Hi fidelity SSB bandwith

Discussion in 'Ham Radio Discussions' started by K1VSK, Feb 27, 2007.

Thread Status:
Not open for further replies.
ad: L-HROutlet
ad: l-rl
ad: Left-2
ad: Subscribe
ad: MessiPaoloni-1
ad: L-MFJ
ad: Left-3
  1. WA0LYK

    WA0LYK Ham Member QRZ Page

    Mark,

    The last sample sounded "pinched" to me. You could certainly pick out the 'sh' and 's' (plural), such as "shows" very well.

    Would it be hard to add a 4th sample to the end with a response just like the third only from 300 to 4000 Hz?

    The extra low frequencys might add a little 'presence'.

    Thanks

    Jim
    WA0LYK
     
  2. N5RFX

    N5RFX Ham Member QRZ Page

    Jim,

    I had to repost the sample because I inadvertently lowered the sample rate which squeezed the bandwidth of the first and last samples. I will add your request.

    73,

    Mark N5RFX
     
  3. N5RFX

    N5RFX Ham Member QRZ Page

    Here is the revised clip with 300hz-4000hz with a 6db per octave preemphasis at the end.

    1st part 14K bandwidth with white noise giving a SNR of 9 to 10 dB.
    2nd part of the sample reduces the bandwidth from 14K to 2.8K (200 to 3000) Hz. 18 dB SNR
    3rd part 500hz-4000Hz with a 6db per octave preemphasis 15dB SNR
    4th part 300-4000Hz with a 6db per octave preemphasis 10 dB SNR
    Revised NPR samples
    73,

    Mark N5RFX
     
  4. K3VR

    K3VR Ham Member QRZ Page

    Interesting stuff. There are a few unknowns with clips taken from the internet: What kind of pre-emphasis has been applied in the studio, and; what kind of compression algorithms have been applied in the analog to digital conversion, and; how much information is lost in the conversion to mp3 format?

    Also: Working with a male voice will give us a better idea about what happens when we cut off the low end, since women usually don't produce much vocal information below 200 Hz.

    .
     
  5. N5RFX

    N5RFX Ham Member QRZ Page

    With all of changes I made, I was afraid that the 4 samples were not correct, so I performed the analysis again.

    1st part 14K bandwidth with white noise giving a SNR of 9 to 11 dB.
    2nd part of the sample reduces the bandwidth from 14K to 2.8K (200 to 3000) Hz. 17 dB SNR
    3rd part 500hz-4000Hz with a 6db per octave preemphasis 14dB SNR
    4th part 300-4000Hz with a 6db per octave preemphasis 16 dB SNR

    The SNR on that last sample is what alerted me to the fact that I had lost control. These all pass the sanity test.
    NPR Bandwidth, Preemphasis, and Noise Comparison
    73,

    Mark N5RFX
     
  6. N5RFX

    N5RFX Ham Member QRZ Page

    Since all samples were derived from the same clip, the pre-emphasis, compression and mp3 loss should not affect the comparison. I agree if the comparisons were made with differing initial pre-emphasis, compression, and mp3 loss, then we would have a problem.

    The female/male voice was interesting to me. Our speaker in these last samples still has most of her voice energy at 500Hz or less. That is why the SNR goes up when the lower cutoff frequency goes down. At 200 Hz our speaker looses a little over 4dB of voice energy.

    200Hz 4 dB
    300 Hz 2.5 dB
    400 Hz 2 dB
    500 Hz 1.7 dB
    600 Hz 1.4 dB

    73,

    Mark N5RFX
     
  7. K3VR

    K3VR Ham Member QRZ Page

    You're right, the pre-emphasis and etc. won't affect what we have to work with. Her clip was not overly processed or compressed prior to conversion as far as I can tell. It's very natural sounding. Here's a link to Ted Koppel's podcast, which has been eq'd, but which also doesn't sound overly compressed prior to conversion:

    Podcast

    Choose the button that says "Download MP3/4"

    Forgot to mention -- the research says we need 12 dB s/n ratio for excellent intelligibility, so keep the peak noise level 12 dB below the audio if possible.
     
  8. N3JI

    N3JI Ham Member QRZ Page

    Tim,

    Yes I have energy down to a bit below 100 Hz with a flat studio condenser mic and no processing. Here's what my voice looks like "naked". [​IMG]

    http://mysite.verizon.net/res1t267/Images/NaturalVoice2.jpg
     
  9. K3VR

    K3VR Ham Member QRZ Page

    Joe,
    In the .jpg you attached, at 100 Hz you're only about 7 dB down from the peak energy in that particular syllable. I hear lots of low end in that clip.
     
  10. N3JI

    N3JI Ham Member QRZ Page

    I found my "Punchy" clip I've been talking about! It's embedded in a PowerPoint slide, so you'll need that to hear it. Double-click the speaker icon to listen. The slide includes a spectrum plot of the clip as well:

    http://mysite.verizon.net/res1t267/Files/N3JI_Punchy.ppt

    Joe, N3JI
     
Thread Status:
Not open for further replies.

Share This Page

ad: hrd-2