
-
As mentioned above, if the 50mV drop is an issue the circuit is poorly designed.
If it's not your design, then there's little you can do, other than use a precision supply. Another option in addition to onboard regulation, is remote sensing, which uses (an) extra feedback wire(s) to compensate for the voltage drop along the cable.
-
Usually the problem is surge current requirements which cause the voltage to momentarily dip.
In any event, a boost switching regulator in one form or another will solve the problem if the input voltage cannot be more tightly controlled.
-
I find it interesting that the OP hasn't responded to any of the questions, comments, or suggestions offered on this thread.
We don't know anything about the application. All we know is that he's complaining about a supply rail that is actually tighter (1%) than the specs for many linear regulators (!). As someone else noted, maybe what he really needs is an onboard precision reference and his rails don't actually matter that much. Or maybe his circuit is misdesigned. Or maybe he doesn't actually understand what's going on. Maybe he believes when someone specifies "5VDC" there's no margin. The voltage drops when he connects the load - but is it due to the load itself? Or the current drawn by the load, which then drops all of 50mV in his EIGHT FEET of wire?
The list of unknowns goes on....
-
The load, whatever it is, is either
a) completely broken by design beyond unusable
b) poorly documented, or,
c) you don't understand the documentation properly.
All circuits with such sensitivity to exact input voltage always use local regulation on-board, and are specified with meaningful margin (e.g., they would specify 7V as the minimum input voltage, not 5V, and you would be stepping up from 4.95V to 7V in this case.)
There is no other way around this than to fix the load.
-
Lots of great practical solutions to the possible problem presented here. The OP hasn’t been back to clarify any of the many questions others have asked for 3 days so it couldn’t have been a very important problem, or they found they made some mistake and the problem wasn’t what they thought it was so they just stopped looking at this thread.
-
Hello all,
Had problem with my internet connection these days and couldn't report back. I really appreciate all of the suggestions you've given. Let me clarify the whole scenario:
I have retaken the project which I had dropped about two weeks ago. Actually, I shouldn't have written this post. Instead, I should go back to this one: Issue with Raspberry Power Supply (https://www.eevblog.com/forum/projects/issue-with-raspberry-power-supply/msg1837253/#msg1837253) where I give more details of the project. My bad for making such a mistake. In that post, it's clear that a solution is given. I know that I can supply a greater voltage and regulate it locally to the Pi. I really was looking for a possible solution before I redesign the entire project again. But I won't lose more time. Now that I have retaken the project, that's what I'll do.
At this point, I didn't understand why the Pi was complaining at such little voltage drop. The IC sensing the voltage triggers at around 4.7V, so it doesn't make sense it to trigger at 4.95V. Ok, but I have new data:
Measuring directly to the Pi, I get 4.4V, now makes sense. The 4.95V measure comes when I measure at the end of the cable directly WITHOUT THE PI connected. When I connect the PI, the voltage doesn't drop at the main board, but it drops at Pi's terminals. So, the PSU can't meet the PI's demand, I think.
-
Measuring directly to the Pi, I get 4.4V, now makes sense. The 4.95V measure comes when I measure at the end of the cable directly WITHOUT THE PI connected. When I connect the PI, the voltage doesn't drop at the main board, but it drops at Pi's terminals. So, the PSU can't meet the PI's demand, I think.
Sounds like really low quality cable with hair thin conductors. Or/and junk psu. If there is like 4.7+ V at the input of Rasberry Pi (under heavy load) there shouldn't be any issue.
-
Sounds like really low quality cable with hair thin conductors. Or/and junk psu. If there is like 4.7+ V at the input of Rasberry Pi (under heavy load) there shouldn't be any issue.
I have changed the cable. Now, I don't get losses at the cable without load. I get a big drop when I connect the Pi. I think it's that little PSU. Check the datasheet: https://www.mouser.do/datasheet/2/468/RAC20-K-1369798.pdf (https://www.mouser.do/datasheet/2/468/RAC20-K-1369798.pdf)
-
Update: I've changed the cable again and it works well now. Apparently, I have serious issues identifying a good cable for my application.
-
I've found the RPi (especially the Pi2 and Pi3) to be very finicky with 5V power supplies. That can be a huge pain.
This RECOM module is fine power-wise, the problem is the transient response. How long is your cable?
-
I've found the RPi (especially the Pi2 and Pi3) to be very finicky with 5V power supplies. That can be a huge pain.
This RECOM module is fine power-wise, the problem is the transient response. How long is your cable?
My cable is 7 feet.
-
Yeah that's long enough to cause issues on transients with a Pi if the cable is low-quality. If you're using an off-the-shelf micro-USB cable, I've found a lot are VERY bad for power delivery. An alternative is to power your Pi not using the micro USB plug. You can either solder wires instead or power it from the GPIO connector (which is what some RPi PSU "hats" do).
-
You can either solder wires instead or power it from the GPIO connector (which is what some RPi PSU "hats" do).
Yeah, that's what I did, power it from the GPIO connector.
-
You can either solder wires instead or power it from the GPIO connector (which is what some RPi PSU "hats" do).
Yeah, that's what I did, power it from the GPIO connector.
Oh ok, so you can use any kind of cable then. Use proper wire diameter and you shouldn't have a problem.
Just be aware that by powering the Pi from the GPIO connector, you're bypassing the on-board fuse on the 5V rail. So I'd suggest adding an external fuse.
-
Why not bring mains voltage 7 feet and use the DC voltage conversion next to the Pi? Voltage will drop on long cables. Google voltage drop calculator to estimate your needs.
EDIT: I edited the original post based on a correction noted by mikeselectricstuff
-
You can either solder wires instead or power it from the GPIO connector (which is what some RPi PSU "hats" do).
Yeah, that's what I did, power it from the GPIO connector.
That's a bad idea, you are bypassing input protection.
(https://www.eevblog.com/forum/projects/how-to-step-up-4-95v-to-5v/?action=dlattach;attach=542360;image)
-
You shall drop idea of 5V PoE because rPI needs up-to 1A or so, voltage drop is too big on Ethernet cable, so you need additional, big& fat power cable. If you need big fat cable - then why don't you just run AC mains to rPI and place power supply right to it? Other option would be to run 24V or 12V PoE and convert to 5V locally using 5V DC-DC switching supply.
-
I've a project in where I'm powering the PI's direcly from the GPIO connector. I've set up the voltage regulators at 5.1V on those boards and that seems to work wonders in avoiding the low voltage warnings. IIRC Maxium specified voltage of the PI is 5.25 so don't up the voltage too much..
-
Buy a PoE splitter with a micro-USB plug on it (https://www.amazon.co.uk/dp/B01H37XQP8).
That's a bad idea, you are bypassing input protection.
If you're careful and don't overload it or connect it backwards then you're fine to do that. It is mentioned as part of the HAT spec. The MOSFET circuit is just to stop backfeeding 5V on the micro-USB power socket. They removed it entirely on the 3B+ and it never existed at all on the Zero boards.
