Skip to main content

Is there any practical reason for a small, home studio to be using 192khz over say 96, or even 44.1? Is that high of a rate really needed or necessary?

Thanks!

Topic Tags

Comments

Cucco Mon, 01/07/2008 - 07:44

Good, simple advice so far and I won't differ with what's given.

For the most practical part, lower sample rates (44.1 or 48 if going to video) are just fine.

Yes, many people can hear the difference between 44.1 and 88.2 or 96 kHz. Many report to be able to hear a difference between 96 and 192. Only in rare instances have I heard the difference in the latter and EVERYTHING must be perfect prior to the Digital conversion for the difference to be audible.

What I've found is that, unless the room in which you're recording is phenomenal, the equipment is top-notch and your chops are in top form, the difference between 44.1 and 192 is not noticable and not worth the HD real estate.

One point I will differ on others about (not yet stated in this topic) is:

"If it's just going to be an MP3 or other lossy compression format, I just record at........"

1 - I don't care how it ends up, I want my work and my product to be as good as it can be. When we start pandering to the lowest common denomenator, we've lost our way.

2 - I've found that the higher quality the work going in, the better the MP3 or other lossy format works.

"16 bit and 24 bit are only different because of their noise spec..."

There are tons of variations to this statement - most of them are wrong. The advantages of 24 bit over 16 bit are more than just SNR. Quantization Errors are the biggest reason to stay at 24 bit over 16 bit. In general (and, IMO, rather blatantly) 24 bit sounds signinficantly better than 16 bit because of the fact that the changes in amplitude are far smoother and more detailed and accurate than 16 bit. Regardless of noise. (Besides, there is no true 24 bit noise figure since analog gear cannot approach the 120dB SNR point).

All that being said, I personally do most of my projects at 24/44.1k and only on special projects where all the variables are right will I go higher up to 88.2, 176.4 or DSD.

Cheers-

J

UncleBob58 Mon, 01/07/2008 - 09:51

When creating sound FX I work at 96kHz/24bit as I seem to I end up with fewer artifacts when doing large amounts of Audio Suite processing on the sounds I record for my sound design work, and I do a lot of that kind of work. It doesn't seem to make any difference or I end up with more artifacts when using library material originally at 44.1kHz/16bit. My main work file (dialog, Foley, sound FX and music), however, is at 48kHz/24bit. Then, of course, I have to dither the final mix down to 48kHz/16bit for layback to FCP or Avid.

BobRogers Mon, 01/07/2008 - 10:41

Basically in agreement with the consensus here. I record most things the will be burned to CD at 44.1/24. I'll use 88.2 on something like a solo classical recording where I won't have a lot of tracks, use a lot of processing power or space. I don't feel it makes much difference in the final product, but it certainly doesn't make it worse.

A few quibbles.

Cucco wrote: ...One point I will differ on others about (not yet stated in this topic) is:
"If it's just going to be an MP3 or other lossy compression format, I just record at........"
1 - I don't care how it ends up, I want my work and my product to be as good as it can be. When we start pandering to the lowest common denominator, we've lost our way.
2 - I've found that the higher quality the work going in, the better the MP3 or other lossy format works.

While I agree that we shouldn't engineer for the worst possible format that the consumer will use the product, we are still engineering for the consumer, so my goal is a good 44.1/16 CD.

I highly recommend the following experiment:
1. Record some sort of reproducible sound - I used a demo program on a keyboard but something analog like a music box would probably be better- at several sample rates and bit depths.
2. Bounce down to 44.1/16 and burn to a CD.
3. Put the CD on "random" and see if you can identify the different tracks.
I found it very hard to tell any differences and it wasn't like the "good" sounding one was 96kHz. Lots of flaws in the design of experiment, but it is worth a try. As I say, after doing it I record at 44.1/24.

...."16 bit and 24 bit are only different because of their noise spec..." There are tons of variations to this statement - most of them are wrong. The advantages of 24 bit over 16 bit are more than just SNR. Quantization Errors are the biggest reason to stay at 24 bit over 16 bit. In general (and, IMO, rather blatantly) 24 bit sounds signinficantly better than 16 bit because of the fact that the changes in amplitude are far smoother and more detailed and accurate than 16 bit. Regardless of noise. (Besides, there is no true 24 bit noise figure since analog gear cannot approach the 120dB SNR point)....

