[linux-dvb] [Proposal] Meaningful reporting of SNR
robert_s at gmx.net
Thu Apr 13 07:01:55 CEST 2006
From: "Rusty Scott" <rustys at ieee.org>
> On Wed, 2006-04-12 at 16:57 +0200, Robert Schlabbach wrote:
> > Instead, I'd propose scaling all signal measurements to _percentages_.
> Currently done. Causes confusion. Percentage is meaningless.
No, not currently done. The current method does not linearize the register
> SNR as a percentage is a useless number (Percentage of what? Largest
> possible value a card can produce?).
Exactly, so it's not useless.
> If it is representative of SNR in some form, the
> manufacturers I've dealt with provide the formula to compute SNR in dB.
> For ATSC and QAM there is an SNR point where errors begin to degrade
> the quality of what you see; the same applies to every transmission
> coding scheme.
Have you ever programmed any DVB demodulator? I have programmed a number
(for all three transmission systems), and I recall only two datasheets
giving information about how to compute SNR from the register readings. The
vast majority had no information on this.
> This does not "boil down to the same thing". If cards are incapable of
> producing meaningful numbers, they shouldn't be throwing out "feel good"
So they shouldn't produce any numbers at all...?
What would you gain from "precise" SNR numbers? Numbers which are
comparable between different frontends? I don't think that'll work out.
You'll still have more and less sensitive tuners, so an SNR of xx dB may be
sufficient for one frontend, but not for another.
And if these numbers aren't comparable anyway, you might as well stick with
the percentages, cause they achieve the same: Comparability across the
_same_ frontend, but not across different ones.
FWIW, in commercial receivers/STBs, signal strength/SNR/quality numbers are
not comparable betwen different models either. At least not for DVB ones.
e-mail: robert_s at gmx.net
More information about the linux-dvb