On 14.06.2014 20:20, Arthur Konovalov wrote:
14.06.2014 17:52, Klaus Schmidinger kirjutas:
On 14.06.2014 16:40, Arthur Konovalov wrote:
I've noticed strange effect with the SNR bar. Every time when UNC counter increased SNR bar becomes more and more red. In the same time hex and percent values are fine. Combined screenshot from femon plugin: http://www.pictureshack.ru/images/95487_unc.jpg It's not plugin depended. The same behavior can be observed on standard VDR's channel LCARS skin info bar.
Why UNC occurred, it's a different story. Currently I can't explain why viewing some (not all!) HD channels lead to the UNC (disturbance are visible too). Is it bitrate or something else, I don't know. It seems only 1080i suffer, not 720p and BER stays always on zero. If somebody have any method to debug and find reason of this, I will be thankful for sharing.
It would appear that UNC is a static counter that is simply counted up with every uncorrected error, and is only reset to 0 on a channel switch.
I don't know about femon, but the way VDR uses this counter to produce its "signal quality" measure is
int b = 100 - (Unc * 10 + (Ber / 256) * 5);
So if the value of Unc increases, the signal quality decreases. Maybe we should only take Unc into account if it has been counted up within the past few seconds?
If anybody can suggest a proper way of doing this, please speak up.
Why just don't use values from device without additional calculation?
Because when I introduced cDevice::SignalQuality() it was agreed upon that SNR, BER and UNC should be taken into account. If the general opinion is now that we should leave UNC (and BER?) out of the calculation, that's fine with me.
Any opinions?
Klaus