Audio Myths & DAW Wars
Three things you need to know about audio quality
- Research has shown that for music distribution, 16 bit @ 44.1 kHz (CD standard) is indistinguishable from 24 bit @ 196 kHz in a sample of over 550 listeners. In other words, more bits and higher bit-rates are not going to improve the quality of your tracks.
- There are many traps for young players when comparing audio from two DAWs, make sure you know what they are, discussed below.
- The world is full of marketing departments trying to convince you that equipment and specifications can substitute for talent & hard work. This is not true, the performance transcends the medium every time.
Audio quality, the eternal quest
Spend time on any forum devoted to any Digital Audio Workstation (DAWs) or music production and you will see users making unsubstantiated claims about the audio quality of this or that DAW. Protagonists will say a given DAW is clearly and audibly superior to another. To be frank, that's just nonsense. Any DAW that uses 32 bit floating point internal processing will be capable of processing audio at a quality where residual errors will be far below the limits of human auditory perception. Today, from an 'audio quality' perspective all DAWs, yes even 32 bit vs 64 bit, are created equal and by audio quality we are talking about audible effects caused by the 'audio engine'. Yes, 64 bit DAWs can be useful, but that is for handling very large audio files in memory, they do not improve audio quality.
So where do these perceptions about differences in DAW quality come from, particularly since you see many 'famous' producers and recording engineers extolling the virtues of one DAW's audio quality or 24 bit @ 96 kHz audio format? Surely these guys know what they are talking about. The answer is a little complex:
First, recording engineers and 'famous people' almost never assess the differences between DAWs or high-end audio formats under controlled conditions. That is, make statistically meaningful numbers of forced choice comparisons under double blind listening conditions, in controlled audio environments. What usually happens is that a company sends their product/s out free to as many industry famous people as they can find. This select group listens to it and a smaller sub-group think it "sounds better than anything they have ever heard". It is this sub-group and their quotes that appear in the marketing blurb for the product/s. Now don't get us wrong, these people really do think the product sounds great, but it's a subjective impression and hardly qualifies as proof that the product is any better than others in the market or even an improvement on what went before.
A second reason is that there are many settings and options that affect the live and rendered audio from any DAW. Its unlikely that 'out of the box' any two DAWs will make exactly the same sound. The following list will help you to understand what these settings and options are, and to give you a broader perspective on what really can make a difference and hopefully protect you from the marketing machine:
1. Live Mixer interpolation - This applies to Sampler Channels when transposing samples from the root note. Plugin instruments may have their own live vs rendered interpolation settings. When sample-rates are converted (when pitch shifting a sample for example) then the DAW may need to call on sample data between the existing points. Interpolation is all about making an accurate prediction as to what that level should be, and so reduce a problem known as quantization errors and results in aliasing and or quantization noise.
2. Rendered audio settings - Including WAV bit-depth setting, MP3/OGG bit-rate setting and Sampler interpolation. The wav bit-depth (16, 24 or 32) won't really have much impact on the sound you hear however, the lossy formats (mp3 & ogg) definitely will introduce audible 'garbling' or 'underwater' sounds when used at bit-rates less than about 190 kbps. These formats are really for music distribution, although can sound spectacularly good at bit-rates of 240 kbps and above. Sampler interpolation is the same feature as discussed in the mixer section, but here applies to rendered files. If you are hearing differences between the live and rendered sound, then make sure the live and rendered interpolation settings match.
3. Your mixing decisions - See the section in the manual on Levels and Mixing. This is where the magic happens. If you can mix well, your music will probably sound great no matter what the technical specifications of the DAW are. Mixing is a craft and takes years to learn, just like any musical instrument. So if your mixes sound bad compared to the commercial mixes, this is, with 99.99% probability, the reason why. The DAW doesn't suck, you do. Also be aware that you don't need any more sophisticated tools to mix than a nice Parametric EQ, Compressor/Limiter and the basic Mixer functions. All those 'mastering' plugins are useful tools that can save time, but no substitute for experience and methodical work flow. If you want to get some idea how the actual sound itself can influence you emotionally, load Harmless in a default project and start working your way through the presets. Some sound thin and reedy others will blow your socks off. It's all about the sounds mixed together (the performance), not the technical specs of the DAW. Now imagine how hard it is to separate the performance from the technical aspects of the DAW when it comes to our emotional reaction to a platform. All too often, great mixing, performance or patch programming is mistaken for product design or specifications.
