Heat and latency/throughput PSA
Posted: Mon May 30, 2022 10:56 am
This is applicable to most of the Quectel line, including the RM502Q, RM500Q, EM160, EM12-G.
When connecting to further towers, more bands, higher bands and such, it is absolutely imperative to find a solution for cooling.
I recently did an outdoor PoE build and was connecting to a bunch of new towers, but they were all unusable due to sky high latency. Even when I found a location where the latency was okay, as soon as I'd start a speed test, latency would shoot up and not come back down.
My observations was that the modem would immediately go into the 70c temperature range, whereas anything over 60c is considered "extended operating range". This causes the modem to try desperately to maintain the connection but throttles itself to try to cool back down.
On a cool night, I was able to connect to a distant band 66/2/n71 tower, and my results were good for a few minutes, until I ran a speed test and the temperature shot up.
Connecting back to my local tower would still be high latency until the temperature dropped back under 60c.
My solution? I bought copper slugs and wedged them between the modem and the WG1608 (the WG1608 has copper pads for this, the modem has copper pads for this, yet there's a good 20+mm space between them that you need to bridge), then got a stick-on heat sink for the other side. It's sandwiched between two different methods of heat dissipation, and the modem stays about 30c under heavy load.
The result is that the modem no longer throttles itself. The distant b66/n71/B2 tower is able to provide 50mbps+ down/1mbps up at times, whereas when it was throttling, never went above 10mbps down and usually failed to even do the upload test.
So if you're doing a build and your tower isn't literally visible from where you're installing it, it's probably a good idea to figure out a way to wick some of that heat away from the modem. Pay attention to temperatures, if you see it creeping into even the 50s under load, you're likely getting higher latency than you otherwise would be. It's worth the extra effort to get that chip running cooler, and could make a huge difference if you're trying to hit a further tower or aggregate a bunch of bands.
When connecting to further towers, more bands, higher bands and such, it is absolutely imperative to find a solution for cooling.
I recently did an outdoor PoE build and was connecting to a bunch of new towers, but they were all unusable due to sky high latency. Even when I found a location where the latency was okay, as soon as I'd start a speed test, latency would shoot up and not come back down.
My observations was that the modem would immediately go into the 70c temperature range, whereas anything over 60c is considered "extended operating range". This causes the modem to try desperately to maintain the connection but throttles itself to try to cool back down.
On a cool night, I was able to connect to a distant band 66/2/n71 tower, and my results were good for a few minutes, until I ran a speed test and the temperature shot up.
Connecting back to my local tower would still be high latency until the temperature dropped back under 60c.
My solution? I bought copper slugs and wedged them between the modem and the WG1608 (the WG1608 has copper pads for this, the modem has copper pads for this, yet there's a good 20+mm space between them that you need to bridge), then got a stick-on heat sink for the other side. It's sandwiched between two different methods of heat dissipation, and the modem stays about 30c under heavy load.
The result is that the modem no longer throttles itself. The distant b66/n71/B2 tower is able to provide 50mbps+ down/1mbps up at times, whereas when it was throttling, never went above 10mbps down and usually failed to even do the upload test.
So if you're doing a build and your tower isn't literally visible from where you're installing it, it's probably a good idea to figure out a way to wick some of that heat away from the modem. Pay attention to temperatures, if you see it creeping into even the 50s under load, you're likely getting higher latency than you otherwise would be. It's worth the extra effort to get that chip running cooler, and could make a huge difference if you're trying to hit a further tower or aggregate a bunch of bands.