I agree, but I'd emphasize headroom over quantization. After all, if someone really good uses all 16 of those 16 bits and I use 16 out of 24 we get the same quantization. However, while the person recording at 16 was riding the gain like a hawk, I was laying back, taking a snooze and letting all those nice zeros protect me from any clipping. In reality, I'll try to use only half those zeros and so I'll get a little smoother quantization as well. But the headroom is what I value most (being the hack that I am).

Cucco Mon, 01/07/2008 - 11:35

BobRogers wrote: I agree, but I'd emphasize headroom over quantization. After all, if someone really good uses all 16 of those 16 bits and I use 16 out of 24 we get the same quantization. However, while the person recording at 16 was riding the gain like a hawk, I was laying back, taking a snooze and letting all those nice zeros protect me from any clipping. In reality, I'll try to use only half those zeros and so I'll get a little smoother quantization as well. But the headroom is what I value most (being the hack that I am).

I see where you're coming from, but there's one fatal flaw in that logic.

The flaw is that the greater accuracy in quantization does not simply come after the headroom point has been reached in 16 bit. 24 bit effects the audio stream regardless of the amplitude coming in. True, it does impact the headroom available to you (in relationship to digital black noise), however, a signal at -50dBFS moving to a signal at -49.9999999999999 dBFS (number of nines in the previous example is random, not a specific quantity...you're a math guy so I thought I'd make that as specific as possible) is technically well within the range of the 16 and 24 bit signals.

However, the 16 bit signal may round the least significant bit (LSB - last bit in the series) up or down depending upon the remaining 8 bits in the signal chain whereas the 24 bit signal will figure this far more correctly or at least with the capability of 8 more bits of accuracy.

In other words, it's incorrect to assume that the quantization benefits occur only past a certain amplitude. Those 8 other bits have a direct impact on the signal from -120dBFS all the way up to 0dBFS.

Cheers-

J.

BobRogers Mon, 01/07/2008 - 12:28

Maybe I'm not thinking of this correctly. My understanding of fast fourier transforms is probably a gross simplification of how they get implemented in the real world. Math guys are like that. But first, dBFS means two different things for the two different situations. The 16 bit guy using all 16 bits is hitting peaks at 0dBFS. The 24 bit guy using 16 bits is hitting peaks at -48dBFS. The least significant bit hits the same fraction of the signal either was - from what I can see.

To put it at a more basic level - If the algorithm for quantizing 24 bit signals with 8 zero bits was better than the algorithm for quantizing 16 bit signals wouldn't we use it and ignore those 8 zeros?

taxman Mon, 01/07/2008 - 13:34

Does the fact that a signal may go through the analog/digital conversion process several times change the recommendation? For example, assume guitar is DI to DAW, then sent out to be re-amped, then sent out again to be EQed or Compressed or whatever. Is there a benefit of having these multiple A/D conversions done at a higher sample rate until you get to the final mix for CD?

BrianaW Sat, 01/19/2008 - 22:06

I believe I do hear a major difference between 44 and 96. Mostly in the form of VST plugins. Try Amplitube in 44, and then switch to 96, or even 48 and you'll see what I mean... big difference. So I use 96 most of the time for that reason. I also think drums sound more natural at 24/96 and I'm not using any fancy mics, just an Audix set and I'm monitoring through a set of ns10's. So to answer your question, it is my opinion that a higher bit rate will yield better results in the sound of your plugins, vsti's, and live percussion... try the Amplitube test, and if the increase in quality is worth it to you, go for it. I'm not sure how many plugs support the 192 rate, but there should be a decent amount out there. Anyone else with me on this?