4. Loudness - In a comparison, louder always sound 'better' than quieter. The louder of two otherwise identical sounds will seem to have more bass and clearer high frequencies. This comes from the way our ears work, not anything in the audio itself. Apart from basic mix decisions there are a number of reasons why sound rendered from FL Studio may be quieter/louder (depending on settings) when compared to the same audio/sample rendered from another DAW.
5. Plugins behaving badly - We have a manual page 'Plugins behaving badly' dedicated to that. Some plugins just sound bad or make strange noises when used with the wrong settings. FL studio has a lot of Wrapper settings to give you the widest compatibility with poorly programmed plugins.
6. Plugins behaving differently - This trips many people up when they render the same synth from two DAWs and compare the waveforms under a microscope. Synthesizers usually have some randomization and/or free running oscillators (meaning the phase of the waveform will change as a function of the note start time), as the point of most synths is to not produce the 'exact' same waveform twice. Make sure to disable any randomization settings and to send the same notes of the same velocity with any of the same modulation settings. A better strategy here is to use a .wav file as a test source, then you know it's identical to start with in each DAW.
7. Marketing has influenced you - Yes it has. Digital audio is just a stream of numbers. Computers add up numbers in a well understood and predictable way, if they didn't we'd constantly have satellites raining down on us from the sky. We are talking here basic mathematics (addition, subtraction and division), there is no magic, there are no secret things that some DAW manufacturers know that others don't. Dithering and interpolation are well understood and there are plenty of options in most DAWs to give you control over them. But understand, that the suppliers of professional and consumer audio equipment have strongly vested interests in convincing you that you need to upgrade to the latest and greatest gear/format, that's how they make money, selling gear based on specifications. Audio quality ceased to be a meaningful differentiators of DAWs once they had all moved to 32 bit float internal processing. The effect of a lifetimes marketing has been so powerful, let's consider three aspects associated with bit-depth and sample rate:
- DAW Bit depth - Anyone who has used a calculator will know that when you perform mathematical operations you only have so many decimal places and so get rounding errors and these errors accumulate in the lest significant decimal place. The same thing happens when processing digital audio. If you have 16 bit numbers representing your data then these rounding errors can creep into the audible realm, particularly with very quiet passages of music. A 32 bit floating point format allows mathematical operations to be performed on audio without rounding errors becoming audible. Before you ask, no, 64 bit is not subjectively better. Yes there are a few exceptional circumstances that can be concocted to create audible artifacts in a 32 bit floating point format, but the same is true of 64 bit float and these cases are not worth considering as a driving factor in perception of 'quality'. So that's internal processing, what about the sound quality of the render/distribution format?
- The weakest link - Human hearing. Surely 24 bit 196 kHz wav files sound superior to the 16 bit 44.1 kHz wav files used on CDs? Brace yourself, this may come as an unbelievable shock, the largest and best study conducted to date (see the reference below) shows that there is no audible difference between 'high-end' audio formats ~24 bit @ 196 kHz and 16 bit 44.1 kHz (CD standard). A user-friendly article discussing the research can be read at the following link: The Emperor's New Sampling Rate.
What they found - From a sample of 554 listeners that included professionals, the general population & young listeners (prized for their high-frequency hearing), those that correctly identified the higher quality audio was 276, or 49.8%. The same number you would get if you just flipped a coin 554 times or asked untrained monkeys to do the task. In summary, 16 bit 41.1 kHz sound is indistinguishable from ~24 bit @ 196 kHz. Yes, 32 bit float is important for audio processing in a DAW, but once it comes time to convert that audio into a format useful for distribution and human consumption, you just can't meaningfully improve on the CD standard.
