This is literally the *FIRST* video I’ve seen any mention of Phoronix and Michael’s tireless work, after so many years and years. Thank you for bringing more attention to this critical world resource ! ♥️
And BTW: it very much is surely back feeding 60v out the Ethernet interface but since 802.3u specifies common mode rejection requirements higher than this, it shouldn’t damage any upstream components. But still: it seems more a first-timer JLPCB tinkerer design than an official corporate offering. So bad…
I think it would be interesting to see the efficiency compared to the original hat when both are plugged into a POE switch instead of a POE+ switch. It sounds like the higher voltage of POE+ is a factor.
@asdrubale bisanzio -- No, the primary becomes a magnet. Calling the ferrite core a "magnet" is like calling the "keeper" placed across the old "horseshoe magnet" a "magnet" since it _conducts_ magnetism. Essentially, calling a magnetic conductor a magnet is like calling the copper wiring in your wall a "generator" or calling the battery cable in your radio (remember radios?) a "battery" since it _conducts_ power from the battery. Or, like calling the subway train "my job" because it takes you to work. N.B.: This does not apply if you happen to be the motorman!
Another quality video, thanks for that. I had big expectations for this new POE hat and am genuinly surprised that the board comes with these really obvious flaws. One would expect that these would pop up during testing.
After watching this, I’m definitely waiting for the next revision. My rack has 4x pi’s. 2 running pihole (doesn’t get hotter than 37c, so no need for a fan) the other 2 run servers so cooling by heatsink is perfect, I really want fans to become more silent. Thx Jeff
"Don't ad-hoc a sentence in the middle of a pre-written sentence"... oh how badly I can relate to that statement. Glad to know I'm not the only one suffering there.
@@chicken_punk_pie True they do have different connotations, but I'm not totally mistaken they are "in true meaning" quite closely related. And an ad-lib can also be "an Ad-Hoc action". E.g. if You, while reading Your script, notice that You have written something totally wrong, then You might try to correct it "Ad-Hoc" by "Ad-Libing". And most "Ad Libs" are intrinsically done "Ad hoc", are they not ? English is (in my opinion) a "curious language" that can be used in all so many different ways while in essence expressing the very same thing. Best regards.
The idle power consumption is probably due to the planar transformer used. I've seen the idle current double when switching from traditional transformers to those, which would match the findings here. I love planar transformers, but the idle power consumption is a drawback.
I was thinking the same things... Since HP (HPE) bought Aruba, they're sticking that label on everything. That's a fairly simple HP Procurve switch, with new vinyl. I have several that have been in continuous use for decades. (one had to be replaced, and one has a bad port.)
Hey Jeff, your PoE+ fan was pretty quiet. I just fitted mine last weekend which sits in an open case. But what I was not expecting was the mosquito sound every 20 seconds then quiet 10-15 seconds and then another 20 seconds of a swarm of mosquitoes - continuously all night long. In the daytime I didn't really notice it but this little chap sits behind a monitor in the bedroom and during the night OMG no sleep that night at all ... So the next day that hat had to be removed :-( Also another gotcha was at the moment there does not seem to be any passthrough GPIO headers for the POE+ hat unlike the old PoE HAT. Due to the POE header is not passthrough. I think all you can do at the moment is to get old POE Long Headers and hack the 4 POE ones down to fit. Btw another great video :-)
Also, the new hat doesn't have a cutout to thread the camera cable through, making PoE webcams a little more difficult. I found it odd that there hasn't been an official case, or third party one that I can find that neatly support the hat and camera.
Great video! I always assumed that POE would not be very efficient with power use at either the router or client device, but I"m looking forward to your video about it.
With PoE using a high enough voltage on the cable, things like voltage drop and associated power dissipation can be minimised, so it comes down to the efficiency of the voltage conversion circuit of the power sink/client device. I could see it being theoretically plausible for a PoE switch's bigger, maybe-higher-quality AC-to-DC adapter to have sufficiently improved efficiency compared to the tiny, as-cheap-as-possible power bricks supplied with random consumer devices, that it works out _more_ efficient to use PoE. However I've not run the numbers, nor experimented with it myself, so this is just an idea :)
Great video, I use a lot of Poe at home, Cisco phones and cctv cameras and back in January I migrated a bunch of services to 4 pi4s and a pi2 this includes homebridge, asterisk and 2 piholes. On my enterprise switch under normal load they pull about 2.2 watts each but when backing up they get closer 5 watts, I don’t have anything plugged in to them except network and the micro sd card. (I don’t need disk performance) since then I have carefully modified the setup using the sd card to boot then creating a mix of sd card ,RAM disk and nfs mounts means backup aren’t needed. The interesting thing is using my Hopi power monitor between the mains and the switch if I keep the Poe power draw for the 4 pi’s under 14w the Poe switch does not seem to draw any more power from the mains compared to no Poe being used. Effectively getting power for free.
