44kHz/24bits "classic" vs 88kHz/24bits

[Protools5LEGuy]

Lots of engineers had told me to abandon 44kHz for 88kHz (HD2 users... )

They say that at 88kHz you need "half" the plugs because the tracks sounds better and do not need so much EQ and compression... Somehow it seems to have some logic... I told them that I prefer to have more tracks at the same time at 44k for tracking drums and they even told me to convert tracks from 44 to 88 after tracking drums! Keeps having some logic...

Moving to 88kHz in OS9 is only possible for me with AP2496, my Standalone Focusrite LS56 and Logic. Bye Digi 001 and Audiomedia III...And bye Protools5LE...

Only with Protools HD I could make 88...I can make Protools HD (9 version) with my Hackintosh for the final mixes...
An PT HD setup for OS9 is STILL $$$$ "Core"HD+96I/O

Now I get the "convert tracks" thing... 

What are your thought about recording 44 vs 88 kHz?

[twokayprod]

I'm not buying it. Going with 24 instead of 16 Bit makes sense, because of the dynamic range. But 44.1 kHz is enough IMO. I can't explain this very good in English, but 44.1 is not a coincidence, it's about twice the highest frequency humans can hear and once it gets converted into analog again it simply fits what's humanely possible. So to me 88.2, 96 and especially 192 kHz is just a huge waste of space.
Maybe, if you track with 88.2 kHz and you have only the best equipment available (and I'm talking Massenburg here), there is a chance of actually hearing a tiny difference on a perfect monitoring system, but let's face it. Your stuff gets converted into mp3 and listened to on earbuds. And even if people listen to the real thing on a decent homestereo or in their car, they won't hear a difference.

Sorry for not being able to explain this better. I think there's a youtube-video explaining this very good, but I don't remember....could be Adam from RealHomeRecording.com or the recordingrevolution-guy or someone else...…


p.s. Oops, actually RealHomeRecording is saying the opposite - that it's best to use the native sampling rate of your interface, because in downsampling nasty things can happen, if your converters aren't very good ones. He's recording at 96kHz because that's the native rate of his interfaces, but there's also people out there who claim that if you need to convert, then it's best to use 88.2 instead of 96 because it's a lot easier to calculate and therefore, less errors occur. The source I got this from was a german podcast. But still, I'm using 44.1/24Bit and I've never noticed a difference in sound. Converting tracks from 44.1 to 88.2 after recording might make sense if whatever you're throwing at it is working at 88.2 anyway. If at some point it will be converted no matter what, then it's good to control where it happens so you can use the best converters available to you.

[devils_advisor]

You waste your time by upconverting your material. Like video you should record in the highest possible quality the begin with and go down from there if/when needed. The opposite doesnt work and you blow up your file but you wont add any significant quality back to the material so take the other route. The next thing once you compress to a final codec like mp3 ( do your research ) you cutting a lot of frequencies out of the track which cant be heard by a human anyway. If you dont belief it do a measurement with the right equipment and see for yourself whats left.

[Syntho]

I'm at 16bit 48khz and it's good enough. I'd go for 44.1khz but I choose 48khz to streamline my entire studio. I could get into a really long rant about all this...

[DieHard]

There was a 3 year period from 2002 to 2005 that I used 88k for a few dozen projects and I really regret it, I will explain...

Firstly, most of us have enjoyed many of the added bonuses of 24 bit and 32 bit depth data recordings. Obvious to some, not to others, the greater resolution in "levels" (the bit depth) give our machines more numbers to work with... more numbers mean faders that don't overload very easily and effects plugins that sound closer to the analog counterparts.  Those reasons alone, IMO, are good enough to record all sources at the highest bit depth possible since hard drive space is not as much as an issues as in the past.

That being said, 32-bit, my favorite bit depth for all of my Cubase VST/32 projects... now has caused a small amount of grief for the past 4 years when moving many projects over to Logic Pro.  Logic imports them, they play, but do NOT produce audio... it seams logic's max bit depth is 24, so you must all convert 32 bit files, not a huge deal, but a inconvenient. Most DAWs (if not all) will be very happy with files that have 16 bit and 24 bit bit depths in the same project... as long at the sample rate if the same

Now, back to my original regret, it was not apparent at the time of creating these 88K sample rate projects, that other material may need to be added later.... like that really cool shaker loop that added so much to the groove of the track.  Now remember, mixing sample rates, unlike bit depths, is another story... obviously depending on the sample rate you pick for the project, the files that do not match your project sample rate will playback at half speed, double the speed, of if you mess with 96K then a fraction there of.  So without discussing all the possible variables of project sample rates mixed with different audio source file sample rates... I recommend against 88K or 96K for most of us, I will leave you with this advice...

