An In-Depth Look into 5G ON-OFF Loops in the Wild

Yanbing Liu, Jingqi Huang, Sonia Fahmy, Chunyi Peng

Purdue University

Paper Dataset

What are 5G ON-OFF Loops?

In operational 5G networks, a mobile device often loses its 5G radio access when there are no noticeable changes to its 5G radio channels in use. In these cases, 5G radio access is often recovered but is later lost again. As a result, 5G radio access enters an ON-OFF loop where radio access oscillates between two states: 5G ON and 5G OFF (figure 1), which significantly hurts data performance and even suspend data service. Even worse, such 5G ON-OFF loops are persistent loops, which repeatedly happen persist under same (or quasi-same) environment (figure 2).

Figure 1: State machine
Figure 1: State machine
5G ON-OFF loops examples
Figure 2: A real-world example of 5G ON-OFF loops

5G ON-OFF Loops in the Wild

Our experiments in the wild covers 11 areas in two cities with three U.S. operators (T-Mobile, AT&T, Verizon) in early 2025.

We ran static experiments with bulk file download at selected locations and use Network Signal Guru to collect cellular messages. In total, we collected more than 2,000 loop instances.

Table 1: Dataset statistics
Table 1: Dataset statistics

5G ON-OFF loops are commonly observed in operational 5G networks. Loops are observed in around half of the runs with all three operators (Figure 3, T-Mobile: 48.8%, AT&T: 51.1%, Verizon: 51.7%), and almost all loops are persistent. Besides, the likelihood of 5G ON-OFF loops exceeds 50% at more than half of the locations in 8 out of 11 test areas (Figure 4).

Figure 3: Loop ratio
Figure 3: Loop ratio
Figure 4: Loop likelihood
Figure 4: Loop likelihood over space

5G ON-OFF loops occur quite often every several tens of seconds. Both T-Mobile and Verizon experience a substantial portion of 5G OFF time, which account for >22% of cycle time for more than half of the loop instances. 5G OFF significantly hurts data performance and may even suspend data services.

Figure 5: Cycle time
Figure 5: Cycle time
Figure 6: 5G OFF time
Figure 6: 5G OFF time
Figure 7: Download speed loss
Figure 7: DL speed loss

Why do 5G ON-OFF Loops Happen?

There are three 5G ON-OFF loops observed in the wild:

  • 5G SA: S1
  • 5G NSA: N1 and N2

These three types can be further divided into 7 sub-types based on root causes (Table 2).

S1 5G ON-OFF loop

S1

N1 and N2 5G ON-OFF loops

N1 and N2

Figure 8 Three 5G ON-OFF loop types

All types of 5G ON-OFF loops share one common cause: the triggers to turn 5G ON and OFF can co-exist under (quasi-) same network conditions. Specifically, 5G cells are added because their RSRP/RSRQ measurements meet the pre-configured criteria. However, turning 5G OFF is not always associated with RSRP/RSRQ measurement events. Consequently, loops are created with inconsistent ON/OFF triggering conditions (Table 2).

Table 2: Sub-types and triggers of 5G ON-OFF loops
Table 2: Sub-types and triggers of 5G ON-OFF loops

Study 5G ON-OFF Loops by Yourself

If you are interested in the problem of 5G ON-OFF loops, you have two options to investigate this problem by yourself:

Option 1: Use open dataset

We released our collected 5G ON-OFF datasets here. It contains the traces of 5G ON-OFF events as well as other important information such as location, radio condition and throughput. Please refer to README in the dataset for more details

Option 2: Collect data by yourself

We strongly encourage you to collect your own 5G ON-OFF dataset for analysis. Below are instructions to run “5G ON-OFF” experiments using mperf-any, one task supported by MI-LAB. If you need any technical support, please contact us at milab@cs.purdue.edu.

  • Step 1: install mperf-any. Please refer to the guideline.
  • Step 2: run mperf-any. Please refer to the guideline.
    In summary, you need to do three (four) things.
    • Start. Start running this task at test locations where you want to conduct experiments over 5G networks (stationary experiments recommended).
    • Stop. Stop this task whenever you want (likely after completing the experimentation time).
    • Check. (optional but strongly recommended). Check if the task runs normally when it is running.
    • Upload. It supports offline data uploading after you finish one or many taskrounds.
  • Tips and more information
    • Network access setup (one-time effort). Please enable mobile data and disable WiFi.
    • Register your own account (recommended). You can use GUEST or usernames provided by the MI-LAB team (username: mssn_5g1, password: MSSN_2025_5Gmeas).
    • Run stationary experiments at selected locations to avoid the impact of environmental and radio dynamics. We encourage to use other apps to check the serving cells over time (to check whether 5G ON-OFF switches occur).
    • Please run each experiment for 3-5 minutes or longer to capture multiple ON-OFF switches if applicable.
    • We recommend using file downloading as the default traffic load. If you want to test with other applications such as video or live streaming, change traffic setting to ping in mperf_any and run the target apps on the test phone.
  • Step 3: view (and download) data. Data logs will be available on your test phone and available at MI-LAB if uploaded
Go Top