5:13 it's not a magnet but core/rod made of magnetic-conductive material in order to receive magnetic field from primary winding and induce voltage in secondary winding.
yeah, I'm kinda surprised at the whole poe hat controversaries. Making a SMPS is very simple these days, even cheap chinese buck converter boards from aliexpress are 90+ (or at least higher 80s) efficient.
Given the large vin/vout delta It would be interesting to see if the effect of changing the buck converter components could improve the efficiency E.g A) A larger value inductor swapped out B) if the switching frequency can be changed
My Pi4 (4GB) with RaspiOS Lite, pi-hole, 256GB Intel SATA SSD with the second PoE hat revision (without the mezzanine board) only draws 5.05W, 94.60mA, 53.46V My PoE switch is the EdgeSwitch ES-8-150W PoE+ switch. Thanks for the warning, I was planing on getting the new PoE hat but now I will wait!
I have a WaveShare PoE Hat for my Raspberry Pi 4B 8GB RAM model and I find it much better. The WaveShare PoE Hat has a better placement for the fan where it does have a slot to feed through the cable for the camera and also a slot to feed through the cable for display(for screens), It also has a switch to turn on or off the fan, it also has a pass through GPIO where you plug the PoE Hat as normal but at the top of the hat it has GPIO pins on top to connect more things to it. Lastly it also has a USB port on it as well to plug something into it. I also have an extension board for my Pi as well which connects to the audio jack, the 2 micro HDMI ports, and the USB-C port and when it is connected, it gives your Pi the audio jack and USB-C as usual, and you can now use regular HDMI cables to the Pi instead of using the micros and you also get an extra USB port too. I recommend those. If you want I wouldn't mind you reviewing the 2 I mentioned just to get your opinion on it.
Great video! Maybe you could test some of the other brands that make PoE hats for the RPI and make recommendations as to which are good and bad? Also you could do a comparison of the best of the other brands PoE hat vs the official rpi brand PoE hats? It would be interesting to know if third party brand PoE hat is better than Official PoE hat.
I like how you said "oldER af" standard versus the "old af" standard. Smart move. BTW, they effectively came out with a higher clock speed rpi4 via the Debian Bullseye release! 21.11.11
To all the folks including Jeff) that are raving about the greatness of Phoronix.. I sure hope some (most?) of you are contributing some goodness and Thanks in the form of a buck or two. ($$). Come on, people, let’s show some greatness ourselves.
My Engineer PA-01 tool (combination crimp tool, insulation stripper, wire cutter, etc.) includes a screw cutter for M2, M2.6, M3, M4, M5 screws, and it sees a fair bit of use with my 3D printer and related projects. www.engineertools-jp.com/pa0103 Sometimes I just don't have the right size of screw on hand, or I need a screw in a weird length. Being able to easily cut a screw to length with a simple hand tool is pretty neat 👍 Although even M3 screws require a concerning amount of force to cut with that tool, so I'm not sure I'd like to try cutting M5 screws/bolts with it! 😐 And it doesn't ruin the screw threads, as most tools would! Engineer tools are definitely not a super budget option in this age of Chinese mass-produced tools (and which are often perfectly adequate for the job), but they're not unreasonably expensive either, especially since all of my Engineer tools are high quality :)
Some switch mode isolated DC-DC converters require a minimum load current to maintain regulation stability. This is usually the case with low cost regulators that don’t have an opto-isolated (analogue) feedback and instead they monitor the current through the primary winding during switching to estimate the output voltage.
That minimum load current could easily be 200mA, so at 5V that’d be 1W. One way to guarantee the minimum load is to place a resistor in series with a zener diode. As the unregulated voltage rises above the zener voltage, it starts to conduct with the resistor becoming a shunt load. This is more efficient than just strapping a load resistor across the regulator output since it only conducts when the voltage drifts too high rather than all the time (which would hurt the efficiency figures under normal loads).
