Preferred Sample Rate for recording.

[eyeteeth]

I completely forgot I used to belong and read through these forums until a recent search.

Anyway... on to business. With the the various Sample Rates... what do you prefer to track with? AND.... what are you primarily tracking?

[stubbsonic]

It mostly depends on what the final format of your project will be. That said, you can always do post-mix processes to get the sample rate and bit depth required of any given project. But I prefer to avoid having to do that step, as I can often hear a difference after that step is finished.

Others will have answers perhaps more informed by current format standards.

For bit depth, I almost always track at 24 bit. It sounds noticeably better than 16 bit, especially when the music has lots of dynamics. It makes the stereo field sound more evenly dispersed. It is more forgiving of slightly lower levels. It records with good amplitude resolution, and gives DP's various functions lots of resolution to work with.

For sample rate, if I know the resulting project will end up on a CD, I will record at 44.1, so I can avoid having to do any sample-rate conversion after the fact. Otherwise, I tend to record at 48K. I don't use the higher rates both because I don't want to have to do SRC later, but also to save disc space which is limited here.

[FMiguelez]

What Stubbsonic said.
24/48 or 24/44.1 KHz is perfectly fine for humans.

Also, remember that, unless you have equipment especially designed to handle ultrasonic frequencies (you don't), throughout your chain (converters, amps, mixer, speakers), you would actually do a disservice to yourself by using 96 and especially 192 KHz sampling rates.
All that distortion that your non-specialised equipment will inevitably give you (because it's not designed to cope with all that high energy -that you can't hear anyway), will reflect back into the audible range and make things worse!

Do some blind tests and see for your self.
Never forget how ridiculously powerful the placebo effect is. Learn to recognize and enjoy how easily our ears (brain) and the other senses are fooled and confused, and how full of cognitive biases we are! It's wonderful and hilarious!

Expert wine tasters, anyone?

Expert: "Oh, this wine is crap compared to this one. It's bland and doesn't have any depth. With the other I taste hints of cherry with wooden barrels from 1725".

Tester: "Yeah, you idiot, just because both glasses of wine come from different bottles doesn't mean it isn't the SAME wine. Sorry for tricking you just to show you how full of •••• you are!"

Ahem!

[MIDI Life Crisis]

I put everything in 48/24.If it ever needs to go to picture (which it almost always does) I'm ready. If not, dumbing down to 44 is not big deal.

[dix]

^ Me also. Since almost everything I do is to picture I default to 48/24 since that’s standard for video work.

However I do my ITB mixing to 32bit. I noticed recently that I can always hear the dif between my tracks running live in my sequence and my mixed track if I record it at 24bit, but not if I print it at 32. Beyond that I can hear the dif between a 24/48 exported file or even a highish res mp3 created from a 24bit track and one created from a 32bit. I’m by no means a golden-eared mastering guy, so I was pretty surprised when I discovered this. Has anyone else noticed this and work this way?

[FMiguelez]

However I do my ITB mixing to 32bit. I noticed recently that I can always hear the dif between my tracks running live in my sequence and my mixed track if I record it at 24bit, but not if I print it at 32.

What do you mean when you say you do your "ITB mixing to 32 bit", dix? Do you mean you record/print your individual tracks at 32 bit FP?

I hope you don't, because DP always processes internally at 32 bFP anyway, even if your audio files are at 8 bits. That would be a colossal waste of space and HD bandwidth.

Also, don't we ALWAYS hear, in the best of cases, a truncated 24 bit (or 20 bit) mix as we monitor it, even if we're working on a 64 or 32 bFP DAW? Unless we have 64 or 32 bit DA converters, of course... Do you guys have one of those?

Beyond that I can hear the dif between a 24/48 exported file or even a highish res mp3 created from a 24bit track and one created from a 32bit. I’m by no means a golden-eared mastering guy, so I was pretty surprised when I discovered this. Has anyone else noticed this and work this way?

I used to do that... until I made a blind test and discovered that my guesses were not better than flipping a coin

Also, when one says: "I can hear a difference between 32 bit floating point and 24 bit integer", or "I can hear a big difference between 24 bits and 16 bits", what one is really saying is that one can hear the difference in noise floors (difference between -192 and -144 dBFS, and -144 and -96 dBFS respectively) that happen to be buried in loud music AND room AND electronics AND atmospheric noise (30-40 dB noise floor in your very quiet room).
And the dither noise is, to top it all off, usually shaped, so we would hear even much less of it! If no dither noise is used, the quantization distortion, while more offensive than the noise, is even lower in level than dither.

