Joined in 2023 
82 
63 About this deal
I don't know what the limits are on the USB current, but would guess 5.08V x 0.46A = 2.3368W is too much against 4.97V x 0.37A = 1.8389W. Perhaps 2W is where the limit is ? The announcement blog post details the major differences, but there were two things I really wanted to dig into: Dominic 1:39: Aha! That takes the volts from the PoE — from your Ethernet cable, so it’s around 48V, and brings it out to those pins there then for the PoE HAT to take that, and convert it to the 5V, right? The input voltage is advertised to go up to 21V, but in reality it cant handle huge differences (21->3.3V for example), the inductor is not sized to do so. So I bought a replacement Raspberry Pi 4. I was able to boot it with the same SD card, with no changes to it's contents. Ubuntu came up just fine. It seemed to work exactly as the previous one had.
Power-sourcing equipment for an 802.3af Power over Ethernet network (such as a PoE Injector or PoE-compatible Router/Switch) Dominic 2:00: That is indeed extra taps on the primary windings. So this is non-isolated volts direct from the cable through the transformer — well not through the transformer, but it’s this tap on the transformer. And then those four pins go up to the PoE HAT, which then gives us the isolation required to generate then the 5V to power the Pi 5. Now how many pixels can we turn cyberpunk purple with the PoE+ HAT? Once I hit 250 pixels, the resistance of the strip became a major factor, and increasing the driven pixels wasn’t really increasing the load. The last pixels were a noticeably different color as a result. To continue the experiment, I switched over to testing at pure white, AKA the individual red, green, and blue LEDs turned to 100% brightness. In this configuration, I was able to drive 140 pixels. The PoE+ Hat reported a maximum current of 5.4 amps, while my PoE switch showed that port pulling 30.6 watts of power, at a respectable 87.9% efficiency. The hard limit I finally hit was 5.5 amps at the HAT, at which point the Pi power cycled. But I tried running the test a few more times, and every time, the Pi would reset after a few seconds at full blast. The current global semiconductor shortage — which you’ll almost certainly have read about by now — is constraining our supply of the original PoE HAT. In general, we’re weathering the shortage very well, and the supply of mainline Raspberry Pi computers, Zeros and our other products have not been affected (we’re very good at pipelining). Unfortunately, the first-gen PoE HAT uses silicon that’s in short supply.Unlike other RPI, on a RPI4 the power LED is fully under the control of a GPIO expander, and when booting Raspbian resets this IO expander so causing the PWR LED to blink off on reboot. On booting the bootloader enables it again. But if the PWR LED goes off (blinks) at any other time it means have an unfit power supply/power cable. in short, the PWR LED should be always on except for a very short time just before a reboot happens.
If your Pi is running full blast with USB devices plugged in, the difference isn't as dramatic. But if you're like me, and you have some Pis running lightweight utilities like a web server or some monitoring utilities, doubling the idle power consumption isn't exactly ideal.
Conclusion
And that's because the first PoE HAT used the 802.3af standard. It only guarantees up to about 13W of power to a "powered device" like the Raspberry Pi. I had my RPi 4 (4GB) running for a close to a year on the older PoE hat. Since the PoE+ released, I got hold of one and plugged it in. Although it was getting power through PoE+, the fan didn't automatically work. So I set the /boot/config.txt to have this line "dtoverlay=rpi-poe". After reboot the fan was working, without doing any update/upgrade to the OS. However, the RPi temperature was higher by a good 15-20 degree Celsius compared to the older version PoE hat and had to blast the fan to max at all times to bring the temperature down. To reduce the fan's noise, I set different temperature zones as shown by Jeff Geerling in one of his posts ( https://www.jeffgeerling.com/blog/2021/ ... essive-fan). It works to minimize the noise but the RPi temperature is hovering around 65-70 degree C at all times now. Previously it was around 45-50 degree C. I can think of one explanation that satisfies all the observations. The original HAT’s hottest point is between the HAT and the Pi itself. This is observable in the EEVBlog video linked above. I tested with the HATs installed on the Pis, making it essentially impossible to get a reading on the underside. Setting that explanation aside, my measurements indicated that the original HAT got very hot at higher power outputs, while the PoE+ HAT stayed quite stable. Above 7 watts of power output, the new HAT ran cooler as per my measurements. Dominic 5:20: So we’ve dropped — it becomes 90% efficient instead of… a lot less percent efficient. Eben 10:15: That’s very neat! Good. And how much more efficient is it than the previous design? Is it more efficient than the PoE+ design? Or is it —
The PoE+ HAT implements the 802.3at standard. When used with a compatible switch or injector this means it can deliver up to 25W, as you can see from this comparison table. If you do the math, 3 + 3 + 9 = 15W, which is more than PoE can guarantee, and doesn't even include the overhead of the PoE HAT's own power draw!
Features
So I don't understand what I'm doing to kill these Pi's. I'm wondering if anyone has any insight here. Pull evenly so that it detaches from all the pins at the same rate; do not pull one end of the connector off before the other
Great Deal 
New Comment