On 02/03/2017 20:40, Black Michael wrote:
We've got 90dB of space to work with. Why would we NOT want the
dynamic range? I'm finding more decodes with the higher level...more
dynamic range means more accurate FFTs and such, doesn't it?
30dB RMS is 40+dB of peak value so I rounded up to 50dB and that still
leaves 40dB headroom for peak value.
I've been running 20M with all sorts of signals splattering and never
saw anything over 80dB peak level running a 70dB peak level. Clipping
have never occurred with these settings (I'm doing some runs at
different levels).
The RMS meter had a kind of peak hold but it was RMS peak hold.
The new meter also shows when clipping occurs...which is something one
really needs to know.
Hi Mike,
ADC characteristics are linear so it doesn't matter where the wanted
signal is in the voltage range except, there are two factors that must
be considered, firstly if any signal exceeds the maximum sampling
voltage it cannot be measured and any other signals mixed with it will
be distorted and, secondly there is a +/- 0.5 LSB quantization error
that must not be allowed to dominate. We need to have the minimum
signals high enough to swamp the fixed quantization error and also have
the input level low enough to avoid any signal exceeding the maximum.
So if we set the noise average to 30 dB (about 31 counts greater than 1
LSB) then we are leaving ~66dB headroom for the strongest signal that
might come along while squashing the quantization error. I believe
modern rigs have greater than 100dB dynamic range so it is quite
possible to have a signal that exceeds that available headroom, but it
is fairly unlikely unless your neighbour is working JT65 or JT9 on the
same HF band as you are.
Now we need to consider two cases where the noise level may be less than
30dB with respect to 1 LSB even when you have set the no signal level at
30dB. Firstly if you have the waterfall, and therefore the decoder,
wider than your Rx passband. It is almost certain that signals in the
stop bands have a considerable quantization error component to their SNR
estimate. Secondly if you have the rig's AGC enabled, in this case any
signal the raises the S-meter will be reducing the gain and pushing the
noise floor back to the level that quantization error starts to
dominate. This second effect also changes the level of all signals over
time and will make it harder for the decoder to do its job, the decoder
assumes that wanted signals have continuous amplitude (with allowance
for QSB).
So if you are finding that you are getting better results with the no
signal average noise higher than 30dB with respect to 1 LSB then I
suggest that either you have the rig AGC switched on or you expecting to
get normal performance from WSJT-X on signals that are outside of your
rig's Rx passband. I don't think we should be advising on any optimum
settings that are compensating for either of those cases.
73
Bill
G4WJS.
------------------------------------------------------------------------------
Check out the vibrant tech community on one of the world's most
engaging tech sites, SlashDot.org! http://sdm.link/slashdot
_______________________________________________
wsjt-devel mailing list
wsjt-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/wsjt-devel
