Noise removal from audio tracks

Collapse
X
 
  • Filter
  • Time
  • Show
Clear All
new posts
  • Dave2002
    Full Member
    • Dec 2010
    • 18057

    Noise removal from audio tracks

    A few weeks ago I posted a comment about a downloaded track from iTunes - Bridge: Summer, RLPO Groves - http://www.for3.org/forums/showthrea...CD-remastering I have since revisited this a few times, and experimented with noise removal using Audacity.

    There is a small gap of "silence" at the start of the track, and a not quite similar one at the end. Listen carefully close to the end and the noise is apparent, but then goes down to zero - or close to it.

    I homed in on a fraction of these two gaps - perhaps a second - maybe slightly less, and did a spectrum analysis. One shows noise at -48dB going down to -87dB out to 3kHz, while the other shows noise from -54dB going over a very similar frequency range. Interestingly both show a peak at just over 15 kHz - 81dB and -87dB respectively.

    I then used the short extract from the end (before the "silence" does drop to zero) as a noise profile in Audacity, and selected the whole track for noise removal.

    This does appear to remove some/most of the troublesome noise. Checking the "silent" parts again after the noise removal shows complete removal of the 15+ kHz peak, and the noise in the two gaps has now shifted down to -66dB (end "silence") and lower and -73dB (opening "silence").

    To check that the levels on the samples had not changed significantly the peak around 9 minutes in was analysed, and contains a fairly wide band of frequencies up to 20 kHz decreasing from -18dB down to -90dB at over 20kHz. The levels were consistent across the versions.

    It is of course possible that some of this "analysis" is influenced by other factors, such as the size of the window used for the spectrum analysis, the type of window, but on balance the measurements suggest a noise reduction of over 10dB - could be as much as 18dB - but I'm conservative. I could also have made mistakes, but I have tried hard not to. I'm willing to repeat this a few more times if needed.

    Listening does suggest a clearer sound - though it may now be more important to listen in a low noise environment.

    If I could find somewhere to post samples of the audio for comparison purposes, I would be willing to do so - though not the whole track. I don't know of any good sites for this purpose right now.

    My concern regarding low frequency noise is partly that under some conditions I can hear it, and it is distracting, and secondly that it sometimes has a definite pitch which interferes with the intended pitches in the music. Perhaps randomly pitched noise is less of a problem.

    My track was downloaded from iTunes, but the effect of the noise can clearly be heard in this track (5) - http://open.spotify.com/track/2XzLNwdooyXy2HDOrzxBub

    It would also be good if others could check this out, and perhaps also find their own examples to process in a similar way. Listeners, such as JLW, would be welcome.

    My opinion is that the RLPO Bridge recording does sound quite a bit better after this process, and other recordings may benefit too. That is why I started the other thread on the quality of remastering. Commercial companies should have the capability to do this kind of work, but I feel that often they don't bother to check the quality of their output. It is, after all, cheaper that way - if nobody complains.
  • johnb
    Full Member
    • Mar 2007
    • 2903

    #2
    I would guess that with any noise or glitch removal there is a balance to be struck between aggressive noise removal and maintaining the tonal quality and the subtle ambient clues of the original. To take an extreme example many prefer the Naxos remasterings of the Schnabel Beethoven piano sonatas (which leave quite a lot of noise in tack) to the more aggressively processed EMI remasterings. According to the information on Spotify the Bridge CD was remastered in 1989 and both computer hardware and software have developed in leaps and bounds since then, so it is possible that if the remastering was done today the 'balance' I referred to would have shifted and the result would be more to your liking.

    I did listen to a little on Spotify but other than the slightly fierce treble, very noticeable on the strings (which IMO so often occurs on remasterings for CD) I couldn't identify the background noise you dislike, though it might be more evident and the start and end of tracks. (Out of curiosity I listened to a snippet of the Chandos/Hickox recording of The Sea (on Qobuz) and the strings sound very much more natural.)

    Comment

    • clive heath

      #3
      If you use the "Magix Audio Cleaning System" after you have isolated the no-music-just-noise section and got the sample, this system allows you to hear what is being removed as you test run the procedure, as well, obviously, as the processed sound. The other cleaning system I use doesn't have this facility, so I don't know how common it is. I prefer my own less noisy Schnabel!

      Clive Heath transcribes 78 records onto CD and gets rid of the crackle.

      Comment

      • richardfinegold
        Full Member
        • Sep 2012
        • 7794

        #4
        Originally posted by clive heath View Post
        If you use the "Magix Audio Cleaning System" after you have isolated the no-music-just-noise section and got the sample, this system allows you to hear what is being removed as you test run the procedure, as well, obviously, as the processed sound. The other cleaning system I use doesn't have this facility, so I don't know how common it is. I prefer my own less noisy Schnabel!

        www.cliveheathmusic.co.uk
        I've been very impressed with Clive's results, Dave. If I was you I would take his advice to the bank.

        Comment

        • Jasmine Bassett
          Full Member
          • Dec 2010
          • 50

          #5
          The character of the original noise sounds consistent with how the original recordings would be made and the resulting master tape prepared for transfer to LP.

          It was common practice for B/W TV cameras to be set up in the studio to allow those in the control room to see what was happening. Depending on the position of the camera in relation to the microphones the approx. 15kHz line whistle from the camera could be picked up by the microphones. If a noisy monitor was in use in the control room the 15kHz whistle getting on to the master tape could often be missed.

          Once the mix / edit was completed, the finished tape would be topped and tailed to add plastic leader tape before the start of the piece and often between each movement. If the piece started with a definite bang the leader would be cut tight up to the beginning of the first chord, otherwise a suitable gap would be allowed for the mastering engineer to quickly fade up at the start of each track and fade down at the end before the ambiance of the recording studio or the background tape noise cut to the silence of the leader tape. I suspect this particular skill has been lost when those same master tapes are transferred to CD.

          Comment

          • Dave2002
            Full Member
            • Dec 2010
            • 18057

            #6
            Originally posted by Jasmine Bassett View Post
            It was common practice for B/W TV cameras to be set up in the studio to allow those in the control room to see what was happening. Depending on the position of the camera in relation to the microphones the approx. 15kHz line whistle from the camera could be picked up by the microphones. If a noisy monitor was in use in the control room the 15kHz whistle getting on to the master tape could often be missed.
            That's interesting and could explain the high frequency peaks. In fairness they would most likely have been over 63dB down relative to the loudest signal - though not necessarily when compared with the quietest signals, and in a range which many of us can no longer hear, but I was surprised that they were so marked. I should check for the presence of those peaks in quiet passages to check that they are still present. Sometimes I think fluorescent lighting, including lights on music stands, can add in HF noise, and I have been to live recordings where I could hear the whistle from various lights which could very well have been picked up for radio broadcast.

            I am surprised at the degree of low frequency noise which does get on to some recordings. Some is more or less unavoidable - traffic noise, tube trains etc., if the recording venue is in a city centre, which is why some recordings are done at night. The particular Bridge recording was probably done in the Philharmonic Hall, which I think was not generally too noisy in those days, and I was unaware of some of the problems which plague other venues, such as air conditioning fans - they probably didn't have any in those days anyway! They might have had heating, though.

            Comment

            • Dave2002
              Full Member
              • Dec 2010
              • 18057

              #7
              Just checked again - and that 15+ (around 15.2kHz) peak is present throughout most of the track but disappears once the noise removal is applied.

              Comment

              Working...
              X