[Wifi link] Connection speed lower with 2.5.0 (Beta 1) than with 2.4.2

[Swiftnesses]

Basic Infos

• This issue complies with the issue POLICY doc.
• I have read the documentation at readthedocs and the issue is not addressed there.
• I have tested that the issue is present in current master branch (aka latest git).
• I have searched the issue tracker for a similar issue.
• If there is a stack dump, I have decoded it.
• I have filled out all fields below.

Platform

• Hardware: [NodeMCU V1.0]
• Core Version: [Latest Arduino IDE build]
• Development Env: [Arduino IDE]
• Operating System: [MacOS]

Settings in IDE

• Module: [NodeMCU V1.0]
• Flash Mode: [qio|dio|other]
• Flash Size: [4MB (No SPIFFS)]
• lwip Variant: [v2 Higher Bandwidth]
• CPU Frequency: [80Mhz|160MHz]
• Upload Using: [OTA]

Problem Description

With 2.4.2, I consistently see connections circa 20Mbps uplink / downlink and pings below 10m. After updating to 2.5.0 (Beta 1), link speed is significantly reduced and ping times often 20-100ms.

Sketch contains the following parameters:

  // Prevent wifi sleeping
  WiFi.setSleepMode(WIFI_NONE_SLEEP);
  // 802.11n
  WiFi.setPhyMode(WIFI_PHY_MODE_11N);
  // Station mode
  WiFi.mode(WIFI_STA);

[devyte]

@Swiftnesses post 2.4.2 there was a change that seems to affect rf calibration, see #5056 .
Does reverting that change work for you?
That PR pulls out a workaround implemented previously. There is reference to this comment to force rf calibration on startup.
Does implementing that work for you?
Alternatively you could implement the following function in your sketch, which executes very early in the bootup process (no access to global objects allowed in there):

void preinit()
{
    volatile uint32_t* rtc_reg = (volatile uint32_t*) 0x60001000;
    rtc_reg[24] &= 0xFFFF;
}

If none of the above works for you, then it is likely that something changed in the SDK which affects you. In that case I doubt there is anything that can be done on our end.

[Swiftnesses]

Looking into this now.

Certainly something has changed which effects sensors at the outer edge of the wifi signal (40% and below). Before they connect at 20+ Mbps TX / RX, now they connect at 1 or 2 Mbps.

The most inconvenient part is the sensors no longer present themselves for OTA updates (even with the new dns update in the loop), I assume they're dropping packets as the ping is also slower (sometimes they drop connection entirely!).

[Swiftnesses]

I will try to revert this:

void user_rf_pre_init()
{
    // *((volatile uint32_t*) 0x60000710) = 0;
    spoof_init_data = false;
    volatile uint32_t* rtc_reg = (volatile uint32_t*) 0x60001000;
    if((rtc_reg[24] >> 16) > 4) {
        rtc_reg[24] &= 0xFFFF;
        rtc_reg[30] = 0;
    }

    system_set_os_print(0);
    int rf_mode = __get_rf_mode();
    if (rf_mode >= 0) {
        system_phy_set_rfoption(rf_mode);
    }
    __run_user_rf_pre_init();
}

WIll update if it helps.

[Swiftnesses]

No luck. Spent 2 hours outside trying various things, using 2.5.0 Beta 1, the sensors in the garden not only connect extremely slowly, they also refuse to advertise OTA. I tried with multiple sensors, wiping flash before, I tried both changes you suggested, they did not help.

Reverting back to 2.4.2, connection is restored and OTA works again.

Tried with a different AP, same behaviour.

2.5.0

2.4.2

[d-a-v]

About OTA, check #5487 or run with latest git.

About connection speed, I cannot say right now.
Are you able to run in arduino IDE, enable all debug options, enable timestamp in serial logger, and copy paste the boot/dhcpclient process with 2.4.2 and 2.5.0-beta1 so we can see firmware debug log with relative durations ?

[d-a-v]

Can you also try with lwIP-1.4 and lwIP-2/no-features ?

[devyte]

No feedback in almost a year. Closing.