Axt1800 running hot, 75C

Johnex · 2022-08-10T12:58:40.533Z

In the end this is up to you. Electronics want to sit as cool as possible. As in the other thread, there is an electrolytic capacitor on the board, those have actual liquid inside. Temperature is directly correlated to the lifetime of the capacitor. So if you can run the fan at max, i would do it (and do currently).

You can see a typical lifetime chart, it’s almost exponential so every degree counts:

image1035×487 70.5 KB

wvlad · 2022-08-10T13:00:24.132Z

I did find the answer myself, just read the entire topic above, provided by sincere @Johnex. So the answer is - optimal T for the motherboard used at this router is 0-65 С.
Fan wont launch until T become too hot - this is a bug at soft, fan should work. If you are familiar with Linux CLI, in order to launch a fan we need to:

1. ssh root@ip-address-of-a-router # use password from 
2. /etc/init.d/gl_fan stop # that command will stop the daemon responsible for fan
3. echo 255 > /sys/class/thermal/cooling_device0/cur_state # that will set up the max speed of fan

I did the steps above and that worked perfectly fan is having its fun now!)

However I think it would be great to have schedule that will track the T and will adjust its speed automatically, wouldn’t it?

wvlad · 2022-08-10T13:03:58.020Z

Totally agree it is a must to have cooling. What do you think about fan speed schedule? Is it possible to make it automatically somehow?

Johnex · 2022-08-10T13:07:55.928Z

Lower down in the thread you can play around with the PID for the fan controller:

I found a way to set the gl fan! Technical Support for Routers

So this pdf on the mango SOM used in AXT1800 mentions the temp range for the commercial boards (which I assume this is) is 0-65C. The gl_fan app only polls the temp and sets the fan every 20 seconds, so If you have a large load spike I suggest bumping the integral -i and derivative -d values of the PID loop as it will not have very many readings to work with to ramp up or smoothly change the speed. In other words- If the load spikes you want the fan to come up to speed quickly. If you only se…

I didn’t really get it to work, for me the fan did not run at full speed even though it needed it to lower the temperature, but i just put the values @nopro404 suggested. Maybe you can give it a try and see if it works for you.

asdffdsa1122 · 2022-08-10T13:16:05.909Z

tl;dr, leave the fan running all the time.

Johnex:

Electronics want to sit as cool as possible

yeah, back in my youth, when not re-aligning the r/w head of 12" floppy disk with an oscilloscope.
the main issue was thermal creep.
if an integreated circuit was not soldered into to its dip package,
then the pins of the IC, would heat and cool over time.
that would cause the pins of the IC to expand and contract, thus causing the IC to creep up and out of the DIP package.

asdffdsa1122 · 2022-08-10T13:38:25.370Z

this is great detailed info,

Johnex:

i just put the values @nopro404 suggested. Maybe you can give it a try and see if it works for you.

i did try this and the good news is that the fan does seem to change speed
but getting errors

/etc/init.d/gl_fan stop
gl_fan -v -t 60 -i 4 -d 20

root@HOME01:/usr/sbin# gl_fan -v -t 60 -i 4 -d 20
2022/08/10 09:34:06 debug (gl_fan.c:148) set_pwm:-12.000000 proportion:20.000000 integration:8.000000 differential:-40.000000
2022/08/10 09:34:06 debug (gl_fan.c:151) current_temp:62 current_error:2 total_error:2 last_error:2 prev_error:0
2022/08/10 09:34:26 debug (gl_fan.c:148) set_pwm:70.000000 proportion:30.000000 integration:20.000000 differential:-20.000000
2022/08/10 09:34:26 debug (gl_fan.c:151) current_temp:63 current_error:3 total_error:5 last_error:3 prev_error:2
^C
root@HOME01:/usr/sbin# sensors && gl_fan -s && cat /sys/class/thermal/cooling_device0/cur_state
ath11k_hwmon-isa-0000
Adapter: ISA adapter
temp1:        +59.0_C

ath11k_hwmon-isa-0000
Adapter: ISA adapter
temp1:       -255.0_C

2220
70

nopro404 · 2022-08-10T14:54:05.510Z

asdffdsa1122:

i did try this and the good news is that the fan does seem to change speed
but getting erros