- Electronics - The best analog to digital electronic circuits that are currently possible to implement in commercially available 'professional' audio equipment are equivalent to, at best, 20 bit. A dynamic range or 120dB. Yes, all those 24 bit recordings are actually delivering somewhere between 18 to 20 bits of real-world precision once they have been mangled by the best converters and room-temperature electronics you can buy. What this means is that even a lowly 24 bit file has outstripped the ability of our electronics to reproduce it, the noise inherent in the electronic resistors and capacitors swamps the remaining resolution. Sample rates, on the other hand can go almost as high as you want, but as we have seen in the study above, more than 44.1 kHz is a waste. Time to stop worrying about the technical specifications of DAWs as a driver of 'quality' and concentrate on the other things in this list that matter more, like mixing and performance.
8. You have influenced you - You simply can't make an unbiased comparison of the audio from two sources, A and B, if you know what source you are listening to at any given time. You can't so forget it, perceptual psychologists realized this over 100 years ago and developed many useful methodologies to work around it. In particular the 'blind listening' experiment with an 'objective response indicator'. Get a friend to play you the two sources in random-order pairs. Your task is to simply identify source A and B, nothing more, nothing less. If you can distinguish source A vs B, at least 8 times, or more, from 10 random-order paired comparisons, then you may be able to hear something. If not, it is likely that you are just guessing. This is probably one of the most enlightening tests any audio-engineer can do, you will learn a lot about perception and your ability to hear things this way. Invariably we are much less sensitive than we think. As the sense of infallibility we have in our own perception is so strong, we have an exercise for you:
An experiment - Render the same ~5 seconds of a project to a 320 kbps mp3 file and 16 bit 44.1 kHz wav (CD format), then make 30 A vs B blind comparisons. This means you are not to know whether your helper is playing the mp3 or wav file to you. You should also avoid eye contact with them and receive no feedback on how well you are doing until the experiment is completed. The helper should write down a list of 30 comparisons randomizing the order to play them to you (wav vs mp3 or mp3 vs wav), they should ensure that you receive 15 mp3 vs wav and 15 wav vs mp3 trials, 30 in total in a mixed up sequence. You don't need to do them all at once, if you need a break do so, but don't confer with your helper about how you are doing. Your task is simply to identify the wav file (the better sounding one!). In order to convince a scientist, with a high degree of confidence you can tell them apart, you need to identify the wav file at least 20 times out of 30. Our untrained monkey fresh from his 24 bit 96 kHz experiment will correct identify the wav file 15 times (chance). Surely, as mp3 is inferior to CD format, you can beat 20 correct identifications or at least the untrained monkey?
9. Your soundcard or Windows mixer & Media player - Make sure you don't have any EAX, Compressors or EQ settings etc., on in the soundcard or Windows settings. Dig deep here, sometimes they are well hidden in 'advanced' options tabs etc.
Some time in the late 1990's we moved beyond the point where technological improvements to 'fidelity' ceased to make any meaningful difference to the audible quality of the music we produce. Further, with the Loudness Wars of the 2000's and widespread adoption of low bit-rate mp3's as a music distribution standard, it's clear that audio quality has been going backwards for a while now, but people are still enjoying their music. In conclusion, we'd like to leave you with a quote from photographer Vernon Trent
Bron: http://forum.image-line.com/viewtopic.php?p=316587
Als ik het dus goed begrijp is het verhaal dat de audio engine van Fruity Loops dus eigenlijk niets slechter is dan die van Cubase, Pro Tools, noem maar op.
Ik ben een Fruity gebruiker en ik heb heel vaak te horen gekregen dat ik beter cubase kan gebruiken of wat ook.. Maar ze konden mij eerst niet overtuigen, ik geloofde er eerlijk gezegd niets van. Maar op een gegeven moment zeiden ook mensen het waarvan ik weet dat ze wel verstand hebben van muziek etc. Dus was ik eindelijk overtuigd en nu ben ik een beetje bezig met Cubase, maar vind het niet zo fijn werken als Fruity. Dus ik blijf liever bij Fruity
Maar ik wil mezelf niet belemmeren door minder goede equipment zeg maar, dus ik was even aan het googelen, en toen vond ik dit stukje tekst! Het is wel door de makers van Fruity Loops geschreven, maar ik geloof wel dat ze gelijk hebben..
