Oh dear, a fan
pity because it is something that moves & so may fail, especially in a dusty environment, where otherwise it might be expected to just work for years.
People are reporting problems with the newly released Power Over Ethernet adaptor for the new Raspberry Pi 3 Model B+, with non-functional USB ports causing much anguished wailing. The Power Over Ethernet, Hardware Attached on Top (PoE HAT) board allows a Pi 3B+ to be powered over Ethernet rather than requiring its own power …
PoE has a lot of plus points. I was looking at Rasp prices a couple of days ago and saw mention of the PoE HAT. I'm only running one 'production' Pi at the moment, that is feeding a TV slide show. I've got the Pi getting updates via ethernet, but I've also got a power supply strapped to the back of the TV. I have another box on the network getting PoE so I could easily add another to the cabinet, which would mean less cabloid mess behind the telly!
"have another box on the network getting PoE so I could easily add another to the cabinet, which would mean less cabloid mess behind the telly!"
I look on it the other way round - what's one more cable given the mess that's there already? I don't think I'd like the fan and SWMBO certainly wouldn't.
There are probably industrial applications where PoE has advantages.
My PoE switches are powered off an UPS, so anything powered via PoE automatically benefits. And the other advantage is that via those switches I can switch off or power cycle all of those devices remotely.
In my house there's at least one Arduino in an inaccessible location that I would not think of powering any other way.
PoE has many uses.
The fans can also be disabled - it was a "freebie" with the board, not required for the PoE to operate. Additionally, the RPi has adequate temperature control (by throttling) without any fans at all. So the "only moving part" can fail and not much will change. Or you can just turn it off.
PoE in an official box means that these things can replace lots of things. Everything from IP phones to IP-based speakers, can be powered throughout a building from a central place (with UPS, etc.) and also be pretty invisible when in situ.
My workplace has PoE phones, wireless (APs and bridges), cameras and speaker/alarm boxes. The Pi can directly replace at least three of those. Not to mention things like thin clients (the free utility rdesktop on a Pi is actually SOLD by a company called nComputing as a thin client for Windows or Linux remote desktops), etc.
I've been thinking about a project where I work where POE powered PIs may be a good thing. We have a lot of rooms that are bookable. The bookings are made largely online, and the timetables are available online.
I was thinking about a 10 inch LCD, with a Pi for each room, displaying that room's timetable, so people can easily look up if a room is booked. I know that there are companies that sell LCDs specially set up to do this, but that's likely to be expensive, and possibly tie us into an expensive maintenance contract. Besides, I'm a technician at heart. I think a large part of me likes the idea of a home grown solution to a problem.
POE could be handy for that, as we don't necessarily have power sockets near where we would want the displays. That said, we don't currently have ethernet sockets near them either, and as we use the same contractor for both, it would be relatively easy to get them to fit both at the same time.
@Stuart Castle - yes a great idea but watch the power budget. The PoE HAT is rated for up to 2.5A at 5V and the 7" Pi Touchscreen needs almost 500 mA alone. A 10" display likely more, mostly into the LED backlight. I bought the PoE HAT with a similar idea using the RPi 7" official touchscreen or a more sunlight-readable equivalent I found from NHD (ordered, still to be tested).
Without a strong plan for a sufficiently low-power display that is still large and legible enough, you're going to end up needing a power socket for that part anyway. Even keeping within the higher currents provided by the later specs, your switch may be limited to how many can be supplied at that current, vs all 12/24/48 ports at a lower standard.
"There are probably industrial applications where PoE has advantages."... yes, a few, like the hundreds or thousands of WAP boxes on college or business campuses, where running power is a real PITA, so a huge gain for PoE. Not to mention mains power supplies are a common point of failure in such deployments: another win for PoE. At the endpoint (in this case the PoE HAT) you still need a switcher from the PoE 48V to the local 5 or 3.3V but that's way easier than a mains switcher. My PoE HAT is still in the box, curious to see if I'm bitten by this bug.
You can turn the fan off if you wish. Usually it's controlled by the Pi itself, so only turns on when required, and will be sped up/slowed down as necessary. If off, then the Pi will use its normal temperature control system, which may throttle the CPU if it get too warm.
Stevie Or is that fast breeder reactors? I always get the two confused (which explains the prompt criticality fiasco when I deployed my home-made set-top box).
For some reason this passage from Good Omens comes to mind:
“He plugged it in to the socket. Then he switched the socket on.
Every light in the house went out.
