Ahhh sampling rates! the inevitable question that bugs a lot of people, here i'm going to give my perspective as a mastering engineer, the short of it is that higher sampling rates does not equal better sound quality, why? you have to understand the sampling theorem first, but without complicating things, higher sampling rates means that your digital processors and any non-linearities will work better, specially when it comes to reducing Aliasing, and this is the prime example of why modern audio plugins have oversampling capabilities, and this is is the main reason why engineers feel that the overall sound improves, because the overall bandwidth increases, this means that you are moving the anti-Aliasing filter higher, this is what produces a better result because the processor is starting to behave more like an analog circuit giving you a more accurate results at the expense of latency, lastly downsampling reduces the quality ? no, if you record your music at higher sampling rates and keep everything high until the mastering stage you'll get a better 44.1Khz file, to better understand this let me explain how a basic oversampling algorithm works, first the audio goes in to your processor then the internal sampling rate in the processor is upsample to your desire sampling rate, for example if you chose the ruff factor of X4 that means that the internal sampling rate of the processor is multiplied four times over moving the anti-Aliasing filter from 44.1Khz to 176.3KHz, it's a simple multiplication operation, and then it down samples back to 44.1KHz and then the signal is output back to the DAW, hope this clarifies a bit.
In the past (10-20 years ago) sample rate conversion with the software available then was not always top quality. Some aliasing could occur. However that has improved over the years. I would not worry about this today. I've seen some testing on the subject on the net supporting this but I can't find it now....
It's up to the engineers making a device how much CPU or DSP horsepower they want to allocate to the sample rate conversion. You can still choose "linear" as an option when developing a new Android device, for example.
Hi Paul I love your channel and seeking out then best sound quality. However here is a fun conundrum ultimate question in my opinion... Given tbe choice / ultimatum of listening experience where it was one or the other😁..would you choose to listen to the best artists in poor sound or the most mediocre or poor artist talent in the finest audio fidelity. ...I'd love to hear your feedback and of course subscribers on this too. All the best! Paul. Dublin IRELAND.
There are DACs that don't support 88.2/176.4 kHz, but do support 96/192 kHz, so offering the latter helps those. Converting 192000 to 44100 Hz could be compared to converting an image with a width of 19200 pixels to a width of 4410 pixels: The conversion process might yield some moiré patterns, unless you're using a good algorithm to reduce these. If you're using a FullHD monitor (equivalent to regular audio equipment) to display the image, it doesn't matter, because this last step yields more errors then the previous steps. But if you're using a high-resolution screen (equivalent to a highly resolving audio system), you might notice these patterns. But to complete this comparison: You might only notice it in certain situations, like color gradients or diagonal lines, which would be the equivalent to artificial test tracks instead of real music. Most of us (including me) will certainly hear no difference at all, but as part of this game is to "feel/know that it is being done the right way", I would recommend to match multiples, because a factor of 4.0 is obviously easier to get right than a factor of 4.35374149659864 ... in other words: It is easier to combine 4 samples into 1 as to combine 4901877145437275 samples into 1125899906842624 ... or 492 samples into 113.00625 and living with the small error :)
The audio quality of digital processing math is a hugely important topic nowadays. Quantization errors and other errors can be audible and when people say "PCM sounds harsh" or something of that nature, it's likely a math issue and not about PCM itself. Sample rate conversion is also done through math and the math can do a more or less precise conversion causing errors that can be audible. For example, if you make an Android device, you can find math options in the Android OS reference code called: "linear", "cubic", "sine with original coefficients" and "sine with revised coefficients". Each of these sound different and generally a compromise can be made between audio fidelity and CPU/DSP load when developing a new Android device. With faster processors nowadays, it is surely a sad place to reduce the processor load.
@@Wizardofgosz Yes, any respectable DAW wiłl focus on audio performance and use better algorithms. It's different on battery powered devices where saving a few percent of such background process can make a few minutes longer battery life and better gaming FPS.
It's a non-issue if you sample rate convert using hardware, which is what most professional big mastering studios do. If you use software, many times you'll find that you lose some high end and stereo image when compared to the hardware conversion.
Hi Paul, this is a odd question. You seem so humble and caring, with strong characteristics of a true believer in Christ. Does God have an influence on your success and helped you to strive where you are now? I am starting my own speaker company here in South Africa inspired by n spiritual happening that I'm holding on to. Thank you for being a inspiring figure. May you prosper even more in your life. Love from SA
A good question, I thought. Does converting, say, 192 kHz sampling to 176.4 kHz require upsampling to the LCM of the two, then downsampling to the new frequency?
