At first let me copy paste the message of Ruchard Langley as posted today in WOR:
https://groups.io/g/WOR/message/153592 in case it is public
Both files with an audio sample rate of 22050 samples/second. That's the rate that the U. Twente SDR receiver uses to record audio files.
For the record, I misspoke. The audio files from the U. Twente SDR receiver are recorded as WAV files with a sample rate that depends on the bandwidth setting of the frequency to which the receiver is tuned. For a total bandwidth of 5.08 kHz, for example, the sample rate is 7119 samples per second. That results in a file size of 0.854 MB per minute of recording. Converting a WAV file to a lossy MP3 file (WAV file is resampled using a complex algorithm) results in a different sample rate and bitrate depending on options selected. In general, a lower sample rate and/or bitrate will result in a smaller file size.
Here is my reply concerning to the resulted audios :
From my 20+ years long practice of audio conversions , the losses between MP3 or WMA and WAV are extremely small with really negligible differences even for studio recordings! As for example 32/32 for WMA is my identical setting for converting standard MP3 files of 128kbit /44 KHz
1458 Lyca audio-span WebSDR vs Wessex kSDR right as shown in freq analysis at max IF bandwidth (13.5 vs 9,7 KHz ) Click for original picture
1170 binaural (q-stereo) reception (mix RNE/Swansea)
max 9.8kHz IF filter, audio on 5kHz Wessex kSDR
Above are two pictures that show the audio span in the maximum filtering settings in WebSdR (twente) and kSDRs (Wessex ) for your info and your understanding . The second picture shows the binaural reception (SAS mode) on Wessex.
Important notice: shortwave recordings from any Kiwi or Twente use much lower bandwidth than any typical studio recording with the highest being 7kHz for the 13.5 kHZ for web SDR and not more than 5.5 kHz for kSDR as can be seen from the 2 pictures.
hope you are quite smart to also understand that
the span of an audio file must be minimum of the twice the audio bandwidth of the recording. That means if the highest audio bandwidth is 6 kHz the container digital format (MP3 or WAV ) must be 12 kHz
That clearly translates, it is not necessary to convert the WAV files into MP3 in higher rates than 32kbit/32kHz That means 2x16kHz of audio that is 2-3 times higher than the required. I have made a few times recordings at 24/11 kHz but some trebles were missing but without any artifacts in the sound in the MP3 format.
Artifacts can start below the 20kbit.
Hope that this explains all the situation now !
from a WOR user
ReplyDeleteA few thoughts from me:
I always make I/Q recordings on KIWI SDRs.
This means you can always record 12 kHz RF (carrier +/-6kHz) and demodulate the sidebands separately afterwards.
In the end, a stereo 16bit *.wav with 12 kHz sampling frequency is always downloaded. (even if you have reduced the online filter to +/- 3 kHz, which makes no sense).
Theoretically you can get an AF signal of maximum 6 kHz, mono, of course. Since I mostly record broadcasts with digital data content (MFSK/PSK/DRM), any kind of psychoacoustic reduction is out of the question.
So I/Q wav or its mono demodulation is saved as *.flac without increasing the sampling rate.
In general: The encoder performance at such low sampling rates in MP3 is pathetic.
WMA? Dead horses should be left to lie where *.vqf, *.ape and *.ra lie. An intelligent alternative to MP3 (single ch for SW rec) these days is m4a/aac, but with reasonable data rates and without spectral band replication.
I hear artifacts not only below 20 kbps, but also with poorly encoded 128 kbps MP3.
This usually happens when using old Blade/Xing MP3 encoders or with cascaded processing, i.e. multiple encoding.
MP3 ==> WAV ===> MP3......
DeleteA.Not all mp3 players support m3a /aac properly. All I tested play them as mono but WMA works properly
B.Both encoders you notice are long ago extinct. I used xing on 2k which really encoded much better then any other I remember the 11kHz /16kb being very linear. Now I use two encoder shells using Lame