Converting to Higher Bitrates??....we goofed.

dudernut

dudernut

New member
My band started recording scratch guitar tracks for a record we are going to attempt. We like to record (as most digital engineers probably do) at 96k. However, we got too eager and did not realize that all 12 tracks were recorded at the default 48k. Naturally, an attempt to convert them results in a drastically sped up version. Obviously, I know that we will not be able to create audio samples from thin air, but since these are scratch tracks, the quality will work fine. Is there a way to convert these tracks to 96k without speeding them up? Is there perhaps a way to use time stretch to my advantage? I thought about exporting them and feeding them in at line level into a new 96k file, but that seems like more work than rerecording, especially trying to realign to same spot on the click track. Any ideas are welcome.
 
Your problem and its fix are DAW specific. I don't know what you're using but it sounds like you just changed the sample rate property of the project or the files and it dutifully played the files at the new sample rate. You likely need to resample the files to your preferred rate. Any 2-track editor should be able to resample.

But I think it's a stretch to say that most engineers record at 96kHz. With modern oversampling converters there is practically no advantage from such high rates, and 96k takes twice the space as 48k. Unless this is a big budget project use 44.1k or 48k.
 
You'll find plenty of people representing each side of the argument, but ultimately, there's nothing wrong with 48 at all.


Time stretching will result in major destruction of your files, and re recording brings in unecessary d/a and a/d conversion, again, doing more damage than good.

Even using a 'proper' method to convert to a higher sample rate won't result in any immediate improvement at all, or any noticeable improvement after any processing you're going to do.

I'd just stay there.
 
That's why I said probably, I really was not sure. I may just keep it at 48k if the consensus here is that it will not matter....I am using Nuendo 4, so I am sure there is a way to resample should I need to. But now I am interested in the 48k vs 96k debate. When it is mastered and placed on CD, it ultimately becomes 48k regardless, right? So what advantage does it provide? In my situation, the only possible advantage would be because we plan to reamp. It will be our first experiment with it, so I am not sure if processing the signal twice through 48k would matter....thoughts?
 
Does up-sampling (if that's the right term) even work? I would think not.

If it did, you could record stuff as an MP3, then convert it to wav and get wave quality. But you can't.

Maybe I'm totally mis-understanding the conversation and just started my own. :eek: :D
 
RAMI said:
Does up-sampling (if that's the right term) even work? I would think not.

If it did, you could record stuff as an MP3, then convert it to wav and get wave quality. But you can't.

Maybe I'm totally mis-understanding the conversation and just started my own. :eek: :D
Click to expand...

I did not expect the quality of the scratch tracks to improve...I just did not want to rerecord them to use as a reference...they sped up when I tried, but it has been pointed out that I was probably not going about it right....Still a newb as the title says :D
 
dudernut said:
I did not expect the quality of the scratch tracks to improve...I just did not want to rerecord them to use as a reference...they sped up when I tried
Click to expand...
Ah, ok. I got it now. I had a feeling I might have been a little out in left field. :)
 
dudernut said:
That's why I said probably, I really was not sure. I may just keep it at 48k if the consensus here is that it will not matter....I am using Nuendo 4, so I am sure there is a way to resample should I need to. But now I am interested in the 48k vs 96k debate. When it is mastered and placed on CD, it ultimately becomes 48k regardless, right? So what advantage does it provide? In my situation, the only possible advantage would be because we plan to reamp. It will be our first experiment with it, so I am not sure if processing the signal twice through 48k would matter....thoughts?
Click to expand...

Ultimately a CD ends up being 44.1k but it's not really a usable debate.
People say "CDs are 44.1 so why would I ever want a higher rate". That's not how it works.


Take time stretching as an example, and Rami, this applies to your question a while back.

If you record at 44.1k, 44,100 samples per second, there's a limit to how far you can stretch that before you hear a granular effect.
96 has a limit too, but should allow you to stretch further before you become aware of it.

It's the same as slowing down a 24fps video. You'll run into problems sooner than if you were slowing down a 48fps video.

