5G home internet keeps dropping

Do yourself a favor and use www.cellmapper.net to learn where the T-Mobile towers are. Record the information about the ones close to or around you with attention to the PCI (physical cell ID). If you go into the router with the web UI at 192.168.12.1 you will see the primary and secondary channels/signals on the overview page. That will show you the connections and RSRP/RSRQ/SNR values to know the signal strength, signal quality, and signal to noise ratio. Then go to the status page and record the band, and PCI information from the primary and secondary signal reporting. If you proceed to the "statistics" page you can select cellular and record the statistics there. See if there are packet errors or packet drops. If you have problems using cellmapper.net open a call with T-Mobile support and ask them point blank where is the tower my router is serviced by. They have the PCI information and can tell you the coordinates of the tower. You can use Google Earth and a drive about to confirm what you know and get a good ideal how to improve matters. Use your phone to locate the tower. If you use an iPhone put it into field test mode and read the PCI value it knows. Chances are both may use the same tower. It is not a given but a good datapoint. 

Con PCI information for the primary and secondary channels and the bands used you can use the information from cellmapper.net to determine where the tower is that sends to you. With this and some testing you may be able to better locate your router. Watch out for metal screens or buildings in the path to the tower. If a window has a screen that is metal either put the router above the screen or remove the screen from the window to prevent shielding of the signal. Don't just rely upon the bars on the top of the router. They are actually rather generic as vendors don't have a MUST directive for how to deal with those but a recommended use. One feature of the LED display on the top of the router is the alarms. If the router does have temperature issues it will or should report an over temp alarm. The only way I have seen to get the alarm is to use the LED display on top of the router. Forget the mobile application it is about as useful as well blah… It is very unreliable and not much value. 

If you only have say the 5G signal dropping from time to time it could well be T-Mobile engineers are working on the equipment on the tower in your area and that is the cause not the router itself. I highly suggest to not just accept mediocre service but talk with T-Mobile support engineers, try the router in different locations, get the information about the tower location and be patient. Record your findings from trying new locations and then dial it in. If you do the same thing, expect the same results. 

If I can help you understand what you see and help you dial it in I am happy to do so.

You can add me to this list as well. I was having drops nearly every 2-3 hours 2 days ago. 

That’s when I discovered this forum and learned that my problem wasn’t unique.

So far what I've done is taken the outside cover of the trashcan off (Similar to the YT videos of people adding an external antenna but not reassembling it), and placed a fan under it (Fan is inappropriately sized) and that had given me a 1D,12HR and 58MIN uptime. After that time, all devices (Two wired and one wireless) said there was no internet connection from the trashcan (but were still technically connected to the trashcan itself). Looking at the gateway panel, it had showed that I had 3 bars connected to my cell tower at the time of the disconnection with a -96 dBm value. I measured the temperature at the top grill at the time of disconnection and I found that it was 88F. The trashcan is placed near a window with a light window curtain, at the time of disconnection it was sunny.
 

I haven't had issues in terms of staying connected to at least one cell tower at all times. However, I do occasionally have issues with the trashcan staying connected to two towers at once, but I have learned that isn't the issue that's causing the drops (AFAIK) and have attributed that to living in the woods.

On a sidenote: I had also removed the battery from the device in an attempt to see if that was causing some type of internal voltage issue as I saw someone somewhere mention that everytime the battery hit 100%, they’d have a dropped connection.

Sorry I did not intentionally ignore the reference to the MTU size that Amith Kamur discusses. He is obviously a very intelligent and talented engineer. His article is well written and offers a solution where the use of the VPN is desired. Encountering MTU issues can be possible, and maybe pretty common, as the VPN adds overhead to the solution with the size of the packet. You have to limit the data contained to have the wrapper small enough so every node along the path can handle the MTU. If you stated you use a VPN regularly I would have taken more focus on that but well. As you state, may be unrelated, and I would not rush down that rabbit holeright away. Analysis of packet captures and looking at the physical layer and moving up the stack from the bottom up would be more productive. 

I had another thought or two about troubleshooting this. You have clients upon the ASUS and still have LAN 2 for leverage. The Roku test was informative but what about connecting a small gigabit switch to LAN 2 and using a client or two multi-homed in effect. You could use the investigation client(s) with wireless to the ASUS and have an Ethernet connection to a switch on LAN 2. If the client has issues hitting the internet through the ASUS then disable the wireless and enable the Ethernet port and check results. Maybe take a client and the Nokia router with the 2.5 Ghz wireless on a different SSID and look at results from that angle as well. El Asus ac68u has four Ethernet ports as well as the 802.11ac so take advantage of the physical ports on the ASUS as well. Test and verify. If you have a client with multiple Ethernet interfaces and plenty of resources stand up some virtual machines as well. Linux  would provide yet another data point and physical Ethernet LAN ports tend to just work with Linux. Lots of tools in the Linux environment to play with. Use different clients, Apple, Windows, Linux, Android, Raspberry PI clients. OK maybe I am making assumptions but today it is not uncommon to have all of the above. 

