Recording at 44 vs 192

What's Hot
24

Comments

  • flying_pieflying_pie Frets: 1806
    JezWynd said:
    And the end point of this is that it'll be uploaded to Soundcloud or YouTube and probably listened to on mobile phone speakers, cheap headphones or a generic car stereo 
    True but garbage in garbage out still applies. The more headroom you give yourself at capture the cleaner the results down the line.
    But that's down to the input level and not the sample rate. Think about it:

    You don't need the massive dynamic range to perfectly record a Jew's harp from the end of the street and a maxed out plexi from 1cm away with the same mic at the same input level.

    If you did, and were able to record it perfectly,  then the equipment playing it back wouldn't reproduce that dynamic range. And even if it could it would sound rubbish going from inaudible to painfully loud.

    This "perfect" recording would need seriously compressed in the mix to work when being listened to. So you'd be better off using different input methods to capture them. So you don't need that massive a dynamic range after all. 
    0reaction image LOL 0reaction image Wow! 0reaction image Wisdom
  • darthed1981darthed1981 Frets: 11669
    ICBM said:
    Admittedly some years ago so the processing power available is different, but I did some comparisons in a recording studio when deciding what rate and depth to use for a project. I’m pretty sure they were using Cubase 5.

    Surprisingly, there was a noticable difference between 44.1KHz and 48, but none that we could hear between 48 and 96. Likewise there was a difference between 16 and 24-bit, but none between 24 and 32. So given that using higher quality slowed the system down drastically when mixing, we used 48KHz/24-bit.

    But that’s for recording, when you need more ‘data headroom’ for processing. On playback media, there’s no evidence that higher than 44.1/16 actually improves the quality, and it could actually make it worse... this is an old article but worth reading:

    https://xiph.org/~xiphmont/demo/neil-young.html

    (Don’t worry, it’s not by Neil Young ;).)
    That article remains superb.
    We have to be so very careful, what we believe in...
    0reaction image LOL 0reaction image Wow! 0reaction image Wisdom
  • ICBMICBM Frets: 71950
    edited June 2018
    darthed1981 said:

    That article remains superb.
    I particularly like this bit

    "With use of shaped dither, which moves quantization noise energy into frequencies where it's harder to hear, the effective dynamic range of 16 bit audio reaches 120dB in practice

    ... 120dB is greater than the difference between a mosquito somewhere in the same room and a jackhammer a foot away."

    Which should illustrate perfectly why anyone who thinks they can hear the difference between 16 and higher bit depths is nuts.

    The only point of recording at higher rates is to allow more headroom and avoid distortion, which *is* audible - and even then 24-bit should be plenty.

    "Take these three items, some WD-40, a vise grip, and a roll of duct tape. Any man worth his salt can fix almost any problem with this stuff alone." - Walt Kowalski

    "Just because I don't care, doesn't mean I don't understand." - Homer Simpson

    0reaction image LOL 0reaction image Wow! 0reaction image Wisdom
  • TheBigDipperTheBigDipper Frets: 4722
    I've been told by my hifi enthusiast mate that, in double blind scientific tests, it shows that very few people (if at all - they could be guessing or just being lucky, the testing can't catch that) can hear the difference between a recording at 96kHz and 44kHz. Nor can they tell the difference between a 16bit recording and a 24bit recording. I'm one of them. A CD at 44.1kHz/16 bit sounds fine to me played back on decent equipment. 

    When recording, I use 24bit just to give me greater dynamic range on the sound I'm recording (as others have said) but I use 44.1kHz sample rate, so there's no rendering to be done when I send my mates a WAV file or cut a CD.
    0reaction image LOL 0reaction image Wow! 0reaction image Wisdom
  • PabloPablo Frets: 38
    Does it make a difference if you are applying effects post recording? Giving the computer more data/information to work with even if some of it is inaudible?
    0reaction image LOL 0reaction image Wow! 0reaction image Wisdom
  • octatonicoctatonic Frets: 33725
    Nor can they tell the difference between a 16bit recording and a 24bit recording. I'm one of them. A CD at 44.1kHz/16 bit sounds fine to me played back on decent equipment. 

    For rock/pop music uses 6db of dynamic range a lot of of the time anyway.
    Classical recordings can be a slightly different case but a different set of problems develop.

    For example, a common complaint of London Philharmonic Chorus and Orchestra / Vladimir Jurowski recording of Holst's 'The Planets' is that the dynamic range is too excessive, even on CD (which has a theoretical dynamic range of 96db undithered and about 120db with noise shaping).

    That recording has such a huge dynamic range (for a CD) that you find yourself having to choose between not hearing the very quiet moments that well or having your head ripped off when the loud bits kick in.
    A 24 bit recording would be, in theory, even worse.