Wondering if Paul has ever tested this with people in a blind sampling test...I would like this kind of test Paul! You could throw some blind samples up on one of your channels so we can try it or see it at least? I recall some insane/ unbelievable bit rates back in the day of Napster...they seemed to trend with favorable results for better quality audio at the expense of larger file sizes thus less compression...? It's extremely interesting to me that audio recordings take up more volume in data than video images.
The best way to put it for those who still may be confused is to have a look at how the same is done with modern day big production movies. The raw footage is shot @ astronomical resolution, while the digital theater copy is scaled down, then even further for TV/Streaming.
Would recording in 48Khz cause the final audio of a production in 44Khz to be of better quality? Something about Headroom? Something about margin? Or not at all? I have no idea.
It's about editing (time & pitch) as well as plugin aliasing and/or internal oversampling artifacts. Yes, you can hear a difference. Ideally, I strive to have my overall mix (volumes and ducking) at 48k with all eq, compression and saturation being "baked in" to multiple track/stem bounces from session for each instrument all done at 96k, bounced to 96k and then converted to 48k. It's a ton of work, but the audible difference I actually hear after a number of tests is qualitatively obvious. It's worth the insane amount of more work for me.
This is conflating sample rates used in the recording process with sample rates in the distribution CD or file. 16/44.1 is perfectly fine for distribution.
No its not. Not anymore anyway. Your Dvd, blu-ray, tablets and smartphones internal clocks all work in multiples of 48khz because they're media players.
Yes, there is a significant audible difference and this is caused by the filtering. At 44.1 kHz the Nyquist frequency is roughly 22 kHz. and all frequencies starting at 22kHz and above must be filtered out. As the upper range of human hearing is supposedly 20 kHz a "brick wall" filter is used to stop the upper frequencies from being recorded and the result is sometimes harsh-sounding. At 88.2 kHz the Nyquist frequency is roughly 44 kHz and a much gentler filter slope can be used to block the Nyquist frequencies but frequencies above 20 kHz, (typically upper harmonics of signals being recorded) can be recorded easily. So the recording and audible playback of an 88.2 kHz recording does sound significantly better, but that quality is lost when converting the file down to 44.1 kHz.
I notice a difference when getting up to 24/48. Any more does not go with my main DAC. I own a DAC that does even 32 bit and 768, ore DSD up to 512, but it is nowhere near as good as the main DAC. People seem to think that higer sample rates equal better quality, yet that is hardly the case. Perhaps my main DAC would sound better at higher sample rates, but there really is no need to me (and it isn't available, the manufacturer hasn't found advantages in higher sample rates)
The main reason for Octave to offer 192khz is as people want it, yet 176 is as good (I think that it is ridiculous that there should be an audible difference (yet degradadion due to digital processing is a question mark). Why not just offer 176 and leave it at that? Consumers have infinitive wisdom for sure, yet it ain't always expert-wisdom.
@@TheDanEdwards I understand fully, yet if the demand is ridiculous and what you are selling is quality based on knowledge, is meeting such a demand a valid reason?
I wonder what Paul thinks of the whole 440 Hz "A" note versus the 432 Hz "A" note debate. I'm not sure I've ever heard him mention anything about that topic.
I went down that rabbit hole years ago and found that everyobody had a different tuning for A. But if you're going to make electric keyboards, wind instruments you have to have a standard for mass production, and so the international standard was agreed to be A=440hz, but sometimes this is too painful for singers so you can tune down to a lower pitch, for example Mozart's tuning fork was measured to be A=421.6hz or something like that. Nothing nefarious about A=440hz. Buy a tuner and tell it A=428 if you want. The real problem is 12-tone equal tempermant. Equal temperament has ruined music and so that's another rabbit hole for you to dive into. Infact there was a book written about this very problem.
We have an artist, Meghan Andrews, on Blue Coast Records that records in 432 at times. We leave those decisions to the artist. My piano stays at 440, though, so in our studio the 432 tracks are generally guitar or harp based.
Aachen ! Where they crown the Emperor of the Holy Roman Empire !!! Visited the NEW part of the church in town: New part was built in the 12th century !
Ya Paul that wasn't very helpful. I've been sort of wondering this since the mid 90's when a studio I worked in had a Pioneer DAT that could record in 88.2 and 96kHz.
