Raspberry Pi PoE+ HAT Review – Part 2: Problems

In Part 1 of this review series, I looked at some high-level changes of the new Raspberry Pi PoE+ HAT. Sadly this second part is more negative as I focus on some problems, from the baffling simple to the more complex and possibly fundamental issues.

After the initial problems with the original PoE HAT, and the public admission they’d not tested the HAT sufficiently, you’d expect that the Raspberry Pi Foundation would be darn sure to ensure that the new Raspberry Pi PoE+ HAT met the highest design and quality standards. Afterall they stated clearly:

If it’s not been tested, it’s broken

James Adams, 9th November 2018 – Anatomy of a product quality issue: PoE HAT

They also provided assurances that the problems of the original PoE HAT would be a thing of the past:

It’s embarrassing to have released a product with a bug like this, but it’s a lesson well-learned, and we will be improving our internal processes to prevent a recurrence.

James Adams, 9th November 2018 – Anatomy of a product quality issue: PoE HAT

I accept we don’t have the same power stability issues as the original PoE HAT, but there are still some problems with this latest offering from the Raspberry Pi team.

Problem 1: Fan Bolt Fouling Camera Connector

As noted by Peter Francis in a comment on the release blog post, one of the bolts securing the fan is too long and hits the camera connector on the Raspberry Pi board. I too experienced this issue with the two HATs I ordered from The Pi Hut. It’s baffling how this problem reached the final release and sold to customers in the first place.

Fortunately, there are a number of simple fixes customers can apply to mitigate the issue themselves:

• Replace the 12mm M2 bolt with a 10mm version
• Add another M2 nut on top of the fan to effectively shortening the length exposed beneath the HAT
• Reverse the direction of the bolt, so that the head of the bolt is under the HAT and the longer length of the bolt is above

Problem 2: Coil Whine from HAT when Powering Pi via USB-C

I discovered this issue when I took the Pi back to my desk for some power consumption testing and with the Raspberry Pi PoE+ HAT installed connected the Pi to an official Raspberry Pi USB-C power supply. I was startled by the noise and immediately pulled the plug! You can hear the noise for yourself in these couple of video clips. One with the USB power measuring device inline (which exacerbates the problem) and with the Raspberry PI 4 connected directly to the official power supply. Warning: you may want to turn your volume down a little bit first!

I struggle to understand how this issue was missed, especially after the lack of testing was identified as a cause of the problems with the 1st PoE HAT. Perhaps the engineers were only powering the Pi from PoE or given the power supply quality appears to makes a difference, were they only using high quality lab power supplies for their Pi source voltage, either via USB-C, or directly with 5V onto the GPIO header. It seems based on the outcome they weren’t using the power supplies they sell for use with the Raspberry Pi.

Update 9th June: I just tested powering the Pi with a portable USB power bank and it had the same results, noise and ~37VDC measured between the pair of tracks on the right of the board heading to the IdealBridge rectifier.

Investigating the Coil Whine

After the initial shock from the horrid noise, I powered the Pi up again via the USB-C power supply and confirmed that the Pi was working. I connected up and ethernet cable and confirmed the networking still worked and that the PoE switch didn’t have an issue with the device running in this state. All good at this point.

In this configuration, I’d only expect the low voltage side of the Raspberry Pi PoE+ HAT circuitry to be powered in order to allow the Atmel ATTiny814 microcontroller to allow the cooling fan still to be controlled, given the reduced airflow caused by the HAT. I didn’t expect to find high (~37.2V) on the test pads next to the 4 pin PoE header socket.

Without a full schematic, a significant refreshing on my electronic engineering and some extra test and measurement equipment, I can’t say exactly why the system is allowing the PoE+ HAT to be driven in reverse. Though I noticed that a voltage was making its way all the way to the 4 pin connector, at least one of the pins appeared to be disconnected, this may be a PD70224 (“IdealBridge™ Dual MOSFET-based Bridge Rectifier”) switching to a mode designed when using external power sources. It appears to protect the upstream PoE switch.

