FlightAware Discussions

Which PoE for RPi4 and Airspy Mini?

Any temp readings with the Fan?

https://www.raspberrypi.org/products/poe-hat/

Recommended power source (injector) should be AF Compliant ( 802.3 af):

Other hardware needed

  • A Raspberry Pi 4 Model B or Raspberry Pi 3 Model B+
  • An Ethernet cable
  • Power-sourcing equipment for a 802.3af Power over Ethernet network

These are two RPI3B+'s in my attic.

This has a POE unit with a fan. It is using an airspy
image

This has no fan, however, I think it has a heatsink. It is using an FA dongle
image

I use Cisco SG350-10P in the basement and SG350-10MP in the attic. They are 802.3AF compliant and can provide ~60-100W of POE power. I am using about 12W to 33W on each switch.

1 Like

The injector I’m using is.
No temperature readings yet. It’s just sitting there idle.

Jon, what’s generating those temperature readings please? I’d like to add that to mine.

It’s using rrdtool much like the graphs.

Whatever it is, it will bite the graphs when installed at the same time.

I use cacti to monitor a few dozen devices via SNMP. https://www.cacti.net/
It uses rrdtools.

Joe’s adsb-receiver script also sets up similar logging.

The snmp configuration came from https://www.seei.biz/cpu-temperature-of-a-raspberry-pi-via-snmp/

sudo vi /etc/snmp/snmpd.conf

For external monitoring I had to change this section
#  Listen for connections from the local system only
**#agentAddress  udp:127.0.0.1:161**
#  Listen for connections on all interfaces (both IPv4 *and* IPv6)
**agentAddress udp:161,udp6:[::1]:161**



extend cputemp /usr/local/sbin/cputemp
extend gputemp /usr/local/sbin/gputemp

sudo vi /usr/local/sbin/cputemp

#!/bin/bash
cpu=$(</sys/class/thermal/thermal_zone0/temp)
echo print $cpu/1000 | /usr/bin/perl
echo

sudo vi /usr/local/sbin/gputemp

#!/bin/bash
gpu=$(/usr/bin/vcgencmd measure_temp)
echo $gpu | /usr/bin/awk -F '=' {'print $2'} | /usr/bin/awk -F "'" {'print $1'}
sudo chmod 775 /usr/local/sbin/cputemp
sudo chmod 775 /usr/local/sbin/gputemp
sudo usermod -a -G video Debian-snmp 

sudo systemctl restart snmpd
sudo systemctl status snmpd
1 Like

You can use 3 back-ticks ``` above and below the code to make it display nicely

No problem, if back-tick ` is not available in keyboard. Use [code] and [/code] instead. See example below:

How typed How will appear
[Code]
text1
#text2
## text3
##tetx4
[/code]
text1
#text2
## text3
##text4
1 Like

Thanks, I’ll take a look at that.

My shack air temperature is around 22°C and with the settings at default, the fan was kicking in every few minutes for twenty seconds or so. The default is for it to start at 45°C.

I’ve tweaked config.txt with this line:

dtparam=poe_fan_temp0=30000,poe_fan_temp0_hyst=5000,poe_fan_temp1=35000,poe_fan_temp1_hyst=2000

Basically it’s on all the time now but once I actually work out what those parameters all are, I may tweak it to be temperature controlled. Or I may just leave it like it is so it’s on all the time!

The temperature is hovering between 30° and 32°C. This is with no heatsink on the CPU.

2 Likes

That’s just the cable that came with the injector, I use CAT6 up the mast.

And replying to myself, just because I can.

The gap is while I was fitting the HAT and setting up the PoE injector. The temperature straight after is fractionally lower because it was only coming on for a few seconds at a time. You can see where I tweaked the settings to be on continually. The slight lift is when I made a change which had the fan cycle off and on but I put that back to be running non stop. I think this is more effective than the fan shim.

This is without any heatsinks.

2 Likes

CAT6 cables come in two varieties:

  • UTP (Un-sheilded Twisted Pairs)
  • STP (Shielded Twisted Pairs)

https://www.amazon.ca/dp/B00LE6JEJC




https://www.amazon.ca/dp/B072Q1CL7J
image

This is what I’m using.
It seems to cope quite happily, even with QRO HF going up the same mast with the coax less than a foot away.

2 Likes

STP is mainly used in Datacentres and really noisy environment as far as I am aware. My company uses it in a D/C in Europe but not elsewhere.

My radarcape came with a cat 5 STP cable.

I have one between the attic and basement, however, it is much thicker than regular UTP. I think I only managed one UTP and one STP in the 1" plastic conduit(one or both were cat 6). I think I could have put four UTP cat 5 in there.I don’t need multiple now because I managed to get power installed in the attic.Before I was powering one RPI(or radarcape) and one odroid via POE.

I was messing regarding the UTP - (Sort of anyhow) But for POE and radio, FTP/STP is the way to go. For one, UTP can get noisy, secondly for POE applications the shielding acts as a heat sink and decreases noise, so two birds with one stone using good shielded cable.

Those temps look great with the original hat. I went the other route with the mini hat and passive heat sink case (Flirc), since I always feel it’s a good idea to have some grounded shield around things. Not sure how much it helps in the end…

2 Likes

Here is a stress test I did on an RPI4 with Raspberrypi.org POE unit in my attic. It does have a CPU heatsink.

/boot/config.txt command

dtparam=poe_fan_temp0=40000,poe_fan_temp1=45000

command to install stress application

sudo apt install stress

Command to run stress application. You can also use another application to stress the CPU.

stress -c 8 -t 1999

You can see the times when the CPU was at ~100%. The first drop is when I enabled the new fan speed settings.
After the change, during the stress test, CPU barely gets above 43C.

An Article on RPI3s benefiting from a fan
https://downey.io/blog/raspberry-pi-3-heat-sink-comparison/

1 Like

Here is another article but with the pi4.


Interestingly, cooling performance was best with only the fan. It degraded slightly when adding both a heat sink and the fan.

1 Like

I’ve done my own stress testing.

First you can see the temperature with the default settings of the PoE hat, this is from the start of the graph to around 08:40.

Then I switched to have the fan on at full speed continually until 11:40

Then I changed the fan configuration to be the same as @jonhawkes2030:

dtparam=poe_fan_temp0=40000,poe_fan_temp1=45000

Translated, this means turn the fan on at the lower of the two speeds at 40°C and then go faster at 45°C.

I ran the same stress command until 13:00 (with a very quick dip when I had to restart the test)

stress -c 8 -t 1999

You can just see that at 10:40 my central heating kicked in and the shack temperature rose by ~1°C.
While the stress test was running, the temperature was around 45°C to 46°C but never higher. It was around the same as the Pi at idle without any cooling.

htop

2 Likes

They also used just the heat sink alone, and that only lowered the temperature just a few degrees.

It was one they sell for their cases, which is not that high but with a footprint way larger than the cpu. Maybe it just is a bad design, and smaller heat sink with the fan directly attached would perform better.

Still, I‘d say any airflow will be enough to keep temps in check.

1 Like

@LawrenceHill,

I did look at that post, however, it was on a commercial site advertising their own fan.
It is not clear if they used heatsink tape or adhesive.
Their tests got a lot hotter than mine. I did not get hotter than 45C. I may not have used the same application.

For airspy and/or attic use, a fan is clearly going to help. I think a fan and heatsink even helped my RPI2 years ago when trying the airspy in the early days.

The heat sink they used was also one their own and uses thermal tape, pre-applied.

1 Like