-
Where's PoE and Ethernet cable suddenly come from? I can't see it mentioned previously in the thread until ogden introduced it. :-//
-
I've a project in where I'm powering the PI's direcly from the GPIO connector. I've set up the voltage regulators at 5.1V on those boards and that seems to work wonders in avoiding the low voltage warnings. IIRC Maxium specified voltage of the PI is 5.25 so don't up the voltage too much..
Same here. On board 5v regulation fed over 50mil 2oz trace to header connected to gpio caused warning. 5.1v, no warning. 3b+ doesn't have that issue though, the 3b did.
-
Where's PoE and Ethernet cable suddenly come from? I can't see it mentioned previously in the thread until ogden introduced it. :-//
OP has another thread (https://www.eevblog.com/forum/projects/issue-with-raspberry-power-supply/msg1836968/#msg1836968). They're using the spare pairs to feed 5V to the Raspberry Pi at the end of the cable.
-
Where's PoE and Ethernet cable suddenly come from? I can't see it mentioned previously in the thread until ogden introduced it. :-//
OP has another thread (https://www.eevblog.com/forum/projects/issue-with-raspberry-power-supply/msg1836968/#msg1836968). They're using the spare pairs to feed 5V to the Raspberry Pi at the end of the cable.
Ah, makes sense now.
-
Fundamentally, the problem here is voltage drop across the rather lengthy cable. That's not a new problem, there are lots of good solutions. Just paralleling more too-small wires is not one of them. The best solution, as recommended several times here by multiple people (including me), is to run a higher voltage over the distance and regulate locally to the load. Not only will this address the voltage drop problem, it will also give you far better regulation performance, lower supply noise, etc. You need to remove that 14 feet of thin wire (yes, twice your seven feet, the *circuit* goes out AND back, remember!) from your "regulated" environment.
-
DC voltage will drop on long cables, that is why AC is used for power distribution.
Completely untrue.
It's about the voltage, not whether it's AC or DC
-
If you need accurate 5V at the device, then the best soution is usually to have a local DC-DC converter at the device, and send a higher voltage (typically 12 or 24V) down the cable. You get less drop due to lower current, and also better tolerance to any drop you do get.
-
DC voltage will drop on long cables, that is why AC is used for power distribution.
Completely untrue.
It's about the voltage, not whether it's AC or DC
Erm, current surely? As in I2R losses.
-
Where's PoE and Ethernet cable suddenly come from? I can't see it mentioned previously in the thread until ogden introduced it. :-//
OP has another thread (https://www.eevblog.com/forum/projects/issue-with-raspberry-power-supply/msg1836968/#msg1836968). They're using the spare pairs to feed 5V to the Raspberry Pi at the end of the cable.
Ah, makes sense now.
It's you. Again. Trying every chance? It's crusade? Shall I become worried?
You suggested me to let it go. Please do the same.
-
Where's PoE and Ethernet cable suddenly come from? I can't see it mentioned previously in the thread until ogden introduced it. :-//
OP has another thread (https://www.eevblog.com/forum/projects/issue-with-raspberry-power-supply/msg1836968/#msg1836968). They're using the spare pairs to feed 5V to the Raspberry Pi at the end of the cable.
Ah, makes sense now.
It's you. Again. Trying every chance? It's crusade? Shall I become worried?
You suggested me to let it go. Please do the same.
It was a simple factual question as I wasn't aware of the other thread, you just happen to the the first, but not only, person to mention PoE. It wasn't even your message I commented to but the second reference to PoE in tsman's. If you think you're being persecuted whenever anybody mentions your name you've got real problems. If you carry on badgering me every time I'm in a thread that you're also in you're going to find yourself pissing off a lot more people than just me - my experience is that the forum doesn't take kindly to that.
-
Where's PoE and Ethernet cable suddenly come from? I can't see it mentioned previously in the thread until ogden introduced it. :-//
OP has another thread (https://www.eevblog.com/forum/projects/issue-with-raspberry-power-supply/msg1836968/#msg1836968). They're using the spare pairs to feed 5V to the Raspberry Pi at the end of the cable.
Ah, makes sense now.
It's you. Again. Trying every chance? It's crusade? Shall I become worried?
You suggested me to let it go. Please do the same.
Some sort of paranoia? I don't get your concern.
-
DC voltage will drop on long cables, that is why AC is used for power distribution.
Completely untrue.
It's about the voltage, not whether it's AC or DC
Erm, current surely? As in I2R losses.
That's a consequence using a higher voltage to send a given amount of power.
-
I'm happy with all of the possible and good solutions given here. Thank you all. I'm really starting to make physical projects, so encountering these kind of issues gives me experience for the projects to come.
-
Some sort of paranoia? I don't get your concern.
LOL, no :D We have history. He is accusing me of "moving goalposts" and derailing discussions (and not only). Please note that he specifically mentioned my nick in his "simple factual question". If you were me, you would suspect that he is coming to accuse me of offtopic talk again. I have rights to be mad as well. I suggested him: let it go. Politely. If needed, I will answer your further questions in PM - if any.
-
I hate the way they used a micro USB connector for power, almost all of the random cables floating around are inadequate. It would be nice if they had at least provided a space for a barrel jack, Molex or other decent power socket. Even mini USB would be an improvement over micro.
-
I hate the way they used a micro USB connector for power, almost all of the random cables floating around are inadequate. It would be nice if they had at least provided a space for a barrel jack, Molex or other decent power socket. Even mini USB would be an improvement over micro.
Cables coming with relatively recent smartphones and decent 3rd party charging/data cables work just fine. If you get very cheap crap which does not work with RPi, such crappy cable will also cause smartphone to drop charging current as well. I'd say that you can get decent cable starting from around $2.5-3 if ordering from gearbest.
It would be nice if they had at least provided a space for a barrel jack
Number of blown up RPi would rise several times. It would be real rarity to get a decent 5V PSU with barrel jack to begin with. I bought cheaply more than thousand of customer returned RPi which came from Farnell. 90+% of failures clearly caused by users, not mfg defect. With micro USB you can at least be more or less sure that PSU will have correct voltage and no reverse polarity.
-
So provide pads for a header, or include an onboard regulator for the 5V like is already there for the core voltage and then it could run from 9-12V adapters. I've had so many micro USB cables that wouldn't work and have seen countless forum posts by users trying to get by with inadequate cables. This is especially true if you want to utilize the onboard USB ports to power devices like hard drives that can draw significant power. I've ended up powering most of my pis by soldering on wires to a barrel jack. I have lots of nicely regulated 5V wall warts.
-
So provide pads for a header, or include an onboard regulator for the 5V like is already there for the core voltage and then it could run from 9-12V adapters. I've had so many micro USB cables that wouldn't work and have seen countless forum posts by users trying to get by with inadequate cables. This is especially true if you want to utilize the onboard USB ports to power devices like hard drives that can draw significant power.
Increased size of the board, increased cost of BOM. And then deal with all those people with non existing soldering skills destroying their devices. 5V usb is something relatively certain, basically plug and play. 9-12 V adapters can have different polarity and often are not even regulated, simply transformer + rectifier. Some "12V" adapters output 19V without load. It's like opening Pandora's box.
