Hey friends,

I produce electronic music. I was wondering how you guys feel about the difference between 44khz - 48khz - 88khz and 96khz and why you prefer one above the other. Also how drastic of a difference do you hear between 16bit (actually 14bit) and 24bit.

Best,

John

 

Over most genre's I don't hear the difference, but if I'm listening, to say Classical that is sampled at 24bit/96kHz or higher, they overall clarity of the mix is much better. Though for most music 16bit/44kHz is suitable, or else the file size, especially when making a project increases dramatically & the tax on your system performance is much higher.
So if your just producing electronic music I'd stick to 16bit/44kHz. That's what I use.

 

Well, Saint, there is difference. At least, in theory. For the real world, whether we hear the differences or not, depends on source material, DA converter, monitors, acoustic treatment of a room and listening experience. Theory-wise, higher bit depths (word lengths) gives us more/better dynamics (more discreet levels), more of "audio resolution", more headroom while higher sample rates extends frequencies that can be present in the recording. Regarding sample rate, for 44.1 kHz sample rate audio, the highest possible frequency that can be sampled can't be higher than half of it, i.e. 22.05 kHz. For 96 kHz sample rate, it's 48 kHz of "frequency response". Why's that while human can hear only up to 20 kHz?

The newest first. Man can't hear but can respond to frequencies up to 50 or so kHz. Assuming you have monitors that play so high.

In order to AD converter proper/better does his job, before conversion a circuit applies a filter which cuts off frequencies above the highest one for the particular sampling rate. No filter can't cut off like a knife, imagine graphic representation like vertical line. It starts to cutting off earlier than 22.05 kHz in case of 44.1 kHz sampling rate and get down to zero at 22 kHz. Some starts at 20, back then in time some starts to cut off at 18 kHz.

When we apply certain processing to an audio signal, many kinds of disturbances occurs and some are shifted towards the end of frequency spectrum in order to be less perceived. So, higher we push them, less audible they are. In case of 96 kHz sampling rate, you have 48 kHz range and if you put them higher, the better. Putting them at say 45 kHz (96 kHz sampling rate case) is better than at 18 kHz (44.1 kHz sampling rate case).

When you down sampling from 96 to 44.1 kHz sampling rate, you get aliasing. Hence, anti-aliasing filter which is not perfect.

Combine all of that and you have a picture.
------------------------------------

Why bother with that if we end up with good old CD-Audio at 16 bit 44.1 kHz? Well, we shift disturbances higher, let's say at 45 kHz and later cut them off when downsampling because we apply filter that will cut off audio at 22 kHz in order to get 44.1 kHz sampling rate. Also, if we record at 24 bit and later change bit depth to 16 bit, due to the better resolution, clearer sound remains. All in all, if you have it, use it.

Also, theoreticians advice that if we have to record at higher sample rate and downsample later, we should do that to record at 88.2 kHz sampling rate because when we convert it to 44.1 kHz it's perfect half so there's no imprecise math, rounding of figures by computer. The same with 176.4 and exactly four times lower 44.1 Khz sampling rates.

Rounding figures occurs with EVERY single thing we do with audio, gain adjusting and everything else because computer have to recalculate value for each sample and rounding numbers is inviability. Thus, micro dynamics is changed. That's why purists insist on "doing nothing" or "as minimal changes as you can" with recorded audio.

Summing (summing-mixing several channels-tracks into one stereo) is another problem of software and such calculation in software often spoil audio a little bit. It's better if you have a multiple output interface with good converters and sum up audio in decent outboard analog mixer or doing all digital with decent digital mixing board.

---------------------------------------------------------
Assuming all other is same, people hear "higher" listening to 96 kHz sampling rate audio than 44.1 kHz sampling rate.

People tend to hear "more dynamic" in 24 bit audio than in 16 bit audio. (Note that problem occurs when lowering bit depth from 24 to 16 bit. Hence, we use dithering).

DVD-Audio can go as high as 24 bit 192 kHz audio in case of stereo album.

Blue Ray Disc (BD) could use 24 bit 96 kHz of lossless compressed audio. Can't remember whether it use 192 kHz or not. I think it does.

Don't get fooled with audio quality of DVD-Video discs. They use lossy compression, just like mp3s. CD-Audio sounds much better.
--------------------------------------------------------

I'm writing from my head, and hurry for get some nutritive elements, so consult Internet for detailed explanation why higher is better. By my opinion, and I do like this, if you record audio, you're fine with 24 bit 44.1 kHz or 24 bit 96 kHz. Only with some critical stuff I record at 24 bit 192 kHz.

 

For some Film Score music, you have to deliver your work in a spécific format, and be compliant whith the needs of post-prod (Dollby 5/1, 24/48)
I also got some tracks mastered from 24/96 production chain,in a good mastering studio using digital tools (Pyramix ) in the process, and heared the Master before the drastic réduction to 16/44,1 .. I've got suicidal tendancies for some seconds ! I heared the difference !

I agree whith Felino's post and i use 24/44,1 for all my "home" work .