Hoe denken jullie erover?
Three things you need to know about audio quality
- Research has shown that for music distribution, 16 bit @ 44.1 kHz (CD standard) is indistinguishable from 24 bit @ 196 kHz in a sample of over 550 listeners. In other words, more bits and higher bit-rates are not going to improve the quality of your tracks.
- There are many traps for young players when comparing audio from two DAWs, make sure you know what they are, discussed below.
- The world is full of marketing departments trying to convince you that equipment and specifications can substitute for talent & hard work. This is not true, the performance transcends the medium every time.
Audio quality, the eternal quest
Spend time on any forum devoted to any Digital Audio Workstation (DAWs) or music production and you will see users making unsubstantiated claims about the audio quality of this or that DAW. Protagonists will say a given DAW is clearly and audibly superior to another. To be frank, that's just nonsense. Any DAW that uses 32 bit floating point internal processing will be capable of processing audio at a quality where residual errors will be far below the limits of human auditory perception. Today, from an 'audio quality' perspective all DAWs, yes even 32 bit vs 64 bit, are created equal and by audio quality we are talking about audible effects caused by the 'audio engine'. Yes, 64 bit DAWs can be useful, but that is for handling very large audio files in memory, they do not improve audio quality.
So where do these perceptions about differences in DAW quality come from, particularly since you see many 'famous' producers and recording engineers extolling the virtues of one DAW's audio quality or 24 bit @ 96 kHz audio format? Surely these guys know what they are talking about. The answer is a little complex:
First, recording engineers and 'famous people' almost never assess the differences between DAWs or high-end audio formats under controlled conditions. That is, make statistically meaningful numbers of forced choice comparisons under double blind listening conditions, in controlled audio environments. What usually happens is that a company sends their product/s out free to as many industry famous people as they can find. This select group listens to it and a smaller sub-group think it "sounds better than anything they have ever heard". It is this sub-group and their quotes that appear in the marketing blurb for the product/s. Now don't get us wrong, these people really do think the product sounds great, but it's a subjective impression and hardly qualifies as proof that the product is any better than others in the market or even an improvement on what went before.
A second reason is that there are many settings and options that affect the live and rendered audio from any DAW. Its unlikely that 'out of the box' any two DAWs will make exactly the same sound. The following list will help you to understand what these settings and options are, and to give you a broader perspective on what really can make a difference and hopefully protect you from the marketing machine:
1. Live Mixer interpolation - This applies to Sampler Channels when transposing samples from the root note. Plugin instruments may have their own live vs rendered interpolation settings. When sample-rates are converted (when pitch shifting a sample for example) then the DAW may need to call on sample data between the existing points. Interpolation is all about making an accurate prediction as to what that level should be, and so reduce a problem known as quantization errors and results in aliasing and or quantization noise.
2. Rendered audio settings - Including WAV bit-depth setting, MP3/OGG bit-rate setting and Sampler interpolation. The wav bit-depth (16, 24 or 32) won't really have much impact on the sound you hear however, the lossy formats (mp3 & ogg) definitely will introduce audible 'garbling' or 'underwater' sounds when used at bit-rates less than about 190 kbps. These formats are really for music distribution, although can sound spectacularly good at bit-rates of 240 kbps and above. Sampler interpolation is the same feature as discussed in the mixer section, but here applies to rendered files. If you are hearing differences between the live and rendered sound, then make sure the live and rendered interpolation settings match.
3. Your mixing decisions - See the section in the manual on Levels and Mixing. This is where the magic happens. If you can mix well, your music will probably sound great no matter what the technical specifications of the DAW are. Mixing is a craft and takes years to learn, just like any musical instrument. So if your mixes sound bad compared to the commercial mixes, this is, with 99.99% probability, the reason why. The DAW doesn't suck, you do. Also be aware that you don't need any more sophisticated tools to mix than a nice Parametric EQ, Compressor/Limiter and the basic Mixer functions. All those 'mastering' plugins are useful tools that can save time, but no substitute for experience and methodical work flow. If you want to get some idea how the actual sound itself can influence you emotionally, load Harmless in a default project and start working your way through the presets. Some sound thin and reedy others will blow your socks off. It's all about the sounds mixed together (the performance), not the technical specs of the DAW. Now imagine how hard it is to separate the performance from the technical aspects of the DAW when it comes to our emotional reaction to a platform. All too often, great mixing, performance or patch programming is mistaken for product design or specifications.