So that is the only possible difference between 32, 24, 16 and 8 bit signals... just the relative loudness of their noise floors (or QD).

Can anyone really hear that in even very loud listening conditions in a super quiet room?
ABX results, please!

Any other difference between those signals is just gremlins/

[mikehalloran]

Since most of my work for others ends up on CDs or radio, I track 44.1k/24bit. If there's a chance it will wind up in any AV format, 48k/24bit. Yes, you can convert but why if you already know.

Of course, I can handle higher sampling rates but that costs more (drive space and time) so I charge enough to discourage most from going there. Nowadays, I get few requests for that as people become better educated.

If ya want analog, I can still do it but my price would include getting my gear out of mothballs ... I would try to talk you out of that.

Speaking of which, there's a well known acoustic musician in the greater San Fran Bay Area whose label is known for doing everything analog up to the CD master. My wife put one of the Christmas CDs in the changer and cranked it up while we were decorating the home... I thought there was something wrong with my system as I heard a nasty 60Hz hum with multiple harmonics (we have 1960 wiring) but it disappeared during the 2 seconds between cuts. Turned out, it was this "audiophile" CD. One of this week's projects is to run that CD through RX 7 and burn a new copy to make it listenable. Yikes!

[mhschmieder]

As with everything, context matters.

If you're soloing cymbals or delicate acoustic stringed instruments, you will get more detailed tails from the higher sampling rates (88.2 kHz and above).

More importantly, if you use soft synths, anti-aliasing artifacts are often present at rates below 96 kHz.

As a result, I record everything at 96 kHz/24 bits, unless it's going to have live drums, which generally get tracked first and in my case involves 20 mics and an ADAT based expansion unit whose channel count goes down above 48 kHz.

For most material, anything above 48 kHz is unnecessary unless you are doing high-precision video edits for dramatic cues that need to maximize their impact. But there are probably other ways to do that without recording everything at 386 kHz.

The main thing is to avoid 16-bit recording as the headroom is insufficient at the input stage and further downstream, forcing you to record to as close to 0 dBFS as you can and then trimming things during the mix.

With 24-bit recording, you can record very low without distortion, such as at -20 to -24 dBFS. As things get louder as you sum during mixing, this means your overall signal path from start to finish minimizes the need for gain and trim along the way, increasing the integrity of the signal overall.

As for 48 kHz vs. 44.1 kHz, I prefer to work at 48 kHz because I can definitely hear the difference, and have taken dozens of double-blind tests that prove it.

But even that is a qualified statement, because filters are getting better and better all the time, and we may eventually encounter a 44.1 kHz slope that is so perfect that there are no anomalies between 14 kHz to 22 kHz.

As much of what I do has the potential to accompany video, and as my own converters clearly avoid upper-frequency issues at 48 kHz but not at 44.1 kHz, I choose to go with 48 kHz when no soft synths will be used.

Saving the conversion for the post-mastering stage, I have yet to notice degradation from 48 kHz to 44.1 kHz compared to what I hear going back down to 16 bits (or especially to MP3).

I use top-end conversion from iZotope, and have a few others that I use that have different attributes, depending on the source material. But I also keep my masters at -0.3 dBFS or lower -- this avoid clipping, overs, errors, etc. as a little bit of headroom is needed even when performing format conversion.

[dix]

FM, clearly you, and everyone else in this thread, have a much better handle on the technical aspects of bit depth than I do, but numbers be damned in DP I’ve always, since I started mixing ITB, been able to pick out the printed track from the sequence tracks 100% of the time. If I print in 32bit I hear less of a difference. I haven’t tried all types of program, but enough that I always switch to 32bit when I go to print a stereo mix....only at that stage btw to save space.

Is it possible there’s more going on in DP than just the math you describe? Have you tried comparing 24bit and 32bit mixes? It’s easy enough to set up a blind test.

[MIDI Life Crisis]

For me, at any rate (pun intended) it's not matter of what sounds better, but what is going to be the most compatible with the majority of the stuff I produce. If I was sending out to the internet (except for film stuff) I'd be at 44.1.

