Home recording from tape: 24/48 vs 24/96

OK. I know that 24 vs 16 gives you much more "information" and a better sound, but what exactly do you gain from boosting the frequency from 48 to 96? Is it worth the extra space that the files take up to do 24/96 vs 24/48. I have read postings elsewhere that say the difference is slight and not worth the space, yet I always lean towards trying to do everything to the max.

Anyone have strong feelings about this one way or the other?

NOTE: I am talking recording tape to be used in DVD-A creation, not CD burning. Home use, not professional stuff (Wavelab 5, etc)

:-jon

 

Hello Jon.

I'd say that if you have the capability why not use it.

After all, at the end you will erase the large files from your hard disk.

If applicable, do remember to restore the original Lp song order, pleeze!

 

Also check:

http://www.ethanwiner.com/BitsTest.html

 

Some people, like me, feel that larger sampling rates improves the sound. The sound is warmer and smoother. With an ideal sampling rate and proper sample rate reduction, one can retain the richness of the original ri sampling rate.

Others feel that there are no benifits to be gained, and large files only waste time, space, and offer no improvement.

 

Another useful tool (some cables are needed and you may be humbled):

http://audio.rightmark.org/manifest.shtml

 

Thanks guys, I value both of your opinions!

 

You'll gain more from the 16 to 24 bit jump than you will from 48 to 96. Still, there's a doubling of sampling information so there's going to be an improvement if the converters are true 96K and if the monitors and signal path are revealing enough. Also, think about the Nyquist frequency. The top frequency is theoretically raised from 24KHz to 48KHz, taking all that noise up there with it (that's a good thing).

I'm like you, I always want to do everything to the max. If 24/96 is available I'd use it. You're making DVD-R's so you can clean your hard drive off after they're made.

 

That's my method also--sample at the highest rate I possibly can, do any "manipulation" (even level changes will benefit), then downsample if I need to burn a CD. Hard drive space is so cheap, as are blank CD-Rs, I don't worry about it.

Take these with a grain of salt, but I made them up awhile ago as a "For Dummies" tutorial to show the effect of different sampling rates. 15,750 Hz sine wave (hey, I had to pick a random frequency, so why not a TV's flyback transformer? ), at various rates. Since low frequencies are more accurately sampled, I used a higher one just to amplify how much different the sampling rate could be on the signal. (So...I don't want this "devolving" into a debate about my sampling methods or digital in general.) These are also raw, unfiltered digital waveforms as well, generated in an editing program.

Sine wave:


192 kHz, 24 bit:


96 kHz, 24 bit:


44.1kHz, 16 bit:

 

Hello Rudy.

What program you used for this?

Thanks in advance.


Note: I've been doing some research recently and found this:

http://www.digido.com/portal/pmodule_id=11/pmdmode=fullscreen/pageadder_page_id=18/

I think many here would like to read these articles.

 

Hi Jon,

Use whatever sounds best to you. Some converters sound best at 44.1/24 and some at 96/24. IMO, 48/24 with a good converter should be enough.

Rudy, your pics are not correct. The waveform should come back as a sinewave, not as some sawtooths which are constructed out of infinite number of sine waves.

Here's what 15,750 Hz really looks like when sampled at 44.1/16:

 

Connect the dots in your image, and you'll get the samples I posted. The samples I posted are synthesized digitally, not run through any filtering.

Pablo: you can use a sound editor like Sound Forge to synthesize a frequency. You tell it what frequency to create, and for what duration. With different sampling rates, you get the different results, which is what I posted.

 

Jon, It is the bandwidth and the filters that also make a difference along with getting twice as much information. Because of the higher sampling frequency you can capture higher frequency ultrasonics that we can hear but they can effect the signal we do hear. Second the filters they use at the higher rates do not have to cut-off until 48KHz (versus 24KHz). This means the filter can be non-existent at the upper audio limit of 20KHz and thus does not have the same effect on phase information that the 24KHz filter does.

Gorts: This does not belong in off topics, either hardware or music., please move it.

 

I'm using WL as you know, and with my E-MU 1820, have recorded reel to reels in at 96/24 for eventual burning on a DVD-A with WL5. The quality is great.

I've also converted some of the material down to 44.1/16 and it sounds great. So I'd go 96/24 if you can, but if you don't need 96, 48 may do you fine. 24-bit is certainly better than 16-bit.

I mean, I have 200 GB hard drives dedicated for this project, so space isn't an issue.

 

Thanks!

 

I was thinking the same thing three days ago.

 

Rudy:

That's not how sampling works. Each sample point undergoes the sinc (x) function.

All that a higher sample rate wins you is the ability to increase the frequency range of the recording.

Read all about is in this previously posted article:

http://www.lavryengineering.com/documents/Sampling_Theory.pdf

 

Rudy, simply connecting the dots as you have done would require much more bandwidth than is available. Give the Lavry paper a read or, better yet, acquire a copy of Ken Pohlman's book; it'll clear a lot of things up.

 

Oblio98: the following thread might help you........or just melt your brain. Either way, I personally found it extremely enlightening to read the pros discussing this touchy subject, and why I no longer care that much about a recording's sampling frequency:

The pros duke it out over high sampling rates.

It looks like the DACs and the digital filters are more important as far as good sound goes.

The thread was started in 2001 but continued for a looooong time as you can see by the enormous page count. "Nika" is one of that forum's tech-heads. This particular forum was closed a couple of months ago--I guess George Massenburg got tired of moderating it (check out this guy's discography-wow).

Have fun!

 

Folks, I am reluctantly commenting on this. For those of you who are not mathematicians, physicist or EE's you will probably think this is way too esoteric and it probably is. I am upset with the computer view taken by Dan Lavry in his paper (I am not sure if it is Dr. Lavry). DL is 110% right that if you bandwidth limit a square wave and then sample it at twice the bandwidth limit that you can reconstruct all of the information as Dr. Nyquist originally proved and anyone who has studied integral transforms knows (mathematically digitization is nothing more than an integral transform). The problem with his computer view is that our understanding of psychoacoustics is still not good enough to tell us how much bandwidth limiting is acceptable in the real world of analog signals which is what our ears hear and our brain interprets. Take a look at the redbook bandwidth limited square wave in his paper. Does it look like a square or an approximation of a square wave? Now if instead of a bandwidth limit of 22.05 KHz it was 96KHz (i.e. a 192 KHz sampling frequency) you would have 96 harmonics instead of 22 and the reconstruction would come a lot closer to a real square wave (it takes an infinite number of harmonics to make a true square wave and there is no such thing as infinite bandwidth in the real world, if you have not studied integral transform, Fourier series, Fourier transforms, etc., trust me on this). I cannot tell you if sampling at 192 KHz is over kill or not enough, but what I do know is it will come a lot closer to analog and I think that is good. This why I think SACD sounds so analog like, very high sampling frequency (at the expense of quantum noise, there is no free lunch, just trade offs). In case any of you are wondering why I am upset with this, when I took Intergral Trnasforms in my senior year of undergraduate education for my mathematics minor ( I am a physicist) I wrote one of my senior papers for that course on Chebyshev polynomials which form the mathematical basis for the Chebyshev filter shown in DL's paper. If any of you have read this far and really want to know why it does not appear to ring I can explain that to you but I hope you have better things to do like listen to good music.

 

Wow, this post has gotten very "deep"! I thank all of you for the input, links, and opinions. It takes time to reply in the depth and detail that some of you have done and I appreciate it. That's what makes this place so great!!!

:-j n