Newton beamed with pride. He was getting better. Last time he’d done it he’d blacked out the whole of Dorking, and a man from the Electric had come over and had a word with his mum.”
If only the folk at HP/Nvidia were as honest as Eben Upton, to own up to their poor product testing, that has left so many Geforce based laptops dead from poor Geforce GPUs/poorthermal design and the use of lead free solder. These companies have just kept kicking the can rather than properly dealing with the problem.
Apple did something at least. Apple dropped Nvidia, though Apple have done the minimum in terms of dealing with the historic problem circa 2010-2012 iMacs, now branded 'Vintage'.
"If only the folk at HP/Nvidia were as honest as Eben Upton, to own up to their poor product testing, that has left so many Geforce based laptops dead from poor Geforce GPUs/poorthermal design and the use of lead free solder."
You literally have no idea about modern soldering.
To get BGA connections at the required density requires the solder has extremely precise characteristics and unfortunately the effects resulted in joint fractures after a combination of time, changes in the solder and temperature cycles which arose from the varying workload.
Yes, I worked at Nvidia (a non graphics division) and Jensen was genuinely upset about letting the customers and users down. I even met him once, a nice guy.
the fan may be a necessity with certain Pi versions, having that circuit board over the top of the 'otherwise open' ventilation space for the CPU...
Otherwise, it's probably cheaper to get a simple $10 PoE adaptor on amazon [several versions exist at around the same price with very good reviews], one of the 'wye splitter' variety at any rate. [yeah this was discussed on IRC last week, along with links to one particular device that fits the category of what I just said].
but PoE is less important (to me) than having a proper power/shutdown switch... and yeah external boards exist for that, too.
> maybe open source the design files on github
You are joking right? The Raspberry Pi foundation is moving further away from being open by the day.
Circuit diagrams are incomplete if any are released at all. Component labels are no longer printed on boards. They are not only taking us back to the glory days of home computing but to the times before when only the privileged boffins in white coats were allowed access to the box of tricks.
The foundation perpetuates an elitist attitude that no one can do things as well as they can so it should be quite instructional when they do screw up but it seems they become even more closed.
Wow. Quite the rant.
Just to inject semblence of truth in, we have never provided full schematics of any of the Pi boards. so to say we are geting less open is an exageration. We have had to change a little due to to becoming more competitive - ie we need to maintain our competitive advantage to remain in business. Easiest way to do that? Keep certain aspected under wraps.
Mising component labels is a cost saving measure, and what we can fit in limited space. Not much more to it that that.
We do not have an elistist atitude. We welcome anyone to contribute to our linux kernel, our documenation, our projects etc. As we do get some really good contributions in all those areas. What we cannot do is open csource the GPU binary blob, because we do not hold the copyright on it - that is held by Broadcom. We release what we can.
We screw up, everyone does. This is a case in point. But we've come clean, which is more than most other companies. Be nice to understand what you mean by your last paragraph, taking in to account this exact example.
@ James Hughes 1
The Pi was meant as an educational computer and it is historically educational about how electronics were in the late eighties when I did my apprenticeships and many of the same problems, solutions and techniques were common.
I'm not being sarcastic, I mourned when 'fault find to component level' changed to scrapping the apprenticeships and having the operators throw the board away and replace it, so it is heartening to read the skill set of your users. One difference is our VMEbus boards were being sold for many thousands of pounds in their thousands back when a pound could buy you a packet of cigarettes and a dram of whisky (and back when dram implied whisky not memory), so you have nothing to be defensive about.
I, for one, welcome our new Pi over-hats. I have a Fluke 187 and I am not afraid to use it!
Just to inject semblence of truth in, we have never provided full schematics of any of the Pi boards.
Were the original Pi A/B schematics not complete?
Those look complete to me. If not then perhaps you could indicate what's missing.
so to say we are geting less open is an exageration.
It does seem later schematics are a whole lot less complete than they were.
> Were the original Pi A/B schematics not complete?
> Those look complete to me. If not then perhaps you could indicate what's missing.
Friends who know more than I claim that the composite video from the original Pi is not fully documented: apparently collectors of vintage televisions would like to be able to use it to generated 405 line signals but can't because the necessary details to hack away at a low level aren't available.
"the composite video from the original Pi is not fully documented: apparently collectors of vintage televisions would like to be able to use it to generated 405 line signals but can't because the necessary details to hack away at a low level aren't available."
Are these collectors aware that lots of obscure stuff that works may not be formally documented (not least, because it might break in the next release of hardware or software) ?