[FMiguelez]

FM, clearly you, and everyone else in this thread, have a much better handle on the technical aspects of bit depth than I do, but numbers be damned in DP I’ve always, since I started mixing ITB, been able to pick out the printed track from the sequence tracks 100% of the time.

Not that I don't believe what you tell me, but how did you compare them, if you don't mind my asking? The comparison must be instantaneous and blind (you can't know what's playing when).

If I print in 32bit I hear less of a difference. I haven’t tried all types of program, but enough that I always switch to 32bit when I go to print a stereo mix....only at that stage btw to save space.

Is it possible there’s more going on in DP than just the math you describe? Have you tried comparing 24bit and 32bit mixes? It’s easy enough to set up a blind test.

A truly blind test for live DP playing VS a printed file, for the reasons I mentioned, is not that easy to implement.

However, a null test would reveal exactly what the differences are, if any.

To test this, I'd suggest something like>
Setup DP in such a way that you can compare the difference between running your sequence live, and a previously bounced (or recorded) mix, so they null. This should be quite easy to set up.

All things being equal, if you put a limiter quantizing to 24 bits in your master bus, you should get perfect destructive interference.
If you do not have a quantizing plugin in your master bus. i.e., the wordlength is still 32 bFP, and if you bounce to 32 bFP, you should also get total silence in your difference meter (make sure it shows down to at least -144 dBFS).

IF, however, you bounce to 24 bits, and the limiter wasn't doing any limiting, and then delete the limiter (quantizer) from the sequence (so you're back to "hearing" 32bFP -which you aren't --> see bellow), then the difference between the live 32 bFP and bounced 24 bit file should show only a) the dither noise you used to bounce down to 24 bits, or b) if you didn't use dither, then the difference would represent the quantization distortion, and possibly c) any minute difference as to how DP might convert from 32 bFP to 24 integer. I'm not really sure about the details of this last one, though.
Either way, the difference, if any, should be down to around -140 dBFS. and THAT's what you would have to be hearing on top of the music and your room noise! For you to hear that, even in an extremely quiet passage, would mean you're listening to the music at almost instant hearing-damage levels!

Also, even if the above weren't true, remember that, if you play DP sequence live and hear a difference against a bounced version, you would still have to explain how you can hear a difference when your DAC doesn't do much better than 20 bits... IOW, you're never "hearing 32 bFP"
What would be that which makes the files different and obviously recognisable to you otherwise?? You'd have to explain that too

That's why I asked you how you checked for these perceived differences you're hearing. Unless you tested them level-matched, and instantaneously changing from one to the other without you knowing what you're doing and which is which (i.e. blindly), you are most likely victim of one of our human typical cognitive biases

It's so much fun (and quite humbling) when one discovers his biases with a test like this that no one can argue with!

Either way, like I read the other day someone saying, no song has ever stopped making it to the top-100 charts due to dither (or lack thereof) or bit depth or sample rate...

[dix]

My test is simple. Print a mix. With eyes closed rapidly toggle the Solo state an unknowable number of times (mix/sequence). With eyes still closed hit Play.

Primitive I know, but if I do that I can always pick the print against the sequence tracks. Am I missing something? Is this not scientific enough somehow?

[FMiguelez]

Yeah, that doesn't sound very scientific, I'm sorry.
There are tools, like Melda's MCompare, that let you run blind tests easier. Or Foobar's blind test plugin.

The less unknown variables, the better. You need to do it many times, say, 20 times, and see how many you get right or wrong.

What is it you're comparing exactly? Is the mix 24 or 32 bits?

I really recommend you do the null test I described! Aren't you curious to see and hear in isolation that which you claim is different and make the music distinguishable? See if you can put a face on it.

You do realize, that you've never heard a true 32 bit file, yes? Your converters prohibit it, so what we're really hearing is anybody's guess.. This is an important point I hope you address.

[eyeteeth]

Thank you everyone for responding. I found it interesting, and not entirely what I was expecting. Several of you went fairly into depth and now I feel like I should supply an explanation.

I am preparing to do a bit of semi serious recording again, and my 'studio' had fallen behind. It still works beautifully in its own eco system... But its a G5, with MOTU 1224s, UAD-1 cards... and lacking the ability to run some software and features I would like on this project. As I've been thinking about the equipment I would like... one item I thought was a requirement today, was being able to record at 88.2 or 96k.

Why?

