Q. Is it desirable to turn the NR off and on periodically to reinitialise it?


A. This is largely  unneeded since the NR logic is only doing its job.

NR in the PRO2  is a time-varying adaptive digital filtering operation which
alters its response according to the statistics of the noise it encounters
within the IF passband. The statistics can and do vary with time and the
noise reduction algorithms react accordingly.

When first activated, NR appears to consider everything "noise" and thus
lowers gain across the entire IF passband considerably giving the perception
of a very quiet receiver . This can be interpreted as effective NR
operation, but that is probably a premature conclusion..

As time goes on, however, the NR logic has more statistical history to work
with and begins to alter its response accordingly. The actual noise within
the passband that meets the rejection criteria of the NR logic continues to
be suppressed while other frequency components are restored approximately to
their original level.

One result is that the perceived "noise" (audio) level increases and one
could  conclude that the NR is no longer functioning. This may or may not be
the case, depending upon the statistical nature of the noise.

I find it difficult to accept the premise that Icom failed to execute the NR
function properly in the PRO2 and earlier models. If one examines the
IF/audio spectral response of the PRO2 with the NR operating, the change in
noise response of the passband can be seen to vary with time.

Initially the whole passband is diminished in level and then usually (per
the type of noise I have locally) the low and high ends of the spectrum
quickly "grow" in level while the midrange tends to diminish. The increased
signal levels at the extremes of the passband appear to account for the
apparent increase in audio level, which some are interpreting as a loss of
effectiveness of the NR function.

As to "defeating" this effect with periodic PTT operation, one could
probably simulate the same outcome by periodically decreasing the AF gain
and then slowly bringing it back up. The PTT operating restores the NR logic
to its original "starting" point so it begins anew "thinking" that most of
what is in the passband is noise and lowering the overall response level.
Continued PTT operation merely restarts the process and keeps the NR logic
from reaching and maintaining its steady state response condition.

Sorry if any of this sounds negative or even argumentative - that is not my
intent - but I think that efforts to circumvent the programmed operation of
the PRO2 NR logic, while subjectively appealing, actually accomplish little
toward improving the noise reduction capabilities of the receiver.

An interesting exercise would be to bury various signal waveforms in "noise"
and to determine the S/N ratios for each as a function of time with NR
activated. The final arbiter is whether NR improves S/N to make a weak
signal more readable, not whether the receiver sounds "quieter" with NR
activated. IF I can ever find time, I would like to conduct such a test.

A final note: the NR function is most effective with the wider IF
bandwidths. As bandwidth is reduced, the noise statistics are changed and
the NR has less to work with. At very small bandwidths, NR is essentially
ineffective since even pure noise appears to be a randomly modulated sine
wave whose frequency is related to the bandwidth. Using NR with a 100 Hz
filter for CW produces little useful effect, for example.