Streaming versus downloads

Collapse
X
 
  • Filter
  • Time
  • Show
Clear All
new posts
  • johnb
    Full Member
    • Mar 2007
    • 2903

    #31
    I've never had any problems with dropouts on the files I have downloaded (I've only used TheClassicalShop, Qobuz, eclassical, and HDTRacks).

    Richard, do the dropouts always appear in the same place when you repeated play the file? If not that would indicate it is a problem with your own system rather than with the file.

    Comment

    • Anastasius
      Full Member
      • Mar 2015
      • 1860

      #32
      Originally posted by Cockney Sparrow View Post
      I've amended my post #28 to clarify I'm talking about downloaded files (e.g. purchased from Hyperion or iTunes).

      Surely, in essence audio files are no different from all the other data (those 0s and 1s) transmitted over the internet. If there were regular omissions in those files, then data exchanged in all sorts of commercial, military, communication etc fields would be unreliable. I presume the software must have inbuilt methods to ensure that files are downloaded intact.
      Exactly so and that's what I meant by my comment regarding checks and balances earlier on. However, data files per se as opposed to audio (and video) files are different in one very important aspect. Realtime! Your normal data file can be sent with great pauses in the data stream...then when the data stream starts up again, the software carries on storing the file. Even if the connection is broken, most file transfer mechanisms know exactly how much of the data file has been transferred and where to start up again once the connection has been restored. [Point A]

      The connection between your router ( actually it's the modem inside it that does the heavy-lifting) and the exchange is pretty clever. As the data is transferred between the two it is split up into smaller chunks on different frequencies, has extra 'recovery' data bits added so that in the event of an error in some of the data stream between the two, the true data can be recovered by using these extra 'recovery' bits. If the error cannot be repaired then the modem asks the exchange to send the bad data bytes again. All of this happens pretty transparently. Unless, of course, you have a bad connection between your modem and the exchange when the errors being introduced overwhelm the inbuilt correction mechanisms and your connection slows right down or even stops completely. ADSL 2 + is nothing short of magic.

      All this is fine for a normal data file. But with music or video you are listening in realtime and so if there are any breaks in the connection (too many errors on the line or a very slow server supplying the music, for example) then the audio buffer in your computer is emptied faster then it can be filled (pace 'slows right down' in the sentence above) and the music stops - briefly. If it is a download or a stream of data from a recording then the music will pick up again but without dropping any notes. My point A above. However, if you are streaming a live broadcast then, of course, the orchestra hasn't stopped (!) and so you'll miss the odd note. And if you happen to be recording this to a file then whenever you play back that piece of music again, that same break will always occur at the same place.

      Hope this helps.
      Fewer Smart things. More smart people.

      Comment

      • richardfinegold
        Full Member
        • Sep 2012
        • 7794

        #33
        Originally posted by Bryn View Post
        If Richard has an audio editor, even a free one such as Audacity, why not open one of the suspect files in that and look for a break in the waveform at the point the hiatus appears?
        I could probably use Garage Band for that. Alas, I have returned to work after the time off from back Surgery and have been putting in some long days so it will have to wait a while. Btw, I am not using a PC for playback. The files are on a USB drive and played back by inserting into a USB input in my Oppo 105 Universal Player. However I previously used a PC for playback with these files and other downloads and the dropouts occur there as well so the Oppo isn't the culprit.

        Comment

        • richardfinegold
          Full Member
          • Sep 2012
          • 7794

          #34
          Originally posted by Cockney Sparrow View Post
          Wouldn't the drop out occur in exactly the same place, time after time, if it was a flaw in the file downloaded?

          If it was occurring in the replay (buffering, the local laptop system being "overstretched" momentarily) then it would occur in different places, and by being able to shutdown other programs using the laptop, it would be possible to eliminate it?

          So if the mp3/FLAC/AIFF file is flawed,so not local replay factors - is it the case that the circumstances of the initial download could be the problem - if one's internet speed is very slow, or connection freezes/drops out** will that compromise the file integrity? If so, I would expect it to be a common problem and commented upon.
          **(e.g. when all one's neighbours are watching a Sports match, people in the house all watching Netflix or whatever)

          PS (on edit) - I am referring to downloads - e.g. a recording purchased from the Hyperion website and downloaded onto my PC, from which i replay it. So the problem is either inherent in the file stored on my hard disc, or in the replay system on the day I replay it).

          Streaming - well with the various links in the chain of replay - e.g. Spotify's servers, the various internet routing/connections, then the wi-fi in my home, then my PC/MAC/ streaming device such as Chromecast audio - its not surprising there might be dropouts, etc - particularly if one's internet connection is slow.
          That perfectly summarizes my situation .
          Btw, fwiw, I saw a Gramophone article from late in 2015 that said that classical downloads were down almost 20 % from the previous year. Probably many factors at work but I suspect streaming is the main culprit

          Comment

          • Cockney Sparrow
            Full Member
            • Jan 2014
            • 2294

            #35
            Anastasius in post #32 above. I'm recalling, vaguely a radio program about the guy who invented or developed into a useable system, packet switching - the practical way of sending bulk data but not having exclusive use of the communication channel (in those earlier days, limited and costly to access) -so in packets and as you describe, with the capacity to check it has the full file, and correct it up to a point. (The import of the programme was that he deserves some, well, a lot of fame as the enabler of the internet, etc. Per Wickipedia it seems he was Donald Davies of the National Physical Laboratory (UK).