Even looking at it that way, there's still a strong debate about whether the 2x storage space is justified by any increase in signal preservation.

I hope that's all correct; It's how I've always looked at it.
 
Steenamaroo said:
Ultimately a CD ends up being 44.1k but it's not really a usable debate.
People say "CDs are 44.1 so why would I ever want a higher rate". That's not how it works.


Take time stretching as an example, and Rami, this applies to your question a while back.

If you record at 44.1k, 44,100 samples per second, there's a limit to how far you can stretch that before you hear a granular effect.
96 has a limit too, but should allow you to stretch further before you become aware of it.

It's the same as slowing down a 24fps video. You'll run into problems sooner than if you were slowing down a 48fps video.

Even looking at it that way, there's still a strong debate about whether the 2x storage space is justified by any increase in signal preservation.

I hope that's all correct; It's how I've always looked at it.
Click to expand...

With time dilation, it shouldn't matter much nearly as much because unlike video, audio is a continuous waveform. With video, someone can slip between the object and the camera and a pixel suddenly shifts from green to blue, and it is a lot of work to mathematically model everything in a scene to try to make that object shift half the distance it did before.

With audio, everything is a simple change in voltage, and halfway is halfway, roughly. I mean, to do it well, you need to turn it into a function that's continuous at the first and second derivatives, ideally, but it isn't rocket science, just simple math.

Where it matters for audio is in a few other places:

  • Your sampling rate can affect what sorts of analog brickwall filters are used, and more to the point, audio interfaces that support higher sampling rates are less likely to have the brickwall cutoff at an audible frequency, choosing to make up for it using oversampling instead.
  • Most software that does DSP does not upsample the sound on the way in. When you're performing complex mathematical transforms like convolution, pitch shifting, pitch detection, anything else involving FFTs/DFTs, etc. on the signal, you'll often get more precise results if you start with a higher sampling rate. This doesn't necessarily require recording at a higher sampling rate, just running the processing chain at a higher rate.

My attitude can be summed up like this: disk space is free. Not literally, but close enough. You can get three terabytes of storage for $125. That's 132 days of continuous 24x7 recording at 96 kHz/24-bit. Since nobody records 24x7, that's likely to last any home recording enthusiast several years. Compared with analog tape of yesteryear, $125 for several years worth of media is effectively free. If you can't afford to buy one hard drive (and, ideally a second one as a backup) every few years, I'd hate to think how shocked you're going to be when you hear how much the average recording engineer's mic locker costs. :D
 
dgatwood said:
When you're performing complex mathematical transforms like convolution, pitch shifting, pitch detection, anything else involving FFTs/DFTs, etc. on the signal, you'll often get more precise results if you start with a higher sampling rate.
Click to expand...

That's pretty much all I was saying, except you said it better. ;)

You're right about storage space.
If the argument is as simple as "why not record at 96k....you have plenty of space", then sure. Go for it.
 
Point taken about storage space, but file size also impacts time to do things like transfer and defrag.

Okay, with oversampling converters, aren't the analog LPFs basically the same as with a higher sampling frequency? And then the remainder of the filtering is done with much higher quality digital filters? So the brick wall filter benefit of higher sampling frequency is much reduced as compared to the old days of 1x sampling?
 
I'd say that, for the dry tracks, there's little or no advantage to using high sample rates these days. However, as others have said, the higher resolution CAN sound better when certain effects are added.

Personally, I've taken the decision that the advantages aren't enough to outweigh the extra system overheads so tend to work at 44.1 or 48 (depending on where the project has to end up) but there might be circumstances where I'd consider 96kHz working. Maybe.
 
bouldersoundguy said:
Okay, with oversampling converters, aren't the analog LPFs basically the same as with a higher sampling frequency? And then the remainder of the filtering is done with much higher quality digital filters? So the brick wall filter benefit of higher sampling frequency is much reduced as compared to the old days of 1x sampling?
Click to expand...