The error in the gl_fan output is basically the ‘deviation’ from where PID control loop target is – how far off from goal it is. It’s not an error per se.

wvlad · 2022-08-10T15:09:08.494Z

@nopro404 Is there any extra option/coefficient to increase fan speed from its “basic” values during that schedule? Where we can find the initial logic of it?

nopro404 · 2022-08-10T18:25:27.763Z

wvlad:

Is there any extra option/coefficient to increase fan speed from its “basic” values during that schedule?

Sorry, I don’t follow. the default values from gl_fan from its usage output

root@GL-AXT1800:~# gl_fan -help
gl_fan: unrecognized option: h
Usage: gl_fan [option]
          -t temperature   # expected CPU temperature, default is 90
          -p proportion    # Proportion parameter in PID algorithm, default is 10
          -i integration   # integration parameter in PID algorithm, default is 2
          -d differential  # differential parameter in PID algorithm, default is 10
          -s               # print fan speed
          -v               # verbose

If you want it to to spin faster earlier you can set the temperature lower, and use the I and D values I posted earlier. Or play with the values to see what you like.

Hope thats helpful.

EDIT: Just updated to 2020-08-08 4.01 beta 5 and they lowered the default temp to 85.

asdffdsa1122 · 2022-08-10T18:51:13.855Z

as i am working on my tmux/bash skills, i created a very simple script,
to display these current values on the tmux status bar, once per second

temperature
speed
state

and scripts to change the fan speed.

— fan.stop.sh	stop the fan
— fan.max.sh	set fan to max speed
— fan.custom.sh	set fan to custom speed, valid values from 0 to 255

and stress.sh to max out the the cpu, which will raise the cpu temp

in .tmux.conf

set -g status-interval 1
set -g status-right "#(~/scripts/temp/fan.stats.sh)"

and for ~/scripts/temp/fan.stats.sh

#!/bin/sh
temp=`cat /sys/class/thermal/thermal_zone0/temp`
speed=`gl_fan -s`
state=`cat /sys/class/thermal/cooling_device0/cur_state`
printf "temp:$temp speed:$speed state:$state"

and four .sh scripts:
— fan.stop.sh

#!/bin/sh
echo 0 > /sys/class/thermal/cooling_device0/cur_state
cat /sys/class/thermal/cooling_device0/cur_state

— fan.max.sh

#!/bin/sh
echo 255 > /sys/class/thermal/cooling_device0/cur_state
cat /sys/class/thermal/cooling_device0/cur_state

— fan.custom.sh

#!/bin/sh
echo $1 > /sys/class/thermal/cooling_device0/cur_state
cat /sys/class/thermal/cooling_device0/cur_state

stress.sh

#!/bin/sh
stress-ng --matrix 0 -t 5m

asdffdsa1122 · 2022-08-11T22:21:18.937Z

now, thanks to your help, i got where i need to be,
to be able to perform a cpu/temperature torture test of the axt1800.

tl;dr - the router passed the torture test.

in a small closed cardboard box, i placed the following.
— at the bottom of the box, as a heat source, a powered-on raspberry pi4 motherboard.
— on top of the pi4 heatsink, i placed the axt1800.

make sure the gl_fan service is stopped
/etc/init.d/gl_fan stop
cool the router to get low end temp.
run fan.max.sh - so that fan speed is approx. 6500RPM and wait 20 minutes
65°C is the temp of the cpu
heat the router to 90°C
— run fan.stop.sh, fan speed is 0RPM
— run stress.sh, to max out cpu and all cores to 100% utilization
now, let’s try to cool the router from 90°C
note: stress.sh is still running, cpu at 100%
— run fan.max.sh - fan speed is approx 6500RPM
and over time, takes 45 minutes to cool down to 70°C

limbot · 2022-08-15T02:42:06.242Z

My AXT1800 is only operating in Access Point mode with my primary workstation running through a wired connection and very few WiFi endpoints connected and was still sitting at 79-80 °