I've ended up powering most of my pis by soldering on wires to a barrel jack. I have lots of nicely regulated 5V wall warts.
Geez, it's not that hard to get a decent cable :palm:... unless you shop for cheapest garbage. Heck, just buy official RPi cable/charger if you are desperate.
-
If you are above doing a dirty trick then there is a quick and easy way out. The trick to a dirty trick is to not talk about it nor put it on your diagram. This eliminates peer pressure with it's long lectures on standard engineering practices. Hey this is a nerd project not NASA. Place any diode on you bench on the 7805 regulator ground. The regulated voltage out should now be 5.6 volts rather than 5 volts. This extra .6 volts should be enough to prevent the pi brown out condition. You did not hear this from me as I would never stoop to such dirty tricks. :-+
-
If you are above doing a dirty trick then there is a quick and easy way out. The trick to a dirty trick is to not talk about it nor put it on your diagram. This eliminates peer pressure with it's long lectures on standard engineering practices. Hey this is a nerd project not NASA. Place any diode on you bench on the 7805 regulator ground. The regulated voltage out should now be 5.6 volts rather than 5 volts. This extra .6 volts should be enough to prevent the pi brown out condition. You did not hear this from me as I would never stoop to such dirty tricks. :-+
Don't forget to cool that poor 7805. And in case of Rpi3/3+ it likely will trigger current protection of 7805, especially if you attach some USB flash.
-
Update: I've changed the cable again and it works well now. Apparently, I have serious issues identifying a good cable for my application.
FYI, cat6 and cat7 cables use thicker wire than cat5. Helpful for PoE — Though as you’ve seen, at 5V, you need really, really thick wire. It’s much better to run a higher voltage at lower current across the wire (e.g. 24V) and then use the DC-DC converter locally.
-
Place any diode on you bench on the 7805 regulator ground. The regulated voltage out should now be 5.6 volts rather than 5 volts. This extra .6 volts should be enough to prevent the pi brown out condition. You did not hear this from me as I would never stoop to such dirty tricks. :-+
That's one trick I've used in the past for or'ing several 5V supplies with series diodes while still getting a steady 5V with no significant drop.
-
That doesn't work for a pi unless you want to ruin it at the same time though.
-
Update: I've changed the cable again and it works well now. Apparently, I have serious issues identifying a good cable for my application.
FYI, cat6 and cat7 cables use thicker wire than cat5. Helpful for PoE
Indeed high power/current PoE benefit from Cat6 cables, they cost more than Cat5 as well. We all agreed that it's better to increase voltage than price of the cable ;) Many may not know that while shopping for lowest cost Ethernet cable, you can get copper clad aluminium cable which may work for some time up-to some cable length (not even close to 100m), but is nearly useless for PoE applications due to high DC resistance. Obviously such CCA cable does not conform to Ethernet specs:
https://www.belden.com/blog/digital-building/not-in-my-network-copper-clad-aluminum-is-a-recipe-for-failure (https://www.belden.com/blog/digital-building/not-in-my-network-copper-clad-aluminum-is-a-recipe-for-failure)
So provide pads for a header, or include an onboard regulator for the 5V like is already there for the core voltage and then it could run from 9-12V adapters.
Increased size of the board, increased cost of BOM. And then deal with all those people with non existing soldering skills destroying their devices. 5V usb is something relatively certain, basically plug and play. 9-12 V adapters can have different polarity and often are not even regulated, simply transformer + rectifier.
Indeed USB shall stay for many reasons. On the other hand I agree that DC jack accepting 5-15VDC with some protection would be nice. Yes, socket and buck regulator with bypass for 5V will add to price. They could spin one more version and name it.. dunno... rPI embedded or something like that :)
-
Options to power the Raspberry Pi properly with a long cable, in approx. the order of preference:
1) Use local mains power brick. (Preferably at least 2A.)
2) Use local buck regulation from higher DC voltage (e.g., from around 10-20V). Again, at least around 2A.
If impossible,
3) Use thick enough cable for the 5V.
Additionally, whichever you choose (and especially for #3), I strongly recommend adding some electrolytic bulk capacitance on the 5V line, because the Raspberry Pi is designed without such capacitance for ultimate cost and size reduction. A high-ESR, large value cap has two purposes:
1) it dampens any inductive ringing which can happen with a long cable, given the ultimately low ESR of the ceramic bypassing on the Raspberry power itself. Without it, a quick 1V drop may be followed by a 1V overshoot, risking blowing something up. Not something I have actually seen on a Raspi, but a typical failure mode in general.
2) The large capacitance also helps with spiky power consumption; without it, while you measure acceptable DC drop (the DC consumption is typically around 1A only), but the spikes may easily take 2A, doubling the voltage drop below unacceptable level. Now, these spikes can be long enough that you need considerable energy storage, the ceramic bulk (on the RasPi itself) is not enough.
A 6.3V 1000uF electrolytic is a cheap and relatively small fix for this.
-
So provide pads for a header, or include an onboard regulator for the 5V like is already there for the core voltage and then it could run from 9-12V adapters. I've had so many micro USB cables that wouldn't work and have seen countless forum posts by users trying to get by with inadequate cables. This is especially true if you want to utilize the onboard USB ports to power devices like hard drives that can draw significant power.
Increased size of the board, increased cost of BOM. And then deal with all those people with non existing soldering skills destroying their devices. 5V usb is something relatively certain, basically plug and play. 9-12 V adapters can have different polarity and often are not even regulated, simply transformer + rectifier. Some "12V" adapters output 19V without load. It's like opening Pandora's box.
I've ended up powering most of my pis by soldering on wires to a barrel jack. I have lots of nicely regulated 5V wall warts.
Geez, it's not that hard to get a decent cable :palm:... unless you shop for cheapest garbage. Heck, just buy official RPi cable/charger if you are desperate.
Providing pads on the board costs nothing, there is space. If someone damages their board by poor soldering or connecting it backwards, who's fault is that? If someone uses an obsolete non-regulated wall wart to power something that needs a regulated voltage again who's fault is that?
Why should I buy a special cable and hope that it's adequate when I have piles of them that have come with various devices? Why should I have to guess if a cable or power supply will work? Out of probably 30 cables and a dozen USB power bricks I have, maybe 1-2 cables have proven to be marginally adequate for the RPi, provided I don't have anything pulling a significant load from the USB ports, and maybe 5 of the power adapters. On the other hand I have lots of 5V and 12V wall warts and bricks that put out their rated voltage, can supply their rated current without any issues. I've never had one single problem with one of them, the USB garbage was headache after headache, it's a stupid way to power something. I would wager that probably 75% of RPis returned as defective actually had nothing wrong with them, but were unstable due to crappy power cables which are practically ubiquitous. Even a mini USB would be a big improvement over the micro USB trash.
Well whatever, I'll just keep modifying my RPi's because once modded they are rock solid, no more headaches.
