I am seeking some insight into the behavior of my Chateau 5G R16 (running RouterOS 7.23 with the internal Quectel RG520F-EU modem).
I am located in a dense urban area (Riga, Latvia) with high tower density. Currently, I am experiencing a counter-intuitive behavior where forcing the router into LTE-Only Mode provides vastly superior stability, better signal metrics, and specifically faster upload speeds compared to leaving 5G enabled.
Symptoms:
The 5G NSA Upload Bottleneck: When network-mode=3g,lte,5g is enabled, the router connects via 5G NSA. Download speeds can spike, but upload performance becomes highly unstable, occasionally dropping packets completely during video calls. When I toggle network-mode=3g,lte (disabling 5G), the connection is rock solid, and upload speeds actually improve and stabilize at around 100 Mbps.
Signal Degradation when 5G is Active: I have noticed that my primary LTE anchor band metrics (specifically SINR) are noticeably worse when 5G is active. The moment 5G is disabled, the LTE-only signal metrics instantly clean up and stabilize.
Aggressive CA/NR Flapping: Looking at the cell monitor, the modem has great LTE coverage (e.g., B3 as primary anchor on PCI 317 at -75 dBm, aggregating B7, B1, B20). However, it sees dozens of weak 5G NR cells (n78/n77 ranging from -100 to -115 dBm). The router seems to aggressively try to add and remove the weak n78 band via Carrier Aggregation on loop, leading to brief renegotiation "hiccups."
Questions:
Why doesn't the modem gracefully fall back to pure LTE upload? If the 5G NR layer is clearly struggling with a poor SINR/RSRP compared to the LTE anchors, shouldn't the modem's internal logic route critical upstream traffic through the cleaner LTE bands instead of forcing it through a failing 5G NSA link?
Why does enabling the 5G radio degrade the LTE signal quality? Is this a known hardware limitation with internal antenna sharing/isolation on the Chateau 5G, or is it a power-allocation issue on the modem firmware level when handling concurrent LTE + 5G NR streams?
Is there a way to optimize CA logic via commands? Since the native /interface lte cell-lock menu isn't supported on this modem/firmware combo, I've had to resort to raw AT+QNWLOCK commands via at-chat to pin a stable cell. Is there a way within RouterOS to tell the modem to only aggregate 5G if it passes a specific signal quality threshold (e.g., 5G NR SINR > 5dB)?
I would love to hear if anyone else has faced this urban 5G NSA penalty on the R16, or if there are specific Quectel modem firmware tweaks that can enforce a more sensible fallback logic.
When it comes to broadband modems, one has to understand that (unlike with WiFi where station acts pretty independent from what AP might want it to do) terminal is tightly controlled by network (cell tower). The only "autonomous" thing that terminal (smart phone, modem) can do is to omit somethings from measurement reports (and this is the way locks work: modem omits measurements and reports of cells/bands/technologies which are "prohibited" by setting). Apart from that, it's network which instructs terminal to use particular technology, particular cell as serving cell and particular cell set for CA.
Then it comes to strategy of particular MNO ... some aggressively push terminals towards technologies/cells with higher capacity (even though some particular terminal suffers due to low or noisy signal), some are more conservative at that ... but then technologies/cells with higher coverage but lower capacity get more loaded and consequently some if not all users get worse service.
While it is possible to solve some of your "problems" using carefully crafted settings (technology selection, band/cell locks), the logic of assigning resources to a particular terminal on the MNO network side is pretty complex and it is likely that your manual settings won't yield very good results ... or not consistently over time of day (load on broadband networks tend to vary with time of day quite a lot. E.g. some MNOs try optimize their OPEX by disabling certain transmitters to save on energy costs and if you manually set modem to work with those cells, your throughput might suffer during some time periods ... or you might even loose service. So be careful.
It could be that network forces modem to connect to different band/cell when modem admits it can do 5G NSA. And that different band/cell could either be a more far cell ... or it can be higher frequency one. In both cases pathloss (signal strength loss while traveling through air) would be higher. The thing is that not all LTE cells can "piggyback" 5G NSA cells. It might be that in your particular spot signal strength of LTE-only cell is stronger than of LTE+5G cell.
So if you're willing to experiment, set modem to LTE-only, record all cells in use (both when modem is "idle" and while actively transferring data), if possible with corresponding signal strengths and band indications ... then set modem to 3G/LTE/5G and record the same data. I guess comparing both data sets might show something.
I don't think so. Broadband modems, built in MT devices, are not Mikrotik's own products, they are sourced from other vendors (such as Quectel). I guess that also firmware is done by modem vendor (it might be partially tailored towards ROS). So it's not really MT's own knowhow inside those devices. OTOH terminal behaviour tightly follows settings in MNO and (as I already explained) they may vary wildly from broadband network manufacturer to another ... and from MNO to MNO.
Tell me, if I set my AP to work with 802.11b only and I even successfully miss the interworking.realms-raw="" bend ... and then you pay me a visit with your super-duper iPhone and try to use my AP ... and then find out that you can only watch 240p youtube videos without excessive buffering, is this MT's fault?
Connecting to MNO is like that: MNO sets their own network (knowingly or not) to some shitty behaviour and everybody using that network has either to follow or stop using that network.
