The audio signals presented here were recorded from various FM receivers and processed according to the patented noise reduction method. The method reduces any kind of interference of stereo signals or any combination thereof to mono quality without reducing the stereo separation.
Each WAV file contains the disturbed signal, followed by the improved signal.
Please download and convert the WAV files to mono if you need a mono version for comparison.
ACI (adjacent channel interference)
ACI + random RF noise
EMI (electromagnetic interference)
EMI 03 is a recording of an outdoor radio play. FM reception is disturbed by a blowdryer. The interference from the blowdryer mixes with the sound of croaking frogs.
As you can hear in the processed version, the frog sounds are clearly reproduced with full spatial information. In the background you can hear water splashing.
EMI 04 was recorded from a HiFi tuner. The audible noise comes from an equalizer positioned directly above it. The equalizer (with DSP) has a not completely shielded housing and radiates into the antenna cable of the tuner, which is also not perfectly shielded.
Mobile reception
Both car radios and smartphones with built-in FM receivers reduce the stereo channel separation when the reception becomes poor to keep the audio SNR at a bearable level. An audio example with current car radios or smartphones can therefore unfortunately not demonstrate that the stereo channel separation can be maintained while at the same time improving the audio SNR.
Instead, the audio examples „mobile fast“ and „mobile slow“ were recorded with an FM receiver that does not reduce the stereo separation and does not change the audio signal when FM reception becomes poor.
To come: audio examples „mobile“
Multipath interference
multipath01.WAV and multipath02.WAV were recorded in hilly terrain.
Please note that multipath interference also partly affects the sum signal (0-15 kHz). Since the noise reduction method cannot reduce interference beyond mono quality, some distortion can still be heard in these examples.
Multipath interference + CCI (co-/common channel interference)
Random RF noise
Digging deeper
In this section, 4 artificially generated audio examples are provided for testing and analysis purposes. In contrast to the audio examples above, by adding noise to a noise-free source signal (from CD), a very high interference level (as below the FM threshold) can be achieved without affecting the sum signal (mono signal). At the same time, we can avoid the crackling sound that occurs below the FM threshold. The noise reduction therefore approaches the mono quality of an undisturbed source signal (from CD), and a comparison with the source signal is possible. This makes it easier to assess the noise reduction capability of the method and also of artifacts at SNRs (avg rms) of around 0 dB (as in examples test#2 – test#4), which would be well below the FM threshold.
test#1 is an undisturbed stereo signal wav#1, with random FM noise added to its difference signal. test#1 corresponds approximately to the reception in the range of the FM threshold:
and here the undisturbed file wav#1 compared to the disturbed and processed file:
test#2 is the same as test#1, but with an 11 dB-higher noise level:
Below you can hear a sequence of the first 3 seconds of wav#1 / test1#processed / test#2processed:
test#3 consists of an undisturbed stereo signal wav#2, to whose difference signal an undisturbed mono signal wav#3 was added. wav#3 represents a strong interference, which does not occur in practice, since it is comparable to wav#2 in terms of rms level:
and here the undisturbed file wave#2 compared to the disturbed and processed file test#3processed:
test#4 uses the same audio tracks, but consists of the undisturbed mono signal wav#4 (= wav#2 in mono), to which the undisturbed mono signal wav#3 has been added as the difference signal:
and here the undisturbed file wave#4 compared to the disturbed and processed file test#4processed:
In this section, 2 audio examples show the behaviour of the algorithm at the FM threshold with a 50 μs pre/deemphasis.
In the first example, a 441 Hz sinusoidal signal is panned from left to right in the presence of noise. Panning is done in 11 steps, the first amplitudes being L = 100 % and R = 0%. Amplitudes are incremented/decremented in steps of 10 %, with the last step being L = 0 % and R = 100 %. The noisy signal is then processed. The first and last steps of the processed signal can be used to evaluate the stereo separation. They also allow the residual signal in the other channel to be identified as a musical noise artifact (the noise spectrum resembles the useful signal spectrum).