    I get a lot of tracks recorded by other people to mix.
    My most common complaint is that they are recording too hot.
    Quite often I get clipped drums, vocals or guitars that I then have to spend hours fixing in Izotope RX.
    I prefer recordings to be averaging around -18dbFS with peaks no louder than -6dbFS.
    0reaction image LOL 0reaction image Wow! 0reaction image Wisdom
  • octatonicoctatonic Frets: 33725
    Pablo said:
    Does it make a difference if you are applying effects post recording? Giving the computer more data/information to work with even if some of it is inaudible?
    Well I certainly think that things like reverb tails sound more realistic in 24 bit when you solo them and listen closely, but does it matter in the context of a finished track with all other elements sounding at once?
    The reality is though that if you are doing label work then you simply cannot deliver 16 bit masters.

    Something else that has not been mentioned (I don't think) is all audio interfaces are not equal.
    You will get prosumer audio interfaces that capture at 24 bit/192khz that will sound obviously worse than 20 year old Lavry/Prism converters that are only 16 bit, 48khz capable. 
    That said, the entry level interfaces sound much better than the entry level interfaces of 20 years ago.
    0reaction image LOL 0reaction image Wow! 0reaction image Wisdom
  • fastonebazfastonebaz Frets: 4065
    ICBM said:
    Admittedly some years ago so the processing power available is different, but I did some comparisons in a recording studio when deciding what rate and depth to use for a project. I’m pretty sure they were using Cubase 5.

    Surprisingly, there was a noticable difference between 44.1KHz and 48, but none that we could hear between 48 and 96. Likewise there was a difference between 16 and 24-bit, but none between 24 and 32. So given that using higher quality slowed the system down drastically when mixing, we used 48KHz/24-bit.

    But that’s for recording, when you need more ‘data headroom’ for processing. On playback media, there’s no evidence that higher than 44.1/16 actually improves the quality, and it could actually make it worse... this is an old article but worth reading:

    https://xiph.org/~xiphmont/demo/neil-young.html

    (Don’t worry, it’s not by Neil Young ;).)
    Excellent and interesting article @ICBM. Cheers
    0reaction image LOL 0reaction image Wow! 0reaction image Wisdom
  • thegummythegummy Frets: 4389
    Something I'd recommend to people who are doing mixing - next time you're sitting tweaking something really subtle like a half db eq adjustment or an emulation of a clean console channel, render a "with" and "without" version and use software to blind abx test it. You might be surprised that what seems obviously different when you know it's changing is suddenly unnoticeable.

    Some people might dislike discovering they can't hear certain things so those people shouldn't do it but for me it allowed me to forget about certain minute details and focus on the important things.

    Same can be applied to the sample rate idea - if you really want to know for yourself if you can hear a difference it has to be blind abx, it's just not possible to judge it reliably when you're aware of what you're hearing.

    There's a common story that I've heard many big time engineers confess as well as experienced for myself - they sit tweaking a compressor or something for ages until it's just right and sounds perfect only to find that it's been in bypass the whole time.
    0reaction image LOL 0reaction image Wow! 1reaction image Wisdom
  • CirrusCirrus Frets: 8481
    edited June 2018
    The higher the sampling rate, the more accurately it will reproduce the analogue input.;
    Technically maybe, but... All sample rates will *perfectly* reproduce any signal of less than half the sample rate ie below the nyquist frequency. So if all you care about is the part of the signal that isn't ultrasonic - you only need to record the stuff below ~20kHz - both 44.1 and 96k will reproduce the signal in the audible range *perfectly*, all else being equal - unless you can hear over 22kHz (you can't). For example, a 15khz harmonic will be recorded exactly the same at 44.1 as it is at 96k sample rate - The latter won't be *higher quality*, it'll just be a bigger file.
    0reaction image LOL 0reaction image Wow! 2reaction image Wisdom
  • StuckfastStuckfast Frets: 2393
    Again... up to a point, m'lud. The fly in the ointment with recording at 44.1 or 48 kHz is that all digital systems need filters on the input and output stages to eliminate frequencies above the Nyquist limit. Since there is no such thing as a filter with an infinite slope, the designers face a choice between using a gentle slope and potentially rolling off material within the audio range, or using a very steep slope and introducing ripples and phase shift into the response.

    So in fact the main technical argument for using higher sample rates is that it largely eliminates this compromise. You can have a nice gentle filter slope that doesn't introduce nasty artifacts, yet still stays out of the audio range.
    0reaction image LOL 0reaction image Wow! 2reaction image Wisdom
  • mr-macmr-mac Frets: 200
    ICBM said:
    Admittedly some years ago so the processing power available is different, but I did some comparisons in a recording studio when deciding what rate and depth to use for a project. I’m pretty sure they were using Cubase 5.