If your DAW is a glorified tape machine and you do NOT plan on adding outside material (loops, samples, etc.) and you are going to mix with mostly outboard gear, then by all means go for 24bit/96K, but remember, if you can't get a "good sound" at 16bit/44k, don't bother recording 24bit/96k.

Now, like most of us, if you plan to add other things later (even if you don't realize it now); stick with 24bit/44K.  all good quality libraries will integrate easily, you will have plenty of headroom, and internal FX and mixing in the box will be simple.

Conclusion: 24Bit/44K for me

PS: One last note, these days I do not records my tracks as "hot" as in the old days.  If there is very little noise in the original recording, then a lower lever is better than a strong signal that has a single overpeak and produces a click. In seams that the summing of 24-bit tracks is extremely efficient and that 16 or 24 tracks recorded at a lower "safer" volume, produce amazing results.

[acelera]

I am with DH on this. Have been 24bit/44.1Khz ever since my first G4 and Digi001 combo.

If I had some kit that had a 48khz preference, I would be in the same boat as Syntho but 44.1Khz it is for me, here.

After all, I am enjoying the convenience of Logic 6.43 as a front-end to my TDM hardware too much and, as much as earlier HD stuff is plummeting in price, I really am not seduced by the idea of starting all over again.

I have tried 88 and 96 for fun on other systems but didn't quite see what the fuss was about. A good recording is a better idea than bitrate overkill.

[Syntho]

I guess I'm going to open a can of worms, but using a higher bit depth doesn't get you more 'volume resolution' in the end. What it DOES get you is a digital noise floor (quantization errors) at a lower level than it would be at 16bit. That's pretty much the only benefit of 24bit. It means you can leave a bit more headroom. I don't worry about a little ol' noise floor that's wayyyy down in level anyway myself, especially since I'm gonna dither, and plus I don't run my stuff through a million digital FX either so the noise won't be stacked.

We could talk about human hearing range and all that for sample rates, but one benefit of a higher sampling rate is that it's easier for engineers to build the filter at the very top more easily than it is to build one so steep for a lower sample rate. The filter is to cut down on the amount of aliasing in the high end. If we have better quality filters, it will equal a slightly better sound, but the higher sampling rate itself is unnecessary. 44.1k is enough, but I'm at 48k due to some of my gear just wanting to be at 48k natively.

Man, I could also get into converters and using a separate master clock and the jitter thing and how all of that is ridiculous, but I'm pretty sure there are gonna be some fights 

This guy actually knows what he's saying for a change: http://xiph.org/~xiphmont/demo/neil-young.html <-- read this stuff, it's the way that it works no matter what.

[GaryN]

I guess I'm going to open a can of worms,

The head worm is named Dynamic Range

That's pretty much the only benefit of 24bit.

And it's a HUGE benefit when recording real instruments playing songs that have soft and loud parts. If your thing is metal, EDM or hyper-compressed pop, you hear less of a benefit.  On the other hand, when recording an orchestra, string quartet, bluegrass, jazz trio etc. etc., having that extra 8 bits is pure bliss.

the higher sampling rate itself is unnecessary. 44.1k is enough,

Enough for what? Enough for who? In 10 minutes I can find 50 kids who think 128 Kbps Mp3's are "enough".
The "can of worms" you talk about gets opened when you declare that what's good enough for you is good enough for everybody else.
I have a friend who thinks her 17" TV is "good enough". She can't understand why I have a 60" HDTV.

There was a time when 78 RPM transcriptions were "good enough" to broadcast Bing Crosby across the nation - according to CBS. Bing financed Ampex because he wanted better. So, analog tape became "good enough". Then, tape with Dolby A and on and on.