            I'm a little bit confused though - can you confirm in your 2nd paragraph you are referring to a download file (e.g. iTunes "song" purchased and stored on my PC) - so it should, reliably have ensured it was the full file in the download process.

            And your 3rd paragraph is referring to streaming - Spotify, You Tube etc? If not then does your description - of the audio buffer being emptied faster than it can be filled - refer to the capacity of my PC to retrieve the data from an otherwise perfect file in my local hard disc, and play it through the sound card.

            Comment

            • Dave2002
              Full Member
              • Dec 2010
              • 18057

              #36
              Originally posted by Cockney Sparrow View Post
              I've amended my post #28 to clarify I'm talking about downloaded files (e.g. purchased from Hyperion or iTunes).

              Surely, in essence audio files are no different from all the other data (those 0s and 1s) transmitted over the internet. If there were regular omissions in those files, then data exchanged in all sorts of commercial, military, communication etc fields would be unreliable. I presume the software must have inbuilt methods to ensure that files are downloaded intact.
              There are differences between streaming and downloading. For downloads, the software may do error correction on each block of data, but in the worst case it might do a request for a resend of one or more blocks. The simplest system is Stop and Wait - so the download "halts" while the dodgy block is fetched again. If there are multiple failures, then usually the whole download would be marked as a failure. A cleverer system has numbered blocks, and the download process tries to keep going - with a process called selective repeats. Actually, for terrestrial communications (which would certainly include transfers locally in a LAN, or even between a server and an end user) there is not a lot of point in doing the selective repeat method, but there are benefits for satellite communications, as otherwise the overall efficiency drops off considerably. Essentially, if it is possible to get the download to work, then the results should be perfect, even if there are many repeats and errors corrected.

              Re streaming, if errors are detected, then if there isn't sufficient data buffered up, then either there will be a gap/glitch of some sort, or the software might interpolate the "missing" audio data - which is what CD replay will do in the worst cases. If there is a big gap, then even interpolation won't work, and the gap will be obvious to the listener.

              Comment

              • Anastasius
                Full Member
                • Mar 2015
                • 1860

                #37
                Originally posted by Cockney Sparrow View Post
                Anastasius in post #32 above. I'm recalling, vaguely a radio program about the guy who invented or developed into a useable system, packet switching - the practical way of sending bulk data but not having exclusive use of the communication channel (in those earlier days, limited and costly to access) -so in packets and as you describe, with the capacity to check it has the full file, and correct it up to a point. (The import of the programme was that he deserves some, well, a lot of fame as the enabler of the internet, etc. Per Wickipedia it seems he was Donald Davies of the National Physical Laboratory (UK).

                I'm a little bit confused though - can you confirm in your 2nd paragraph you are referring to a download file (e.g. iTunes "song" purchased and stored on my PC) - so it should, reliably have ensured it was the full file in the download process.

                And your 3rd paragraph is referring to streaming - Spotify, You Tube etc? If not then does your description - of the audio buffer being emptied faster than it can be filled - refer to the capacity of my PC to retrieve the data from an otherwise perfect file in my local hard disc, and play it through the sound card.

                The 2nd paragraph applies to all data...everything, emails, web, files, photos, you name it.

                3rd paragraph really only applies to streaming now I come to read it through again. It was not really aimed at any issue local to your system but the theory still applies nevertheless. In the case of your local PC then if, say, your hard drive is dying and the operating system is struggling to read the data correctly in a timely manner then, yes, your audio buffer will empty and you will hear a break or stutter in the audio. But if the 'break' occurs always in the same place then it's more likely to be a 'break' that occurred during the streaming or download and got recorded as a permanent 'break' in the file ..if you follow me.
                Fewer Smart things. More smart people.

                Comment

                • Anastasius
                  Full Member
                  • Mar 2015
                  • 1860

                  #38
                  Thinking a little bit more, even your hard drive in your PC behaves in a similar way. They only manage to squeeze so much data into such a tiny space by packing it ever more densely. That then puts more severe constraints on the ability of the read/write heads to read the data in a constant steady stream. To help the drive do this, the data is stored on the hard drive platters with protection bits, extra bits to aid recovery in the event that the data skips a beat 'as it were'. But recovering from 'bad' data takes time and so hard drives also have buffers in them - both read and write. If the disk is starting to die then the number of times a data block might need to be read in order to constitute all the data correctly - which means that the read/write head has to wait for that data block to come round again as the disk spins...and again...and again..that that buffer might not get filed up quite as quickly as the operating system is emptying it. However, all you will see is a slowness in your PC doing things.

                  This buffer also has the added benefit of speeding up the effective data transfer because the operating system can anticipate to a certain extent what data is needed or might be needed next and so pre-load the buffers with what it thinks is needed next. Same thing happens in CD players.
                  Fewer Smart things. More smart people.

                  Comment

                  Working...
                  X