New Hubsan Upgraded Version on the way

The for( i=0; i<20; i++) is called every ms (7 times), 3ms after the packet is sent.
But that doesn't explain why this stop com with the quad.

Well ... maybe I should wait that I have a H107D before continuing but BNF is out of stock everywhere I've looked at.

Yes

Thanks,

in fact issuing A7105_ReadReg(A7105_00_MODE) commands is enough to lose com, after a random amount of time...

you cannot spend significant abmounts of time in the interrupt handler. You need to create new states to handle this.
after transmission, go into Rx mode, then return something like 8000 to return 8msec later.
Now check if you got any packets. If so, read them in. Then return 2000 and setthe state to be ready for the next iteration.

I haven't been following this thread closely enough to know what the proper timing is, but you should never busy-wait in an interrupt handler, and all protocol code is done in interrupt context.

The Devo code is a better example of how to do telemetry, even if the commands themselves don't apply to the A7105.

Note that the return value is the time from the last interrupt that you want to return. So regardless of what you do in the callback, the total sum of the return values should be 10000 so that you poll at a regular cadence. And make sure you never spend more time in the interrupt handler than the value you return (though generally it should be much, much less)

PhracturedBlue wrote: The Devo code is a better example of how to do telemetry, even if the commands themselves don't apply to the A7105.

My code is more like the Devo one, well... I'll find what's wrong

If you need a hand, let me know, and I'll spend some time actually going over it, but probably not today.

PhracturedBlue wrote: If you need a hand, let me know, and I'll spend some time actually going over it, but probably not today.

Thank you, if you don't see anything obvious in the code I posted I'll scratch my head a little more, that's nothing urgent anyway, my priority was the vTX channel selection, who needs voltage telemetry when we have timers ?

SeByDocKy wrote: Me ... and you never know the real state of your LiPo ... So a double security is better than a unique one

That was a lazy dev joke only

Sorry, I got confused and was looking at midelic's code. I don't see anything wrong with your code. But my guess is that somehow you are going more/less than 10000us and are eventually getting out of sync. Looking at the timimg on a logic analyzer would likely be informative.

midelic wrote: I remembered now why I didn't copy PB code...I didn't get it at that time(even now is difficult ) how his timer interrupt function works.Of course in a timer interrupt routine you have to stay the smallest amount of time possible.

Thta's why I use a for( i=0; i<20; i++) loop only (every ms) to check A7105_MODE register, we can't afford spending 10ms in the callback !

The only thing I see is that you don't check for write completion before switcing to Rx mode. After 3msec that really shouldn't be an issue, but it would be safer to verify that transmission has completed before strobing the A7105.