fix for incorrect decimal shift?

[SybexX]

could be a fix for the following problems:
#3439 (reply in thread)
#3445

The error occurs when the following conditions exist:

• CNNType = DoubleHyprid10 or Digit100
• Transition from 9 to 0 or a NaN of the last digital digit
• Extendedresolution disabled

One solution (if you only have digital digits) was/is to activate the Extendedresolution.

Unfortunately, it is impossible to fix the problem correctly, since a number of the last digit greater than 9.994999 is a correctly recognized number for some user/condition and a NaN for others user/condition?
With this fix, a number greater than 9.994999 is always classified as a NaN and the correction is made in the further steps (findDelimiterPos() / checkDigitIncreaseConsistency()).

#3445

Read: 43643.610 - Pre: 4363.904 - Rate: 39279.706,3.8,3.0,6.0,4.0,3.0,7.0,10.0

should be "Read: 4364,369"

#3439 (reply in thread)

[0d00h09m23s] 2024-12-15T14:28:15 <DBG> [CNN] getReadout _analog=0, _extendedResolution=0, prev=-1
[0d00h09m23s] 2024-12-15T14:28:15 <DBG> [CNN] PointerEvalHybridNew - No predecessor - Result = 10 number: 9.996094 number_of_predecessors = -1.000000 eval_predecessors = -1 Digital_Uncertainty = 0.200000
[0d00h09m23s] 2024-12-15T14:28:15 <DBG> [CNN] getReadout(dig100) prev=10
[0d00h09m23s] 2024-12-15T14:28:15 <DBG> [CNN] PointerEvalHybridNew - O analogue predecessor, >= 9.5 --> no zero crossing yet = 4 number: 4.863281 number_of_predecessors = 9.996094 eval_predecessors = 10 Digital_Uncertainty = 0.200000 result_after_decimal_point = 8
[0d00h09m23s] 2024-12-15T14:28:15 <DBG> [CNN] getReadout#PointerEvalHybridNew()= 4
[0d00h09m23s] 2024-12-15T14:28:15 <DBG> [CNN] getReadout#result= 410
[0d00h09m23s] 2024-12-15T14:28:15 <DBG> [CNN] PointerEvalHybridNew - NO analogue predecessor, no change of digits, as pre-decimal point far enough away = 5 number: 4.796875 number_of_predecessors = 4.863281 eval_predecessors = 4 Digital_Uncertainty = 0.200000
[0d00h09m23s] 2024-12-15T14:28:15 <DBG> [CNN] getReadout#PointerEvalHybridNew()= 5
[0d00h09m23s] 2024-12-15T14:28:15 <DBG> [CNN] getReadout#result= 5410
[0d00h09m23s] 2024-12-15T14:28:15 <DBG> [CNN] PointerEvalHybridNew - NO analogue predecessor, no change of digits, as pre-decimal point far enough away = 2 number: 2.000000 number_of_predecessors = 4.796875 eval_predecessors = 5 Digital_Uncertainty = 0.200000
[0d00h09m23s] 2024-12-15T14:28:15 <DBG> [CNN] getReadout#PointerEvalHybridNew()= 2
[0d00h09m23s] 2024-12-15T14:28:15 <DBG> [CNN] getReadout#result= 25410
[0d00h09m23s] 2024-12-15T14:28:15 <DBG> [CNN] PointerEvalHybridNew - NO analogue predecessor, no change of digits, as pre-decimal point far enough away = 7 number: 6.984375 number_of_predecessors = 2.000000 eval_predecessors = 2 Digital_Uncertainty = 0.200000
[0d00h09m23s] 2024-12-15T14:28:15 <DBG> [CNN] getReadout#PointerEvalHybridNew()= 7
[0d00h09m23s] 2024-12-15T14:28:15 <DBG> [CNN] getReadout#result= 725410
[0d00h09m23s] 2024-12-15T14:28:15 <DBG> [CNN] PointerEvalHybridNew - NO analogue predecessor, no change of digits, as pre-decimal point far enough away = 3 number: 3.000000 number_of_predecessors = 6.984375 eval_predecessors = 7 Digital_Uncertainty = 0.200000
[0d00h09m23s] 2024-12-15T14:28:15 <DBG> [CNN] getReadout#PointerEvalHybridNew()= 3
[0d00h09m23s] 2024-12-15T14:28:15 <DBG> [CNN] getReadout#result= 3725410
[0d00h09m23s] 2024-12-15T14:28:15 <DBG> [POSTPROC] handleAllowNegativeRate for device: Gas
[0d00h09m23s] 2024-12-15T14:28:15 <ERR> [POSTPROC] Gas: Raw: 37254.10, Value: , Status: Rate too high - Read: 37254.10 - Pre: 3725.48 - Rate: 33528.62

should be "Read: 3725.49"

[caco3]

@SybexX can you please add a description to the pull request?

I think this solution is better than the initial proposal with (int) ((int) trunc(round((number+10 % 10)*1000)) ) / 1000;. Because if I understand it right, that one would again be an issue if somebody would have 5 digits behind the decimal point.

I added some minor changes to the comments.
I will load it onto my devices and let it run for a while to check for regressions.

[caco3]

Thanks for the detailed description and the fix of my faulty comment. I must have ben sleepy already 😴

[caco3]

@SybexX Since I test this branch I again saw several errors with Neg. Rate - Read: - Raw: 0731.13571 - Pre: 731.13574 "! Differenz (Raw - Pre): -3.000000003794412e-05. I am not sure if this is a coincidence. Before I have not seen them for a longer time.

Notes:

• Rate Threshold is enabled and set to 2
• Extended Resolution is enabled.

I am now going to increase it to 5 to work around this.

[SybexX]

This change only affects the last digital digit or do you not have analog digits?

[caco3]

Yes, I have 3 digits + 4 pointers. That's why I also thought it might be a coincidence.