6Ghz Roaming Problems

[jlumby@eng.utoledo.edu]

I am testing out a setup before permanent installation.  I have 4x CW9166I-B access points, and 1x C9120AXP-B with an external outdoor antenna running on a Catalyst 9800-L Wireless Controller software version 17.13.1.  When I am running only 2.4 and 5Ghz radios everything works as expected, and I can roam between access points dropping just one or two pings.  When I enable the 6Ghz radios I drop a minimum of 4 pings, and often times far more as the laptop often times disassociates, and takes up to a minute to re-associate.  Manually turning the wireless off and then back on connects things back up immediately.  I only have one 6Ghz capable device to test with, it is an Intel(R) Wi-Fi 6E AX210 160MHz running driver version 23.20.0.4 on Windows 11.  I have run https://cway.cisco.com/wireless-config-analyzer/ and do not see anything potentially related to this issue.  A debug trace on the client shows some interesting errors I cannot quite explain:

2024/01/18 10:34:53.159977223 {wncd_x_R0-0}{1}: [dot11] [15753]: (info): MAC: 3089.4a58.57cd DOT11 state transition: S_DOT11_ASSOCIATED -> S_DOT11_DELETED
2024/01/18 10:34:53.160215217 {wncd_x_R0-0}{1}: [client-orch-sm] [15753]: (ERR): MAC: 0000.0000.0000 Unable to determine vlan and switching mode. fsm_ctxt not found for client
2024/01/18 10:34:53.160289513 {wncd_x_R0-0}{1}: [client-orch-sm] [15753]: (ERR): MAC: 0000.0000.0000 Unable to determine vlan and switching mode. fsm_ctxt not found for client
2024/01/18 10:34:53.160363000 {wncd_x_R0-0}{1}: [client-orch-sm] [15753]: (ERR): MAC: 0000.0000.0000 Unable to determine vlan and switching mode. fsm_ctxt not found for client
2024/01/18 10:34:53.160441434 {wncd_x_R0-0}{1}: [client-orch-sm] [15753]: (ERR): MAC: 0000.0000.0000 Unable to determine vlan and switching mode. fsm_ctxt not found for client
2024/01/18 10:34:53.160783697 {wncd_x_R0-0}{1}: [caaa-acct] [15753]: (info): [CAAA:ACCT] Not a webauth call
2024/01/18 10:34:53.160926754 {wncd_x_R0-0}{1}: [dot11] [15753]: (debug): MAC: 3089.4a58.57cd Dot11 SAE finds ipsk progress as 0, not in pending state
2024/01/18 10:34:53.161060433 {wncd_x_R0-0}{1}: [auth-mgr] [15753]: (info): [3089.4a58.57cd:capwap_90000014] Stopped Persistent Reauth Timer for Client 3089.4a58.57cd
2024/01/18 10:34:53.422086830 {wncd_x_R0-0}{1}: [site-manager-ap] [15753]: (ERR): Failed to get flex interface name vlan table root
2024/01/18 10:34:53.436263826 {wncd_x_R0-0}{1}: [site-manager-ap] [15753]: (ERR): Failed to get first entry for vlan name-id table
2024/01/18 10:34:53.449453412 {wncd_x_R0-0}{1}: [sisf-packet] [15753]: (ERR): Unexpected IPv6 packet glean in Local-switcing mode fe80::b644:bd1:6b47:41e6 : 3089.4a58.57cd in at vlan

[JPavonM · Accepted solution]

It's not that much the difference in signal range from swapping from 5- to 6-GHz, but there are huge differences when it comes to client's adapters using 6-GHz band (https://www.cisco.com/c/en/us/products/collateral/wireless/catalyst-9100ax-access-points/ghz-unlicensed-spectrum-reg-wp.html#:~:text=The%20output%20power%20in%20the%206%20GHz%20band%20varies%20greatly%2C%20from%204%20to%2016%20dBm%2C%20depending%20on.... Notice that there is a 4dBm reduction in the maximum TX power a client can use from 5- to 6-GHz, and then add the quality of the NIC and the maximum power that is allowed to be used from the manufacturer (Check this out from Intel https://fccid.io/PD9AX211D2L?utm_content=cmp-true)

[jlumby@eng.utoledo.edu]

Those links were especially helpful.  I had no idea about the lower power levels in the 6GHz band.  I should have not relied so heavily on 5GHz experience.

[Leo Laohoo]

Did someone do a wireless AP design?

[jlumby@eng.utoledo.edu]

Only based upon 5GHz. JPavonM's post above clearly points out the differences in power levels that lead to the problem even though the frequency/penetration between 5GHz, and 6GHz is minor.

[JPavonM]

In addition, and after you complete all tests and put the APs into production, I would recommend you not to use the same SSID for all bands as roaming between bands require full re-auth so potentially leading devices to packet drops, specially if they use near-real-time apps like conferencing.