Another parallel investigation you can do is leverage your cell phone and if it is an Apple iPhone put it into field test mode. See if it communicates to the same tower as the Nokia router by confirming both report the same PCI values for the tower signals. Android phones have applications for tower location so yet another option. If you have not yet used cellmapper.net to validate the location of the tower your router connects to i highly recommend doing so. It is simple enough to use and very informative. What cellular channels are you linking to? How strong and clean are the signals? Do the cellular signals bounce/change from one channel to another? Do any of the local devices record errors or drops? Profile the behavior in detail. Focus on any device that can be influential and stands out but keep an open mind on other actors in the path.

Use the web searches to get ideas but focus on the facts. The values and behaviors you can confirm. Don't overlook all the tools you can put to work. Keep notes and analyze the operation in a systematic periodic manner. 

The objective: Determine with more certainty where the problem resides. Look at the physical layer before you go up the stack. 

OK you are troubleshooting intermittent connectivity. Let's break this down just a little. Your solution with the ASUS and the Pi-Hole network-level advertisement and internet tracker blocker opens a whole plethora of variables. The Pi-Hole is a creative, interesting software solution. It acts as a DNS server but what big boy DNS server upstream does it use? I would probably test with pings to the Google, Quad9 and CloudFlare DNS servers and pick the one with the lowest latency to point the Pi-Hole at to feed it what it needs. Given the Pi-Hole sits in the middle maybe in the tools folder there are some useful trinkets. Even without clever tools in the Pi-Hole I would speculate you have  a Linux client or two, I do. Using the terminal and "netstat" there are various options for how to get the facts and build what we know vs. what we think. Taking packet captures and doing analysis on the functions of sessions is another option. Using Wireshark or taking a tcpdump can provide important clues. I don't have a Pi-Hole but I can download it and install it on one of my Linux clients to tinker with. Since I have not used the Pi-Hole software I am not the best resource on that solution but I find it very interesting. If you suspect an incompatibility between the Nokia router and another router/device then look at the communication across these devices. The answers are quite possibly in the frames/packets, session information, from the TCP/IP stack etc… There could be some issue that looks like an incompatibility but is really just a fundamental bug. Con datos you will have the silver bullet to put the wolf down. 

On a side note. I observed first hand that if I leave a HTTP session with the Nokia router too long, I have not timed this, the web UI of the Nokia router will report no network connection. I used my MacBook with CAT 6 direct to LAN 1 on the router and a Linux client with a wireless connection. I opened three terminal windows and ran simultaneous pings to the Google DNS, Quad9, and CloudFlare DNS servers. All the pings went right on through and the router was really connected to the tower but the web UI reported incorrect information. I closed the tab and opened a new session to the Nokia router and taadaa the router reported the real facts. Sometimes it takes multiple forks to get proper confirmation and it takes time. I am sure you are putting considerable time into your solution given it is not a garden variety solution. You stated the ASUS was reporting the internet down but how do we quantify that? It seems similar to when I observed, on multiple occasions, the web UI of the router HTTP server being unresponsive. Sure I should have taken packet captures to confirm which side of the session was being mute but it was not that important to me as I know how to work around it. It may be how the Nokia router caches information and when the session fails it acts like Jack Oneall in Stargate, "I got nothin." I could be accused of that at times. :-)

I have spent most of my time investigating the network disconnects with ONLY the T-Mobile Nokia router and my local clients. I know it seems rather rudimentary but it has worked for me quite well. I did 3rd level network troubleshooting for +20 years on enterprise networks so since I retired I have taken a little more relaxed approach to my current network solution. I currently have over 10 days of uninterrupted uptime on my network with just the Nokia router so the KISS principle does have some merit. There are some interesting posts out on the web but don't let all that chatter take you too far down the rabbit hole. Look at the basics before jumping to the more complex thoughts. Taking packet captures will provide hard data points so you know what you know and not think what you know. My father always told me if you know the why you can figure out the how. It works for me. Around 20 years ago a network escalation engineer told me, "I don't want to know what you think. I want to know what you know." I listened to Dave and became a much better engineer. I ended up taking his seat in the office in CA and did the job for +20 years. The key was to open my mind, eat humble pie, and dig for facts. I was also fortunate to have landed in a position with some very creative team members who all became great friends. 

The T-Mobile "router/gateway" may not be God's gift to mankind but it is not a "trashcan". Don't let others put a negative spin in your head. Think for yourself and determine if you really think it is a trashcan or not. I have found it very limiting for visibility to the actual network operations but in the end with careful investigation working around its limitations I managed to obtain a stable solution. The problems I had with these same issues were not with the Nokia router but with the cellular equipment on the tower. By working with the T-Mobile support engineers I was able to see the problem through. It did take a couple of weeks before things became stable but they fixed the problems. I get 10X the download and upload speeds I got in CA for less money using the T-Mobile home internet solution. It was a huge win for me. If it ain't broke don't fix it. If you can get some hard data and I can help out I will try. Please just call it a router and keep it neutral. Best wishes!