-
Why should I buy a special cable and hope that it's adequate when I have piles of them that have come with various devices? Why should I have to guess if a cable or power supply will work? Out of probably 30 cables and a dozen USB power bricks I have, maybe 1-2 cables have proven to be marginally adequate for the RPi, provided I don't have anything pulling a significant load from the USB ports, and maybe 5 of the power adapters. On the other hand I have lots of 5V and 12V wall warts and bricks that put out their rated voltage, can supply their rated current without any issues. I've never had one single problem with one of them, the USB garbage was headache after headache, it's a stupid way to power something. I would wager that probably 75% of RPis returned as defective actually had nothing wrong with them, but were unstable due to crappy power cables which are practically ubiquitous. Even a mini USB would be a big improvement over the micro USB trash.
I' wager that those cables just come with devices that are less tolerant to power drops and/or just use it for charging so voltage drop does not bother them. On top of it, various "fast" charging standards just put more voltage on it so it is less affected, Dropping 1V at 9V/1A is much easier to tolerate than dropping that at 5V.
And let's not forget USB standard was originally 5V/500mA, a lot of cables you have might be just fine standard-wise
Even a mini USB would be a big improvement over the micro USB trash.
Mini-usb is worse in almost every way. Mini-USB is mechanically flawed and it just degrades much faster than micro-usb, on top of being bigger and having worse latching. And (at least few I've checked) micro-USB connectors are rated@1.8A so current should not be the problem either.
-
And let's not forget USB standard was originally 5V/500mA, a lot of cables you have might be just fine standard-wise
Yep. AFAIK, only USB 3.0 introduced more than 500 mA max (900 mA) for non-charging use. So, many cables are just designed and tested for 500 mA. Getting more than that requires a proper cable and the right identification (if you're using a PC or a true phone charger), which I believe is some combination of pull-up/pull-down resistors on the data lines.
I you tear any micro-usb cable down, you'll often see tiny conductors for power (+5V, GND). No way they can hold even 2A without a massive voltage drop and possibly worrying heating.
-
I only power the Raspberry Pis through the GPIO header. There are two dedicated 5V pins and at least 3-4 GNDs fairly close to them.
Never had an issue with "lacking input protection". It's such a cheap device; if you mess up and blow it up, just get a new one. I have installed dozens and never blown any.
Cumulative cost in sourcing proper USB cables (i.e., specified for USB3 charging) that are not con artist junk, and paying premium for such cables - worst case, chasing problems for weeks due to poor power -, quickly exceeds the cost of a single board blowing up once due to your wiring mistake.
Note that the Raspberry really needs over 2A during the short peaks. Standard USB cable rated for 500mA is going to have quite a lot of voltage drop. Even the cables designed for higher current are designed for charging devices that have internal battery; if you charge at 2A, and there is too much voltage drop, the charging current is just reduced slightly, and the charger IC is fine with this condition. This is completely different to powering a computer with no battery and miniscule amount of capacitance on its power lines. Even the shortest peaks must be within specs, otherwise it'll crash; and often do it in the most peculiar ways.
-
Even a mini USB would be a big improvement over the micro USB trash.
Mini-usb is worse in almost every way. Mini-USB is mechanically flawed and it just degrades much faster than micro-usb, on top of being bigger and having worse latching. And (at least few I've checked) micro-USB connectors are rated@1.8A so current should not be the problem either.
Absolutely. The micro-USB connector was designed to explicitly improve on mini-USB’s shortcomings, not only to be smaller.
-
I would wager that probably 75% of RPis returned as defective actually had nothing wrong with them, but were unstable due to crappy power cables which are practically ubiquitous. Even a mini USB would be a big improvement over the micro USB trash.
IME (>1000 of returned RPi), less than 30% were non defective.
Well whatever, I'll just keep modifying my RPi's because once modded they are rock solid, no more headaches.
You should have fantastic inability to buy a decent cable :palm:.
Even a mini USB would be a big improvement over the micro USB trash.
:palm: |O Micro USB is one of the most reliable connectors. Order of magnitude more reliable than barrel jack. Connector itself almost never fails. Intermittent connection (with slight wiggling) in barrel jacks is so widespread that it eventually developed in like every third device I've seen. Very rarely happens with micro USB.
-
Cumulative cost in sourcing proper USB cables (i.e., specified for USB3 charging) that are not con artist junk, and paying premium for such cables - worst case, chasing problems for weeks due to poor power -, quickly exceeds the cost of a single board blowing up once due to your wiring mistake.
Note that the Raspberry really needs over 2A during the short peaks. Standard USB cable rated for 500mA is going to have quite a lot of voltage drop. Even the cables designed for higher current are designed for charging devices that have internal battery; if you charge at 2A, and there is too much voltage drop, the charging current is just reduced slightly, and the charger IC is fine with this condition. This is completely different to powering a computer with no battery and miniscule amount of capacitance on its power lines. Even the shortest peaks must be within specs, otherwise it'll crash; and often do it in the most peculiar ways.
There are no "500mA" cables among those which are barely decent. No smartphone in the last 5 years uses only 500mA for charging. They normally go 1A+, and drop current when detecting voltage drop. USB3 has nothing to do with charging current. Computer port is not a primary source for charging to begin with. Many phones just won't sink more than 500 mA when detect they are connected to computer.
-
:palm: |O Micro USB is one of the most reliable connectors. Order of magnitude more reliable than barrel jack. Connector itself almost never fails. Intermittent connection (with slight wiggling) in barrel jacks is so widespread that it eventually developed in like every third device I've seen. Very rarely happens with micro USB.
The reliability issue I'm seeing with microUSBs is not the connector itself failing, but the solder joints failing. Stress reliefing the power input cable is the fix - or just being careful.
Still, microUSB was not developed to supply 2A with minimal voltage drop for non-charging purposes, i.e., where 2A delivery without significant voltage drop is critical, and not only a factor of reduced rate charging.
Really, the use case for powering the Raspi is completely different from what these USB cables are typically used for:
1) data communication
2) low current (typically <500mA) power supply
3) higher current charging, where it's enough for a linear reg or buck converter to be able to convert to li-ion voltage, down to below 4V, where a voltage drop isn't critical, and worst case, the rate is reduced. And indeed, people do report different charging rates for fast charging using different quality cables!
Instead of going circles of babbling about "good" cables, maybe you should give us some exact recommendations of your known good cables? You know, the requirements for the Raspberry are well outside the USB specification, so having a "good compliant cable" isn't necessarily enough.
-
There are no "500mA" cables among those which are barely decent. No smartphone in the last 5 years uses only 500mA for charging. They normally go 1A+, and drop current when detecting voltage drop.
Exactly this. You have completely proved my point. 1A+, and dropping current when detecting voltage drop.
Guess what? Raspberry Pi hasn't got a battery to power itself from. It absolutely requires -- not 1A, not 1.5A, but at least 2A (depending on what you connect to it and how you configure it). When it detects a voltage drop, it... wait, can it "drop charging current"? Well, usually it flashes the voltage drop warning. Sometimes, it just corrupts the file system, or just crashes.
