Convert SSB recording on tape to AM?

Discussion in 'Microphones, Speakers & Audio Processing' started by KA2PTE, Aug 12, 2019.

ad: L-HROutlet
ad: l-rl
ad: Left-3
ad: Subscribe
ad: K5AB-Elect-1
ad: Left-2
ad: MessiPaoloni-1
ad: L-MFJ
  1. KA2PTE

    KA2PTE Ham Member QRZ Page

    Im converting some analog cassette tapes into mp3's for someone and one of the tapes
    is a radio broadcast in SSB recorded in AM.

    I have used Audacity for Windows to clean up some of the other tapes, but I dont think
    the SSB can be "clarified" with digital software?
     
  2. W7UUU

    W7UUU QRZ Lifetime Member Number 133 Administrator Volunteer Moderator Platinum Subscriber Life Member QRZ Page

    Feed it into a spare transceiver and transmit in SSB on a dummy load - receive on another SSB receiver nearby and record.

    Easy peasy... or not :D

    Seems like it should work....

    Dave
    W7UUU
     
    K0UO likes this.
  3. K7TRF

    K7TRF Premium Subscriber QRZ Page

    Maybe I'm misunderstanding his situation but from what I read that won't work.

    If his recording is what came out of an AM receiver when tuned into an SSB signal all you'll have is the intermodulation products between the various vocal or musical tones and you won't really have accurately recovered audio. IOW, it sounds like the OP has a recording of what amounts to the Donald Duck sounds you get when trying to demodulate SSB on an AM receiver. Not a lot you can do with that sound track. Feeding that back into an SSB modulator and then demodulating it on another SSB receiver won't help.
     
    ND6M, W2VW and W7UUU like this.
  4. KA2PTE

    KA2PTE Ham Member QRZ Page

    I was thinking of transmitting it in AM mode then listening on a side band. Heres a sample mp3.
     

    Attached Files:

    • SSB.zip
      File size:
      120.7 KB
      Views:
      21
  5. KA0HCP

    KA0HCP XML Subscriber QRZ Page

    Low quality audio is low quality audio. You can't restore full audio that was removed in.the original transmitter audio chain. So Single Side Band isn't Double Side Band (AM). Shrug.
     
    K3UJ and K7TRF like this.
  6. K7TRF

    K7TRF Premium Subscriber QRZ Page

    Yeah, as posted above the original information has already been lost and you're just left with a bunch of mixing products that got through the AM receiver's audio chain. Modulating audio in SSB and then demodulating with an AM receiver is not a reciprocal function as it's highly nonlinear. Once the information is lost you can't restore it by applying the opposite modulation. All you'd get is a double side band version of the garbled received audio along with a carrier signal into the SSB receiver.

    There's probably some crazy theoretical approach where you could try to model the combined transfer function of an SSB audio drive chain and modulator followed by an AM demodulator and its audio chain and then derive the reverse transfer function. If you could do that with any accuracy you could try to feed your incorrectly demodulated signal through that reverse transfer function. IOW, given the audio mixing products that came out of the AM receiver what original audio stream would have created those given the model of the transmission to reception transfer function. I wouldn't hold my breath on getting decent results as it would require a lot of assumptions to come up with that transfer function.
     
    Last edited: Aug 12, 2019
    K3UJ likes this.
  7. KA2PTE

    KA2PTE Ham Member QRZ Page

    Another ham on our Liberty net said its basically like using an AM discriminator to hear the SSB. Pretty much a diode and earphone
    is a basic AM discriminator, so I look at it like rectifying AC in a DC power supply, and being left with only the half wave rectified DC
    and trying to take that and come up with the original AC level with frequency without knowing its parameters, so yea pretty much a guessing game.

    However, with computer technology brute force, thought perhaps that could overcome , but I guess no one has tried such a venture.

    Since SSB has no duty cycle / carrier, the modulation is whats forming a rudimentary carrier so to speak, and that would be in the voice range of I guess
    20-20kc , the range of the human ear. I think most voice modulation fluctuates around 3 to 7kc from what I read someplace and if its USB the frequency is
    changing from the tuned freq +3 kc, and the reverse for LSB.

    The sample recording sounds like LSB. Hard to know what the original radio frequency was, but as it was done on an airplane (you can hear the 400hz AC humm)
    its possible the pilot may know that info.


     
  8. K7TRF

    K7TRF Premium Subscriber QRZ Page

    Rather than looking at an AM detector as a power supply rectifier, think of it as a mixer. It's easy to see that the only audio that comes out is the mixing products between the instantaneous tonal frequencies at their respective amplitudes. The trouble with trying to reconstruct such a signal is that there's an infinite number of input frequencies each at an independent amplitude that could create the recovered mixing products that you hear. You can bound that problem a bit as you've hinted by restricting the possible input frequencies to typical communications bandwidth voice signals.

    Seems to me it would be an interesting machine learning project (with limited application). IOW, come up with a parametrically adjustable model for the AM demodulated - SSB modulated chain then capture a bunch of voice waveforms intentionally run through SSB modulation and AM demodulation as labeled sets of training data (you have both the original audio waveform and the incorrectly demodulated output) use that to train the model. If you get decent model convergence over many training runs then try running your original recorded AM output signal through the same model.

    So yeah it's basically a big hypothesis testing problem with a transfer function with somewhat known characteristics but highly nonlinear so many potential input waveforms that could have created the captured output waveform. The trick is to see if you can let a good set of machine learning algorithms converge on a plausible transfer function that restores at least some of the lost information given some of the known boundary conditions such as the range of human communications voice in amplitude and frequency given things like comms systems filtering. Where's an eager grad student when you need one ;)
     
  9. KF5FEI

    KF5FEI Ham Member QRZ Page

    Could you not inject a signal like a BFO does on a cheap receiver?
     
    KW6LA likes this.
  10. K7TRF

    K7TRF Premium Subscriber QRZ Page

    You could have done that before the signal was demodulated, but not after it's already been passed through an AM receiver that didn't have a BFO. Once it's out of the AM receive chain you've lost the original signal and all you have left is a garble of intermodulation products.
     

Share This Page

ad: M2Ant-1