Announcement

Collapse
No announcement yet.

Blip in Waveform ?

Collapse
X
 
  • Filter
  • Time
  • Show
Clear All
new posts

  • Blip in Waveform ?

    Group,

    I have been working on some acoustic Edison Diamond Disc records and have noticed something "strange" to me.
    After applying filters I usually place the entire song on the amplitude vs time display to look for "spikes" from scratches that are left over. Of course, you can look for the spikes before the filtering also.

    I use the "gain normalize" function to have the big spikes show up since the spikes set how far the music's amplitude can be boosted and so after you use the gain normalize, the spikes are going to the top of the display.

    Now, when I zoom in on the spikes I can clean up the signal using the "I" key to remove the impulse...BUT....sometimes, the spike turns out to be part of a real music note. What is different, is that the waveform is nice and smooth with JUST ONE CYCLE HIGH ? This one cycle causes a big spike upwards and then the rest of the waveform is back down.

    What I do, is highlight the one spike in the waveform around the zero crossing and then press the I key. The spike is removed and the waveform then has a nice smooth shape. The gain/normalize function then increases the song and "life is good".

    Where does the one cycle of extra high signal come from ????? The cycle is not a scratch because the scratches looks very much different from the normal song waveforms.

    This "single cycle spike" seems to occur more often on acoustic waveforms than electric recordings.

    Comments ?

    Marc
    Last edited by Craig Maier; 05-07-2019, 05:51 PM.

  • #2
    Hi,

    It sounds like you are seeing a wrap-around error which is most likely a bug in the soundcard driver. Try this - use the gain "Normalized Gain Scaling" function instead of "Gain Normalize". Gain Normalize to something like -0.3 dB. That will give the soundcard a little extra headroom to deal with clipped spikes. BTW - I have heard of that happening before (if I have it right). The positive clip "wraps-around" to the negative side of the waveform and it has something to do with certain drivers.

    Craig
    "Who put orange juice in my orange juice?" - - - William Claude Dukenfield

    Comment


    • #3
      Craig,

      Thanks for the information. I'll run some experiments and let the forum know.

      By the way, for a 16 bit waveform, I take it that 0.0 dB corresponds to +/- 2^16-1 ?

      Marc

      Comment


      • #4
        Hi,

        Quoting Marc:

        "By the way, for a 16 bit waveform, I take it that 0.0 dB corresponds to +/- 2^16-1 ?"

        Yes.

        Craig
        "Who put orange juice in my orange juice?" - - - William Claude Dukenfield

        Comment


        • #5
          Well, I tried using -0.3 dB for a gain normalize setting on a previous "problem" song. I did NOT have any strange blips in the waveform.
          I'll continue to use a lower setting than 0.0 dB.

          If my memory is correct, I do remember having this same effect show up on another audio system of mine. Could it be that having the sound cards receive the maximum digital input to the D/A converter is not recommended ?

          Marc

          Comment


          • #6
            Quoting Marc:

            "Could it be that having the sound cards receive the maximum digital input to the D/A converter is not recommended ?"

            I have not seen the problem, per se. But, Rick has seen it so I guess some sound card drivers can have an issue with a full count converter. The system that I am writing you from does not do that, but that means little.

            Craig
            "Who put orange juice in my orange juice?" - - - William Claude Dukenfield

            Comment


            • #7
              I haven't seen the problem in my sound card either, but I think there's a post on this board that recommends normalizing slightly less than 0dB. I usually use -1 dB.
              Dan McDonald

              Comment


              • #8
                I haven't seen this either, but If you have it available, try using the spectral view for manual declicking. That way clicks and pops appear as spikes in the frequency spectrum and are easily visible regardless of their amplitude. You'll have to set up the display frequency range (~100Hz - 12kHz) and amplitude scales to something suitable.
                At work I may look like I'm doing nothing, but at the cellular level I'm actually quite busy

                Comment


                • #9
                  Hi Doug,

                  Yes, that is a good suggestion - - - I would add - - - use the spectrograph view in Classic Edit mode so you can see both the time and spectrograph domains at the same time. That way, the time domain is in the source display and the spectrogram is in the destination display area and they are in time sync.

                  Craig
                  "Who put orange juice in my orange juice?" - - - William Claude Dukenfield

                  Comment


                  • #10
                    With regard to the spectrograph view....I haven't used that tool much at all.

                    Could you suggest some settings to get me started ?

                    Marc

                    Comment


                    • #11
                      Under preferences and the spectrogram tab I use the following settings:

                      Frequency Axis: Linear
                      Amplitude Axis: Gamma, 0.5
                      FFT Size: 2048

                      Display Frequency Range: 10Hz - 12000Hz, default not checked.
                      Display Frequency Labels: Checked
                      Collor Palette: "White-Yellow-Red-Black"
                      Invert Palette: Checked

                      I then use the sliders on the right of the spectral view to control the brightness/contrast and how much surface noise I can see vs impulse noise/distortions. Hope this helps.
                      Although the imiage is a bit dark, I use the spectrogram looking something like this:
                      Click image for larger version

Name:	selective1.jpg
Views:	1
Size:	177.4 KB
ID:	42578
                      At work I may look like I'm doing nothing, but at the cellular level I'm actually quite busy

                      Comment


                      • #12
                        Thanks Doug for the info. I'll try it on a noise problem I haven't been able to fix (Little Girl-Freddie Rich's Radio Orchestra).

                        Marc

                        Comment


                        • #13
                          Hi Marc,

                          Have you ever tried the DSE (Direct Spectral Editor found under the Edit Menu)? That is useful for fixing long-lived events that need manual interpolation. Anyway, if you want, send me a sample of the file you are trying to fix and maybe I can make a suggestion.

                          Craig
                          "Who put orange juice in my orange juice?" - - - William Claude Dukenfield

                          Comment


                          • #14
                            Craig,

                            File on the way.

                            Thanks

                            Comment


                            • #15
                              Hi Marc,

                              Got the email; I will spend some time with the file this weekend and get back to you.

                              Craig
                              "Who put orange juice in my orange juice?" - - - William Claude Dukenfield

                              Comment

                              Working...
                              X