@@alexscarbro796 If voltage rise is the issue, a zener load that only draws current above 5.2V might (or might not) stabilize the smps. A more complex solution would use a current sensor to reduce or disable the dummy load as actual current draw approaches the minimum. Anyway a well designed PoE+ circuit should have enough complexity to avoid having a minimum load or wasteful own load.
You can not safely provide DC positive energy from two different sources without connecting the ground (negative) together. Think of it as a voltage difference that provides current. When you connect grounds it ensures the same "voltage difference" is established between the two sources.
Thanks! This is a better explanation than others I have seen, and makes more sense. Intuitively, I know that the cheap non isolated adapter I got to power a pi zero from 24vac would cause blue smoke when connected to a DC power supply, but couldn't quite understand why.
Thanks for the heads up. I'm about to rig up something similar, and I've been curious if I should wait for the POE+ hat or just grab POE's. I see now I'll need to figure out my power requirements, and capability of my switch.
They don't think much about their users and their products reflect that, ivory tower syndrome. They had a nice official case with swap out side panels yet they flipped the ethernet and USB position on the Pi4 making not only their own but lots of third party cases obsolete and they switched to the awful micro hdmi ports that literally nobody likes.
Having the same layout for like a decade is pretty good already. Sometimes, designs need to be changed in order to improve a device. They switched to dual micro hdmi ports because a lot (i assume. they wouldn't change it for just a dozen buyers) of customers use rpis for digital signage and wanted to drive two monitors at once, and couldn't fit two full-size ports. I personally think this wasn't a great idea, as the rpi is supposed to be an entry-level computer for learning and hobby projects. Businesses that need to drive multiple monitors for signage can afford more expensive options. Unfortunately, if they change back to regular HDMI now, we'll have a *third* layout that's not compatible with neither 1-3 or 4 cases, so we're probably stuck with those shitty ports for a while.
The issue here is lack of competition. The Pi is pretty much a monopoly when it comes to SBC, so there's no incentive to even listen to users. Whatever they do, people will complain, but buy it anyways, defeating the purpose of complaining about it.
Almost 100% of changes that the Pi has gone through were the results of USER REQUEST. “Faster CPU” “More/faster RAM” “Faster USB” “Faster network (gigabit)” “Better/Faster Video out” “4K Video!!!” Well, folks, all these changes required faster chips, and faster chips required PCB lay-out changes.. And some of these layout changes required re-locating traces and chips and ports. Some changes were dictated by physics (signal, current, AC-coupling, DC-coupling, inductive coupling, capacitive coupling, thermal coupling, etc & etc & etc) requirements.. PCB layout can be fiendishly difficult. If the “deaf, ivory-tower snobs” had never listened to users, we would still be rockin’ Raspberry Pi 1A, 256M, 500MHz CPU, USB2, 100Mb network, 800x600 video…..
@@ernestgalvan9037 How long it took for them to let us boot from USB again? People have been requesting that since the first model. Same for having the source code for the GPU.
Wow, I'm starting to learn about the PoE HATs and now I know that I have to choose an overpowering hat or a normal hat that can fit into a normal build.
When I first saw the PoE hat I thought the price was too good to be true, and it was lol. It has terrible coil whine when the Pi is off and the threaded spacers it came with weren't long enough so it's really easy to overtighten the board and cause pretty bad flexing but if it's too loose it won't turn on because of the tiny socket header they used
7:55 "It get's up to 54 decibels compared to the old fan at 45 decibels. That's a pretty significant difference." Me * sitting right next to my AC blasting at full speed *: "Yeah that's way too loud."
Jeff, thanks for you videos, especially those on the Raspberry Pi. And, I really appreciate you putting a part of the failed recordings in the end. I always thought, I am the only one to produce such nonsense when recording a video :-)
You hope a Pi revision could output more USB power so this thing is worthwhile. To quote Eames, "You mustn't be afraid to dream a little bigger, darling." Why not a Pi version that just includes PoE, so all heatsinks and accessories are compatible with it?
The only actual problem I saw in this video was the screw thing, and maybe the SMT connector... but like you stated, rarely would be an issue. I especially found the USB power test with the hat on to be a weird thing to test for, not that it's bad to test for it, but the way it was framed felt like you were trying to say it was an issue. By the way, I'm sure it was done as a joke, but seeing you saw into that screw still attached drove me insane hahaha.
