HF SSB vs. Digital Voice

ssb_snr_n0.4

Introduction

In a recent YouTube video I saw a demonstration of a commercial HF radio communicating over a distance of 90Km using NVIS and digital voice. Using SSB the received signal was barely above noise level, and hardly decipherable. However, when the digital voice modem was switched on, the signal was very clear. I decided to do a simulation of how HF SSB compares to an Amateur Radio Digital Voice modem FreeDV (Ref.1).

SSB Simulation

Fig.1 Block Diagram for SSB SNR Measurement

Figure 1 shows the Scicos block diagram used to evaluate the SSB SNR Signal to Noise Ratio (Ref.2/Chap.6). A 15 second voice wav file snr_ssb_wav is read into the console editor and converted to a structure V. A variable noise source is generated and added to the voice structure. The S+N structure A is converted back in the Console editor to a wav file snr_ssb_out.wav. Performance is measured for various noise levels. Figure 2. shows the results.

Fig.2 SSB SNR vs. Noise_n_sigma

DV Simulation

Fig.3 FreeDV SNR Testing Block Diagram
Fig.4 USB External Sound Card
Fig.5 FreeDV Scicos SNR Testing Block Diagram

FreeDV is amateur radio software used to convert an analog mic signal to an analog digital waveform suitable for an SSB transceiver. The software is designed to work with a PC sound card and a transceiver sound card. The connection block diagram is shown in Figure 3. For a stand alone test, I used an external USB sound card to emulate the transceiver. This is shown in Figure 4. FDV_700d is used for this simulation as it is designed for low SNR.

The voice test signal is converted to its FDV digital equivalent and sent out to the sound card Speaker/Headphone port. Audacity is used to record this signal as fdv_snr.wav. This signal is now used as before with ssb, converted to a structure and read into the Scicos model, where noise is added. The output of the Scicos model produces a structure which is converted back to fdv_snr_out.wav. This is shown if Figure 5. With various noise levels, this file is read into FDV and the readability determined.

Fig.6 Scicos Scope Display FDV_700d S+N, N, S, n=0.1
Fig.7 Scicos Spectral Display FDV_700d S+N, n=0.1
Fig.8 FDV_700d Decoding Audacity fdv_snr_out.wav to FDV

Figure 6 shows the Scicos graph of S+N, N, S for n=0.1. Figure 7 shows the output spectrum and Figure 8 shows the decoding of FDV_700d.

Input Level
Vp2p
Noise
n=sigma
Noise Level
Vp2p
SNR(dB)
20log(S/N)
Readability
1-5
2 x 0.6 = 1.2n = 0.10.215.65
2 x 0.6 = 1.2n = 0.20.3810.05
2 x 0.6 = 1.2n = 0.41.01.65
2 x 0.6 = 1.2n = 0.51.25-0.45
2 x 0.6 = 1.2n = 0.61.4-1.35
2 x 0.6 = 1.2n = 0.81.6-2.55
2 x 0.6 = 1.2n = 1.02.0-4.470% words missing
Fig.8 FDV_700d SNR vs. Noise_n_sigma

Figure 8 shows the FDV_700d results. Note that digital voice has a hollow computerized sound, so what we are considering here is whether it is understandable. The message lost no words up to n = 0.8, beyond that at n = 1.0, about 70% of the words in the message were lost.

Conclusion

Comparing the SSB and FDV_700d results we see that the threshold for readability for SSB is between 3 & 9dB and for FDV_700d < -3dB which is a considerably lower SNR. Thus we can say that FDV_700d gives much better performance than SSB at very low SNR.

Fig.9 YouTube Video HF SSB vs. Digital Voice

Please send your comments, questions and suggestions to:
contact:

YouTube Channel
YouTube Channel

References

#1 – “FreeDV: Open Source Amateur Digital Voice”
https://freedv.org/

#2. – “HF High Frequency Radio Learn by Simulation”
https://www.clarktelecommunications.com/simulation.htm