2012.04.15. 19:26 keltezéssel, Felix Fietkau írta: > On 2012-04-15 7:21 PM, Gabor Juhos wrote: >> Hi Felix, >> >> Thank you for the comments. >> >>> On 2012-04-14 10:01 PM, Gabor Juhos wrote: >>>> The REDUCE_SCALED_POWER_BY_THREE_CHAIN symbol is >>>> defined in different eeprom files, and the value >>>> varies between the different files. >>>> >>>> In eeprom_def.c and in ar9003_eeprom.c the value >>>> of the symbol is 9, however the comments in these >>>> files indicates the value should be 10*log10(3)*2 >>>> which is 9.54242509439325. Replace the the value >>>> to 10 in these files. >>>> >>>> Also add comments to eeprom_9287.c. >>>> >>>> Signed-off-by: Gabor Juhos <juhosg@xxxxxxxxxxx> >>> I think we should keep the value 9. >>> If I understand the logic behind the power increase through chain >>> combining properly, this value only describes the worst-case (wrt. >>> regulatory compliance) upper limit of tx power, whereas in practice the >>> measured combined power output will be much lower than that due to >>> signal/phase differences. >>> Regulatory compliance is already properly tested on all devices with the >>> truncated value 9, so I don't think that we need to be even more >>> conservative and round up here. >> >> Ok, I will rework the patch. >> >> Apart from the inconsistency between the REDUCE_SCALED_POWER_BY_THREE_CHAIN >> values, there is another thing which confuses me. >> >> In the 'ath9k_hw_set_*_power_cal_table' functions the driver uses the >> REDUCE_SCALED_POWER_BY_THREE_CHAIN constant to reduce the power values. >> In the 'ath9k_hw_update_regulatory_maxpower' function it uses the >> INCREASE_MAXPOW_BY_THREE_CHAIN constant to compensate the reduction. However the >> two constants are different. I might be wrong, but in my opinion the driver >> should use the same constant in both functions. > I just checked with Atheros sources, and it seems I was wrong about > sticking with the value 9. > It appears that Atheros changed it to 10 at some point, and ath9k > contains a mix of the old and the new version. Consistently using the > value 10 for 3 chains for both increase and reduction seems to be the > right thing to do. Thank you for the information. I will add it to the commit message. -Gabor -- To unsubscribe from this list: send the line "unsubscribe linux-wireless" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html