Beta Address: http://blish.org/sdrdxdoc/tuning2ab4.html

5.11.7 - Dynamic Noise Reduction

DNR Settings

DNR works by attempting to discriminate between noise and speech, and then using that information to reduce the audio gain in the presence of noise, and turning off that reduction, or increasing the gain if you like, in the presence of speech. Noise usually doesn't have the spectral character of speech, so this is possible to attempt.

You can adjust the settings with →Right-click DNR

A good starting point is to set all four controls to 75. Experiment from there.

Two of the S-Meter models have DNR indicators that tell you the instantaneous DNR level, when on.

DNR metering (the blue bargraph)

DNR operation is affected by differences in demodulation bandwidth from the actual signal bandwidth. If the bandwidth is wider than the received signal, and no other signal exists within the extra bandwidth (in other words, the "overage" only adds noise), then DNR operation will generally be enhanced. Finding the perfect balance is somewhat of an art, because the spectral character of the noise also affects operation.

DNR is also affected by AGC condition; if you have your AGC set optimally, this means that the RF gain is dropped, and hence so is the base noise level, in the presence of signal and across the normal breaks in speech.

Since when set high, DNR attempts to drop the volume of the inter-syllabic and inter-word spaces, and since AGC keeps those at least somewhat quieter, the combination of the two serves to drop the noise floor even further. Try it with a strong signal while adjusting AGC to get a feel for what the combination does, then listen carefully to a weaker signal while doing the same.

Note: DNR has to be able to respond very quickly — this is the only way it is able to modify the gain back upwards in the presence of a faint syllable using aggressive settings. One side effect of this is that a short noise burst with peak components in the vocal range can trigger a gain increase, which we hear as a somewhat bursty noise pulse. Because it is noise, even though it has voice tones it, it will have a spectral character such that it has a fair amount of high frequency content.

Consequently, if you use LPF with DNR, with Lpf set to about 1000, you can take the edge off those noise bursts and further enhance intelligibility. This works well because the low pass filter has a very gentle slope, and doesn't do the kind of damage to speech content a sharper filter would — yet it will cut the harsh tones substantially.

SSB reception, between words,
DNR

SSB reception, between words,
DNR