[David Ritter]

Ah yes, cell sizes shrink with each iteration of our normal wifi standards.  wifi 6e and 7 are up close and personal.  whereas 802.11f/h is for the long haul, slow as you go.

[jlumby@eng.utoledo.edu]

I agree with your observation of the trend.  I just wonder if we are getting to the point where the cell sizes are so small that the returns have diminished to where things are becoming pointless.  But that may just be because my personal opinion of WiFi is that it is just a nice extra convenience, and I hard wire anything I can.  I value reliability much higher than convenience, or bandwidth.

[Rich R]

Actually the reliability is getting steadily better.  Many offices are ditching LAN now because WiFi has matched or bettered the reliability (correctly designed of course).  The convenience and flexibility is highly valued by most users these days.

The trend is an inevitable consequence of increasing data rates = higher frequencies.  We see the same in mobile - it's not just WiFi.  So the AP or mobile/cell base station density increases.

[JPavonM]

See a comparison from Ekahau between 6-GHz and 5-GHz, cell size is more or less the same, so try tresting with different client adapters (Intel AX211 (https://fccid.io/PD9AX211D2L), Qualcomm WCN985x for Windows laptops (https://fccid.io/J9C-QCNFA765), or Apple silicon from iPhone15 Pro (https://fccid.io/BCG-E8437A)). As you can see the Intel chipset is the only one with reduced power (up to 21 mWatt) in comparison with Qualcomm (up to 156 mWatt) or iPhone 15 (up to 75 mWatt).

[jlumby@eng.utoledo.edu]

Thanks for that info.  I will see if I can get my hands on a different Wireless NIC to test with.

[eglinsky2012]

@JPavonM, thank you for sharing those Ekahau screenshots. What dBm was each radio set to at that time?

[JPavonM]

That screenshot is from a design using Mist APs with 8dBm.

The added "problem" with modern APs is that vendors tend to change the radiation pattern from previous generations to focus the signal more in the downward direction (hemispherical pattern) than to the sides (toroidal pattern), thus reducing the coverage cell.

[train00wreck]

---

@JPavonM wrote:

That screenshot is from a design using Mist APs with 8dBm.

The added "problem" with modern APs is that vendors tend to change the radiation pattern from previous generations to focus the signal more in the downward direction (hemispherical pattern) than to the sides (toroidal pattern), thus reducing the coverage cell.

---

Why do they do that?

[Ayaan-kid]

1. **Roaming Explanation for 6G and 5G with Different SSIDs and Different Encryption Methods (6G with WPA3, 5G with WPA2):**

1) Strictly speaking, switching between different SSIDs is not considered roaming. Generally, when the signal of the original SSID (A) is insufficient to provide service, and there is no better quality signal from the original SSID (A), the terminal may choose another SSID (B) that it had previously connected to. This process happens because the terminal has not scanned a better signal from the original SSID (A) in time, and instead connects to the new SSID (B) with a stronger signal that it has previously connected to. This scanning and decision process takes longer than roaming between the same SSID. Therefore, when switching between different SSIDs, the signal quality of the original SSID (A) may have already deteriorated significantly, leading to potential packet loss.

2) The encryption methods differ between SSIDs: 5G uses WPA2, while 6G uses WPA3 (restricted by the protocol to only support WPA3). When switching between SSIDs, the encryption keys need to be renegotiated, and this process also takes longer than roaming with the same encryption method, leading to more packet loss.

In addition to these two reasons, even with the same SSID and encryption method for both 6G and 5G, the roaming time between them is still longer. So, this extra time must also be accounted for, as described below.

2. **Roaming Explanation for 6G and 5G with the Same SSID and Same Encryption Method (WPA3):**

1) The scanning of 6G channels involves both active discovery and passive assisted discovery. During active discovery, the terminal can only scan the preferred 6G channels, while for non-preferred channels, the terminal relies on beacon frames or probe response frames sent by the 5G and 2.4G radios of the same AP to discover available 6G wireless services. Therefore, the signal scanning process for 6G is longer than for 5G. Additionally, the 5G radios of the AP can only carry information about the AP's own 6G wireless services. When roaming, the terminal needs to discover 6G information from non-local APs in order to roam to the target AP, which takes longer than traditional roaming processes.

2) 6G has more channel resources, and the roaming process needs to scan each channel before determining which signal to roam to, making this cycle longer than normal roaming.

3) The current scanning order for terminals is to first scan the 5G channels, then the 2.4G channels, and finally the 6G channels. As a result, the scanning order for 6G signals is also later.

4) When choosing between 5G and 6G signals, terminals currently do not seem to prioritize 6G. In fact, they may even be more inclined to select 5G signals. This is a terminal selection behavior and is not mandated by protocol.