Comparison of background noise

Here is my attempt to better illustrate the difference in signal-to-noise of the two recorded signals.

First I performed a high-pass filter operation on the waveforms of the first two transmissions. This was to focus a little more on the noise and less on the complex voice waveform. Analyzing the resulting waveform did not reveal a visually significant difference between the two operators' signals.

Next I selected a segment of the waveform between words and performed an FFT operation. The resulting displays show the levels vs frequency. The top graph is from the recording of WA5MLF's transmission, and the bottom is from KB4PYR.

The sharp peak around 6 to 7 kHz is from aliasing and should be ignored. Looking at the graphs in the 1200 to 2400 Hz range reveals a higher level on the recording of WA5MLF's signal, which I think represents its higher noise level relative to the voice peaks. Click on the figure to expand it for better viewing. I believe the noise is higher for WA5MLF due to the receiver's AGC allowing more gain on average during the voice transmission, allowing greater amplification of the background noise.

The FFT graphs shown here were produced by the Cool Edit 2000 program which was acquired by Adobe and is now part of their Audition product offering. An equivalent FFT graph can be produced with the free Audacity program, which also provides a text export of the data to be plotted with the software of your choice.

Analysis of received audio

KB4PYR audio

In reference to Carl's audio recording, here are histograms of each station's initial transmission, as analyzed by Cool Edit 2000 software.


WA5MLF audio

KB4PYR's audio sounded like it had better signal-noise ratio, but it is hard to draw conclusions from the analysis since voice is a complex waveform and the stations have different hardware and different operator's voices. The presence of more lower-level content in WA5MLF's audio suggests that it may have more noise or maybe that it is not as strong, but the AGC action of the Tuscaloosa receiver may obscure the comparison.

FFT analysis shows some differences, but mostly in the low-frequency end of the spectrum. A better frequency analysis should be obtained by first doing a high-pass filter operation on the recorded audio file. I'll report on that later.

Carl's antenna comparisons

Carl's posting provided a recording of HF receiver audio while switching every 10 seconds between two different antennas. Since the audio level on the first antenna was a bit lower than the second, I thought it might be helpful to make the two cases approximately equal for a better listening comparison. Below is my software's view of the edited recording, with the 1st, 3rd, and 5th segments boosted by 40%. A copy of the edited recording is available here.

My impression while listening to this copy of the audio is that the 2nd segment has slightly stronger voice audio but with background noise that has more of a power line noise character.



Using my copy of the Cool Edit software I ran some Fast Fourier Transform analyses on the edited copy of the recording as well as the original. A transform of the complete audio recording, based on 8192 points, is displayed in the graph below.



Transforms of the 1st segment of audio (antenna 1) and the 2nd segment (antenna 2) looked fairly similar. Let me know if you'd like to see those graphs.

Due to the varying frequencies and levels of the voice, it was not possible to pick out its characteristics easily from the broadband background noise. A better test would be to transmit one or more steady audio tones and record the received results on the different antennas for comparison using the FFT analysis.

John