For example have they tried searching youtube for videos relevant to "old composite video pi". Based on what I found there seem to be quite a few things that can be changed relating to composite video output, seemingly changed enough to work with various old-school monitors, and/or on Pi variants without official composite video out.
Whether the possible changes are sufficient to work with a line rate of just over 10kHz (405 lines at 50Hz) isn't immediately obvious, and I've no way to try it. But others interested in 405 line TV presumably do have skills (and time and motivation?).
Have a lot of fun.
James, I would hope you could stop the MPEG2 ripoff, although I know it is Broadcom pulling this one off, not the foundation, pretty please!
Do you really think you have competition ? I mean, you have a great ecosystem of boffins and amateurs hacking on your stuff creating the most incredible things, as long as you copyright the design and layout, I guess you should be fine, right ?
And more RAM would be nice as well, I understand that would require quite a re-design ... but, you know, I think you will have to at some point, competition already sport 2Gb, admittedly at another price-point ...
Hove you considered adding PCI-E lanes for an M.2 port ? Killer feature, that!
The MPG2 thing is a PITA. Unfortunately there are still two regions where the licence is required, and since we do not region encode the Pi we cannot guarantee that any Pi will not be used in those regions. We would dearly love to get rid of the whole thing - it costs us more than we make back, to run the licencing system.
"The foundation perpetuates an elitist attitude that no one can do things as well as they can"
this IS a discussion about the Raspberry Pi, right?
I haven't found any of what you said to be true. Maybe some time surfing around the web site would be a good idea.
Sometimes the info you want _IS_ hard to find, but if you surf around a bit you'll eventually get to it. "Poorly indexed" might be a good criticism to use, and it's obviously not done deliberately.
Some links on their web site lead more down rabbit holes than to useful information. Yeah, that's irritating. Eventually you find the right set of pages and links and VOILA it's all there! Or at least most of it. Then you download the broadcom CPU docs and surf around other sites like elinux.org and sometimes their info isn't 100% accurate, so you occasionally tear out your hair over it, but in general, it's pretty good docs for something this inexpensive and flexible.
as well as the long geeky reads and of course the commentards!
Especially the quality of commentards, most notably James Hughes, THE Raspberry Pi engineer for VideoCore! Ok, I did not upvote James' post, because I think his claims are not 100% correct, however, I understand corporate-speak and cannot blame him for that, don't we all do it with our customers ;-)
I do wish to congratulate the Raspberry pi team for what they have done, I have all the B+ and the first B, big fan!
Diagnosing, openly admitting and replacing.... basically treating their customer's as mature adults!
There are very few organisations that would do this... most would deny the issue, take months to admit the cause and liability, then put in place a complex bureaucratic process that causes people to say "f*ck it!"
"TBH, we should have spotted this in testing"
Well yeah, that would have been nice. But the fact that you saying that in a public forum is not a career-limiting move is what we are applauding.
... although it would have been nice to see something about this in my RSS feed from the RPi blog (I'll go and sit in the nit-pickers corner shall I?)
Are we sure its not down to fake sd cards, or our psu's, or all the other excuses rolled out for the original Pi's by the foundation on its forums before they mostly fudged the timing issues on the usb bus bottleneck. Congratulations on admitting it this time though.
Mines the coat with a pocketbeagle in the pocket, because once bitten forever unsmitten.
PSUs with a high current capability but a ripply output are still not very good on the Pi. But that's not just a Pi problem.
Fake SD cards are also crap whatever you put them in.
USB bus is a bottleneck, but I don't know what the timing issues referred to are, never heard about these despite being a Pi user since the first 10,000 batch.
Mines the cupboard with the Beagleboard somewhere at the back of one of the drawers because let's face it, the OS implementation is a bit shite.
The thing is, when you're selling what is effectively a (massively overpriced) glorified breakout board for an MCU, there's remarkably little to truly screw up. I'm not saying there never was a notable issue with some of their hardware, but by and large Arduino being indeed more than a bit shite is down to basically everything else about it _except_ the boards themselves.
There are PoE solutions for the Pi that have been available many years. I had one from Adafruit which powered a Pi wanted on a mast with a short and very particular length of antenna cable run. That worked for a long time until this recent hot summer did for it.
It was, Upton observed, "dumb luck" that heavy load testing was done with one brand of switch while lighter testing occurred with the other.
Where does luck come into it? Given the historical problems with power on Pis I'd have expected anything in that area to be fully tested with all possible build configurations.
