need an explanation
okay, i have been working in 24/44.1 for quite awhile now. 24/96 seems to be the “standard” that everyone boasts when talking about digital recording (maybe “standard” isn’t the right word, but it seems very common). i have a good understanding of what 24-bit does for me, as opposed to 16-bit. i feel i need a clearer understanding of what benefit there is to recording at a higher sampling frequency.
for one, you will want to google “nyquist theory”…
higher “sample rate” means more “listens” per second… this results in a wave pattern that is closer to the same wave pattern in analog… the benefit is often argued, especially since if cd is the target medium, then it will be downsampled anyway to 16/44.1… but recording at 24/96+ means that your recordings will be ready for more defined media, such as dvd should you want to go that route… upsampling generally isn’t a good thing…
how would this affect my use of an application like fruity loops. if i was to record in n-track at 96khz, obviously i would want to export my drum track at the same sample rate. since all of the drum sounds i use are 44.1khz, this is effectively upsampling to get the final drum mix. bad, no?
i should clarify that upsampling certainly won’t ruin your tracks… i doubt you’d notice the difference having upsampled only one part of a mix… and unless you are using 24/96 samples, you’ll have to at least upsample - all of your tracks must be at the same sample rate, although the bit depth can be mixed…
i wouldn’t worry so much about the drum samples being upsampled, as i said earlier, if in a mix, i doubt that you’d really notice… just keep your other instruments recorded at 24/96…
Keep in mind that the size of your files and the load on your processor increases with higher sample rates. These cons may outweigh the pro of slightly (if at all) audible change in the outcome of your projects. You might also do a search on this board or over on Audiominds for nyquist, or sample rate. I know that Learjeff has posted some pretty detailed explanations of how all this works.
i have to slightly disagree earache… listening to audio recorded at a higher sample rate on consumer grade equipment may not present an audible difference, but with proper equipment, the difference is astounding…
however, i do feel that using longer word lengths (bit depth) may be more beneficial than higher sampling rates… the difference in dynamic range dramatically increases with deeper bit rates… nonetheless, i think it best to record at the highest quality possible and use quality resampling software to output to the standard 16/44.1…
but, that’s just my opinion… …
yeah, i really do prefer to have my source material at the higest quality possible, even if the difference is only subtle to me now. i’ve gone back to old projects and wished i had done this or that, but my knowledge was limited and so was my computer and recording equipment. both my pc and sound card are capable of 24/96 and can more than handle any extra cpu load or disk space requirements. that is why i want to learn about it and at least try it out at this point.
I suspect that for most folks in home studios, there are so many things that make a much bigger difference than 96k, you’re better off just using 44.1k and focusing on more important things. Especially if low latency is important to you (only if you need live FX during recording, like using GuitarRig, or if you’re using plugin synths).
There are some effects that work much better on an 96k signal than 44.1 – for example, many chorus plugins that use variable delay lines. In this case, it would be nearly as true using upsampled tracks as using tracks recorded at the high rate (assuming good upsampling software).
I used to believe that human hearing did indeed end well below 20k, so there was no point recording higher frequencies. I’ve since read a good study that showed that some people can detect the difference (using gammelon music, with people who were used to listening to it – lotsa HF there) and prefer it. That study included two different kinds of objective results (MRI and EEG) that showed significant differences. So, some people can hear higher frequencies (somehow). I doubt I’m one of them.
Regardless, using 44.1 we can record truly excellent tracks, and with care and taste, make excellent mixes from them. For anything but the simplest mixes, my feeling is that only the best engineers mix well enough that the difference between 44.1k tracks and 96k would make a significant difference in the overall impression of quality on the part of a discerning listener. Unless you’re doing very simple mixing (ideal for acts like a string quartet), or you’re a very good mixer for pop/rock music, there will be flaws in your results that are FAR more noticeable to the trained ear, and the extra time you spent waiting for 96k mixdowns would be pretty much wasted.
When talking about mixes from high quality pro engineers, there’s no question that they should start out with 96kHz or 192kHz tracks. Why not use the best? If you’re good, why limit yourself? Not to mention that you’d most likely have the most blazingly fast computers, or else time to wait for mixdowns that take more than twice as long (or methods that minimize the need). Or, you might want an excuse for a coffee break.
