SX1262 DutyCycledReceive question

[tshcherban]

Hi
Can someone help me with startReceiveDutyCycleAuto? I found in comments that According to Semtech, receiver requires 8 symbols to reliably latch a preamble. however i failed to find that info in a datasheet the only mention was for FSK modulation preamble length (im using SX1262, so used according pdf from Semtech's site). Just for experiments i run down to minSymbols = 3, which works quite unreliably, 4 - ~40% packets are missed, and minSymbols = 5 works good.
Later i found this article. It states that 2 symbols rx window works good.
Is working with minSymbols < 8 under specs and should be avoided?
UPD
also library sets radio awake for 2x minSymbols, but in the article above only for minSymbols amount

[jgromes]

This function was contributed by @BarryPSmith in #84

Unfortunately I don't know where the information about 8 symbols being required comes from.

[BarryPSmith]

I believe the 'according to Semtech' was in response to a question I posed on their developer forum. Unfortunately they have restructured their forums and so the original post is no longer available. I searched through my emails, and found this that I received (I think that the responder was Seb @ Semtech. Emphasis added by me):

Hi Barry,

I’ll try and explain the concepts:

• Preamble length: all sensitivity tests are run with programmed 8 symbols
  of Preamble, which in turns is 10.25 symbols if you add the built-in
  additional 4.25 symbols signalling the end of preamble. This explains the
  discrepancy you saw. The demodulator works happily with this size, and down
  to overall 8.25 symbols, and we don’t recommend using a lower value. This
  being said, correlation may work with 6 symbols, but the probability of
  latching on to preamble with decrease dramatically, as you have observed,
  and will depend on when the demodulator actually starts
• Preamble length setting on the receiver: there is a built-in timeout and
  restart, whose duration (in symbols) is set by the preamble length setting
  in the receiver. Say your system is synchronous, and uses short (8 symbols)
  preamble; then assume that another system on the same channel/SF/BW is being
  heard by one of your receivers, and its preamble length is 10000 symbols.
  The receiver will latch onto the first 8 symbols of Preamble, and not find
  the end of preamble delimiter. It will timeout and restart its search,
  resetting its accrued frequency and time error estimations. This is the aim
  of setting the preamble length to a size equal or greater than the size of
  the transmitted preamble of the expected message. If you don’t know, and
  have a variety of sizes incoming, set it to the largest value.
• RxDutyCycle: that is a good approach, use a long preamble in Tx and latch
  onto it in the receiver.

Best regards

(I can't remember the exact question I asked to get that response.)

Asides:

• startReceiveDutyCycleAuto is supposed to err in favour of catching all the packets rather than the absolute minimum possible power consumption.
• Can you point me to where in the linked article they say that 2 symbols works well? I find parts saying that they set it to two symbols, but don't see where they justify that.
• My network uses startReceiveDutyCycleAuto for end nodes, and unfortunately they still sometimes miss packets. While they don't need 100% reliability, I find this a little frustrating because as far as I can tell (using a scope to look at current consumption) the radios are asleep and awake for periods that should ensure they never miss packets. I do use a preamble length of 384 symbols though, which might be an unusual use case.