T
twangbuck
New member
Quick question: will an audio clip recorded at 48khz/24 bit rate and then converted to 44.1khz/16 bit rate sound as good as a clip originally recorded at 44.1/16 bit? I'd bet probably not. Thoughts?
ashcat_lt
Well-known member
All other things being exactly equal, and assuming a decent sample rate conversion process, for a single sound that is just recorded and played back without any additional processing in between, you probably won't actually hear the difference, and of course "better" is a subjective term that each person would have to decide for themselves. There will be measurable artifacts (distortion and noise, changes in frequency response and possibly phase relationships) in the converted file. The real difference comes when you try to record a multi-track arrangement, and process things, either before or after conversion. Then all of the artifacts start to stack up and/or accentuated by the following processes and it can start to get ugly. Honestly, though, it's usually pretty subtle stuff we're talking about.
As I understand it, the biggest offender is the sample rate conversion. Bit depth conversion is pretty simple, and dither usually undoes any distortion that might happen in the process. SRC, though, can cause all kinds of spectral weirdness. Things are getting better every day, and the topic is debatable, but the arguments I've heard have convinced me that SRC should be avoided whenever possible.
As I understand it, the biggest offender is the sample rate conversion. Bit depth conversion is pretty simple, and dither usually undoes any distortion that might happen in the process. SRC, though, can cause all kinds of spectral weirdness. Things are getting better every day, and the topic is debatable, but the arguments I've heard have convinced me that SRC should be avoided whenever possible.
T
twangbuck
New member
Thanks for the quick reply. So when people record at 48khz or higher and then convert it to 44.1 when doing a final mix for cd purposes, wouldn't that create the kinds of problems you're referring to?
ashcat_lt
Well-known member
Yep, and believe me that's the beginning of a long nasty thread there.
The thing is that it's kind of a trade off. 44.1K has problems in that its Nyquist frequency is so close to the top of the audible band, and the anti-alias filter needs to be really tight in order to roll off enough by 22.05K without affecting anything we can hear, and it's hard to create that kind of filter without introducing other problems. You have to weigh that against the possible damage that could be inflicted in the conversion process.
Either way, I think the effects are actually pretty subtle. Like I said, you probably won't barely hear it on a good monitors in a good room, and forget about it in any real world listening environments. It's only a problem if it's a problem.
The thing is that it's kind of a trade off. 44.1K has problems in that its Nyquist frequency is so close to the top of the audible band, and the anti-alias filter needs to be really tight in order to roll off enough by 22.05K without affecting anything we can hear, and it's hard to create that kind of filter without introducing other problems. You have to weigh that against the possible damage that could be inflicted in the conversion process.
Either way, I think the effects are actually pretty subtle. Like I said, you probably won't barely hear it on a good monitors in a good room, and forget about it in any real world listening environments. It's only a problem if it's a problem.
bouldersoundguy
Well-known member
That was a problem with ADCs before oversampling, but I don't think it's relevant today.
Farview
Well-known member
I'm not sure it is a big problem when mixing down a song. Even though the separate tracks are at 48k, when you are mixing down to 44.1k, you are creating something that didnt exist before and the conversion just becomes one of the millions of calculations in the process.
If you mixed down to 48k and then converted that file to 44.1k, that would cause artifacts.
If you mixed down to 48k and then converted that file to 44.1k, that would cause artifacts.
bouldersoundguy
Well-known member
In Pro Tools there's an option to convert during or after bounce, but it's not a sonic difference, it's to save processing power if there are glitches in the real time bounce. Artifacts such as...?
ashcat_lt
Well-known member
bouldersoundguy
Well-known member
That certainly makes it a lot clearer than "artifacts".
http://src.infinitewave.ca/ said:
bouldersoundguy
Well-known member
Andy Peters on Tape Op said:
Here's an interesting piece of a discussion you may remember. It's about hardware SRC, but I think it's still relevant. I've always found Andy Peters' positions on digital processing to be robust.
Original is here: Tape Op Message Board :: View topic - best sample rate for tracking and which for mixing?
T
twangbuck
New member
This is all good stuff, and I appreciate the insight! Here's the original intent behind my question. I'm interested in reamping guitar parts. I've read that the DI tracks should be recorded at at least 48 khz if not higher. I can do that, but I want to record the reamp'd guitar parts at 44.1 to save processing power, otherwise the system's going to start choking once I get a bunch of other parts recorded.
My original thought was to record the reamp'd parts at the higher sample rate and then convert them to 44.1 before recording other instruments, vocals etc. I tried that and I did hear a few audible pieces of noise in the playback, though they weren't recurring. In other words, I'd hear something, stop playback, go backward and play the same portion of the clip again and it wouldn't be there.
Anyway, that has me a little nervous about doing conversions.
Now I'm wondering if the version of cubase I"m using (LE 5) will allow me to change the project settings when recording the reamped parts. In other words, record the DI tracks at 48 khz (or 64 or whatever), then change the project settings to 44.1, playback the DI track and record the reamp'd track at the same time.
So (hopefully) that would mean I'd be playing back the DI track at the higher rate (for the sake of sending a high quality DI part to my amp) and then recording the new (reamp'd) track at the lower rate.
Would that work? Cubase would still play back the DI part at the higher rate after I changed the project settings, correct?
My original thought was to record the reamp'd parts at the higher sample rate and then convert them to 44.1 before recording other instruments, vocals etc. I tried that and I did hear a few audible pieces of noise in the playback, though they weren't recurring. In other words, I'd hear something, stop playback, go backward and play the same portion of the clip again and it wouldn't be there.
Anyway, that has me a little nervous about doing conversions.