 

@digital.musician imo there's a big difference between 16 and 24bit ... bit depth plays a huge difference and for a while I used to also produce at 96khz but it also depends on whether the samples you use were recorded 96/24 cause if they weren't I don't see any gain from converting from 44/16 up to 96/24... unless you pull some missing bytes out of your ass lol I don't see how you can improve on for example a picture that's 300x300 to make it crystal clear at 2000x2000

@felino thanks a lot for the clear explanation! sounds like 88.2/24 is the way to go!

@Myr LOL

 

Is anyone using 48k 24bit instead of 44k 24bit and if so can you tell the difference? I am being given vocals at 48k 24bit but i am running ableton 8 at 44k 24bit. Wondering if its worth switching up to 48k even though all my sample library is 44k? Cheers

 

Supposedly from what I've read 48khz is used more in film application... that's all I can say about it

 

@SAiNT and @John C. You're wellcome!

@John C. For the following advice I admit that I obey it only in restoration work. You gain something if you up convert to 96 or even higher because of what I wrote in my previous post here. You get rid of the artifacts later, when you down convert. Artifacts that occur when you do some heavy processing.

@lekb I think you won't hear much of a difference between the two. 48 kHz sampling rate is in regular use for audio for video. If you work with 48 kHz only, than, stay wit it. I guess you can switch Ableton to 48. If you want to use 48 kHz audio alongside with 44.1 kHz audio, than you have to decide what is more important - not to downsample vocal in order not to ruin it more than you have to or downsample it because you don't want to affect the rest by upsampling it to 48 kHz. Most of all, what is the requirement for delivery? If you have to deliver your finished work to 48 kHz and I guess that vocal is the most important element, than I'll stick to 48 kHz.

The worst you can do is, if delivery is in 48 kHz, to downsample vocal to 44.1 kHz, mix it with the rest at 44.1 and than upsample the whole mix to 48 kHz in order to meet the requirements. This way, you'll affect the vocal changing the sample rate twice, not to mention processing. But hey, some people do it like this. You can figure out the other combinations, I hope.

 

As I was surfing some Reason Forums I found another person's take on the 44.1/48KHz versus 96KHz debate. I am not the author of what follows but despite the fact that some of the content is explained poorly, and some of it is downright garbage, I tend to agree with him about the sampling of unwanted frequencies above and beyond 20 KHz far more than I do with some of the things posited above. For one: I do not believe humans have any ability to respond to anything higher than about 24 KHz as suggested by the content of Felino's post.

I would definitely be very interested in the source of that statement

Here is the article from Reasonfreaks.com from May 24th, 2010

 

This question becomes annoying.
1. You have to figure out what is it this what sells your music
2. If fact from 1 can be better after changing sample rate – change it.
3. Simple people call as sound engineers, not sound artists. Maybe because all the way of production process we are measuring something. The most precise instrument which we are using is our ears. So use it and if you are satisfied on the end of the road this is it. Put your sound together with others and you will know immediately
4. Today sound standard is not connected with music anymore, depends of games and video. Standard setup on Windows 7 is 48 KHz 24 bit. I can’t remember when I so last time that someone use CD for listening music.
5. Finally, who cares about technical data if everybody Jumps when I say Yea :dancing: .

 

48 khz i used in video in order to synchronize the sound with the framerate/timecode better but i am no pro.

Thanks for the knowledge people! looks like a lot of experienced people gather here!

 

Why is this question annoying to you? If you don't like the subject feel free to skip the discussion.

Changing the sample rate for the sake of setting it higher because "higher is better" is an uneducated approach that's typical of the masses that have been given "too much tool while possessing too little skill and knowledge."
Sound engineers are almost always "sound artists" being musicians themselves, that was always a requirement before one could even become a sound engineer, so thank you for the derogatory insult.

Surprising as it may be to you, the Redbook standard (44.1 KHz sample frequency) is the defacto Audio CD standard and will remain that way for a while yet. 48 KHz is the standard for DAT and DVD Audio. Your views on the issue are completely subjective and wrong in fact, specifically your point 4.
You may sell some of your stuff online but you would never be able to put it to the CD medium if you wanted to. And there are still many people that buy CDs.

This is a discussion between people who cherish good quality audio and if that bothers you why not just keep quiet?

 

Further on the issue of sampling at higher than the Nyquist frequency I notice here and in my travels in audioland that there is an awful lot of confusion or misunderstanding about the sample rate and the bit depth of samples, so I would like to offer a segment from the book Sound Synthesis and Sampling 3rd Ed. by Martin Russ. See Chapter 1, pp. 61-66: (This is quite a read so ADD sufferers and those possessing attention spans shorter than 17 seconds, please don't bother. )

 

So Caithleann what is it exactly that felino said that you don't agree with? I read your article but couldn't really contrast it against what felino said, to me it sounded along the same lines.

 

I guess I must not be speaking English, John...

I wrote: "I tend to agree with him about the sampling of unwanted frequencies above and beyond 20 KHz far more than I do with some of the things posited above. For one: I do not believe humans have any ability to respond to anything higher than about 24 KHz as suggested by the content of Felino's post."