Well.. way back when, there were several articles regarding sample rates, and while we are not even close to being capable of hearing anything near 44.1 it doesn't mean those frequencies and higher aren't present as overtones in the material we are recording. It had been shown, when those higher frequencies (say of a cymbal) are cut off, they manifest themselves at a corresponding frequency in the range was CAN hear... creating a type of harmonic digital distortion. By charts and scientific studies... doubling the sample rate gave enough 'headroom'.

Since I have thought higher sample rates were crucial, I have been disappointed, and assuming Presonus missed the mark making the new 3 series mixers 44.1 & 48k only. Actually, currently, 48k only.

As I have been catching up a little, I have been finding most still record at 44.1 for CD, or 48 for picture which really surprised me. BUT! It makes me REALLY start to contemplate the current StudioLive series as the centerpiece for a studio. They are also claiming they are adding HUI and Mackie Control to the console... knowing DP can work with either of these AND 44.1 & 48 are still the norm... The price point to get a bunch of Mic Pre's, D/A A/D conversion, AND a control surface just seems obvious. If it all works.

I have a SL 16III that I have for live situations, that was upgraded from a 16.4.2 classic... I really liked the boards and the features they have. Enough that it has me contemplating a 24 or 32 for the studio. Hence my questioning everyone's preferred sample rate, and even thought I had a specific reason in mind, I didn't want to taint ot color anyone's response.

Again... thanks for the responses.

[FMiguelez]

Guys, I hope I don't come out as stubborn, but right now I do have time to write, and I happen to LOVE this kind of topics (and I want to learn more, so here we go...)

As I've been thinking about the equipment I would like... one item I thought was a requirement today, was being able to record at 88.2 or 96k.

That's what the marketing departments want you to think, that it's a requirement and anything less than 192 KHz will make listeners protest in disgust.
It is NOT a requirement. How can it be, when there is no clear evidence that people actually can tell them apart? The only reliable evidence I've seen shows people can NOT tell them apart. My own tests show me and my friends the same thing (I really enjoy embarrassing them in the studio).

Well.. way back when, there were several articles regarding sample rates, and while we are not even close to being capable of hearing anything near 44.1 it doesn't mean those frequencies and higher aren't present as overtones in the material we are recording.

They are there, but that doesn't mean you can hear them. If you could, you could hear the difference between a sine wave and a square wave at 14 KHz, yet they sound the same.

It had been shown, when those higher frequencies (say of a cymbal) are cut off, they manifest themselves at a corresponding frequency in the range was CAN hear... creating a type of harmonic digital distortion.

You mean aliasing? That's what our great and modern anti-aliasing filters are for. There should be very little signal, if any, above Nyquist that would alias after such filter. And the level would be so low it would be easily masked by the music.

By charts and scientific studies... doubling the sample rate gave enough 'headroom'.

Not really... Headroom/dynamic range have to do with the bit depth, not the sampling rate. All the sampling rate deals with is high frequency content.

Since I have thought higher sample rates were crucial, I have been disappointed, and assuming Presonus missed the mark making the new 3 series mixers 44.1 & 48k only. Actually, currently, 48k only.

Crucial??? No, they are not crucial, otherwise amazing music couldn't have been recorded digitally until now. The difference between 44.1 and 96 KHz productions would be so obvious that anyone could easily tell them apart reliably. Can we? It isn't self-evident, is it?

I don't think it's even a luxury, but more like fear-based stubbornness.

Don't forget working with those SRs also means doubling or quadrupling your computer's processing load, as well as a 4-fold increase in disk bandwidth usage. Is that HF noise worth it?

Like Mark mentioned above, there are certain specific situations where using higher SRs might be justified, but for regular mixing? It's an exaggeration.
And you will have to eventually down-sample to 48 or 44.1 anyway, so your listeners won't enjoy those high freqs when they hear your music in their iPhones and laptops...

Does using high SRs have advantages? Sure! I just think the disadvantages outweigh the good things about them.



Remember that ALL your (and your audience's) equipment would have to be specially designed to handle all that extra energy from those super high frequencies. EVERY component, otherwise, it will distort and reflect back into the audible range and make things sound worse than necessary (what you mentioned above, only in your detriment).

Finally, the only thing that matters is... can YOU hear a difference reliably and constantly between regular and high SRs?
Do some blind testing. IF you decide to get 96 or 192K stuff, do it because you can hear a worthwhile difference, not because marketing departments tell you you should hear it.

These determinations should not be done Pascal's Wager style!