PD70224 / PD70210A – Reference Design (Raspberry Pi PoE+ HAT)

The PD70224 – IdealBridge rectifier (datasheet) is designed to be paired with the PD70210A – Power over Ethernet 802.3af/at PD Controller (datasheet) as it has been on the Raspberry Pi PoE+ HAT. The datasheet for the PD70224 shows an example configuration where a device or “Application” is able to be powered by either an external data source or from a Power over Ethernet connection:

The example shows a diode in line with the external power input, ensuring that the “input” doesn’t become an “output” when powered by PoE. However, when powered via the external power source the output of the DC-DC converter will have the external input voltage across it, whilst also powering the device “application”.

When running in this mode I believe that there should be logic to disable the PD70224 and PD20210A. Though I’ve been unable to confirm if this logic is in place, as I don’t have suitable equipment to ensure I can probe the precisely (I’d ideally need something like the PCBite system and a digital microscope). It does appear that power isn’t sent to the Ethernet Data lines. So maybe this is all working as it should, in terms of the design anyway.

MP8007 – Reference Design (Raspberry Pi PoE HAT)

Compared to the original PoE HAT, with its much simpler circuitry using an MP8007 – 802.3af-Compatible PoE PD Interface (datasheet) also has “wall adapter detection” which seems to shut down more of the circuitry, as I’m not seeing any significant voltage on the higher voltage side of the transformer.

The typical application circuit given in the datasheet shows a Schottky diode on the VOUT, which would protect the circuit from being back powered. I wonder if a similar strategy could be applied to the new PD70224 / PD70210A based HAT, or if the issues is with some other aspect of the design? Hopefully, we’ll get some answers from the Raspberry Pi Engineers soon.

Problem 3: Power Consumption & Heat

The power usage as reported by my Ubiquiti UniFi USW-24-PoE switch shows that the Raspberry Pi PoE+ HAT uses 5.2W, an increase of over 62%. These results are with a Raspberry Pi 4 8GB booted and automatically logged into an idle desktop, not under any significant load.

When powered via a Raspberry Pi 4 USB-C power supply, the Raspberry Pi with PoE+ HAT uses an additional 1.5W (60%) higher than without a HAT or even compared to the original PoE HAT. This extra power usage is being used to power large unnecessary parts of the PoE board, as covered in Problem 2 above.

Thermals

This increase in power usage also manifests itself in an increase in heat across various components on the new Raspberry Pi PoE+ HAT. The thermal image of the HAT included in the Raspberry Pi PoE+ HAT release blog shows that area marked as “Bx1” has a max temperature of 67.9°C, under a 2.5A load. The hot spot appears to reside between a couple of MOSFETs on the top side and a pair of diodes on the underside of the board.

Certainly from my own testing various areas of the board got hot to the touch. The MOSFET in area “Bx3” was the first is just above the array of ports on the side of the Raspberry Pi 4 and was the first one I accidentally touched as I went to plug in a connector and it was uncomfortably hot.

Testing Thermals

To better investigate temperatures I bought a UNI-T UT325 Thermometer, which allows me to capture data over extended periods of time using a couple of thermocouples. One measuring the ambient temperature, the other a specific area or component.

Testing was performed in a number of phases:

• Phase 1 “Shutdown” State: Power on the Raspberry Pi and issue the command: sudo poweroff The Pi doesn’t completely power off. It can be restarted by power cycling the PoE port or removing and reinserting the network cable. However in this state, the Raspberry Pi and PoE+ HAT combination still consume 4.6W, but the cooling fan on the HAT is no longer functional.
• Phase 2 Powered “On”: Power on the Pi to determine how quickly the temperature reduces.
• Phase 3 Powered “Off”: Remove power completely, to measure how quickly the temperature falls.

Test configuration:

• Raspberry Pi 4 8GB powered over PoE using a UniFi USW-24-PoE switch.
• Thermocouple 1 is secured with Kapton tape in the middle of the hotspot of area Bx1.
• Thermocouple 2 is ~15cm away from the Pi measuring ambient temperature. (Important as my desk is far from a controlled lab environment)
• All components are not within any enclosure.