The longer I've worked in small IT companies, the more I've come to accept that testing is something that just doesn't seem to come naturally to most.
I've been gleefully handed hardware which is "Done" and "Fixed" which would have been borderline fatal if I were of the disposition to trust everyone not to be a complete idiot. "Completed" software which won't even compile, let alone install, is all too common.
I'm almost envious of this ability to be so slap-dash. My own perfectionist OCD would have me traumatised if I handed out such half-baked offerings.
RPi Foundation probably had a bit of a communication issue, in that the testers didn't know there were variations in the board. Sure, someone, somewhere should have know, and should have created a proper testing schedule, but with finite resources you have to draw the line somewhere, and you wouldn't naturally think a USB chip will have any interaction with a PoE interface. (Although it does sounds like it needs a bit more decoupling).
Yeah, it's a cock-up. Pity they happen, but they do. Publicly putting your hand up and confessing should be the norm. Now sort out the mess and move on.
"Where does luck come into it? Given the historical problems with power on Pis I'd have expected anything in that area to be fully tested with all possible build configurations."
Unfortunately switches that offer the same standards often vary on load.
Personally I don't trust TPLink for PoE, one minor surge near the threshold and I have to power cycle the one I have.
Yep, they work very nice (mine does pass Ethernet correctly, so, yay).
I'm thinking of building a widget that takes the USB plug (or any of the other plug variants) and feeds 5V into the IO connector in such a way that you don't get that USB plug sticking out sideways.
Thanks, that's a handy looking thing. Is it compatible/compliant with any particular PoE standards, or is it something you have to use with dedicated injectors (rather than a proper standards-compliant PoE distribution setup).
I could see a few uses for something like this, and a few more if it was compatible with some of the real PoE standards. Not all of the uses would be particularly IT-related either. E.g. various low power consumer gadgets powered (and/or recharged) via MicroUSB - everything from 10W LiIon-powered worklight to 5W DAB radio, and lots in betwen. Oh, and obsolete tablets/smartphones repurposed as small networkable computers. And.., well you can probably think of your own examples.
Something like this wired around the home/office/garage/dungeon could suit some applications rather more than the cheap/nasty BS1363 sickets with built in USB power which have been all the rage in the UK for the last couple of years.
Have a virtual beer from me, and another one for the correct answer (or equivalent) to the standards-compliance question.
Has BigClive pulled one of these PoE things apart yet? He's done various mains-to-USB gadgets, sometimes with very 'interesting' results, especially if they're based around badly implemented capacitatititative droppers.
"They are 802af compliant. I'm running Netgear GS724 PoE switches, and found no problems with them."
That's the correct answer :) (as in, the one I hoped I might get).
I've had various ProSafe PoE low-end brothers (802.3af compliant, afaik) for a while, scattered around the place for various explicitly-PoE reasons (network gear and cameras in places inconvenient for mains, that kind of old-school thing). Now it looks like I can easily get MicroUSB power from them too, for LED work lighting, mini-fans, Android gadgets, powerbanks... could I even use a decent battery or a small UPS to feed the DC power into the central 802.3af kit, maybe?
feel the power... ohhh, what's that burning smell :)
could I even use a decent battery or a small UPS to feed the DC power into the central 802.3af kit, maybe?
If you manage to find a PoE switch of the capacity you need, and then a telco version: those have a 48VDC input (in some cases instead of the mains power input, but usually next to one). But UPSes with busted batteries often go cheap-ish, and a lot of them run on 12V 7Ah or 12V 12Ah AGM batteries.
The smaller NetGear desktop switches tend to run off an external 48V power brick/wall wart, depending on size. Easy enough to splice a stack of 12V AGM batteries in.
It's not random, it's if you short out certain pins on the GPIO its can kill the PMIC. Its more comon on 3B+ becuase its the only one that uses a PMIC.....
Just dying is extemely rare, but can happen. Just like any other equipment. If it dies under warrantee, get it replaced.
We have 8 devices that claim to be 802.3af compliant power providers. Most of them work with most devices we plug in but there appears to be no rhyme or reasons why some devices won't talk to some providers. There have been situations such as 20 supposedly identical devices plugged into switch A and 16 work but 4 don't. Plug them into vendor B's switch and a different 4 don't work. We even have a Mobotix camera that used to work on most switches but now only works on just one port on one Cisco switch.