Charging a battery of a device powered by a battery is completely different beast from powering a device with very little bypass capacitance. Even if the voltage drops by a volt, it's going to work. Even if it browns out for a second, nothing bad happens.
Current ratings of good quality USB cables may be based on:
1) thermal ratings: it won't melt or the plastic won't degrade
2) useful charging - because that's what these cables are specified to do: charge the device: the voltage drop at that particular current typically doesn't result in reduced charge rate - at least to non-full battery (with more dropout to work with)
On low quality products, it's of course just marketing. But even if the quality is OK, and the specs hold, the specs are designed for a completely different, and much less demanding, use case.
An example:
Charging:
0A: 5.0V - all fine
1A: 4.5V - all fine
2A: 4.0V - charge to the empty battery still ok at almost 2A. With full battery, charging rate is reduced. No big deal.
Raspberry Pi:
0A: 5.0V - all fine
Powered up. WiFi or any USB device tries to sink 2A for 1 millisecond. Supply brown-outs to 4.0V. May work or may not. Reliability issues seen by the user.
-
Guess what? Raspberry Pi hasn't got a battery to power itself from. It requires -- not 1A, not 1.5A, but at least 2A (depending on what you connect to it and how you configure it). When it detects a voltage drop, it... well, usually it flashes the voltage drop warning. Sometimes, it just corrupts the file system, or just crashes.
Raspberry pi 2 IIRC from my measurements don't go over 600mA and pi 3 don't go over 900mA (without overclock). All that exceeds this is from additional load you attached. Most recent phones use more than 1.5A for charging.
-
Guess what? Raspberry Pi hasn't got a battery to power itself from. It requires -- not 1A, not 1.5A, but at least 2A (depending on what you connect to it and how you configure it). When it detects a voltage drop, it... well, usually it flashes the voltage drop warning. Sometimes, it just corrupts the file system, or just crashes.
Raspberry pi 2 IIRC from my measurements don't go over 600mA and pi 3 don't go over 900mA (without overclock). All that exceeds this is from additional load you attached.
These are DC measurements, I guess. Because I have measured around 1A DC from Pi3 as well.
DC measurement is completely irrelevant, as explained above, because there is little bulk capacitance. Any computer device is going to use spiky currents; think about how WiFi works, for example, or accessing an SD card. The capacitance on Raspberry3 board itself can supply the required load for hundreds of microseconds, maybe a millisecond max.
This is why I recommended adding a big bulk cap on the supply; it will help on the borderline cases. But adding a decent power supply without excess resistance (long, poor cable) is of course better; then the load variation is accounted by the feedback loop of the supply, which easily corrects in millisecond range.
-
There are no "500mA" cables among those which are barely decent. No smartphone in the last 5 years uses only 500mA for charging. They normally go 1A+, and drop current when detecting voltage drop.
Exactly this. You have completely proved my point. 1A+, and dropping current when detecting voltage drop.
Most phones just take nominal current and do not mind voltage drop at all - because voltage of fully charged battery is around 4.2 volts. It does not matter you put in 4.5V or 5V - phone will happily charge your battery. BTW most phone chargers have quite poor voltage regulation, if it is written 5V 1500mA then you get one or another, but not both in the same time :D This of course in addition to cable drop!! rPI on the other hand is very picky about voltage. So here we are.
Yes, I agree that micro USB is way more reliable than mini USB. Thou today I am fan of USB-C connector, USB PD and/or QC2.0/3.0
-
Charging:
0A: 5.0V - all fine
1A: 4.5V - all fine
2A: 4.0V - charge to the empty battery still ok at almost 2A. With full battery, charging rate is reduced. No big deal.
Nonsense, no barely decent cable has 0,5V voltage drop at 1A.
Raspberry Pi:
0A: 5.0V - all fine
Powered up. WiFi or any USB device tries to sink 2A for 1 millisecond. Supply brown-outs to 4.0V. May work or may not. Reliability issues seen by the user.
Nonsense again. You may not know but Raspberry pi has enough bulk capacitance for such short bursts of power. There are 47uF MLCCs on 1.2V, 1.8V and 3.3V power rails and 3 of such in parralel on USB. Also it's not like WIFI has short power bursts of high power, it's not GSM.
-
The reliability issue I'm seeing with microUSBs is not the connector itself failing, but the solder joints failing. Stress reliefing the power input cable is the fix - or just being careful.
Well that’s not a property of the connector design, sorry! Look at how many good designs implement their connectors: heavy metal or plastic with a precision cutout for the plug, so that the plug is well-supported by said thick metal or plastic. (Look at how Apple machines the various jacks in a MacBook. By shifting strain relief to the housing, connector failure is far less common than in older machines where the PCB took the brunt of the strain.) Another good example is the cord on my Ultimate Ears Boom 2 speaker, which is recessed (so that it can take a rubber bung to waterproof it), so when the cord is plugged in, the jack isn’t seeing any strain at all. The downside is that most other micro-USB cords won’t fit in the recess at all. (The UE cords are expensive but lovely (https://www.amazon.com/Logitech-Ultimate-Bluetooth-Speaker-charger/dp/B018EC8ZS4) — they do not tangle. The UE charger is 2.1A, so I would assume the cable can handle that as well.)
Instead of going circles of babbling about "good" cables, maybe you should give us some exact recommendations of your known good cables? You know, the requirements for the Raspberry are well outside the USB specification, so having a "good compliant cable" isn't necessarily enough.
If their Lightning cables are any indicator of the quality of their micro-USB cables, the AmazonBasics cables are superb and don’t cost an arm and a leg. (For sure, I know they’ve considered voltage drop, because their 2 meter Lightning cables use thicker cable than the shorter ones, and the 3 meter Lightning cables use far thicker cable still.)
Charging:
0A: 5.0V - all fine
1A: 4.5V - all fine
2A: 4.0V - charge to the empty battery still ok at almost 2A. With full battery, charging rate is reduced. No big deal.
Nonsense, no barely decent cable has 0,5V voltage drop at 1A.
It’s not nonsense, in that a TON of people don’t realize that many cheap gadgets come with obscenely bad cables, and that most cheap cables they buy separately are equally horrific. They see the $2.99 one and the $11.99 one at the store, and don’t realize the cheap one (whose eBay price would be 2 for 99¢ shipped) is garbage, so they buy it. And then they have no idea why their phone takes longer to charge than before...
-
Most phones just take nominal current and do not mind voltage drop at all - because voltage of fully battery is around 4.2 volts. It does not matter you put in 4.5V or 5V - phone will happily charge your battery.
Except that most phones detect said voltage drop and back off their charging rate at some point. The exact voltage where that happens varies, but it’s not going to be down at like 4.3V or so, because by that point it’s obvious that either the charger or cable is severely overburdened.
BTW most phone chargers have quite poor voltage regulation, if it is written 5V 1500mA then you get one or another, but not both in the same time :D This of course in addition to cable drop!! rPI on the other hand is very picky about voltage. So here we are.
