1. Register NOW and become part of this fantastic knowledge base forum! This message will go away once you have registered.

How does file conversion effect pitch?

Discussion in 'Recording' started by ClarkJaman, Feb 12, 2016.

Tags:
  1. ClarkJaman

    ClarkJaman Active Member

    Hey ya'll! Long time no see. For the past year or two I haven't been doing much recording or spending time online, but the past few weeks have brought major changes to my life, so I hope to be around more. I'm living in Canada again, and I'm hoping to stay here for a while.

    Anyways...

    I guess because I'm rusty I failed to make sure that all my files in this project were properly converted into the correct sample and bit rates when I first set out. After spending a dozen hours or so working on it I noticed that the notes didn't quite sound right when I pitch corrected them. That's when I realized that my sample rates were all messed up. Now I'm trying to sort this all out and learn from my mistakes.

    So my question is this: converting from 44.1khz to 48khz and from 16 bit to 24 bit seems to have raised the pitch 1.3 tones, as best as my ears can tell. But is there a way I can calculate exactly how much the pitch was changed?
     
  2. Boswell

    Boswell Moderator Distinguished Member

    It's purely ratiometric, Clark.

    The ratio 48/44.1 is 1.0884, or 8.84% higher. A semitone is an interval of about 6% (this is by how much the 12th root of 2 is greater than 1), so replaying a 44.1KHz file at 48KHz without any sample rate conversion will raise the pitch by a bit less than 1.5 semitones. The playing time will reduce by the same ratio.

    The wordlength does not play a part in this.
     
    kmetal, audiokid, dvdhawk and 5 others like this.
  3. Kurt Foster

    Kurt Foster Distinguished Member

    i love reading your replies Bos ..... you're one smart cookie
     
    kmetal, audiokid and DonnyThompson like this.
  4. ClarkJaman

    ClarkJaman Active Member

    You are indeed a smart cookie Boswell. I understand everything except this:

    How can I calculate exactly how many semitones I need to lower the pitch? Right now I have it lowered 1.3 semitones and it sounds close but not quite perfect.
     
  5. bouldersound

    bouldersound Real guitars are for old people. Well-Known Member

    I'd probably reverse engineer it: tune a guitar by ear to the song and let the tuner tell me how off it is.
     
  6. bouldersound

    bouldersound Real guitars are for old people. Well-Known Member

    Or I'd create a 44.1k file with a 440Hz tone, change the flag to 48k and analyze the spectrum (which reads out the the most prominent note).
     
    ClarkJaman likes this.
  7. ClarkJaman

    ClarkJaman Active Member

    Ah! Yes now that is a good idea! I should have thought of that! ;P
     
  8. DonnyThompson

    DonnyThompson Distinguished Member

    Bos is an audio God. ;)
     
    audiokid likes this.
  9. Boswell

    Boswell Moderator Distinguished Member

    Sorry, I thought I mentioned it in my previous post: 1.5 semitones (1.487 if you want it more exactly).

    Calculation is really the only useful technique for this type of problem, as you won't get the necessary precision out of any of the standard spectrum analysers. A guitar tuner that reads out in cents would give you a better experimental result, but even then you would have to correct the reading once you had subtracted the whole semitone from it. The required reading is a semitone plus 27.3 cents of the next semitone, not 48.7 cents.
     
  10. ClarkJaman

    ClarkJaman Active Member

    Ok great. When you said "a bit less" than 1.5 semitones you really meant "a bit!" I was almost 20cents off. :/
     
  11. dvdhawk

    dvdhawk Well-Known Member

    With all of that genius applied to give a definitive answer to your actual perfectly good question, may I ask a few dumb questions of my own as a follow up? ( hopefully you don't mind, because here they are )

    What's the end goal of the project? (CD, video, mp3)
    What program / process are you using to convert them?
    What DAW are you using?

    As much as I enjoyed learning the answer, your original post made me wonder if it was worth all the trouble - which depends on what you want to do with it. CD quality is 16-bit 44.1kHz. Often conversion(s) are only going to degrade it, even if it's up-converting to theoretically higher rates. It can't make data where there was none, so the algorithm has to calculate a best guess. There are worse things than working in 16-bit 44.1kHz (CD quality).

    And I'm currently using StudioOne (and before that an LE version of ProTools), both could import standard CD files into a session with a higher bit-rate and sample-rate without affecting the pitch. It still has to do the up-conversion during the import process, but it also factors in the exact sample-accurate length of the file ( in terms of minutes:seconds:samples) so it doesn't alter the pitch. So your difficulties might be the difference between opening the files in a different format on the DAW vs. importing the files into a different format session.

    Boswell has brilliantly covered the 'how', I'm left wondering about the 'why'.
     
    Brien Holcombe, kmetal and audiokid like this.
  12. ClarkJaman

    ClarkJaman Active Member

    Yeah, I totally understand what you mean. I am not actually going to use the converted files in the final product. The final product is a youtube video, which I understand is best done in 24bit with 48khz (correct me if I'm wrong, or if this has changed? ) But the real answer to the question of WHY is that I was also experimenting with getting around the copyright detection system on youtube and just trying to learn by doing. :p
     
  13. dvdhawk

    dvdhawk Well-Known Member

    "trying to learn by doing." I can certainly appreciate that! (y)(y)

    You're right 48kHz is the preferred sample-rate for video editing. So you're working in a video editor rather than a DAW? Again from first hand experience, Premiere, Final Cut, or even iMovie will both import CD files into a 48kHz project. I'm sure most, if not all, of the other NLE programs will too.

    If your video is for educational purposes you should be OK under the fair-use provisions of the law.

    Good luck!
     
  14. ClarkJaman

    ClarkJaman Active Member

    Ok I'm glad that hasn't changed. :p I'm using Cubase 8.5pro for the audio, and once the audio is finished I upload it to FCPX.
     
  15. kmetal

    kmetal Kyle P. Gushue Well-Known Member

    That was my first thought. Why not just work w what's there. Any processing, effects, conversion is always at a price analog or digital, it's always a matter of is it worth it, and diminishing returns.

    Last time a checked sound quality didn't matter to YouTube viewers. ;)
     
  16. John Santos

    John Santos Active Member

    Wow. That's an awesome engineer-y way to explain this.

    A bit difficult to understand for me though. Haha :cry:
     
  17. paulears

    paulears Well-Known Member

    Just a thought - you've changed the pitch AND the time, so running the thing through a pitch shifter will still sound a bit wrong because the speed is wrong too. Why don't you just re-sample them. Bring them in at their recorded sample rate, the output it at the new one. Pitches and tempo will then be fine!
     
  18. audiokid

    audiokid Staff

    :D
    oh ya... :love:
     
  19. paulears

    paulears Well-Known Member

    Just thinking about what you said about pitch and Youtube's copyright system. I doubt that will work. You tube is pretty good at detecting well known songs even in different keys and with different tempo settings. I have no idea how it does it, but it's flagged a couple of ours recently and while nice to think it's close enough to be detected, still a bit of a pain. It even spotted a few seconds of a popular song being played right in the corner of the frame on a screen in the room with playback through a tiny speaker. So loads of people talking in a busy room, and the TV in the background was playing a well known song, that you could just hear - and it found that! Not bad programming by somebody.
     
    Sleepyhollos likes this.
  20. ClarkJaman

    ClarkJaman Active Member

    Actually, it is quite easy to fool Youtube's copyright detection. Just shift the pitch and time a little bit. Those examples you gave are probably not the copyright detection software, but rather a human being who flagged the video as a violation of copyright. Did the video get flagged as soon as it uploaded, or was it later on, after it got some views?
     

Share This Page