Hacker News Reader

Comment by vitovito | Hacker News Reader

From my notes on a radio archiving project I'm doing:

> Could you record the entire radio spectrum and extract stations and broadcasts later?

> In the USA, the AM radio band is 540-1710 kHz, a spread of 1170 kHZ,

> http://rtl-sdr.better-than.tv/?page_id=237 states that spectrum recordings are also a function of how many samples per second you choose to record:

> 2.8msps - 44.8mbps = 5.6 MB/sec.

> 2msps - 32mbps = 4 MB/sec.

> 1msps - 16mbps 2 MB/sec.

> http://www.myradiobase.de/perseus/ has sample files. 3.5 minutes is ~360MB. 60 seconds of 1500 kHZ is a gig. All too much. 24 hours at 1MB/min is 1.5GB, but 17MB/sec. is 1.4TB. Spectrum recordings are out.

View on HN

Assuming most of this is dead air, don't you think this would be a highly compressible data set (even lossless)?

Highly is perhaps an overstatement, but compressible like most clear signal data, behaving similar to PCM audio.

I compressed '20090922_950_GLFS_outdoor.wav' (150 kHz - 1750 kHz) from the Perseus site [1] with some lossless compressors. In each case, I independently round-tripped the de/compression and the results were confirmed to be lossless.

The compression ratios are in line or slightly (but not overwhelmingly) better than what you'd expect [2] from audio:

  original:		 1.000
  7zip-lzma2-xz-ultra:	 0.820 (not a signal compressor)
  wavpack-normal:	 0.724
  wavpack-high:		 0.691
  wavpack-hhx6:		 0.593

I would've liked to test additional compressors. In addition to WavPack, FLAC, TAK, ALAC, OptimFROG, TTA, WMA Lossless all purport to have some level of high-resolution support, but I'm not sure how many kHz they go up to. FLAC wouldn't take the file, and the rest I don't have access to right this moment.

[1] http://www.myradiobase.de/perseus/

[2] http://wiki.hydrogenaud.io/index.php?title=Lossless_comparis...

At least in HF, this "dead air" is anything but dead (lightning, channels with lots of fading, things you only catch due to ionospheric effects, etc). Noise is basically any signal you don't care about in lower frequency bands. Also many transmissions look like wideband noise at higher frequency due to direct sequence spreadspectrum type techniques.

Reply on news.ycombinator.com