People are still debating the Nyquist Theorem. It looks good on paper (the way mathematicians work) but it seems obvious that the best way to build a proper brick wall low-pass filter to avoid aliasing is to not need one in the first place. 88khz and up is the simple answer to that issue: Lots of margin = simple filter = less phase shift and associated nasties.  Think about tweeters… the good ones have response beyond 20khz so they operate in the smoother part of their range rather than right at their limit.
(I know that's analog apples and digital oranges, but still)

How far will it go? The tweeks are already in love with 192khz. Is that enough? Jeez, you think?. My money is on 96khz - at least for a long while.
It's just data… CPU clock speeds, storage and transfer rates increase almost daily it seems. I think 96 is a good balance between audio quality and what's manageable in the real world. Especially with the incredibly shitty internet in the world and particularly in the U.S.

Who knows… only time will tell.

[Syntho]

I knew that was coming and that's why I didn't want to get into this. What you say about the required sampling rate being 'enough' for someone or not personally just plain isn't true, unless you factor in the placebo effect, which musicians seem to suffer from heavily. I don't have time to argue it right now since I'm at work, but again, PLEASE see this page about bit depth and sampling rates http://xiph.org/~xiphmont/demo/neil-young.html Above 48k or so is just plain unnecessary for us musicians.

[MacTron]

IPLEASE see this page about bit depth and sampling rates http://xiph.org/~xiphmont/demo/neil-young.html Above 48k or so is just plain unnecessary for us musicians.

I have known this page for long time and It always make me feel furious about it, because he doesn't explain a shit. 


That's how the things go at high frecuency, despite what this guy is telling.
And if most of us are OK with 16b/44khz (and with our 5 megapixel camera) it doesn't mean that 24b/96khz aren't better. (as a camera is best at  12 Mp than at 5 Mp)

[dr bu]

i love this discussion. makes me feel human. honestly. please do not ever stop!

[DieHard]

Dear Syntho,

As you know, many pro products were made at 16bit/44K and you can still get pro results as long as you use very good converters and mostly outboard gear.  Back in the day, as I mentioned before, many studios would track to 24-track analog and simultaneously track to computers using Apogee Rosettas @ 16bit/44k; the computers were mostly used as a virtual tape machines, although some processing was done when needed. The tracks from the apogees sounded, IMO, virtually identical (minus the very slightest amount of low mids that even good A/Ds somehow loose).

The 24Bit/32Bit advantages I mentioned come into play for those who have decided to mix virtually in the box.  It is a KEY advantage to have these higher bit files when applying digital FX and mixing virtually. 32-bit floating point DAW engines have no trouble manipulating huge numbers; the reverbs will sound lusher, the compressors will pump less, the delays will sound more real, and controlling peaks on the overall summation of tracks will much, muck easier.

In today's world, I also mentioned, if the user is not mixing in the box and using the mac as a virtual tape machine, then by all means, go for the highest (24bit/96K) that your interface can handle. This will get the tracks closer to using a better quality interface at a lower bit depth and sample rate.  I know you use the Tascam DA-38 converters, but if you ever get a chance to A/B DA-78HR converters @ 24 bit, you will hear a difference. I have used both and you can definitely "hear" the difference on playback.

But in conclusion... don't bother wasting HD space if the track already sounds shitty at 16bit/44K. There is no magic, as I am sure you know, that higher bit depths and sample rates, can apply to a source with too much noise, low end rumble, poor EQs, or low-end Mics.

[Syntho]

I agree with what you're saying DieHard -- I even pointed out the software FX and bit depth thing in an earlier post (I don't have to worry about that myself though, and only will if I'm recording an orchestra), but the main thing I'm on about is if a higher sample rate (88/192 etc) is beneficial or not.

MacTron: there's actually a video on the page. See here: http://xiph.org/video/vid2.shtml Go to about 4:45 or so. He pushes it up to 20khz and what was put into the converter is what comes out. This is why I don't worry about sample rates at all. 44.1k will capture all you need. Plus, there really aren't any jaggies there. It's just the visual representation of the waveform that's confusing. A lollipop graph of the waveform makes more sense.

[DieHard]

From Syntho...

but the main thing I'm on about is if a higher sample rate (88/192 etc) is beneficial or not.

Yeah, like I mentioned above, I am really mad at myself for ever deviating from 44K... I am paying the price now with a bunch of projects that are very difficult to expand upon with so many cool apple loops available

[MacTron]

MacTron: there's actually a video on the page. See here: http://xiph.org/video/vid2.shtml Go to about 4:45 or so. He pushes it up to 20khz and what was put into the converter is what comes out. This is why I don't worry about sample rates at all. 44.1k will capture all you need. Plus, there really aren't any jaggies there. It's just the visual representation of the waveform that's confusing. A lollipop graph of the waveform makes more sense.

