DK8KW Longwave Information
Slow-Voice - Transmitting compressed analog Audio Signals on LF

The first Succesful 2-way QSO in Slow-Voice

After we developed the idea of tranmsitting compressed audio on LF using a limited bandwidt, Markus, DF6NM has send a short test message (MP3-file, 12 kB, with Winamp for example you are aple to listen to MP3-files) using a bandwidth of 500 Hz.

December 31, 2000, Markus, DF6NM and I have conducted the first succesful 2-way LF QSO in slow-voice. Markus used a software named Goldwave (don't know where that can be found but Markus said it is shareware, so have a look in the Internet search machines ...). I used Cool Edit 2000, that can be downloaded at (it is also shareware, the demo version comes with two functions only, you need to register to have full access to all features).

We started at 137.350 to 137.600, where Markus was able to read my signal (8 times slower than normal voice) with 57. When I tried to receive him I found that the 1.7 kHz filter that I use in my MV62 Level Meter was too wide to prevent DCF39 from causing heavy QRM. The build-in 100 Hz filter is too narrow for the 250 Hz bandwidth we used. The 250 Hz filter that I use in the IC-746 that comes behind the level meter on 200 kHz is narrow enough, but it comes too late, the band is already spoiled by the Burg transmitter.

The solution was to move 500 Hz further down, with our transmission spectrum situated between 136.850 and 137.100 kHz. I hope we did not cause any inconvenience to anyone tranmitting there. We have accompagnied our slow-voice transmissions with CW to steer the QSO. All in all it took about 2 1/2 hours to complete the QSO, whereby my report for Markus was 35, however, in a later transmission I could read him better than that (probably 45).

The QSO mainly too that long because I had to fiddle around with the appropriate adjustments. For the transmission I used macros, so it took only one or two minutes to be able to transmit a 3 minute over. For reception I did not prepare anything and had to do some experimentation.

For future transmissions we will need to find another way or a better clear spot, maybe even further reducing the bandwidt to 100 Hz and make use of the narrow filter ...

MP3-Sound Samples of this QSO

DK8KW transmitting
DF6NM transmitting
an initial test using a sentence that Thomas Edison used to check his first recodring device (10 kB)
Markus' answer with my report, very hard to read due to QRM by DCF39 (24 kB)
after changing the frequency, Markus can be understood much better, giving me 57 (11 kB)
I give Markus a report of 35 (40 kB)
Markus' final (13 kB)
My final (11 kB)


Slow-Voice data of the QSO

- transmitted audio bandwidth 300-2300 Hz (2000 Hz total bandwidth)
- speed: 8 times slower than the normal voice
- transmition mode: USB
- total radio frequency bandwidth 250 Hz with 300 Hz audio equal to the lower frequency of the spectrum, 2300 Hz to the upper freqeuncy
- QSO control by CW in the middle of the spectrum used

Advantages of Slow-Voice

I know that SSB QSOs have been performed in the band some years ago in the U.K., but this slow-voice transmission helps us to

- overcome the bandwidth limitation in Germany (max. 800 Hz)
- helps to prevent QRM aking best use of the availabel spectrum
- increases the signal-to-noise ratio relative to the bandwidth used
- ... and is a lot of fun ... hearing the other's voice clearly readibly despite the low signal strength!

First Conclusions

A new and interesting field of experimentation. We will conduct further tests and improve the operating procedures. Maybe someone who has fun programming the Soundblaster card might want to develop a slow-voice software that allows more-or-less realtime QSOs ... at least considering the low field strength a fascinating ways of transmitting audio on LF.

Best 73, thanks to Markus, DF6NM for the nice experience, to Walter, DJ2LF for his reception report and a very happy New Year 2001 to you all!

Some further comments and hints, collected from the RSGB LF-Reflector

Hallo to All,
   received Markus, DF6NM, yesterday in the evening with slow-voice. The
time was stretched 4 times, the bandwidth reduced to around 500Hz, but I
used SSB-USB. This morning at 9.45 UTC received him again, but now through
the 500Hz-filter. Yes, it is his voice!
   Also heard both, Markus and Geri, in PSK31 - no Problem, good copy.
All the best for 2001.
73 Walter DJ2LF

Dear Geri, Lionel, and LF group,

 playing with slow voice, I found that the main problem with simply
time-stretching ("transposing") the audio spectrum is the slope of the
transmitter's SSB filter. It is not steep enough to pass a lower band edge of
about 75 Hz. One solution is to digitally shift the whole passband to the
middle of the normal speech spectrum, eg. a 1/4-rate bandwidth of 1450 ...
2000 Hz above 1378 Hz zero beat. Such a function is not normally present in
sound editing software, but by a combination of undersampling-aliasing and
bandpass-filtering, an SSB mixer can be emulated. In the appended .txt file
there is a description of how to (ab)use the GoldWave shareware sound-editor
for this purpose.

For receiving, you can reverse the procedure. An easier way for analog-minded
people like me is to record the slow SSB signal from the RX (eg. Windows
audio-recorder at 22050 samples per s), couple the PC's audio output to the
antenna input of your VLF converter (tuned to 5.5 kHz USB, plus 4x any offset
from the LF link), and replay the recording 4 times fast.

Locally, the scheme worked amazingly well, now I'd like to try it out on the
air. My preferred freq would be 136.88 kHz shifted zerobeat, 136.95 to 137.5
USB. With 16 mW radiated PEP this experiment will be rather restricted, so
who of my neighbours would like to try?

73s de Markus, DF6NM

How to produce time-expanded, frequency-compressed and passband-shifted SSB
using the GoldWave sound editor

by DF6NM, 26.12.2000

Basic principle:

- start with 5512 Hz sampling,
- equalize and filter,
- resample to 22050 Hz causing aliasing around n*5512 Hz,
- filter out USB spectrum above 5512 Hz,
- reduce speed to 1/4.

Working steps in GoldWave:

Record voice using 11025 Hz sampling rate

 1. 11025 sps 8bit -> 16bit
 2. Lowpass 400 Hz (steepness 1, 6dB/oct to equalize sinc response of alias)
 3. Bandpass 300...2500 Hz
 3. (Volume *2)
 4. Resample 5512 Hz
 5. Resample 22050 Hz  (causes the desired aliasing above 5512 Hz)
 6. Bandpass 5800..8000 Hz (steepn. 20)
 7. Volume *4 *4 *4 *4
 8. Playback rate 11025 sps
 9. Speed *0.5
10. Bandpass 1400...2100 Hz (steepn. 20)
11. (Volume *1.2 or nonlinear dynamics + Bandpass)
12. Save (8bit)

Resulting SSB spectrum 1453...2003 Hz, zero beat 1378.12 Hz.
Using Windows audio-recorder "sndrec32.exe" does not interrupt spectrogram while playing.

Please send comments & suggestions to:

copyright (c) 2001-2006 by dk8kw, all rights reversed