Now I'm wondering if the version of cubase I"m using (LE 5) will allow me to change the project settings when recording the reamped parts. In other words, record the DI tracks at 48 khz (or 64 or whatever), then change the project settings to 44.1, playback the DI track and record the reamp'd track at the same time.
So (hopefully) that would mean I'd be playing back the DI track at the higher rate (for the sake of sending a high quality DI part to my amp) and then recording the new (reamp'd) track at the lower rate.
Would that work? Cubase would still play back the DI part at the higher rate after I changed the project settings, correct?
bouldersoundguy
Well-known member
I wonder if you're confusing advice about "reamping" with a sim and reamping for real. There's little processing needed to simply route the DI out to an amp and re-record it. Recording at 48 vs. 44.1 might offer a tiny improvement at the highest frequencies, but that range doesn't seem especially relevant for electric guitar.
Ethan Winer
Acoustics Expert
Exactly. Further, anything that will end up on a CD should be recorded at 44.1 KHz to start with. I never use a higher sample rate anyway, except when the intended medium is a DVD. That's the only time I'd use 48 KHz.
--Ethan
ashcat_lt
Well-known member
So...
Your guitar doesn't put out anything at all anywhere near the Nyquist frequency even at 44.1K. Hell, you could get everything you need from a DI guitar sampling at 22K. There might be a difference between 44 and 48 if you were recording cymbals or piccolos or something, but there is nothing to be gained there for an electric guitar.
Next, once the signal leaves the DAC it is an analog signal and even if you plug it right back into an ADC running at a different sample rate it'sgoing to sample that analog signal just like it would any other. This is not any type of sample rate conversion. What we were talking about above was assuming that you were going to use some program ITB to convert the digital files. The trip through meatspace alleviates all of the issues of SRC. Instead you have all the issues of a trip through meatspace.
If your machine has trouble at 48K, then just do everything at CD speed. If you're losing anything, it cant possibly be worth the hassle!
Edit - Though I actually totally agree with Ethan on this. I just didn't want to be the one to kick off the 26 page debate/flame war which is likely to follow.
Your guitar doesn't put out anything at all anywhere near the Nyquist frequency even at 44.1K. Hell, you could get everything you need from a DI guitar sampling at 22K. There might be a difference between 44 and 48 if you were recording cymbals or piccolos or something, but there is nothing to be gained there for an electric guitar.
Next, once the signal leaves the DAC it is an analog signal and even if you plug it right back into an ADC running at a different sample rate it'sgoing to sample that analog signal just like it would any other. This is not any type of sample rate conversion. What we were talking about above was assuming that you were going to use some program ITB to convert the digital files. The trip through meatspace alleviates all of the issues of SRC. Instead you have all the issues of a trip through meatspace.
If your machine has trouble at 48K, then just do everything at CD speed. If you're losing anything, it cant possibly be worth the hassle!
Edit - Though I actually totally agree with Ethan on this. I just didn't want to be the one to kick off the 26 page debate/flame war which is likely to follow.
T
twangbuck
New member
I'm pretty sure the thinking of the comments I had read in other posts/sites was that it was better to record the guitar's DI track at a higher sample rate in order to ensure having the highest quality track possible to feed back out to the amp. Basically in order to give the amp a signal that was as true to an actual guitar signal as possible. That way the 'reamped' sound would be as true to life as possible. One person's comment I read said he usually wouldn't even consider reamping unless the original DI track was recorded at at least a 48khz sample rate.
I'm not saying those folks were right because I have little experience in this, but I did see that comment in several places, and it seemed to make sense at the time. But I hear what you're saying in that there's just not enough detail there in an electric guitar to worry about.
bouldersoundguy
Well-known member
Maybe you should just try things and decide for yourself. That's how we used to do things before the internet.
T
twangbuck
New member
totally hear ya. but I'm sure you know how it is when you have a wife and a kid and rarely anytime alone to crank things up and hit the record button... it's nice to prepare ahead of time to maximize the available time! 
MrWrenchey
Member
I'm going to try and keep this from any derogatory comments to keep from the flame war everyone else is talking about. xD
This is basically what everyone else has been saying, but to reiterate, even if having that extra 3.9kHz in sampling DID affect the sound, it's outside of our range of hearing, and certainly outside of the reproduction of a guitar cabinet's speaker.
If I'm correct in saying so, the speakers associated with guitar cabinets and such (10" to 12") are too big and heavy to reproduce much above around 8kHz reliably. Sampling rates will capture anything up to half the sampling rate "perfectly," so having it at 44.1 vs 48 is basically giving you an improved frequency range of 22.05kHz to 24kHz, neither of which you're really going to be hearing on a guitar anyway. So having a higher sample rate really doesn't improve anything in this case.
The reason the guy was probably saying that he wouldn't reamp anything under 48kHz is probably because of what the big debates on here are always about, and that's that an analog waveform has infinite points, while a sampled one can only have a finite amount, so it's not REALLY the original signal. But as I said, I don't think it affects it, particularly in this case when you're going to be sending it out to a cabinet with speakers that can't practically reproduce all the frequencies humans can hear up to anyway. But that's my personal opinion on the debate. =]
ashcat_lt
Well-known member
Yes, that, but since he's looking to record the DI at a higher sampling rate it's important to keep in mind that the high-Z inductive nature of the pickup and the capacitive nature of the cable combined mean that the guitar itself will not put out anything above that 5k-8k range, so there's no help there either.
Farview
Well-known member
The only thing a higher sample rate gets you is the ability to capture higher frequencies. The di signal of an electric guitar wont come close to using the frequency range that 44.1k affords you, so any more is pointless. Besides, the difference between 44.1k and 48k is neglegable anyway. (nyquist of 22.5k vs. 24k)
A lot of the "record at a higher sample rate" stuff is superstition, especially when talking about the difference between 44.1k and 48k.