For the significant majority of home studios, I think it’s pretty much wasted disk space and CPU time, with certain specific exceptions, like when using one of the plugins where sample rate matters significantly. (Most plugins aren’t this way.)
If you can’t make a professional-quality sounding mix using 16/44.1 tracks, you can’t do it with 24/192 either. But the overhead for using 24 bits is low, and the advantages are obvious (more digital headroom, now you only have to really sweat your analog signal chain rather than making sure that both analog and digital domains are optimized).
thanks jeff, alot of good info there.
again, i think it is about having the best quality source available. maybe i’ll notice a difference right away, but probably i won’t… but down the road, it could be a benefit to have that extra quality if i ever come back to older recordings. maybe at a time when my ear is a little better and computer a little faster
of course, it has to be weighted against the speed of the computer and the disk space now, to decide whether it’s worth it. i think i’ll give it a try on a song or two and decide if the benefits are worth the extra time and cpu power.
Just keep in mind that free stuff might be worth what you paid for it!
Sure though – I sampled my Rhodes and now I wish I could have done it in 96k mode, even though I wouldn’t end up using that rate in the end. But it would be nice to have them on file just in case (and there is a particular reason, which I discovered later, why the higher rate would be better for early stages even though I’d sample down to 16/44 for soundfont format in the end anyway). Instead, I do still have the Rhodes packed away in the attic … not having sampled it thoroughly enough is as good an excuse not to sell it as any!
What it really means to me … the difference between 44.1 and 96 is merely one single octave way beyond the frequencies that most of us can’t hear anyway …
It’s like the difference of having a playground of 10 octaves or a playground of 11 octaves …
Just like LearJeff commented : it’s different for “sound samples” which are re-sampled/interpolated for different frequencies/notes.
Some of the benefit of 96khz on cheap convertors is that it smashes all the low pass filtering into the super high range where we can’t hear it rather than at the top of our hearing where we might with less than stellar filters and convertors…
Right, Bubba. You see, in order for sampling to work without nasty aliasing, it’s necessary to squash all frequencies over half the sampling rate (this is called the Nyquist ratio or limit). So, to record up to 20kHz but kill everything at 22.05 kHz, you need a filter that drops by over 130 dB in the space of a few halftones. That’s like what – a 30-pole filter? ( ). Regardless, it’s very hard to do will with analog ("IIR or infinite impulse response) filters. Sampling at 96k, soundcard designers get to pick anywhere between two extremes: recording the extra HF, or leaving all that extra room for the filter.
This filter used to be the single most critical element in an A/D converter. However, with the advent of so-called “one bit A/D” technology, analog is converted to digital at much, much higher frequencies (like 3 MHz) – a 1-bit delta modulation signal. The good news is that they can then use a digital FIR (finite impulse response) filters. For reasons I don’t comprehend, it’s easier to do this way, so it’s not nearly the bugbear it used to be. But I bet that the DSP algorithm for filtering is still an important design point and one of the big differentiators between different levels of quality. And allowing more spectrum range for it probably still allows it to do a better job with fewer tradeoffs. However, that’s mostly speculation on my part.
Holy cow am I confused now! I have the 16bit version of N and record at 44.1.
If I listen to one opinion, I should upgrade to 24bit and record at 48 or higher (96 being overkill perhaps).
If I listen to the other, I should leave it 16/44.1 and focus on other things (in other words, “walk then run”).
One more factor, I’m trying to bone up on dither and noise shaping (different thread). Currently I don’t dither at all. What factor does this have. I don’t wish to derail the discussion by bringing up dither, but they are related, no?
|One more factor, I’m trying to bone up on dither and noise shaping (different thread). Currently I don’t dither at all. What factor does this have. I don’t wish to derail the discussion by bringing up dither, but they are related, no?|
If you record at 44.1/16 you don’t need to worry about dithering. CD format is 44.1/16. You only need to dither if you are going from a higher bit depth to a lower one - eg 24 bit to 16… Dither is just a technique for dealing with the data that is stored in those bits which need to be got rid of.