Thermal Testing Results

Thermal Comparison

In the release blog it was stated that the new PoE+ HAT: “runs cooler, thanks to various design improvements”. Under low load this claim seems far from the truth. With two Raspberry Pi’s side by side, one running the original PoE HAT and the other the new PoE+ HAT, both booted to a command prompt only and left idling for around 6 hours, the Pi with the new PoE+ HAT is running hotter, likely due to the increase in power consumption.

• RPi 4GB PoE HAT: 39.4°C – PoE Power: 1.43W
• RPi 8GB PoE+ HAT: 46.2°C – PoE Power: 3.41W
• Ambient Temperature currently: 22.3°C
  In this instance the, RPi Temperature measurements are those of the Raspberry Pi 4 SoC.

It’s unclear if this “runs cooler” claim relates a specific test scenario, i.e. under maximum load. I just can’t tell and I don’t yet have a good way to test the maximum power capacity of the new HAT.

Thermal Testing Summary

During the “Shutdown” phases of testing where the PoE HAT doesn’t provide any active cooling and load should be at its lowest, the temperature increased to ~66°C, (41°C over ambient). I’ve yet to test this in an enclosure, but I suspect this scenario is a worst-case scenario and great care should be taken. It’s probably not an all that likely scenario either, but it can happen. At least if using a case fan powered from the Pi GPIO header, the fan would continue to run in this state and help keep things cooler.

Once the system is fully powered on the fan on the HAT quickly reduces the temperature to ~26°C above ambient with the fan speed varying during this time. Once power is removed the system cools quickly dropping to 5°C over ambient in 22 minutes and continuing to fall.

Problem 4: Raspberry Pi Enclosure Compatibility

The original PoE HAT has a notch cut out along the left edge. This meant it could be installed in the official Raspberry Pi 4 case (though I’d recommend against using that case). The Raspberry Pi PoE+ HAT doesn’t provide that same cutout, meaning it’s no longer compatible with the official Raspberry Pi case. There will be some cases that require changes to their design I’m sure, though some cases like the Argon NEO will continue to work, though you’ll need to leave the top clamshell off in order to allow the fan to breathe.

Problem 5: Fitting Issue With Compute Module Carrier Board

(Added 26 June 2021) A commenter on this post has noted that the standoffs supplied with the new PoE+ HAT are slightly too short when used with the official Raspberry Pi Compute Module Carrier Board. One of the inductors on the underside of the PoE+ HAT would be crushed/damaged against one of the full-size HDMI video connectors on the carrier board.

Like the wrong length bolts on the fan, it seems the Raspberry Pi team didn’t test actually using PoE+ HAT with their own products. As the commenter has noted a simple “fix” was to add a washer between the standoff and the board to provide the extra needed bit of clearance.

Conclusions

I’m not convinced the Raspberry Pi team have fully embraced the feedback and lessons learned from the previous HAT release. This new product doesn’t appear to be sufficiently tested and doesn’t look to have used the opportunity to utilise beta testers to cover scenarios and environment the Raspberry Pi engineers might have overlooked.

Unlike the original HAT, this new Raspberry Pi PoE+ HAT does “work” but there remain simple issues (screw length) and more complex areas to investigate, i.e. powering the Pi with USB with HAT attached, which is the most worrying of the problems above. The thermals of Raspberry Pi products seem to be an ongoing concern, this HAT is no exception, though something might be able to be done in some scenarios to reduce power consumption (and heat). This is compounded by claims in the release blog like “runs cooler” which under basic tests have been shown not to be true.

Given time, I may go on to evaluate using this PoE+ HAT within an enclosure, as the increase in overall power usage and resultant heat might be a challenge for some enclosure designs.

Product Links

• Raspberry Pi 4 (Amazon) (Pimoroni) (ThePiHut)
• Raspberry Pi PoE+ HAT (Pimoroni) (ThePiHut)
• Raspberry Pi PoE HAT (Amazon)
• Raspberry Pi Compute Module 4.0 IO Board (ThePiHut)
• Ubiquiti UniFi USW-24-PoE (Ubiquiti EU | USA)
• Uni-T UT325 Thermometer (Amazon UK | USA)

Amazon links, are affiliate links that help support the site, where possible I’ve used links that should take you to the product in the Amazon store in your region. Links to other suppliers are included for your convenience.