4. Loudness - In a comparison, louder always sound 'better' than quieter. The louder of two otherwise identical sounds will seem to have more bass and clearer high frequencies. This comes from the way our ears work, not anything in the audio itself. Apart from basic mix decisions there are a number of reasons why sound rendered from FL Studio may be quieter/louder (depending on settings) when compared to the same audio/sample rendered from another DAW.
5. Plugins behaving badly - We have a manual page 'Plugins behaving badly' dedicated to that. Some plugins just sound bad or make strange noises when used with the wrong settings. FL studio has a lot of Wrapper settings to give you the widest compatibility with poorly programmed plugins.
6. Plugins behaving differently - This trips many people up when they render the same synth from two DAWs and compare the waveforms under a microscope. Synthesizers usually have some randomization and/or free running oscillators (meaning the phase of the waveform will change as a function of the note start time), as the point of most synths is to not produce the 'exact' same waveform twice. Make sure to disable any randomization settings and to send the same notes of the same velocity with any of the same modulation settings. A better strategy here is to use a .wav file as a test source, then you know it's identical to start with in each DAW.
7. Marketing has influenced you - Yes it has. Digital audio is just a stream of numbers. Computers add up numbers in a well understood and predictable way, if they didn't we'd constantly have satellites raining down on us from the sky. We are talking here basic mathematics (addition, subtraction and division), there is no magic, there are no secret things that some DAW manufacturers know that others don't. Dithering and interpolation are well understood and there are plenty of options in most DAWs to give you control over them. But understand, that the suppliers of professional and consumer audio equipment have strongly vested interests in convincing you that you need to upgrade to the latest and greatest gear/format, that's how they make money, selling gear based on specifications. Audio quality ceased to be a meaningful differentiators of DAWs once they had all moved to 32 bit float internal processing. The effect of a lifetimes marketing has been so powerful, let's consider three aspects associated with bit-depth and sample rate:
- DAW Bit depth - Anyone who has used a calculator will know that when you perform mathematical operations you only have so many decimal places and so get rounding errors and these errors accumulate in the lest significant decimal place. The same thing happens when processing digital audio. If you have 16 bit numbers representing your data then these rounding errors can creep into the audible realm, particularly with very quiet passages of music. A 32 bit floating point format allows mathematical operations to be performed on audio without rounding errors becoming audible. Before you ask, no, 64 bit is not subjectively better. Yes there are a few exceptional circumstances that can be concocted to create audible artifacts in a 32 bit floating point format, but the same is true of 64 bit float and these cases are not worth considering as a driving factor in perception of 'quality'. So that's internal processing, what about the sound quality of the render/distribution format?
- The weakest link - Human hearing. Surely 24 bit 196 kHz wav files sound superior to the 16 bit 44.1 kHz wav files used on CDs? Brace yourself, this may come as an unbelievable shock, the largest and best study conducted to date (see the reference below) shows that there is no audible difference between 'high-end' audio formats ~24 bit @ 196 kHz and 16 bit 44.1 kHz (CD standard). A user-friendly article discussing the research can be read at the following link: The Emperor's New Sampling Rate.
What they found - From a sample of 554 listeners that included professionals, the general population & young listeners (prized for their high-frequency hearing), those that correctly identified the higher quality audio was 276, or 49.8%. The same number you would get if you just flipped a coin 554 times or asked untrained monkeys to do the task. In summary, 16 bit 41.1 kHz sound is indistinguishable from ~24 bit @ 196 kHz. Yes, 32 bit float is important for audio processing in a DAW, but once it comes time to convert that audio into a format useful for distribution and human consumption, you just can't meaningfully improve on the CD standard.