The Ozone doc recommended in the other thread should put you right.
Sorry, Mark – wrong answer. You would be correct if (and only if) you recorded a single track at 16/44.1, and did no adjustments to it at all before burning to CD.
As soon as you make any adjustments (eq, fx, or adding two tracks) you effectively expand the bit width, up to a maximum of n-Track’s internal format of 32-bit float (roughly equivalent to 24-bit format for most purposes). Even a simple fader adjustment and you should dither (in theory, anyway). Let me know if you want me to explain that in detail.
For a set-and-forget approach, use 1 bit with noise shaping. Later, when you’re getting picky about subtleties, experiment and choose your own settings based on listening tests. Be sure you’re using a good monitoring system.
About 16 vs. 24 bits:
1) Use what ya brung. Don’t postpone doing stuff while waiting for better stuff. If all you have is a built-in soundcard, go for it and start recording! Meanwhile, consider the advantages to upgrading.
2) If you already have a 24-bit soundcard, then definitely pop the extra $25 for 24-bit n-Track. You’ll never regret it. For one thing, it makes setting recording levels a lot less fussy. You can leave more headroom when recording and if you don’t peak at -12dB or above, who cares? Push the fader up. With a 16-bit soundcard, if you peak at -12dB, you’ve wasted two bits so now you’re really using a 14-bit soundcard. Still sounds OK, but not ideal by any means, especially if that happens on lots of tracks. With a 24-bit card, you sweat the ANALOG sound chain, make sure the gain is optimal at every step of that. In the digital realm, just get it in the ballpark and you’re good to go. The noise in your analog chain is way higher than the digital quantization noise, even if you leave plenty of headroom in case of peaks.
If you don’t have a 24-bit soundcard yet, start thinking seriously about it. 24 bits isn’t the only advantage – the quality of conversion will be noticeably better. Monitoring will sound better too, and you’ll hear more of what you’re doing in the mix.
Final bit of advice for newbies (sorry if I’ve already said this): don’t start by recording that song you’re just dying to do. Start with easy songs, even covers. You’ll make mistakes, and you’ll get burned out before long on the first couple projects. Wait until you’re getting good recordings & mixes with simple tunes before starting your magnum opus, or that gift song for your true love. You’ll be glad you did.
Here are my $.02 on the subject. I have been doing digital recording since 1991. Back then I bought a Sony DAT deck. The DAT recorded at 16bit 44k and at the time sounded pretty good. The A/Ds and D/As used today are so far superior to the converters back then it is astounding. Even recording at 16bit 44k on a cheap sound card today sounds better then the DATs back then.
That said, it is always best to record at the greatest bit depth that your sound conversion device will allow. You will get a much more accurate recording and even dithering down in the final process will allow a much more accurate picture of the original sounds.
|Quote (learjeff @ July 15 2005,15:18)|
|If you already have a 24-bit soundcard, then definitely pop the extra $25 for 24-bit n-Track.|
|Quote (DrGuitar @ July 15 2005,15:38)|
|always best to record at the greatest bit depth that your sound conversion device will allow|
Ummm… … how do I know if my soundcard can do 24bits? I use the M-Audio Fast Track USB interface, and that can do 24bit/48k, but I don’t know about my laptop.
Or does the Fast Track effectively take the place of my laptop soundcard making this question irrelevant?
Sorry about all the questions. I’m actually starting to annoy myself!!
"Final bit of advice for newbies (sorry if I’ve already said this): don’t start by recording that song you’re just dying to do. Start with easy songs, even covers. You’ll make mistakes, and you’ll get burned out before long on the first couple projects. Wait until you’re getting good recordings & mixes with simple tunes before starting your magnum opus, or that gift song for your true love. You’ll be glad you did."
Thanks for this - I’m going back to recording a generic boogie and putting aside the major fusion intrumental till I get a lot better at this recording/mixcing stuff…also forces me to get more practice playing cleanly and in time…it is funny how that stuff seems to slide by when you have not been listening to yourself recorded. I mean I knew I sucked but the digitally recorded truth is very humbling.