With what part of this do you have a problem understanding? Perhaps I should have just said "more" instead of "far more" ?

Also, this article states quite clearly that you don't gain more fidelity by using a higher sampling rate than 44.1 KHz whereas Felino stated that besides more dynamic resolution aka headroom you also get more "audio resolution" which is pertinently not true. He wrote literally:

While the latter part of that statement is actually true, a less knowledgeable reader could well misread that statement and infer that those "extending frequencies that can be present in the recording" are a good thing, thinking that the sound is more crisp or you can hear more as many people believe but the reverse is true, you don't hear more and those extra frequencies included from beyond the audio spectrum muddy up the sample manifesting themselves as aliasing which is bad.

 

Thank you ladies & gents for Clarifying things up for Us noobs! :dancing:

 

@Caithleann just perplexed that you would believe something "explained poorly" and "downright garbage" from a secondary source over the explanation of a primary source with obvious experience in domain. Plus the first article you posted is indeed complete garbage with an unfounded example of running 30,000 cycles where half are sampled erroneously @ 44khz? Prolly cause the guy who wrote the post (article is a bit of an overstatement) build the Analog Digital Converter himself out of hay, sand and some paperclips. That "article" is wrong, you do gain "fidelity" when sampling (which means /to record/) at higher sampling rates! WHY? It's quite f*cking easy to understand that 88khz samples store twice the amount of information that 44khz samples do (twice the number of bits)... when you go from analog to digital the sampler takes your raw (analog) wave and passes it thru the ADC which then from a perfectly round (lossless) signal fragments it into rectangular little slices to resemble the original signal as close as possible. The higher the sampling rate the thinner the slices are ... when you have twice the amount of information your digital wave is closer to the original one

(THIS DOES NOT APPLY IF YOU CONVERT FROM 44khz TO 48khz or higher in my opinion ... "up-sampling" IMO is useless because there's no information to fill the gaps... unless the converter averages the value of 2 bits to get the 1 in between but that's just an assumption... i'm sure there's a lot of rounding involved and in the end it probably depends on the algorithm. still once you downsample the additional information which gives you the extra "clarity" and "audio resolution" it is lost when trying to take that sample back to a higher sampling rate - this aside is just my opinion).

And when I say lossless, don't think of FLAC or WAV cause those formats are obviously already digital when you have'm on your computer... think of the sound in its natural state.

So felino is very much right that you'll get better "audio resolution" when sampling at higher rates. It's simple math, a little bit of physics, and understanding of how DAC/ADCs work. Aliasing is prevented by your converter which applies a bandpass for frequencies above the Nyquist frequency which is half of your sample rate. That's so completely beside the point of why I originally started this thread.

This a graph of signals sampled at different rates, showing how the character of some signals changes dramatically when the rate is too low. Here's your aliasing



To give you a better understanding of how ADCs work in context to my reply:

The red sine wave is in its natural form while what you see in steps is the digital signal.

 

With natural sources (recording with microphones) added samplerate adds frequencies. We can hear up to 20KHZ (with young ears) and up to 200KHZ phase. So a 40khz tone may have lower octaves wich distort lower frequencies. We will notice. As this is what helps us localize sources.

In our 44.1khz cd decoders there is a highpass filter wich filters out noise. This damages the lower frequencies. So 44.1khz samplerate equals 22 khz of sound. (Every two samples is one wave form). When we up the samplerate the highpass filtring takes place higherup so we do notice less 'harshness'.

Bitlength from 24 bit give a more accurate number. Meaning more accurate samples. When we process in the digital domain with compression, reverb larger bitlengths (read 64bit processing) helps us achieve more accurate processing. If we add 24 + 24 +24 + 22,212222222 +77,123123123123123 calculating this will result in a 64bit environment in a much cleaner way. Way less round offs in the sound.

So when recording with DPA microphones (wich will provide source material up to 50KHZ) we will have more data. So a samplerate of 96khz will sample more of that source then with a shure SM58 wich will sample up to 16khz. (the membrane is too thick to vibrate faster and thus capture content above 16khz.) For 16khz we need 32khz samplerate. With the right (small membrane condenser) microphone we can capture (and sample) a high hat with much more accuracy. With an audio analyser we will see that a high hat will provide frequencies above 20KHZ. We need a samplerate of 40khz for that and something higher to avoid phase artifacts caused by the low pass filter). So a samplerate of 48khz will give a better reproduction of the source. 96khz may be even safer. 96khz allso makes sure we can process the file more accurate. "we have more meat to work with".

Downsampling;

If 44,1 is one of the spin off's of the recording then 88,2 is handy. Because 88,2 divided by 2 is 44,1. Sampling at 96khz divided by 2 gives odd numbers meaning we could damage the audio content in ways we won't notice right away, but will be there.

Going from 24bit to 16bit gives odd numbers. So if you have 24bit audio, rerecord it analog to 16bit with a master recorder. This will sound truer to the source then any downsampling.

An impulse response from a studer 24 track recorder will give back content up to 50KHZ.

I hope this has helped.