DFU becomes significantly slower on Android 15

[Teruyama]

When performing DFU with the same Android device + nRF Connect on the same target device, the speed becomes significantly slower on Android 15.

The attached image captures the graph of DFU performed on the same device using Pixel 6a + Android 14 or 15, and the transfer speed is clearly different.

The device in use is an nRF52840. However, we have confirmed that the DFU speed decreases on other devices as well.
Both tests used nRF Connect version 4.28.1.

Is it possible to improve the speed by changing the settings in nRF Connect or the usage of the DFU library?

[Vidar Berg]

Hello,

Is your FW based on the nRF Connect SDK or the nRF5 SDK? In case it's the latter, please ensure the 'packet receipt notification' option in nrf connect -> settings -> nRF5 DFU options is disabled. If it is necessary to troubleshoot this further, it would be helpful to have a BT sniffer capture of the DFU transfer with Android 14 and 15.

Best regards,

Vidar

[Teruyama]

We are using nRF5 SDK version 17.1.
I tried disabling PRN, but unfortunately, it did not improve the issue.

[Yukinori Nakamura]

> But this time, only test the bootloader by programming the bootloader and SoftDevice.
> When only the bootloader and SoftDevice are programmed, the bootloader will start in DFU mode, ..

Does that mean “rewrite the device program and try it, not the phone” ?

[Vidar Berg]

You will need to reprogram the device with only the bootloader and SoftDevice. Also ensure that the bootloader is configured to not require bonding. After that, you can test the DFU of the application using your phone.

[Yukinori Nakamura]

> EDIT: could you try changing NRF_DFU_BLE_MIN_CONN_INTERVAL and NRF_DFU_BLE_MAX_CONN_INTERVAL to '24' in your bootloader's sdk_config.h in case the Android phone is rejected the request for 15 ms?

I tried this.
But I did not see any change in DFU behavior.

[Vidar Berg]

Did you test this with the SDK 17.1.0 bootloader as I suggested?

[Yukinori Nakamura]

> Did you test this with the SDK 17.1.0 bootloader as I suggested?

I'm not in charge of the device.
It's difficult to do because I have to ask the person in charge..

[Vidar Berg]

I understand. Do you have a nordic development kit you can test on? The problem is that I don't have Pixel 6a to test with here. Otherwise, I could have tried to reproduce this myself.

[Yukinori Nakamura]

> Do you have a nordic development kit you can test on?

I have a nRF52 DK. (for sniffer)
but can I try it with this?

>The problem is that I don't have Pixel 6a to test with here. Otherwise, I could have tried to reproduce this myself.

It is happening on Android15.
It is happening on Pixel6a and 7 as well,
so it seems to be OS dependent rather than device dependent.

[Vidar Berg]

Yes, you can use the nRF52 DK for this. Easiest way to test this is to use the pre-built DFU test images located in nRF5_SDK_17.1.0_ddde560/examples/dfu/secure_dfu_test_images

Test steps:

 - Program '/secure_dfu_test_images/ble/nrf52832/sd_s132_bootloader_buttonless_with_setting_page_dfu_secure_ble_debug_without_bonds.hex' to your board using nrfjprog: nrfjprog --program sd_s132_bootloader_buttonless_with_setting_page_dfu_secure_ble_debug_without_bonds.hex --chiperase --verify -r

- Upload the hrs_application_s132.zip DFU package via the nRF connect app.

[Yukinori Nakamura]

Thanks for the detailed instructions.

I tried it and the DFU speed was the same on Android 14 and 15.
However, the test app you gave us is Bondless, right?
The actual device starts DFU in bonded state, so I don't think the situation is the same.



logs_nrfdk52.zip

[Vidar Berg]

Yes, the link is not encrypted in this case. However, this should have no impact on the throughput. You may use the test images listed as having bonds to confirm this.