Well Paul you would be right that a person would want to record in multiples of the lowest common sampling frequency. The reason is one of simple mathematics. A file being converted from 88.2 kHz to 44. 1 kHz will simply use every second sample. The same for converting from 96 kHz to 48 kHz. However, when converting from a file sampled at 96 kHz to 44.1 kHz the conversion process either samples must be skipped, or created because of the uneven division (44.1 goes into 96 2.18 times) This does create issues in dithering. But wait there's more! One must consider the final output when doing the initial recording. Is CD really the final destination or will this recording wind up streaming or being downloaded? Will this recording eventually wind up on or in a video? All the above formats except the CD would be better done in 48 kHz or a multiple thereof and not in 44.1 or 88.2 kHz.
Hi,i know is it ; 44.1Khz - 88,2Khz(44,1x2^1) - 176,4Khz(44,1x2^2) 48Khz - 96Khz(48x2^1) - 192Khz(48x2^2) and so the mathematical series continues,is the Digital Universe,based on number 2. 2-4-8-16-32-64-128-256-512-1024-......... 3-6-12-24-48-96-192-384-...........
How i love people, who do not afraid to say ''I do not know'', but share their opinion and experience. Thank you!!
I used to record bands. I chose 24/192. The better the source recording, the better anything derived from it will be.
Ahhh sampling rates! the inevitable question that bugs a lot of people, here i'm going to give my perspective as a mastering engineer, the short of it is that higher sampling rates does not equal better sound quality, why? you have to understand the sampling theorem first, but without complicating things, higher sampling rates means that your digital processors and any non-linearities will work better, specially when it comes to reducing Aliasing, and this is the prime example of why modern audio plugins have oversampling capabilities, and this is is the main reason why engineers feel that the overall sound improves, because the overall bandwidth increases, this means that you are moving the anti-Aliasing filter higher, this is what produces a better result because the processor is starting to behave more like an analog circuit giving you a more accurate results at the expense of latency, lastly downsampling reduces the quality ? no, if you record your music at higher sampling rates and keep everything high until the mastering stage you'll get a better 44.1Khz file, to better understand this let me explain how a basic oversampling algorithm works, first the audio goes in to your processor then the internal sampling rate in the processor is upsample to your desire sampling rate, for example if you chose the ruff factor of X4 that means that the internal sampling rate of the processor is multiplied four times over moving the anti-Aliasing filter from 44.1Khz to 176.3KHz, it's a simple multiplication operation, and then it down samples back to 44.1KHz and then the signal is output back to the DAW, hope this clarifies a bit.
It's hard make a summary this good. Well done. I can now stop biting my lip. :-)
How can I record at 192khz when all the microphones I have put my hands on only capture up to 20khz?
In the past (10-20 years ago) sample rate conversion with the software available then was not always top quality. Some aliasing could occur. However that has improved over the years. I would not worry about this today. I've seen some testing on the subject on the net supporting this but I can't find it now....
It's up to the engineers making a device how much CPU or DSP horsepower they want to allocate to the sample rate conversion. You can still choose "linear" as an option when developing a new Android device, for example.
2:23 Though at Octave Records we do offer 192Khz Sampling... (One Octave higher) - LOL🤣🤣🤣
Hi Paul I love your channel and seeking out then best sound quality. However here is a fun conundrum ultimate question in my opinion... Given tbe choice / ultimatum of listening experience where it was one or the other😁..would you choose to listen to the best artists in poor sound or the most mediocre or poor artist talent in the finest audio fidelity. ...I'd love to hear your feedback and of course subscribers on this too.
All the best!
Paul.
Dublin
IRELAND.
There are DACs that don't support 88.2/176.4 kHz, but do support 96/192 kHz, so offering the latter helps those.
Converting 192000 to 44100 Hz could be compared to converting an image with a width of 19200 pixels to a width of 4410 pixels: The conversion process might yield some moiré patterns, unless you're using a good algorithm to reduce these. If you're using a FullHD monitor (equivalent to regular audio equipment) to display the image, it doesn't matter, because this last step yields more errors then the previous steps. But if you're using a high-resolution screen (equivalent to a highly resolving audio system), you might notice these patterns. But to complete this comparison: You might only notice it in certain situations, like color gradients or diagonal lines, which would be the equivalent to artificial test tracks instead of real music.