That is my understanding. With oversampling, the actual sampling occurs at a frequency which is an integer multiple of the desired frequency, and will be in the MHz range. The Nyquist frequency in this case is now much higher than the audio frequencies, which relaxes the requirement of a steep analog filter. In fact, input filter of the ADC can be a simple RC filter with a corner frequency high above the audible band but well below the over-sampling Nyquist frequency (for example, corner frequency at 400kHz).

The over-sampled audio is then down-sampled to the desired sample rate with the so-called digital decimation filters.

Paul
 
Would there be a time when the default WOULD be to record at 96kHz? Just throwing some crap answer, but like 10 years from now?
 
What's the distribution format going to be in the future? We worry so much about pristine quality on our recordings and mixes then the only way others hear our product is via crappy MP3. Yeah, I'm a boring old git but I lived though decades of music technology getting better and better--mono to stereo, cheap "record players" to proper hifi, noisy tape to dolby, etc. etc. Then, all of a sudden when recordings sounded almost perfect, we stopped demanding the best and chose MP3 for quantity over quality.

A rant, I know--but the distribution (at least for me) affects the recording needs. I choose 44.1 as my sample rate if I'm going to end up on CD (or 48kHz for video) to avoid sample rate conversions later on. My rationale above for "maybe" using 96kHz would be if I knew I was going to use effects that change the duration/speed or other ones that will sound better with the higher resolution.

So? Ten years from now? Well, there a good chance I won't even still be breathing but, if I am, the mix plans and the distribution plans will affect my decision. Effects aside, I trust Nyquist.
 
Bobbsy said:
Then, all of a sudden when recordings sounded almost perfect, we stopped demanding the best and chose MP3 for quantity over quality.
Click to expand...

That's true. One thing I DO like about this whole "hipster" thing, is that they think they are being innovative for wanting .WAV files. Hell if I'm a hipster for wanting .WAV's, then I'm the biggest hipster in my college. :D

I hope it does make a turn around. Not only are people wanting mp3's, but they're not even good mp3's either. If I'm taking an mp3, 320kbps would be the only option. People are settling for 160, or lower. :facepalm:
 
I agree with Bobbsy with the trivial exception that I just do 48k because one sample rate conversion at the end doesn't bother me and I would rather my recordings sound better in the event they end up in a movie.

As far as 96k for the future, to me that's like saying you're going to start using video that captures infrared and ultraviolet in case it's needed in the future. Humans can't see those colors so why record them? To impress aliens who might arrive in the future?

Maybe 96k has some advantage with effects and processing, but it's not enough to warrant the extra resources (mostly my time) required.
 
The reduced quality of mp3 doesn't bother me either. To me it's the modern equivalent of radio. FM degrades the sound a bit at best and a lot at worst, and AM is even worse. And as much as sound got better in my lifetime often peoples' playback systems didn't always keep up. It didn't seem to prevent people from enjoying the music. People were just used to how their systems sounded and often weren't particularly impressed with better ones. There were lots of people who were perfectly happy with junky little portable record players and a few who had to have something better. Things have not changed that much.
 
I've been on 88.2 for the past 10+ years....it was just the rate I picked at the time since the debates about higher sampling rates were still pretty fresh and ongoing.

I've done stuff at 44.1 too...just 'cuz it worked out that way...and wasn't too hung up about it.
I mean, to me it's just a setting at this point, and I tend to leave it at 88.2
While questions about sound quality may continue to be discussed for years to come...for me (since I mix OTB from the DAW), the only reason NOT to record at the higher rates was for the sake of hard drive space...which I have plenty of.

For those that do it all ITB...the much higher rates might pose a CPU strain if you do lots of plugs and whatnot, and you have huge track counts...but I think with today's computers, even that's not as much of a an issue any more.
In my case, I'm still using 5-10 year old systems...but they work fine since I mix OTB out of the DAW, and the DAW just plays back files...no heavy track processing going on, just minor stuff from pre-mix editing.
 
You must log in or register to reply here.
Back
Top