- Electronics - The best analog to digital electronic circuits that are currently possible to implement in commercially available 'professional' audio equipment are equivalent to, at best, 20 bit. A dynamic range or 120dB. Yes, all those 24 bit recordings are actually delivering somewhere between 18 to 20 bits of real-world precision once they have been mangled by the best converters and room-temperature electronics you can buy. What this means is that even a lowly 24 bit file has outstripped the ability of our electronics to reproduce it, the noise inherent in the electronic resistors and capacitors swamps the remaining resolution. Sample rates, on the other hand can go almost as high as you want, but as we have seen in the study above, more than 44.1 kHz is a waste. Time to stop worrying about the technical specifications of DAWs as a driver of 'quality' and concentrate on the other things in this list that matter more, like mixing and performance.
8. You have influenced you - You simply can't make an unbiased comparison of the audio from two sources, A and B, if you know what source you are listening to at any given time. You can't so forget it, perceptual psychologists realized this over 100 years ago and developed many useful methodologies to work around it. In particular the 'blind listening' experiment with an 'objective response indicator'. Get a friend to play you the two sources in random-order pairs. Your task is to simply identify source A and B, nothing more, nothing less. If you can distinguish source A vs B, at least 8 times, or more, from 10 random-order paired comparisons, then you may be able to hear something. If not, it is likely that you are just guessing. This is probably one of the most enlightening tests any audio-engineer can do, you will learn a lot about perception and your ability to hear things this way. Invariably we are much less sensitive than we think. As the sense of infallibility we have in our own perception is so strong, we have an exercise for you:
An experiment - Render the same ~5 seconds of a project to a 320 kbps mp3 file and 16 bit 44.1 kHz wav (CD format), then make 30 A vs B blind comparisons. This means you are not to know whether your helper is playing the mp3 or wav file to you. You should also avoid eye contact with them and receive no feedback on how well you are doing until the experiment is completed. The helper should write down a list of 30 comparisons randomizing the order to play them to you (wav vs mp3 or mp3 vs wav), they should ensure that you receive 15 mp3 vs wav and 15 wav vs mp3 trials, 30 in total in a mixed up sequence. You don't need to do them all at once, if you need a break do so, but don't confer with your helper about how you are doing. Your task is simply to identify the wav file (the better sounding one!). In order to convince a scientist, with a high degree of confidence you can tell them apart, you need to identify the wav file at least 20 times out of 30. Our untrained monkey fresh from his 24 bit 96 kHz experiment will correct identify the wav file 15 times (chance). Surely, as mp3 is inferior to CD format, you can beat 20 correct identifications or at least the untrained monkey?
9. Your soundcard or Windows mixer & Media player - Make sure you don't have any EAX, Compressors or EQ settings etc., on in the soundcard or Windows settings. Dig deep here, sometimes they are well hidden in 'advanced' options tabs etc.
Some time in the late 1990's we moved beyond the point where technological improvements to 'fidelity' ceased to make any meaningful difference to the audible quality of the music we produce. Further, with the Loudness Wars of the 2000's and widespread adoption of low bit-rate mp3's as a music distribution standard, it's clear that audio quality has been going backwards for a while now, but people are still enjoying their music. In conclusion, we'd like to leave you with a quote from photographer Vernon Trent
Bron: http://forum.image-line.com/viewtopic.php?p=316587
Als ik het dus goed begrijp is het verhaal dat de audio engine van Fruity Loops dus eigenlijk niets slechter is dan die van Cubase, Pro Tools, noem maar op.
Ik ben een Fruity gebruiker en ik heb heel vaak te horen gekregen dat ik beter cubase kan gebruiken of wat ook.. Maar ze konden mij eerst niet overtuigen, ik geloofde er eerlijk gezegd niets van. Maar op een gegeven moment zeiden ook mensen het waarvan ik weet dat ze wel verstand hebben van muziek etc. Dus was ik eindelijk overtuigd en nu ben ik een beetje bezig met Cubase, maar vind het niet zo fijn werken als Fruity. Dus ik blijf liever bij Fruity
Maar ik wil mezelf niet belemmeren door minder goede equipment zeg maar, dus ik was even aan het googelen, en toen vond ik dit stukje tekst! Het is wel door de makers van Fruity Loops geschreven, maar ik geloof wel dat ze gelijk hebben..Hoe denken jullie erover?
