Lost "good" packets when refreshing Lua on some STM32

[CapnBry]

Current Behavior

When refreshing Lua, the bad/good number goes down, e.g. from 0/250 to 0/246. This is apparently only present on FM30 STM32 TXes which worked perfectly on 2.0.1.

Steps to Reproduce

1. Install master (2022-Jan-13)

2. Go into Lua

3. Watch as the bad/good number dips as the Lua loads

Possible Solution (Not obligatory)

This starts happening with 7f25cd8 STM32 platform upgrade (#1169), the previous commit is fine.

Details

There's no CRC errors reported from our debugging, and no errors reported from the OpenTX debug log side either. The platform was updated from ststm32@8.0.0 to ststm32@15.1.0 so there's a lot that could be wrong here.

• R9M apparently works fine (same STM32F103)
• Ghost apparently works fine (different STM32)

Your Environment

• TX hardware: SIYI FM30
• RX hardware: None
• Handset model: Taranis Q X7 ACCST
• OpenTX version: 2.3.14 (115200 and 400000)
• ExpressLRS version (TX & RX MUST MATCH): master as of today
• Packet Rate: any except 50Hz
• Telemetry Ratio: doesn't matter
• user_defines: defaults + DEBUG_LOG

[zandorsp]

Is this related to my case?
https://youtube.com/shorts/3WoXJ2juR-E?feature=share

I'm using a Siyi FM30 TX with ELRS 2.4.0 on Flysky NV14 with EdgeTX 2.7.0

[pkendall64]

Is this still a problem on your Siyi @CapnBry ?

[CapnBry]

Is this still a problem on your Siyi @CapnBry ?

I'll check and see when I have a chance to put together a test tomorrow

[zandorsp]

I tested today a SIYI FM30 TX (master 82f623) module on a RadioMaster Boxer (EdgeTX 2.8.2) and I see 2 things:

• The max functional baud rate is 921k
• Even on this or lower baud rate the status error counter keep raising, as same way of the video I send before (I don't have NV14 radio anymore)

[CapnBry]

Yeah it still does it. Tested with ExpressLRS 3.2.0 and OpenTX 2.3.14 as well as EdgeTX 2.8.0 on packet rates 500Hz / 333Hz / 250Hz. 400k and 921k baud tested. I dunno, maybe we can just flag this as won't fix if you want to close it. I'm not sure I want to dig into it deeper to figure out why it is happening, but leave it open if you think someone else might want to.

@zandorsp This is always the case with a FM30 TX that the err count keeps going up. This is because when the half duplex is flipped from RX to TX it frequently generates an edge which the handset takes as incoming data. That garbage is discarded and the whole packet is usually received so it isn't a big deal. There might be a way to flip the RX to TX that won't do this, but I could not figure out any combination that did not do it. See the original FM30 TX PR for the schematic of this circuit if you'd like to take a crack at it. :-D

[pkendall64]

Marked as closing as no one seems that bothered by it and STM probably have a limited life span.