If people are needing more than 13 watts to power whatever setup they are running may be the raspberry pi is not the best option. After all, you can get other options in the soc space that offer decent power efficiency with more performance and support for more peripherals without needing to kludge a solution together that may or may not run for the long term. At 15-20 watts idle you are already into intel atom microserver space. The other thing to keep in mind is most rackmount switches, especially PoE ones, consume quite a bit of power as well.
I was happy when I managed to get a PoE HAT for my Raspberry PI 3. That thing whined worse than my kid having to do chores ! I just abandoned using PoE for a Pi after that. Looks like I will have to wait for longer, and a usb-c power adapter will have to remain in my mini rack for my HASSIO Pi4.
Nope. In fact, if you need to shorten a screw or bolt, *you do it with the nut installed*; that way, when you back the nut off, it cleans up (mostly) any burrs you left by cutting.
I don't recommend it either yet, mine makes that noise as well, and since I have it setup in my bedroom, while it's operating, it makes another hard to hear noise. Not to mention that if you plan on leaving this running 24/7, it will get really hot. They should've put in a bigger fan
It would be better if the regular PoE version was a bit cheaper than this PoE+ version. If I'm ignorant of the efficiency issue I would just see two options for the same price, one providing 2.5 amps and the other supplying 4 amps. Then I would buy the "obviously" better version because it appears more versatile. If the price were different I would be more likely to stop and consider if I actually need the extra power and the associated inefficiency.
@@JeffGeerling Not sure if there is demand for that, but I would love towatch a review about those Blinksticks. Maybe with some examples on how to configure/use them :)
I finally worked it out, The new Raspberry Pi PoE+ HAT use for DVR Security Surveillance to control the PTZ Cameras, It will not work on any ADSL or NBN Broadband Router and also TP-Link Powerline Wi-Fi Kit Extend Wi-Fi to Every Room by plugging in the Blue or Yellow Network Cable but I was hoping that I was going to use it to power my Raspberry Pi. Obviously not gonna workout for me, not what I was hoping for :-(
@11:46 That sound is the sound of the magnetics (flat plane transformer, most likely) either saturating or starting and stopping. For the hey of it, use a pencil eraser to press down on the transformer's core. That should at least attenuate the sound if not actually extenguish the sound.
I have a Pi4 ( Raspberry Pi 4 Model B Rev 1.2) with a single 5 TB USB3 spinning hard drive. Both are running off a POE hat and the switch shows 5.7-5.9W when idle, and up to 8.something W when actively doing a backup. I do not have a " /sys/devices/platform/rpi-poe-fan@0/power_...." to show that number.
I have the new Poe+ on my cm4/io board. Admittedly not running bunches of peripherals. Guess it's time to see what happens.🤓 Oh and my fan throttles with the temperature and I didn't do anything 🤨
It should throttle, but if you use the defaults the fan starts kicking in around 45 or 50 (can't remember the default). I'm okay with the pi running at 60 or so, so I keep mine set to only start fan at 60 and ramp up around 75
@@JeffGeerling Here's the rub, on a cm4/ioboard the Poe+ with fan isn't near the cm module. So, use in the efforts to control cpu temp is questionable. And while I have your attention, I can't get BT working on the cm4 iob. It worked 1 time for just long enough to start a yt. The BT speaker works fine on most of my other devices including pi 4 and 400. Any suggestions?
Here's one to learn from my mistakes, I made a fool out of myself by not doing any calculations before testing. I'm testing out a standalone SDR project that will be in an enclosure located next to the antennae, obviously for a clean one cable solution I went with a PoE hat. Turns out running 3 simultaneous SDRs (with one supplying bias power to an LNA) was asking too much and I bricked a Pi4 which thanks to my aforementioned foolishness was one of my deployment 4gb models instead of the 2gb model I have reserved for testing so I wasted even more money. I only really had trouble with NESDR nanos I was testing, if you've never used them they actually get too hot to touch which should be a hint at the power consumption. Running 3 SDRs at once isn't a stretch as I plan to log and plot ADS-B, OP25 trunk tracking, and another for SDR listening so it's possible I might hit 3 at once. One thing I did learn from all this is that low current supplied Pis behave in an insidious manner, running headless and/or in CLI mode you get no gui based low power warning. I worked on an issued I had with audio playback for days even bought a few different USB audio dongles thinking it was just an issue with linux audio drivers which are notorious for being finicky. I was getting skips and lagging in audio playback of a network VLC stream and also when listening in GQRX. If you're having issues where your pi seems abnormally slow with random hiccups you might be having a power supply issue.