Then you’re using garbage chargers. There are websites with tons of tests of USB chargers, and the original name brand chargers (Apple, Samsung, Microsoft, HP, etc.) all perform at least to spec, and often quite a bit beyond. Quality third-party chargers (Anker, IKEA, Belkin, etc) do equally well. All of these will produce clean output without voltage sag to at least their stated current.
Meanwhile, the chinesium garbage death-chargers are often 500mA beasts no matter what the stated current is, and often plummet to below 4V on full load, or lose regulation and go ripply, or shut down.
-
The reliability issue I'm seeing with microUSBs is not the connector itself failing, but the solder joints failing. Stress reliefing the power input cable is the fix - or just being careful.
There is no such issue with raspberry Pi. There are 4 through hole solder joints. I never seen solder joints failing, though I several times seen manufacturing defect when connector was soldered lifted on on side and several pins were not touching pads. I also have seen a few RPi with connector broken off together with pads but it requires extreme level of abuse. There are micro usb connectors which do not have through hole terminals, those are much easier to break off but it has nothing to do with RPi.
-
Well, usually it flashes the voltage drop warning. Sometimes, it just corrupts the file system, or just crashes.
True. My raspberry just kept crashing at booting when I was getting the undervoltage warning. Well, if the voltage drop is critical, it does that. Sometimes it runs "normally" even with the warning.
-
Most phones just take nominal current and do not mind voltage drop at all - because voltage of fully charged battery is around 4.2 volts. It does not matter you put in 4.5V or 5V - phone will happily charge your battery.
Buck converter is used for charging. There is still need for input voltage to exceed voltage on the battery. 0.1V or 0.2V above output is not enough for proper operation. And as tooki already said, phones do detect voltage drop and reduce charging current.
-
The reliability issue I'm seeing with microUSBs is not the connector itself failing, but the solder joints failing. Stress reliefing the power input cable is the fix - or just being careful.
Well that’s not a property of the connector design, sorry! Look at how many good designs implement their connectors: heavy metal or plastic with a precision cutout for the plug, so that the plug is well-supported by said thick metal or plastic.
Properties aside: USB specification requires 5000 mating cycles for mini USB and 10000 cycles for micro USB. Those who are in doubt can check molex datasheets: mini (https://www.molex.com/pdm_docs/ps/PS-56579-015-001.pdf) micro (https://www.molex.com/pdm_docs/ps/PS-47346-001-001.pdf). Other option : look for usb.org specs.
-
Aaaand...? So what? Point is, the micro-USB was designed to significantly surpass the mini-USB performance. I would say that doubling the mating cycles counts as “significant”.
-
There are 47uF MLCCs on 1.2V, 1.8V and 3.3V power rails and 3 of such in parralel on USB. Also it's not like WIFI has short power bursts of high power, it's not GSM.
We engineers tend to calculate things instead of sentimental "nonsense" shoutouts. I based my estimate on around 100-200uF MLCC, which seems to be fairly close. Please do the math yourself. 1 F = 1 As/V, which gives you the easy-to-remember, 1mF (1000uF) drops by 1V when drawn 1A for 1ms. So, depending on which rails are exactly the weakest links (i.e., the exact amount of voltage drop allowed), we are talking about the territory between 100us and 1ms there.
Any microcomputer is going to be a tricky on AC consumption, which is why a lot of local bypassing and local regulation happens. Local regulation happens on Raspi, as well; it's just that it's still a computer that is often powered by substandard supplies and cables.
What's actually tricky in the AC consumption is that there are many very low-duty loads that very seldomly appear all at once. It's well possible that in a typical use case, a Raspi3 eats:
1) 0.8A DC current with fairly high CPU duty cycle (i.e., when not idling)
2) 1ms peaks varying between 1 and maybe 1.5A
3) Once a year there is a 2A peak, because of the exact combination of timing and changing variables (such as WiFi signal strength, causing automatic adjustment of the transmit power).
And a 100us brown-out is enough to crash it. Completely unlike charging.
I don't know, I have quite limited specific experience on Raspberry Pi, the example values are more about the general principles. I have fixed a few Raspberry Pi2&3 issues by adding bulk capacitance and/or replacing the wiring. IMO, having to shop for specification-exceeding cables on the market literally saturated with variable-quality crap is a real risk (and tedious job IMHO!), when you can just use a bog standard cable of your choice of AWG, and connect it to the header directly. But I understand you skip the input protection "ideal diode" that way.
But I understand you desperately need to have this fight, and desperately need to be right. So be it.
I just find it very funny you guys keep repeating how the phones detect voltage drop and back off charging current. That's exactly proving the whole point; these cables have significant voltage drops happening, and the charging case is proving it. With charging, it's non-critical, because the device is powered by the stable battery supply.
-
IMO, having to shop for specification-exceeding cables on the market literally saturated with variable-quality crap is a real risk (and tedious job IMHO!), when you can just use a bog standard cable of your choice of AWG, and connect it to the header directly.
Absolute good sense.
But I understand you skip the input protection "ideal diode" that way.
You can either assume the risk as very low, or if you're really worried, add the protection outside the board between your PS and the GPIO header. Either way, it's really no big deal.
-
Charging:
0A: 5.0V - all fine
1A: 4.5V - all fine
2A: 4.0V - charge to the empty battery still ok at almost 2A. With full battery, charging rate is reduced. No big deal.
Nonsense, no barely decent cable has 0,5V voltage drop at 1A.
It’s not nonsense, in that a TON of people don’t realize that many cheap gadgets come with obscenely bad cables, and that most cheap cables they buy separately are equally horrific. They see the $2.99 one and the $11.99 one at the store, and don’t realize the cheap one (whose eBay price would be 2 for 99¢ shipped) is garbage, so they buy it. And then they have no idea why their phone takes longer to charge than before...
I said "barely decent", not cheapest junk. I don't mean cheapest ebay cables by that. BTW you certainly can buy decent cable for $3. But I won't look on ebay for that. IME when buying on gearbest you should receive satisfactory cable at such price. When I was selling RPi on ebay, when people had power/reboot issues, the first thing I suggested was to change cable/PSU and it solved most of the problems.
-
Aaaand...? So what? Point is, the micro-USB was designed to significantly surpass the mini-USB performance. I would say that doubling the mating cycles counts as “significant”.
Just info about how many mating cycles one or another is specified for. If you are not interested - ignore it. No reason for "so what" posts, especially if I provide additional info that agree to your point.
-
It’s not nonsense, in that a TON of people don’t realize that many cheap gadgets come with obscenely bad cables, and that most cheap cables they buy separately are equally horrific. They see the $2.99 one and the $11.99 one at the store, and don’t realize the cheap one (whose eBay price would be 2 for 99¢ shipped) is garbage, so they buy it. And then they have no idea why their phone takes longer to charge than before...
I have seen totally junk cables in nice retail packaging for 11.99€-like pricing too many times. For example, hair-thin "USB 3" cables, barely able to power a 500mA device, or completely impedance-mismatched SATA cables basically out of coat hanger wire.