SomeoneSomewhere wrote:

I am currently troubleshooting the same issue. My set up is the trashcan connected to the WAN, fire walled, side of an Asus ac68u. The Asus is handling all LAN Wi-Fi DHCP etc. I just did a test hooking up a Roku directly to the trashcan (using the 2nd Ethernet connection) at the same time that the Asus was showing the Internet down, The Roku was able to still stream on it's wired connection.
 

My current working hypothesis is there is some incompatibility between the trashcan and a separate router (LAN side) vs the t-mobile is dropping the internet (WAN Side) my prior working hypothesis). 
 

This seems unrelated, but I wonder if there is some MTU incompatibility between the WAN side of the asus and the LAN side of the trash can. https://amithkumarg.medium.com/resolved-t-mobile-home-internet-vpn-issue-2f5ca594c23e Does not seem too easy to change the MTU on the asus though. 

Update on troubleshooting so far. Noticed that the WAN DNS on my asus was pointing at my LAN pi hole….and DHCP on ASUS was giving out itself as DNS server to clients This was working OK when with Xfininity for several months. Have changed so that ASUS DHCP is giving out the Pi Hole as the DNS server - how it should have been all along. This has also had the effect of improving the statistics from the pie hole has it now no longer shows all queries coming from the Asus router. WAN DNS server on the ASUS is now Default, meaning is getting it from the Trashcan. I also turned off the 5Hz WiFi on the asus and turn it on on the trashcan, with the theory that if I use that while working from home & VPN in to work it should be more stable. That combo seems to of worked OK, meaning I've had about 12 hours of what I think uninterrupted uptime. Not sure if fixing the DNS server, or just changing the DNS server helped.. The advice to change the DNS server came from "Nater Tater"'s you tube channel - he has a bunch of guides on the t-mo trashcan. 

The Nokia router has 4 U.FL connectors, one for each antenna. The U.FL connectors are small and making the connection takes great care. When seated properly there is a distinctive click feel as the U.FL connector seats. This is a mechanical connection. The location of these U.FL connectors in the router does not offer a huge amount of real estate so making these connections would take care and focus. Maybe some of these are not connected firmly. It would take 5-10 minutes to confirm this and it might explain signal connectivity issues if this were the case. It might be a long shot but maybe not. If the cellular statistics report packet errors or drops I would investigate this. The guide waveform has is excellent so checking this is super easy. One point I would like to clarify is that IF you ever need to remove the SIM be sure to have the router horizontal with the SIM card holder so that the retention screw is to the right of the SIM when removing the SIM. There is nothing holding the SIM in the holder so if it is in a vertical position when the SIM tray is removed the SIM will fall out of the SIM tray. Just pay close attention to details as progress is made. If the U.FL connectors are checked and they look to be seated properly I would not disconnect them as they are rather delicate but careful examination would not hurt anything. Removal of the shell of the can is easy. Just take a look at the guide and read each step. If you don't feel comfortable doing so or don't have the tools that is fine. Getting a replacement router might be possible, in time, but also offers the possibility of introducing new variables. 

Link to waveform guide: t-mobile-5g-gateway

I know when the power is not to the router it turns off the WIFI radios and only maintains the cellular radios. I still am not convinced a software reboot would be as effective as using the switch on the back of the router. A software reboot via a script may only perform a soft reload not a hard reload. My guess is the switch on the router is there for that very reason, to allow for a hard reboot. If the battery is not disconnected to prevent power to the board then the soft vs. hard reboot are not 100% equivalent. I believe it would actually depend upon how clean a reboot must be to achieve a total restart. Sometimes a soft reboot is enough and other times a hard reboot is required. It comes down to objectives. If the LAN ports stay up when external power is removed then using a hard wired client to run a script could be used or in the case of external wireless connected to the router via the LAN with Ethernet cabling then running a script from a local client to reload the router should work fine. I run all my equipment on UPS backups to avoid possible damage during power outages due to spikes so I can maintain internet for a good bit of time even should a power outage take place. In rush of power after an outage can do serious damage to equipment. I know from experience with the problem from living in CA and being stuck with PG&E service. Outages with PG&E in CA have been a frequent and constant problem. Using UPS systems on expensive gear is really helpful.

TucsonJim wrote:

I discovered that my attempt to autoreboot the trashcan via an Alexa smart power strip will not work since once the internet is off, so is Alexa services.  I am not surprised but sad and still looking for a solution.

Since once the power goes off it switches to battery anything that interrupts the power will not work well. I mean yea after a few min off it will boot to battery then if you put the power back it should boot but not sure this is a good idea. Also yes you drop the internet so that is a monkey wrench too.

I think a small script that logs in and executes the reboot is probably the best amd clean.

I discovered that my attempt to autoreboot the trashcan via an Alexa smart power strip will not work since once the internet is off, so is Alexa services.  I am not surprised but sad and still looking for a solution.