    Surprisingly, there was a noticable difference between 44.1KHz and 48, but none that we could hear between 48 and 96. Likewise there was a difference between 16 and 24-bit, but none between 24 and 32. So given that using higher quality slowed the system down drastically when mixing, we used 48KHz/24-bit.

    But that’s for recording, when you need more ‘data headroom’ for processing. On playback media, there’s no evidence that higher than 44.1/16 actually improves the quality, and it could actually make it worse... this is an old article but worth reading:

    https://xiph.org/~xiphmont/demo/neil-young.html

    (Don’t worry, it’s not by Neil Young ;).)
    Again depends on hardware but 48khz takes the processing sound further away from audio frequencies.  Most hardware can't resolve the advantage 192 might theoretically offer.  48/16 48/24 and 96/24 are where I'd be looking to use.

    there can be some advantage in efx applied in the daw software at the higher rates.  So there can be an advantage adding reverbs and other efx in 24/96 or 24/192 or higher but bring down to a 24/48 when creating final mixdown.

    agree it doesn't need to be high numbers but do think a lot of hardware produces better results at 48 than 44.1 but that's not because of the extra samples 
    0reaction image LOL 0reaction image Wow! 0reaction image Wisdom
  • mr-macmr-mac Frets: 200
    Most of problem with red book cd 44.1/16 is stupid levels of compression and sometimes a crap final mix down.

    high def doesn't solve the ridiculous levels of compression used.  
    There are plenty 44.1 and 48khz recording out there that prove the problem isn't inherent with that sample as they sound incredible.

    one thing to do if you capture and edit at 96 don't make final music avail at 44.1 make it 48.  If you plan on releasing a cd don't do all work at 96 or 192 etc etc but do it at 88.2 or 176.4.

    Why? We'll its very easy to accurately convert 176.4 or 88.2 to 44.1 with very little error.  But to change from 192 or 96 to 44.1 requires the software to interpolate the data (ie best guess of where wave would have been at point). 

    So if you wanna do a CD and HD use 88.2 or 176.4.... If doing video work use 48, 96 or 192.

    0reaction image LOL 0reaction image Wow! 1reaction image Wisdom
  • mr-macmr-mac Frets: 200
    edited June 2018
    And whatever you do dont record or do the work in DSD even though some editors are available that are compatible now.  Almost every operation to change things or add efx requires the DSD data to be converted to multibit have work/change applied then converted back (ok happens in background but still two codec conversions just to do a basic operation on data isn't optimal and will degrade it).
    0reaction image LOL 0reaction image Wow! 0reaction image Wisdom
  • CirrusCirrus Frets: 8481
    edited June 2018
    Stuckfast said:
    Again... up to a point, m'lud. The fly in the ointment with recording at 44.1 or 48 kHz is that all digital systems need filters on the input and output stages to eliminate frequencies above the Nyquist limit. Since there is no such thing as a filter with an infinite slope, the designers face a choice between using a gentle slope and potentially rolling off material within the audio range, or using a very steep slope and introducing ripples and phase shift into the response.

    So in fact the main technical argument for using higher sample rates is that it largely eliminates this compromise. You can have a nice gentle filter slope that doesn't introduce nasty artifacts, yet still stays out of the audio range.
    Absolutely - I wouldn't argue against that. So the real-world answer is that the 96kHz recording will be as much better than the 44.1kHz equivalent as the nyquist filters are audible to you. Personally, I can't really tell the difference on my RME Fireface. I think most converters are good enough in that regard these days to make that pretty much a non issue - not always the case, I remember the old Motu converters I used in 2003. You could hear what they were losing on playback.

    So, in practice, it comes down to how well the converters were engineered. Personally, I have no problem with my signal being, say, 1dB down and a few degrees out of phase by the highest frequency I could possibly hear. It pales in comparison to the things I'll do to the music by the time it's mixed down to stereo.  , and it pales in comparison to things like the quality of the clock, the op-amps on the way in and out...

    There's also the issue of aliasing and extra inter-modulation distortion if you record ultrasonic signals you don't actually need - record your project at 96k and you might have ultrasonic stuff that takes up headroom, interacts with signals in the audible range etc. At some point you down-sample the signal and you hope that stuff is going to be dealt with properly by the software.

