Anyone record at 24/96?

[Leh173]

Hey just wanted to ask if anyone has tried or works at 24/96? Usually for music I work at 24/44.1, at work 24/48(post production sound). I'm re-recording my old songs for fun, and wonder if it's worth it. I did read the last Van Halen record was recorded 24/44.1 but mixed down onto another system at 24/96 which makes sense I guess. What do you guys reckon?

[Al Coda]

Hey just wanted to ask if anyone has tried or works at 24/96? Usually for music I work at 24/44.1, at work 24/48(post production sound). I'm re-recording my old songs for fun, and wonder if it's worth it. I did read the last Van Halen record was recorded 24/44.1 but mixed down onto another system at 24/96 which makes sense I guess. What do you guys reckon?

 

20Khz sine wave @44.1K = square wave

You have only 4 sample points in that case.

That´s the theory ...

 

When I do pop stuff and/or go back to old songs from the past not knowing anymore which samples are in the project (there were often samples @44.1K in use), 32Bit float/44.1K is what I´m using.

Otherwise,- recording real and/or acoustic instruments, 32Bit float/48K is my choice.

Up to now I never recorded @96KHz or higher SRs (192/384KHz).

I doubt it´s essential for most signals.

I don´t do classical music and don´t have an audio optimized room w/ a grand piano and hi end mics.

I doubt I´d be able to hear the difference of 48K and 96K in my room even my speakers and amps are very good.

 

A.C.

 

[Leh173]

Good advice. I am skeptical that it's worth it with the stuff I'm recording at home, I'm sold on the 24bit part as I use that at work. When I was first working recording and in post production the bit depths and sample rates were completely fixed so you had to choose one. Pro Tools is smart and can handle different things at once at least with bit depth anyway. Using Logic at home so not sure how flexible. I was sticking to 24.44.1 for music due to using samples which would be 44.1 anyway and not sure how the Fantom would react for digital connections as it's fixed at 24/44.1 as far as I know. I recorded one song at 24/96 as an experiment and its sounds fine, computer handled it fine too, but might go back to 44.1 on the next one. I guess I can always downsample (convert ) this song Interesting stuff.

[Al Coda]

Good advice.

 

Can be,- or not ...

 

There´s lots of discussion about fixed point vs floationg point calculation.

fixed vs float

 

Same for 48K vs 96K SRs.

48 vs 96

 

But I myself, I´m more a keyboardplayer than a (mastering-) engineer and until I have read all that, I´d probably have recorded 3 albums incl. composition/arrangement.

I´m lazy too,- so I prefer using what works for me and sounds good to my ears.

Digital DAW is editing heaven, but when I have good players, electric guitars and acoustic drums sound best on tape for me p.ex..

96KHz+ sample rates produce a lot more data to deal with, much larger files,- and is wasted disk space for many if not most signals IMO.

What´s the range of the mics you use ? 16, 18 or 20KHz ?

What´s the range of the voice you record ?

What´s the range of the guitar amp´s speaker you mic ? 7KHz ?

What about your ears ?

Do you hear more than 20KHz ?

 

Electric keyboards:

I myself, I prefer dull warm sounds over harsh and bright stuff.

What´s the highest frequency a Hammond tonewheel organ can produce (thru a Leslie) and what are the most used drawbar settings ?

I think we´re talking about 8KHz max..

What about warm synth pads or buttery Minimoog leads ?

 

Well, there are other sounds too offering lots of overtones, but these aren´t loud in a mix all the time if at all and the bells and whistles get lost anyway.

 

Digital recording is sampling and I deal w/ it like I did w/ samplers in the past.

Use what the project needs, analyse the project before choosing higher sampling rates.

I think most is well covered w/ 48KHz because the Nyquist frequency is about 24KHz then.

 

A.C.

[Theo Verelst]

I regularly record with 192kH/24bits to a 192kHz 32 bit float digital signal path, using pretty good computer-mother-board AD converters.

 

Of course there's an easy and a hard part to understand about simply the sampling considerations. The simple part is what was indicated: 2 samples for frequencies near the Nyquist rate can easily be imagined to be very, very inaccurate.

 

The solution is harder: DA converters will be able to reconstruct the recorded signal accurately to be close to the recorded analog-signal if they're really good, and the samples didn't get messed with. Up-sampling to 192 from 44.1 with a computer and a computation intensive up-sampling filter (load on your processors should go high, filter should be called "sinc") is the closest to this pure reconstruction people can normally get in this time. That's worth it, but specialist at the moment.