With significantly more expensive products, chances of getting junk may be somewhat lower, but it's far from guaranteed. At the other end of the spectrum, there are the audiophoolery thingies at ridiculous prices. Everybody agrees they are scams.
I never shop by price only. I do the homework beforehand. Often, a relatively cheap (usually not the absolutely cheapest, nor the most expensive) product is just fine.
I wouldn't blame the consumer for choosing the $3 cable vs. $12 cable at the store. A proper cable can easily be manufactured and sold, all profits included, for $3. But as you say, the $3 cable would often be $0.50 on Ebay, shipping included, so by sourcing them directly for $0.10 ea at 1000000pcs, they are making good profit out of crap. But why stop there? You can as well sell the same $0.50 Alibaba special for $11.99 or $24.99, and it will sell just fine given a bit of brand design and a nice packaging, with claims of high performance. Which is borderline scamming, but it happens.
For Raspberry Pi, I still power it using standard wire which I know the cross-sectional area of, and add the 1000uF bulk elcap, if nothing else, for damping, because, as a designer for similar complexity designs as the Raspberry Pi, that's how I'd design my product.
I know this is boring as it's too easy, doesn't require the secret knowledge of good cable brands, and doesn't generate nice forum fights about who or which component to blame when things don't work out after all.
-
Buck converter is used for charging.
Thank you for reminding obvious. Anyway main thing I wanted to tell - phones does tolerate input voltage drop much better than rPI and hopefully nobody will argue that.
And as tooki already said, phones do detect voltage drop and reduce charging current.
Phones (in a sense their firmware) does not need to. If input voltage of buck converter drops meaning it cannot draw required power anymore, it just draws less power. IMHO intuitive enough.
-
I have seen totally junk cables in nice retail packaging for 11.99€-like pricing too many times. For example, hair-thin "USB 3" cables, barely able to power a 500mA device, or completely impedance-mismatched SATA cables basically out of coat hanger wire.
Of course. Most of those cables are made with the lowest manufacturing cost possible. Half the diameter of a copper wire will equate to approx half the price. Copper is not cheap. That's as simple as that. A $5 to $10 cable (retail) may just cost 20 to 50 cents out of factory. What can you really get for this cost?
What is actually "junk" or not is very difficult to assess before testing. Final pricing means squat. The manufacturer reputation may mean a bit more. But I defy people to actually *find* the name of a manufacturer of USB cables with a particularly good reputation and which is a cable specialist (having a decent reputation when you sell just anything is not enough either). Good luck.
-
Phones (in a sense their firmware) does not need to. If input voltage of buck converter drops meaning it cannot draw required power anymore, it just draws less power. IMHO intuitive enough.
It would need to be rather smart IC. AFAIK all modern smartphones have software controlled charging. If battery was completely discharged, it is charged by low current until voltage increases enough so the phone can boot.
-
If someone needs cheap but good cable, https://www.blitzwolf.com/5.9ft/1.8m-PVC-2A-Micro-USB-Charging-Data-Cable-p-283.html (https://www.blitzwolf.com/5.9ft/1.8m-PVC-2A-Micro-USB-Charging-Data-Cable-p-283.html)
https://www.blitzwolf.com/3.28ft/1m-PVC-2A-Micro-USB-Charging-Data-Cable-p-281.html (https://www.blitzwolf.com/3.28ft/1m-PVC-2A-Micro-USB-Charging-Data-Cable-p-281.html)
Their chargers also are cheap but rock solid.
-
It would need to be rather smart IC. AFAIK all modern smartphones have software controlled charging.
Sort of. The later USB spec revisions specify a handshake protocol for the current. 500mA is the default, but the load and the source can negotiate for higher currents up to the limits of the less capable end. Plug in a "dumb" device and the source is supposed to limit its current to no more than 500mA. All of this is normally handled by a dedicated USB interface IC, available from multiple manufacturers and costing very little thanks to their huge consumption for cellphones and tablets (just like accelerometer IC's). Once the max current has been agreed upon, the load can then make its own decisions about actual charging currents as the battery charges. Obviously the consumed current will drop as the battery nears full charge - the negotiation doesn't specify a minimum current, only a maximum current.
-
If someone needs cheap but good cable, https://www.blitzwolf.com/5.9ft/1.8m-PVC-2A-Micro-USB-Charging-Data-Cable-p-283.html (https://www.blitzwolf.com/5.9ft/1.8m-PVC-2A-Micro-USB-Charging-Data-Cable-p-283.html)
It's great to have actual numbers available. 22AWG (0.33mm^2) wire has resistance of 53mOhm/m; hence, a 1.8m cable has a typical resistance of 190mOhm, so +/-20% manufacturing tolerance assumed, let's say worst case is around 230mOhm. Combined with the max contact resistance (e.g., for a Hirose microUSB, http://www.symmetron.ru/suppliers/hirose/files/pdf/hirose/hirose_ZX.pdf (http://www.symmetron.ru/suppliers/hirose/files/pdf/hirose/hirose_ZX.pdf) , 30mOhm) - two contacts at both ends - we are at 350mOhm.
So, given my earlier exemplatory numbers (from the top of my head, to illustrate a concept, or point) equaling to 500mOhm total resistance you called "nonsense", they were not very far off from your specimen of a "good" cable of 350mOhm!
So, using this "good" cable alone will cause a 700mV voltage drop on my claimed 2A peaks (which should be actually verified), browning the power out to 4.3V. A short 4.3V valley is still probably OK for the Raspberry - this would need to be measured as well -, but even if it works, I know for sure there's not much margin left.
Now, as for the smartphone comparison, a charger buck converter IC might typically have maximum duty cycle somewhere in the 90-95% range, and power stage losses included, maximum output for 4.3V input would be around 3.9V, which would equal about 50% SoC while charging. This agrees with what the people are often actually seeing: poor cables charge very slowly (sub 1A), while "good" cables such as this charge at about full rated current (1.8A here) to an emptyish battery, but tend to taper out somewhat. This is often acceptable: we want fastest charging to fill the emptyish battery.
It's worth noting it's a normal trick for a 5V USB supply to compensate the output voltage assuming a certain wire resistance. Such a supply might also help on the Raspberry, but the response time of such compensation may not be enough; this feature is designed for phone charging, after all. Many 5V chargers also output 5.1 to 5.3V. This would probably "mask" the resistance of the cable so that the "full" 1.8A charging current is seen over most of the cell state-of-charge range.
So the user of this cable would be completely happy using it for charging.
For powering a Raspberry, I'd expect, given the information available and my design experience, that it's borderline failing. It's likely to be OK, but the margin would be poor. So either the user would be happy; or the user would experience very weird, random and possibly rare reliability problems. Now changing to a different "good" cable could either fix it for good - for example, if the better cable used 20 AWG wires, or is shorter - or then it just borderline fixes the issue by being just marginally (say 10%) better. Or, the "new" better cable could just be worse.
When it comes to reliability, this number game is getting really nasty. Simply put, way too many variables, way too little margin, way too much uncertainty.