BobRogers Tue, 01/22/2008 - 03:50

BrianaW wrote: I believe I do hear a major difference between 44 and 96. Mostly in the form of VST plugins. Try Amplitube in 44, and then switch to 96, or even 48 and you'll see what I mean... big difference. So I use 96 most of the time for that reason. I also think drums sound more natural at 24/96 and I'm not using any fancy mics, just an Audix set and I'm monitoring through a set of ns10's. So to answer your question, it is my opinion that a higher bit rate will yield better results in the sound of your plugins, vsti's, and live percussion... try the Amplitube test, and if the increase in quality is worth it to you, go for it. I'm not sure how many plugs support the 192 rate, but there should be a decent amount out there. Anyone else with me on this?

Are you talking about comparison between the sounds before or after conversion to final format, say 44.1/16 for CD? My experience is that after dithering down to 44.1/16 the differences are very hard to detect, though I admit that I have not done any careful tests with either Amplitude or drums.

BrianaW Tue, 01/22/2008 - 05:20

Are you talking about comparison between the sounds before or after conversion to final format, say 44.1/16 for CD? My experience is that after dithering down to 44.1/16 the differences are very hard to detect, though I admit that I have not done any careful tests with either Amplitude or drums.

Sorry, I forgot to mention that. I was referring to the sound I perceive after the dithering down to 44/16. But in order to get that sound I have to print the effects before the conversion. So if I have a guitar track with Amplitube as an insert, I'll apply that amp sim to the track immediately before doing the final mix and downsampling. I usually pop it (the 16 bit 44k 2 track mixdown) into Wavelab after the conversion to give it a little more presence and punch too. I know that's probably more processing than a lot of people like to do, but I do find that it helps to get me the sound I want. I think in general, it's just that many VST's and VSTi's sound cleaner and more similar to hardware when using higher sampling rates... so printing channels with those settings seems to retain a lot of the higher quality (especially with amp sims). This is just my opinion of course, everyone has their own way of getting the sounds they like. :)

BobRogers Tue, 01/22/2008 - 09:30

Well, as I say, I've done some pretty careful tests on this subject and come to the opposite conclusion, but (1) the test was on a different kind of sound source and processing regime and (2) your ears may be better than mine. It's important to be careful in a situation like this. When there is a stage where the audible difference are clear, it's easy to fool yourself that they persist throughout the process. If you haven't already done so, I'd encourage you to design some careful tests of your conclusion. I think what I would do would be to record the guitar at the highest sample rate and then downsample the naked guitar track down to the lower rates. Apply identical settings of Amplitude to the various tracks. Downsample all instances to 44.1/16. Put two or three copies of each sample on a CD. Put in the CD player and hit shuffle. See what your batting average is at identifying the tracks.

For something like classical or other acoustic music where we deal with smaller numbers of tracks and (hopefully) minimal processing the costs of using higher sample rates is often so small that it isn't worth spending a lot of (my) time worrying about the benefits. But for rock recording with a lot of tracks and a lot of processing the costs in disk space and processing power are very real limitations, so it pays to be very sure of the benefits.

BrianaW Wed, 01/23/2008 - 02:19

Hello.
That is a great idea! I just finished the listening tests. I used Nuendo 3, Amplitube 1, and the Waves Stompbox EQ. I saved the plugin presets, downsampled the dry track to 44/16 and imported it into a new 44/16 project. I then inserted those plugs in the same slots/order, and loaded the presets. I exported both versions to stereo interleaved 44/16 wav files. I then put them on shuffle, and I was able to identify which track was playing every time for about 10 mins... the track itself being 17 seconds. Obviously I'd look only after I made the call, but usually there was no doubt. It was harder for me to discern the jump from 24/96 to 44/16 at first, but every time the 24/96 came on, it was very obvious to me. Sounded clearer, the gain sounded more crisp and natural (more like natural tube breakup), and the track itself just seemed to have more high end sparkle. Maybe I'm doing something wrong, and again there isn't so much of a difference that I'd recommend it to someone with less HD space (these projects are massive!), but I really do feel that I hear enough of a difference to make the sacrifice worth it. I do not know if the jump to 192 would be noticable to me though. It could even just be the plugs I'm using, who knows. That was definitely fun though! It's so nice to geek out about this kind of stuff with other audiophiles! :D