 

Close to perfect reconstruction from let's say CD or 44.1/24bit is possible when the frequencies you've recorded do not go above 22kHz, but strictly so, which is hard but a lot of equipment can do a reasonable job of making that so. It is possible new Digital to Analog converters become available to replay CDs and other digital audio that has been properly recorded with much better signal coming out of the DA converter, but the much more perfect DA converter will exhibit delay, up to seconds (!) of delay to do so.

 

Lots of mess is going on in recording, production and post-production (IMO that's objective fact), with people just messing about until they get some sort of monitored signal that pleases them a bit, or makes them feel superior.

 

Doing everything right like in the (also older, from the 60s onward) A grade products isn't easy and includes all kinds of complicated audio considerations, which can also be sampled and digitally created, but almost nobody works on this, mainly the sampling problem is that also in the mid-frequency range (few kHz) there are only so many samples to contain a wave-form, and that gives (for me very) clear distortion which can easily be measured, too.

 

There (complicated, currently little known) production tricks to prevent huge sampling errors and also to an extend dangerous sampled loudness wars. I work on some of that, but using not 96kHz but 192 kHz to do processing.

 

Recording at a higher sample rate can allow you to make a better down-sampled product. It's certainly useful record in 96kHz, use plugins and softsynths at this more accurate sampling frequency, and after you've monitored the master mix to sound ok, make a CD track of it. In fact when I use my complicated mastering signal paths, I regularly record a 44.1/16 track of what I'm doing, and when I do everything right, those CD tracks sound great on anything between mobile phones and my big monitoring system.

 

It's also possible to (multi-track) record with CD quality, and properly upsample to do a mix, should be better. I think it is useful, depending on what instruments you record and the professionalness of the result required, to test out what the difference is between a pure analog signal path, and the digital path of choice. For me, even blowing a humble enough AT2020 microphone signal through a pretty decently spec-ed 192/24 AD/DA conversion isn't perfect. Everything lower certainly needs work for e to sound ok. But it depends on what you record: a bass and an instrument which is already digital isn't the same as a great female voice+ Lexicon, a shaker and an earth shattering orchestra by a potent and complicated digital power machine.

 

Working on the perceived quality of your tracks may do more than simply switching to 96kHz without knowing the tricks built in the signals you use, and what to listen for. Experimenting, without leading to a loudness war that can turn ugly when messing with samples, should be interesting.

 

T.

 

[Dave Bryce]

I regularly record a 44.1/16 track of what I'm doing, and when I do everything right, those CD tracks sound great on anything between mobile phones and my big monitoring system.

Yup.

 

I almost always work at 16 bit/44.1, because that's the hghest rate it'll be played back at (if I'm I'm lucky and it makes it that high - I firmly believe that most people listen to stuff at mp3 resolutions these days)....and, as far as I can tell, it sounds just fine.

 

dB

[zoooombiex]

I'm with the OP on the 24 bit being a pretty substantial jump from 16 bit and definitely worth it. IMO, it makes a pretty big difference in terms of noise floor and the overall approach to recording, even if it is later played on a 16 bit CD or even mp3.

 

As for 44.1/96, my old interface was 24/44.1 and it was fine. I upgraded to something with more options, but haven't gone above 24/96. I wouldn't begrudge anyone who wanted to go beyond that, but I haven't found it necessary. I have weaker links in the chain.

 

Also, I don't have the background to evaluate it, but I've seen at least one article suggesting that 192 can - on some systems - be less accurate than 96 because of how much it taxes the system. Something about how, in trying to keep up with the flood of incoming data at 192 there end up being more sampling errors than at 96. That was from a manufacturer who stopped offering 192 on his products. And another company (Echo) recently removed the ability to record at 192 from an existing product. Regardless of all that, I know I won't personally benefit from 192 without a substantial increase in skills and gear...

 

 

EDIT: here is the thread I was referring to: http://www.gearslutz.com/board/so-much-gear-so-little-time/568180-192khz-sampling-loss-accuracy.html

 

Reminds me a little of a quantum uncertainty between location and momentum...

 

Another EDIT: the link for the underlying white paper on sampling theory has change and is now here: http://lavryengineering.com/pdfs/lavry-sampling-theory.pdf

[Theo Verelst]

Certain speakers/amps/mixers respond to certain sampling errors, and there are people who therefore cannot easily upgrade their pack of tricks to another sampling rate. Or to a decent listening setup, for that matter.

 

Computers not keeping up is an entirely different matter, preferably there should be a clear plan to indicate to the user when that is the case.

 

T.