I'm not at all surprised at your earlier comment:
"when people had power/reboot issues, the first thing I suggested was to change cable/PSU and it solved most of the problems"
But change to what? It's clear that only the very best supplies and cables could give a good reliability margin. With medium-class stuff people would be most likely to find, even when shopping for "good", the devil's fully in the small little details and environmental variables with poor margin for error.
Now, the OP's problem required a 8' (2.4 meter) cable. If 22AWG wire was barely working / on the brink of failing at 1.8 meters, at 2.4m it won't cut it anymore. I'd expect the available market options for a 2.4m microUSB cable with at least 20AWG power wires, is going to be harder and fairly expensive.
OTOH, if powering through the GPIO header pins (30mOhm max is BTW also typical for said connector type; but they have paralleled two to halve the effect of contact resistance!), it's utterly trivial to just use your go-to 16 AWG or 1.0mm^2 or whatever wire you are using for power in your projects, and easily get the total down to less than 100mOhm, even with long wires up to several meters. No issue whatsoever! Only after about 3-4 meters, you start worrying about the copper cost and the heavy cables, and need to start thinking about power distribution with higher voltage, and local buck regulation.
It's also worth noting that the assumption that you "shouldn't" consider connecting anything to the Raspberry is unrealistic, and serves no purpose (except to keep discussion going). It's a computer with USB and Ethernet connectivity, of course people are going to connect things to it, such as keyboards, mice, 4G/LTE modems, or external wi-fi because the Raspberry doesn't have a connector for an external antenna. While it wouldn't make much sense to calculate for the absolute worst case of loads, it's not acceptable to completely ignore very typical loads either.
This being said, by all means do ignore the real-world usage patterns, and you are guaranteed to see problems, and you are guaranteed a job of fixing issues, or have endless discussions :horse:
-
It's great to have actual numbers available. 22AWG (0.33mm^2) wire has resistance of 53mOhm/m; hence, a 1.8m cable has a typical resistance of 190mOhm, so +/-20% manufacturing tolerance assumed, let's say worst case is around 230mOhm. Combined with the max contact resistance (e.g., for a Hirose microUSB, http://www.symmetron.ru/suppliers/hirose/files/pdf/hirose/hirose_ZX.pdf (http://www.symmetron.ru/suppliers/hirose/files/pdf/hirose/hirose_ZX.pdf) , 30mOhm) - two contacts at both ends - we are at 350mOhm.
So, given my earlier exemplatory numbers (from the top of my head, to illustrate a concept, or point) equaling to 500mOhm total resistance you called "nonsense", they were not very far off from your specimen of a "good" cable of 350mOhm!
You just made up that 350 mOhm figure based on worst assumptions. GND wires + shield likely have lower resistance than +5V wire too. Most likely it should be somewhere around 200-250 mOhm. In any case, I would not advise using 2m cable if you take 1A over USB/GPIO additionally to what RPi consumes itself. And If you need long distance, just move mains voltage closer to the device, extension cord is not something extraordinary. Also 350 vs 500 mOhm is hell a lot of difference. BTW good chargers often slightly increase output voltage under the load, IIRC RPi official PSU is among such.
-
Of course reliability analysis is done using the worst case values (possibly neglegting the absolutely worst case, very rare combined conditions for non-mission critical equipment; such a rare combination was not present in my resistance calculation.). Unless you want to have problems measured in several percents over the units on the field, of course...
Especially since I didn't include some possible error sources such as connector wear and contamination at all, but instead, used the brand-new connector maximum initial contact resistance value...
There isn't even a derating factor or a safety margin factor involved in my numbers. They are so optimistic in many regards that you would want to seriously do a better job than I did to find more possible error sources - or add a generous safety factor.
It's no wonder people think Raspberry Pi is "unreliable". It's almost always about the power source, and both the wrong assumptions and wrong concepts (let's use typical values instead of the worst case; or let's use DC current instead of AC when the ~1kHz AC matters because there is no significant energy storage (below about 1kHz or so)).
There might be reason why the Raspberry Pi Foundation themselves specify: "steady 2.5A it needs for proper performance" for their recommended power supply they sell, while people seldom are able to measure much over 1A of DC consumption, even with extra peripherals connected. And, there might be a reason it has a fixed cable.
Active "negative resistance" feedback loop is great in chargers, but does it have the bandwidth around 1kHz? This parameter is unimportant for charging these supplies are designed for. I wouldn't trust it, but would calculate the required cable based on an assumption that this compensation does not exist.
-
It's no wonder people think Raspberry Pi is "unreliable". It's almost always about the power source, and both the wrong assumptions and wrong concepts (let's use typical values instead of the worst case; or let's use DC current instead of AC when the ~1kHz AC matters because there is no significant energy storage (below about 1kHz or so)).
It is "unreliable", and has several reasons for that. Micro SD card is one of them. With many of them data will be corrupted over the time. From my experience some are relatively robust but others are not. Some of them will corrupt data on first power loss, while others are relatively robust against that. If you need high reliability, you shouldn't run 5V over 2m of cable in any device unless it's low current.
but instead, used the brand-new connector maximum initial contact resistance value...
From that spec initial resistance must not increase by more than 10 mOhm after 10k mating cycles.
-
It would need to be rather smart IC. AFAIK all modern smartphones have software controlled charging.
Sort of. The later USB spec revisions specify a handshake protocol for the current. 500mA is the default, but the load and the source can negotiate for higher currents up to the limits of the less capable end. Plug in a "dumb" device and the source is supposed to limit its current to no more than 500mA. All of this is normally handled by a dedicated USB interface IC, available from multiple manufacturers and costing very little thanks to their huge consumption for cellphones and tablets (just like accelerometer IC's). Once the max current has been agreed upon, the load can then make its own decisions about actual charging currents as the battery charges. Obviously the consumed current will drop as the battery nears full charge - the negotiation doesn't specify a minimum current, only a maximum current.
Not exactly. From the very beginning, the USB standard allowed up to 100mA current draw during device enumeration and negotiation. Any current above that was supposed to occur only after power negotiation. But this was when USB was strictly on computers, and USB chargers didn't exist yet.
When USB charging started to become a thing, companies began implementing various nonstandard ID schemes so that simple "dumb" chargers wouldn't need a whole USB interface IC. So we had the various types of resistor-divider systems putting voltages onto the data pins for the gadgets to sense. These schemes grew more complex as vendors implemented higher currents above 500mA and wanted an ID for each, so the gadget could identify the charger wattage.
Eventually, basic charging got standardized in USB (the shorting of the data pins), and most gadgets since have been made to accept both that and their respective vendor's proprietary ID scheme as well. (For backward compatibility, vendors have tended to continue using their proprietary schemes in their chargers in order to work for their older devices. Or they put in smart ID chips in the chargers that dynamically figure out what ID scheme to use.)
Then we got new nonstandard power standards like Qualcomm's Quick Charge to allow higher voltages after negotiation.
And eventually we got USB Power Delivery, which formalizes various higher charging voltages and currents, again with negotiation.