Most of us (including me) will certainly hear no difference at all, but as part of this game is to "feel/know that it is being done the right way", I would recommend to match multiples, because a factor of 4.0 is obviously easier to get right than a factor of 4.35374149659864 ... in other words: It is easier to combine 4 samples into 1 as to combine 4901877145437275 samples into 1125899906842624 ... or 492 samples into 113.00625 and living with the small error :)
The audio quality of digital processing math is a hugely important topic nowadays. Quantization errors and other errors can be audible and when people say "PCM sounds harsh" or something of that nature, it's likely a math issue and not about PCM itself. Sample rate conversion is also done through math and the math can do a more or less precise conversion causing errors that can be audible. For example, if you make an Android device, you can find math options in the Android OS reference code called: "linear", "cubic", "sine with original coefficients" and "sine with revised coefficients". Each of these sound different and generally a compromise can be made between audio fidelity and CPU/DSP load when developing a new Android device. With faster processors nowadays, it is surely a sad place to reduce the processor load.
@@Wizardofgosz Yes, any respectable DAW wiłl focus on audio performance and use better algorithms. It's different on battery powered devices where saving a few percent of such background process can make a few minutes longer battery life and better gaming FPS.
"48 times 2, which is a simple division"? I believe that is an example of multiplication.
... inverse division... ;)
It's a non-issue if you sample rate convert using hardware, which is what most professional big mastering studios do. If you use software, many times you'll find that you lose some high end and stereo image when compared to the hardware conversion.
Hi Paul, this is a odd question. You seem so humble and caring, with strong characteristics of a true believer in Christ. Does God have an influence on your success and helped you to strive where you are now? I am starting my own speaker company here in South Africa inspired by n spiritual happening that I'm holding on to. Thank you for being a inspiring figure. May you prosper even more in your life. Love from SA
A good question, I thought. Does converting, say, 192 kHz sampling to 176.4 kHz require upsampling to the LCM of the two, then downsampling to the new frequency?
To make sure to lose as little as you can yes the upsampeling has to be done (in a good way)
Aww man that Audeze LCD in the background
Wondering if Paul has ever tested this with people in a blind sampling test...I would like this kind of test Paul!
You could throw some blind samples up on one of your channels so we can try it or see it at least?
I recall some insane/ unbelievable bit rates back in the day of Napster...they seemed to trend with favorable results for better quality audio at the expense of larger file sizes thus less compression...? It's extremely interesting to me that audio recordings take up more volume in data than video images.
These tests was done many times (including AES) - yes even average person can easily tell the difference
Im transfer my reels 7ips to 24 bits 385khz and the sound is just perfect
The best way to put it for those who still may be confused is to have a look at how the same is done with modern day big production movies. The raw footage is shot @ astronomical resolution, while the digital theater copy is scaled down, then even further for TV/Streaming.
Scaled down?
@@PetraKann Yup, They start @ 8k all the way down to 1080i for TV
@@AllboroLCD TV?
@@PetraKann Television?
Would recording in 48Khz cause the final audio of a production in 44Khz to be of better quality? Something about Headroom? Something about margin? Or not at all? I have no idea.
Thanks a lot for your videos. I have a question: Does recording to DSD and then convert to 96 KHz sound better than recording to 96 KHz straight away?
Yes. But i can't really understand why as well
Such better outro music! the previous felt to me lethargic, and always had be stopping the clip before it started.
PS Audio Reference Music downloaded and converted to 16 bits/ 44,1 kHz with «sox» sounds great.
From the '80's:
44100 Hz CD
48000 Hz DAT
32000 Hz DBS
It's about editing (time & pitch) as well as plugin aliasing and/or internal oversampling artifacts. Yes, you can hear a difference. Ideally, I strive to have my overall mix (volumes and ducking) at 48k with all eq, compression and saturation being "baked in" to multiple track/stem bounces from session for each instrument all done at 96k, bounced to 96k and then converted to 48k. It's a ton of work, but the audible difference I actually hear after a number of tests is qualitatively obvious. It's worth the insane amount of more work for me.
This is conflating sample rates used in the recording process with sample rates in the distribution CD or file. 16/44.1 is perfectly fine for distribution.
No its not. Not anymore anyway. Your Dvd, blu-ray, tablets and smartphones internal clocks all work in multiples of 48khz because they're media players.
Is there a genuine difference between 44.1 and 88.2 can you really hear the difference?
Yes, there is a significant audible difference and this is caused by the filtering.
At 44.1 kHz the Nyquist frequency is roughly 22 kHz. and all frequencies starting at 22kHz and above must be filtered out.
As the upper range of human hearing is supposedly 20 kHz a "brick wall" filter is used to stop the upper frequencies from being recorded and the result is sometimes harsh-sounding.
At 88.2 kHz the Nyquist frequency is roughly 44 kHz and a much gentler filter slope can be used to block the Nyquist frequencies but frequencies above 20 kHz, (typically upper harmonics of signals being recorded) can be recorded easily.