    These are all very minor points, and my feeling is that the downside of capturing unnecessary info in the first place pretty much balances out the minor improvement in the nyquist filter situation.
    0reaction image LOL 0reaction image Wow! 1reaction image Wisdom
  • CirrusCirrus Frets: 8481
    mr-mac said:
    Most of problem with red book cd 44.1/16 is stupid levels of compression and sometimes a crap final mix down.

    high def doesn't solve the ridiculous levels of compression used.  
    There are plenty 44.1 and 48khz recording out there that prove the problem isn't inherent with that sample as they sound incredible.

    one thing to do if you capture and edit at 96 don't make final music avail at 44.1 make it 48.  If you plan on releasing a cd don't do all work at 96 or 192 etc etc but do it at 88.2 or 176.4.

    Why? We'll its very easy to accurately convert 176.4 or 88.2 to 44.1 with very little error.  But to change from 192 or 96 to 44.1 requires the software to interpolate the data (ie best guess of where wave would have been at point). 

    So if you wanna do a CD and HD use 88.2 or 176.4.... If doing video work use 48, 96 or 192.

     agree with the first half - 44.1/16 is enough to reproduce fantastic audio quality.

    Re' the sample rate conversion, that seems intuitively right but apparently it's not at all. Whether you're downsampling to 1/2 rate or some very complex ratio, the maths is sound. I've seen enough clever people say it that I'm inclined to believe them - including someone who designs A/D and D/A converters.
    0reaction image LOL 0reaction image Wow! 0reaction image Wisdom
  • mr-macmr-mac Frets: 200
    Cirrus said:
    mr-mac said:
    Most of problem with red book cd 44.1/16 is stupid levels of compression and sometimes a crap final mix down.

    high def doesn't solve the ridiculous levels of compression used.  
    There are plenty 44.1 and 48khz recording out there that prove the problem isn't inherent with that sample as they sound incredible.

    one thing to do if you capture and edit at 96 don't make final music avail at 44.1 make it 48.  If you plan on releasing a cd don't do all work at 96 or 192 etc etc but do it at 88.2 or 176.4.

    Why? We'll its very easy to accurately convert 176.4 or 88.2 to 44.1 with very little error.  But to change from 192 or 96 to 44.1 requires the software to interpolate the data (ie best guess of where wave would have been at point). 

    So if you wanna do a CD and HD use 88.2 or 176.4.... If doing video work use 48, 96 or 192.

     agree with the first half - 44.1/16 is enough to reproduce fantastic audio quality.

    Re' the sample rate conversion, that seems intuitively right but apparently it's not at all. Whether you're downsampling to 1/2 rate or some very complex ratio, the maths is sound. I've seen enough clever people say it that I'm inclined to believe them - including someone who designs A/D and D/A converters.
    Ears tell me different. And a lot of very clever audio recording people do it that way.  Downsmspling  in a sympathetic rate (with clock dividable) takes zero guesswork or computer heavy lifting.  interpolating to a non sympethetic freq requires the software to do heavy calculations and interpolate date points.  I'm not saying difference is massive but it does bring in potential for not being close to 1:1.

    Things change though and all software is different so millage may vary.  Seems a no brainer to avoid a conversion stage that requires calculations to be made. 

    like i say on high end audio gear in past i have been able to tell which was interpolated and which wasn't..  Bit my ears now and software now may have got to stage its not really a big consideration.
    0reaction image LOL 0reaction image Wow! 0reaction image Wisdom
  • normula1normula1 Frets: 639
    I do a fair bit of digitising Vinyl for use on my phone which has a HiRes DAC. My turntable also has a built in HiRes DAC which connects via USB and records at 192k/24bit. I then convert down to 48k/24bit and can definitely hear a difference between the two. However given the huge file size of the 192k recordings I tend to use the 48k ones.
    1reaction image LOL 0reaction image Wow! 0reaction image Wisdom
  • mr-macmr-mac Frets: 200
    normula1 said:
    I do a fair bit of digitising Vinyl for use on my phone which has a HiRes DAC. My turntable also has a built in HiRes DAC which connects via USB and records at 192k/24bit. I then convert down to 48k/24bit and can definitely hear a difference between the two. However given the huge file size of the 192k recordings I tend to use the 48k ones.
    The difference heard could just be how well the dac handles each rate though rather than there being an audible difference.  The chip in dac could perform better at 192 than 48.
    0reaction image LOL 0reaction image Wow! 0reaction image Wisdom
  • StuckfastStuckfast Frets: 2393
    It's pretty unlikely that there is any useful information coming from your turntable that can't be captured at 16-bit/44.1kHz. If I remember right the typical frequency response of a good vinyl master is about 30Hz - 15kHz and the dynamic range maybe 65dB at most.
    0reaction image LOL 0reaction image Wow! 2reaction image Wisdom
Sign In or Register to comment.