I haven't used PoE miself, but for what I can read on Wikipedia 802.3af and at lack power saving features, and it's a common complaint that there's a 4.5w per port power consumption with 802.3at, so there's not much to do on that front afaik. On the other hand, the rest on the issues on the PoE+ hat are hard to justify after having issues with the previous version, specially with glaring mistakes like the screws that could have easily been fixed. RaspberryPi is not a new company anymore, and I'll expect a bit more quality control, specially since they're now supporting industrial use cases.
Anyone buying a switch today should consider a POE++ or bt standard switch to get lots more power per port. What are your thoughts of the external POE power adapters? I'm thinking of getting something for a Pi4 project and won't have room for the POE hat.
The PoE (802.3a*) specs account for power drop across long distances (up to the rated distances for each type of wire). PoE+ is rated up to 25W at the Powered Device, whereas PoE is rated up to 12.95 at the Powered Device (though you could receive a tad bit more if you're closer or use better cabling, assuming your switch/injector is pumping through the maximum spec at 48v).
I slag off raspberrypi all the time but really, this PoE hat will do even if there are some ways the other PoE is preferable. Just glad to see this thing exists tbh. I am however kind of happy that its power drain reduces the only useful thing about the pi3A (its low power) because that low-RAM 512MB thing is ridiculous especially seeing as it has the tasty protection rings of the 3b and ironically you cannot use them in a VM because you're out of RAM. Of course you'd want the protection rings for a webcam in a VM and it would use power. This PoE gets and extra point from me for kicking the Pi3A in the arse, exposing how stupid it is. Damn, just make more 3b or 3bplus.
I would love to see your thoughts on some of the other PoE hats for Pi. There are two versions of the LoveRPi Power-Over-Ethernet that might be interesting on Amazon.
@@LiLBitsDK 'magnetic material' isn't the same as 'magnet'. For more information of ferrite materials: en.wikipedia.org/wiki/Ferrite_core But to be specific: a ferrite core is usually not much of a 'magnet', put a piece of steel near it and you won't get much of a reaction (possibly none). The point of the ferrite material is it has a very low magnetic permeability (basically amplifies any applied magnet field, which massively improves the coupling), but has very high resistance to electrical current (induced eddy currents result in heat buildup and energy loss).
@@LiLBitsDK Yup, IF the material maintains a magnetic field without the presence of a current, or other factor. Take an iron nail: rub a magnet along it in the same direction multiple times and you've 'magnetized' that nail, making it a magnet. That is the basic principle of magnetic recording. The core of a transformer OTOH has an alternating magnetic field passing through it during operation. This, by it's very nature, means when you turn off the current, there is basically no lasting magnetic field. If fact, if you used a 'magnet' as the core of a transformer it wouldn't work right (if at all, depending on the currents involved).
Guess it would be more efficient to use one poe+ to usb/ 5V Out switcher-pcb and go to multiple raspis from there. When you optimize for high loads, the low end always gets less efficient and vice versa. Coil inductance and frequency dont fit for all loads.
I feel, if you're adding peripherals to your Pi, you can afford having a dedicated power supply too. The only Pis I run on PoE are the one's where I can get away with a single cable to them
I would actually like to see Raspberry Pi's support a simple power rail connection - even if just a screw. This will provide a lot more options for building clusters.
why aren't we just using QC or PD standard but over the ethernet cable? the devices can already communicate, the cables should definitely be able to carry the 20V5A...
Higher voltage and more power = more heat. Mikrotik stuff that are powered with poe+ run really hot for somereason. The high voltage increases their tempature almost doubled compared to normal when running their 24volt powerbrick or injector.
To those comparing this to dumb "PoE splitters". Those don't deal with the specific voltage range used by 802.3af, nor do they do the handshake to agree with the switch about permitted power draw etc. Adding support for non-standards like Cisco PoE adds even more complexity. Ultimately, each signal pair and it's signal transformer acts as a logical DC wire and the PoE extraction circuit must determine and negotiate polarity and current draw.
I have the POE+ hats (4 of them) and they buzz like an angry hornet. Thoroughly vile. Definitely use the Poe hat unless there is no other option… now where did I put those aspirins….