Have just implemented the changes as per nopro404 's suggestion and it’s now running at 60 ° with no excessive fan noise ( 1530rpm) sitting less than 1M away

jgutman · 2022-08-21T06:46:01.715Z

wvlad:

2. /etc/init.d/gl_fan stop # that command will stop the daemon responsible for fan
3. echo 255 > /sys/class/thermal/cooling_device0/cur_state

I did these two commands and the fan started running at a much quieter level than using a value of 255. And the router was running much cooler. HOWEVER, after two days I noticed the fan was turned off and not running. It seemed to turn off by it’s own. Anyone seeing this type behavior?

I reissued the two commands and will start watching it closely, but I am curious if anyone else is/has experiencing this.

TIA, Jim

wvlad · 2022-08-21T07:43:25.981Z

That is probably because you need to “commit” these settings via luci, otherwise any reboot of a device will clean that settings up. Nevertheless, I am using another solution, suggested in neighbor thread: it persistent at router’s memory after reboots, launches fan dynamically, adjusts entire system temperature to desired level on the one hand, and not forcing fan to work at max speed all the time, which increase its life a bit on the other. I am using it at my router with active Wireguard client, Synology NAS and about 5-6 devices connected via Wi-Fi, the average temperature of router ~52C which is quite good.

I wrote to support team that suggested threshold at 85C is much more than technical T ranges for motherboard used at AXT1800, so lets see how they handle it. Until that, you can resolve it with commands:

/etc/init.d/gl_fan stop
uci set glfan.@globals[0].temperature='60' 
uci set glfan.@globals[0].integration='4'
uci set glfan.@globals[0].differential='20' 
uci commit glfan 
/etc/init.d/gl_fan restart

jgutman · 2022-08-21T10:15:00.226Z

wvlad:

uci set glfan.@globals[0].integration='4'
uci set glfan.@globals[0].differential='20'

@wvlad
Thanks and this looks good. Can you tell me what these two lines do and what are the ranges of the numbers 4 & 20 and what these numbers specify?

uci set glfan.@globals[0].integration=‘4’
uci set glfan.@globals[0].differential=‘20’

Thanks!

wvlad · 2022-08-21T10:38:54.713Z

Please, refer to this post. As far as I understand - these parameters are responsible for “polling” frequency of T in order to “perform” fan to work. I suppose @nopro404 could shed some light onto it.

nopro404 · 2022-08-21T11:35:39.346Z

gl_fan is hard coded to check the temp and set the fan speed only every 20 seconds. For a PID control loop to be effective it needs to read it’s input and adjust it’s output much more frequently. An RC drone for example will sample 1000x a second to be able to stay in flight.

If you want to understand the I and D values read this

en.m.wikipedia.org

PID controller | Control loop example

Consider a robotic arm that can be moved and positioned by a control loop. An electric motor may lift or lower the arm, depending on forward or reverse power applied, but power cannot be a simple function of position because of the inertial mass of the arm, forces due to gravity, external forces on the arm such as a load to lift or work to be done on an external object. 1.The sensed position is the process variable (PV). | 2.The desired position is called the setpoint (SP). | 3.The difference ...

jgutman · 2022-08-21T18:24:18.520Z

Thank you both, @nopro404 & @wvlad. I’ve got it running and stable now.

shawnl · 2022-08-28T20:42:20.898Z

the cpu governor on glinet firwmware defaults to performance–change it to ondemand and you will be fine.

Helminthe · 2023-06-22T08:19:35.365Z

Hi - tried ondemand, doesn’t make that much of a difference, maybe 1 deg C… Opened Slate AX (AXT1800) CPU throttling feat req · Issue #76 · gl-inet/gl-infra-builder · GitHub