I have seen the this video too. And the lollipop model doesn't show a shit.
If the original tone is a sawtooth, a square wave or whatever at 20 khz, will the oscilloscope recreate exactly the same signal with just a few lollipops?
Any way, as so many more lollipops are available, (bigger sampling rate) the better accuracy we have into restoring the original signal.

So easy .

[Syntho]

The lollipop graph explains a lot actually, the jaggy graphs confuse everyone because that's not what's happening. In the end of it a 20khz saw/square/whatever -- once converted back to analog -- will be reproduced transparently and the same by a 44.1 or 88.2 or 192 sampling rate. I understand what you're saying but there is no benefit for us. There just isn't.

You can't reconstruct a 20khz waveform any more accurately with a higher sampling rate, or more 'points', because doubling the frequency (the sample rate) ensures that there are enough for that. If you decided to anyway, you'd add more of those points, sure, but the signal will be the same with either.

[MacTron]

 We can discuss the convenience or not about the use of sampling rates and bit depths over the CD standard (16 bits and 44khz), which I have no a definite opinion. But the full explanation in "Digital Show & Tell" is a fraud.

The real background of this, is that it is really unusual to use musical notes with a fundamental frequency over 2 khz. Over this frequency there is only "some" partials and mainly pitchless instruments like Hi Hats.



So if we are sampling at 44 khz we have at least around 22 samples (lollipops)  for this high notes -the worst case- to define accurately the timbre of the most of instruments (the over 2 khz partials).

The higher the sampling rate and bit depths is, the better accuracy you get. But OK, It's important to ponderate the drawbacks too.

[MacTron]

... using own Monty's pictures ...
If we are sampling a square wave:

At 22 khz we can't have any partial over 11 khz, so this is what we have once converted to analog signal:

If we are sampling at 44 khz we can't have any partial over 22 khz, so this is what we have once converted to analog signal:

So as we were increasing the sampling frequency more fidelity we achieve to the original.

Even though at 44 or 48 khz sampling rate we are at the top of human sensitiveness, using sampling rates over this, adds and extra fidelity to the audible frequencies also and allows more possibilities of audio processing while preserving audio fidelity.

[Syntho]

You're saying that ultrasonic frequencies somehow interact with the frequencies below that? You're either alluding to general, unintended distortion that comes along with higher sampling rates, or maybe the effect you're referring to in particular is intermodulation distortion which happens in the analog domain. Gear usually can't handle ultra high frequencies very well and that's one important reason for having a filter on the converter. Another thing you may be alluding to could be how when you hear overtones themselves without the fundamental, our brain thinks we're hearing a fundamental when we're not.

All of those things are generally considered no good though, perhaps minus the phantom fundamental thing, which I don't really see the big deal in considering that we won't be able to hear the ultrasonic overtones in the first place, consequently no phantom fundamental (our analog gear can't reproduce it, and our ears can't even hear it).

If ultrasonics are thrown into the mix, yes, technically a higher sampling rate will preserve the original ultrasonic overtones more accurately, but again our ears can't hear it and our gear can't reproduce it. We use a filter on converters for the distortion reasons I mentioned above. Once you get to the point of about 48khz, or even some some stuff like a Synclavier which can get up to 50k-60k if I remember correctly, you're basically going to have to squint and also have golden ears to hear any benefit other than that.

The only other thing I can muster up from what you're saying is that perhaps software FX have been programmed these days to produce audible tones from ultrasonic frequencies. I've never heard of that, but maybe it exists.

If I'm missing something here about how ultrasonics somehow modulate frequencies within human hearing range other than the three or four scenarios I listed above, please link me to something so I can correct myself. I have no problem admitting it if I'm wrong, I just haven't seen any evidence to support what you're saying (yet).

[MacTron]

My main point into this is to show that Monty's claims in his video are false.
At first he claims that a 20 khz a tone can be digitized and "analogized" with full fidelity to original a 44khz of sampling rate. And that's no true.
To show this, He uses a 20 Khz sine wave.
But at the end of the video it uses a 1 khz ! square wave and...  wait! ... it cannot be reconstructed with full fidelity to original ... unless ...