Thank you for the clear explaination
Leif and for such a great piece of software.
Jim, KM5PO, EM12ko
4 x 2M18XXX
8877
.11 NF LNA
Leif Asbrink <leif.asbrink@xxxxxxxxxxxxxxxx> Sent by: owner-linrad@xxxxxxxxxxxxxxxxxxxxxx
11/12/2004 04:35 PM
Please respond to
linrad@xxxxxxxxxxxxxxxxxxxxxx
To
<linrad@xxxxxxxxxxxxxxxxxxxxxx>
cc
Subject
[linrad] RE: signal strength
Hi Jim,
> To get a rough idea of S/N for the arriving target signal,
> is it safe to simply compare the middle S meter value to
> the bottom S meter value (subtract the values) and is this
> value in decibels?
> I have read the definitions for these items and I understand
> what Leif says to do with a 1 Khz sample to determine noise
> floor but is this shortcut valid?
What do you mean by S/N?
I think the most common interpretation would be: "S/N is the
number one can read off the S-meter on a conventional radio."
You then have to do two measurements, one for the noise only
while the signal is absent and another when the signal is
present.
Another definition is S/N is the RMS power ratio between the
powers of S and N as they come out of the loudspeaker (headphones)
of a receiver.
The second definition is the "correct" one while S-meter readings
are use fast attack, slow release detectors that respond differently
to sine-waves and white noise. I do not know how much this effect
amounts to but as one example the log detector of conventional
spectrum analyzers differ from a true RMS detector by 2.51 dB
if the signal is white noise. Spectrum analysers show the true
RMS value of steady carriers only.
It is far more complicated than the above however;-)
For the noise floor, here is another problem:
If you look up the ARRL Lab measurements for a receiver like FT1000D
you will find a MDS (minimum discernible signal) that is -126 dBm
in a bandwidth of 250 Hz at 14 MHz. I have a SM5FRHs FT1000D here
and I get results in close agreement with the ARRL Lab data when
following the ARRL Lab test procedures. -126dBm in 250 Hz is
the same as to -150dBm/Hz which means that the noise figure NF
equals 24dB. (kT at room temperature is -174dBm/Hz).
The FT1000D result is misleading, the radio is actually about 4 dB
better because 50% of the noise is wideband noise. The measurement
of N in the loudspeaker output is done from 10Hz to 40 kHz and 60%
of the noise is well outside the 250Hz passband. The human ear has
built-in selectivity so when listening to a CW signal through the
250Hz filter of the FT1000D the effective noise floor is the noise
power density at the passband center normalized to 250 Hz bandwidth
and that corresponds to a noise figure of 20dB.
S-meters show S+N so you can get S/N from the S-meter directly only
if S is much larger than N. The way the S-meter sums S and N is
complicated because the waveform gradually changes from white noise to
a sine wave so the correction due to the use of a peak detector
depends on S/N.
Even at large S/N there are complications in case you are not
measuring a steady carrier. What is the S you want to read off your
meter on an SSB signal? Is it PEP (peak envelope power) or is it
RMS power? Presumably you want the PEP value since it a well defined
power at the transmit side while the RMS power depends on the
different sounds. Likewise you presumably want the "key down power"
rather than the average transmitted power from a CW signal.
Typical S-meters with fast attack and slow release come rather close,
but a conventional receiver may be designed for saturation close
above the point to which AGC regulates the signal level in order
to make the output saturate for a short time while the AGC voltage
(S-meter reading) grows. The reason is to not have the AGC voltage
increased due to short interference pulses (with a dead period
after each pulse).
-------- o ------------
The S-meter is a complicated thing. Even if you make a calibration
table for sine waves using a signal generator you may arrive at
different conclusions when comparing two different receivers that
compare the signal strengths from the same antenna. It does not
mean that the receiver that seems to give the best (calibrated)
S-meter reading actually gives the best intelligibility.
In Linrad it is very simple. The bottom S-meter is true RMS power
and the middle meter is true peak power. Get the noise floor from
the bottom meter when the signal is absent, then get the peak power
from the middle meter when the signal is present. Contrary to
conventional S-meters, the result you get is accurate and well
defined. The peak power of white noise is about 10 dB above the
average power of white noise so (S+N)/N that you get as the
difference between the two readings starts at 10 dB for S=0.
Conventional S-meters that have fast attack, slow release detectors
have the same problem (of course) but it is not something one
usually worries about since S-meters typically are very inaccurate
anyway.
To get an idea, look at this table of Linrad S-meter readings.
It is a single carrier in white noise.
The peak detector gives an accurate result for signals that are 30dB
or more above the noise. For such strong signals the noise does not
affect the reading. When S/N is 10 dB only, the noise peaks are of
similar amplitude as the sinewave and they add to about 6 dB above
the reading for noise only. An unprocessed SSB signal that is at its
peak power only a small fraction of the time will give an accurate
reading for S/N above 30 dB also, but such an SSB signal with pPEP
10dB above the RMS noise will give a reading about 3 dB below the
reading for a carrier with the same PEP. The reading will be noisier
though and it may occasionally go up when a noise peak happens to
coincide with maximum power of the SSB signal. It all depends on the
decay time of the peak detector. All fast attack, slow release detectors
have the same problem. We usually just ignore them.....
So back to this question:
> To get a rough idea of S/N for the arriving target signal,
> is it safe to simply compare the middle S meter value to
> the bottom S meter value (subtract the values) and is this
> value in decibels?
No. You get a PRECISE measurement of (S+N)peak/Nrms but you have to
remember Nrms from what it was when the signal is absent. The value
is in decibels.
> I have read the definitions for these items and I understand
> what Leif says to do with a 1 Khz sample to determine noise
> floor but is this shortcut valid?
The value you get is precise for the bandwidth you have selected.
In case you want to compare it to S-meter readings of a conventional
radio, just select the bandwidth that the conventional radio uses.
The conventional radio has the middle meter only so you have to
compare the reading of the middle meter with and without the
target signal. The meaning of the result is not too well defined
but it is the same for Linrad and a conventional radio. Simply
a fast attack, slow release detector.