I've been using el-cheapo POE to network/5V microUSB adaptors for my Pi and usb powered mikrotik hap-lites. Just a little brick that plugs inline and converts incoming 802af to data and 5v usb, they're about $20usd for 4. Not as neat as the pihat, but no fan and they seem to be working well for what I'm doing.
to the unhappy days of early-Noughties wifi adapter purchasing for Linux, where not just white-box jobs but most big brands were prone to having their innards completely re-jigged without so much as a telltale hardware rev to warn buyers of its now paperweight status. Traumatic.
@James Hughes 1 if you're still hanging around, is there any way to identify the faulty boards? Are any steps being taken to withdraw them from sales channels? (I don't remember seeing anything about this in the article, although that was a good 5 minutes ago now #NanaMoon)
I can remember a severe problem with Western Digital ISA Ethernet cards at one of our customer sites. There were at two completely different implementations of one board - WD8003E if I can recall it properly.
Later versions of WD's own official IPX and NDIS drivers would work with all card revisions. Our system, though, relied on monolithic drivers from the network vendor at the time, and our customer's entire set of new desktop PCs suddenly stopped working after a client software update as the newer drivers didn't work with their version of the hardware. As the client software had to match the version running on the server, the site had to downgrade all of their severs to get their desktop PCs back.
I'm sure something similar also happened with a popular SCSI adapter around the same time.
Unfortunately its not easy to identify the problem boards, the chips don't have much in the way of ID. Keep an eye out on the blog over the next couple of days - more detailed information to be published soon.
Not sure about the channel - certainly the next production run will have the problem fixed, IIRC it's delayed until the fix is finalised.
As I said when it was first announced, a 3rd party PoE hat has been available for Pis for years. And, what's more, it's not a terribly difficult thing for a competent electronics hobbyist - granted, one with probably too much time on their hands - to cobble together for themselves if they really wanted it and had the networking equipment to support it. What advantage does the "official" PoE support - that still requires a hat - have over the one that's already out there? No one yet has been able to give me an answer to that simple question.
"What advantage does the "official" PoE support - that still requires a hat - have over the one that's already out there? No one yet has been able to give me an answer to that simple question"
Have a look at the Raspberry Pi official blog post on the PoE Hat problems:
https://www.raspberrypi.org/blog/poe-hat-revision/ (9 Nov 2018)
"it's not a terribly difficult thing for a competent electronics hobbyist - granted, one with probably too much time on their hands - to cobble together for themselves if they really wanted it and had the networking equipment to support it."
In the blogpost is a phrase along the lines of "if it's not properly tested it's almost certainly broken".
A DIY hobbyist (or similar) design might or might not expose the design issues which ended up affecting the official PoE Hat. Stuff that works fine in multiple sets of circumstances doesn't always work fine across the whole range of the intended application.
I have used this one from pisupply for years with my Pi 2. Never had any issues. It even survived a close lightning strike that killed the ethernet port on the PI, but replaced that with a USB ethernet adapter and the PI still works fine. The Pi lived in a waterproof box outside as well. Very tough little blighters!
The PoE HaT is massively over priced, instead of re-inventing the wheel they should have just purchased one of the existing PoE modules that are freely available on the market and left headers for that on the Pi (or built a cheap Hat for them).
I think this smells like an engineer designing in-house because they can, not looking at the market and finding the best value for the customer. "NIH"?
Last I checked, there are feed-through PoE adapters. Seriously...search Amazon for "PoE Adapter". They're even cheap. They output 12V; make a
small module bring that down to the USB level, and you're done. What is
this Hat thing of which you speak? Yes, the HAT would be great, fan and
all, but it it's the PoE functionality that's prime, a solution is
What I'm surprised at is that Raspberry peeps didn't give a replacement
switching regulator oscillator cap value to simply up the switching
frequency. Sure, will lose efficiency through the filtering, but it
will give better load regulation, if the chopping frequency is the issue.
What exactly has this function--Power-Over-Ethernet--got to do with the lofty intention of the Raspberry Pi's existence being the teaching of computer programming to small children?
Eben Upton has yet to admit to, or have fixed, one of the most egregious mistakes one could make on an electronic product made for public consumption, to say nothing of the "8-year-old masses" target audience; and one which has existed since the introduction of the very first Raspberry Pi: a SIMPLE, rational shut-down procedure which does not corrupt the mass-storage device. It is unimaginable that one would design an electronic product, especially one designed for this one's purported market, without a simple "on-off" switch.
But I forget...this has been explained previously as a 'benefit': a 'real-world' teaching experience. Silly me.
Biting the hand that feeds IT © 1998–2019