So the recording and audible playback of an 88.2 kHz recording does sound significantly better, but that quality is lost when converting the file down to 44.1 kHz.
Short answer - No, you will not hear a difference.
@@JonAnderhub "audible playback of an 88.2 kHz recording does sound significantly better" LOL
I notice a difference when getting up to 24/48. Any more does not go with my main DAC. I own a DAC that does even 32 bit and 768, ore DSD up to 512, but it is nowhere near as good as the main DAC. People seem to think that higer sample rates equal better quality, yet that is hardly the case. Perhaps my main DAC would sound better at higher sample rates, but there really is no need to me (and it isn't available, the manufacturer hasn't found advantages in higher sample rates)
@@JonAnderhub What if the filter for 22khz would be of real high quality, then there would be no need for a higher sample rate?
The main reason for Octave to offer 192khz is as people want it, yet 176 is as good (I think that it is ridiculous that there should be an audible difference (yet degradadion due to digital processing is a question mark). Why not just offer 176 and leave it at that? Consumers have infinitive wisdom for sure, yet it ain't always expert-wisdom.
"Why not just offer 176 and leave it at that?" - why one is in the business of selling products the idea is to meet the demand.
@@TheDanEdwards I understand fully, yet if the demand is ridiculous and what you are selling is quality based on knowledge, is meeting such a demand a valid reason?
I think 24/96 is a very good format for Hires recording.....
48khz are the old studio standard in the 80s and before, right ?
imma sticking to 48khz v0 mp3..48khz is also the sole output samplerate of Opus..
DAT.
No
@@380stroker yes shuppa
I wonder what Paul thinks of the whole 440 Hz "A" note versus the 432 Hz "A" note debate. I'm not sure I've ever heard him mention anything about that topic.
Oh boy
I went down that rabbit hole years ago and found that everyobody had a different tuning for A. But if you're going to make electric keyboards, wind instruments you have to have a standard for mass production, and so the international standard was agreed to be A=440hz, but sometimes this is too painful for singers so you can tune down to a lower pitch, for example Mozart's tuning fork was measured to be A=421.6hz or something like that. Nothing nefarious about A=440hz. Buy a tuner and tell it A=428 if you want. The real problem is 12-tone equal tempermant. Equal temperament has ruined music and so that's another rabbit hole for you to dive into. Infact there was a book written about this very problem.
We have an artist, Meghan Andrews, on Blue Coast Records that records in 432 at times. We leave those decisions to the artist. My piano stays at 440, though, so in our studio the 432 tracks are generally guitar or harp based.
Aachen ! Where they crown the Emperor of the Holy Roman Empire !!! Visited the NEW part of the church in town: New part was built in the 12th century !
Ya Paul that wasn't very helpful. I've been sort of wondering this since the mid 90's when a studio I worked in had a Pioneer DAT that could record in 88.2 and 96kHz.
Well Paul you would be right that a person would want to record in multiples of the lowest common sampling frequency.
The reason is one of simple mathematics.
A file being converted from 88.2 kHz to 44. 1 kHz will simply use every second sample.
The same for converting from 96 kHz to 48 kHz.
However, when converting from a file sampled at 96 kHz to 44.1 kHz the conversion process either samples must be skipped, or created because of the uneven division (44.1 goes into 96 2.18 times)
This does create issues in dithering.
But wait there's more!
One must consider the final output when doing the initial recording.
Is CD really the final destination or will this recording wind up streaming or being downloaded?
Will this recording eventually wind up on or in a video?
All the above formats except the CD would be better done in 48 kHz or a multiple thereof and not in 44.1 or 88.2 kHz.
11289600 Hz. 🙂
DSD256.
PCM, get out!
go analog and this issue goes away, and the sound is superior. life is analog. musical instruments are analog.
First to comment
*stands up and applauds*
Wow! Great job!
Amazing!
Congratulations on this amazing performance you get 🥇🏆🍾🥂👏🇺🇸 You were so fast to click !
You forgot to like your own reply.
You forgot to say something audiophile to make yourself sound good
Hi,i know is it ; 44.1Khz - 88,2Khz(44,1x2^1) - 176,4Khz(44,1x2^2)
48Khz - 96Khz(48x2^1) - 192Khz(48x2^2) and so the mathematical series continues,is the Digital Universe,based on number 2.
2-4-8-16-32-64-128-256-512-1024-.........
3-6-12-24-48-96-192-384-...........