Record at 48kHz Mix down 96khz

[msmcleod]

I don't see any downsides to recording at 48K or even 44.1K.  The vast majority of people over 21 are not going to hear any of the harmonics above 22K.

The issue comes with using any effects that add harmonics to those recordings.

As pointed out in the video I posted earlier, unless the plugin employs oversampling, errors can occur when a plugin tries to generate harmonics of a higher frequency than the sampling rate supports, resulting in lower frequencies being added in the audible range.

I did a test myself using an 8KHz saw tooth wave and running it through various effects such as saturation or exciter plugins, and then on to an analyser plugin (Melda MAnalyser).

Without oversampling, frequencies as low as 50Hz to 400Hz were being generated, which on an 8KHz signal is pretty crazy. 

Upping the sample frequency, or using oversampling (which most of the Melda plugins support) reduced this significantly. Using the "x2" button within Cakewalk also produced a significant improvement.

Now I can't say I could hear a clear 400Hz signal in the non-oversampled version (unless I compressed the hell out of it and wacked the level up), but I could definitely hear the difference in quality between the oversampled version and the non-oversampled version.

That being said, we all use our ears when tweaking the parameters of these effects... so if it sounds fine to our ears, then it's probably not a big deal.

[azslow3]

6 hours ago, Gswitz said:

So this shows that through this process, most of the audio was retained. This defends that the RME UCX will send the higher frequency content to your speakers.

If you were questioning whether my speakers do a good job up there... haha... not worth testing. ? I have cheapie charlies... but I'm sure you can find speakers that attempt it anyway.

As an interesting aside, while the first and final spectral analysis bits are almost identical, at one moment you see a tiny jump at 40 (otherwise quiet between 20k and 40k like the original). So, not perfect null test as I said. Something funny in there.

No other gear or interface in the middle. Out of the RME UCX on channels 1 and 2, in on channels 5 and 6. No pres on those channels.

Because that final 40 was after slowing it back down, I suspect something to do with the loop construction view slowing algorithm. Who knows what. ?

I have to apologize. I have used the word "driver", but I had to explicitly write "speaker driver". My fault.

Thanks for making the test. So, RME works fine with high frequencies, recording them correctly and delivering them correctly to the output.

My point was about the next step, what happens with these frequencies in your or my  "cheapie charlies" or super "high end" towers? They can be discarded prior speaker drivers or try to drive them. In the last case, they will produce some distortion. Is it "aliasing" as in digital domain or other kind of noise? I have not seen plausible test results so far. But I have seen a test with general conclusion "there is audible distortion" (which is sometimes used as a prove for "better sound" in "high end", when used with 96kHz source... so in reality the sound, while really different, is just worse...).

Returning to the topic. If we mix in 96kHz and send output without LPF to our monitors, can we fall into the same "trap" as "high end" lovers? So we listed the mix with some extra audible distortion produced by monitors from existing (and may be even real world correct) high frequencies?

 

[sjoens]

Heh heh... More important than kHz is, are you using gold plated cable connectors? 

Edited January 23, 2020 by sjoens

[azslow3]

48 minutes ago, sjoens said:

Heh heh... More important than kHz is, are you using gold plated cable connectors? 

I use Kimber Kable Hero... Not intentionally, the reason is in fact very sad...

But I have seriously tried to listen/measure the difference to "normal" XLRs. So far without success ?

[azslow3]

Quote

...there is no single thing that makes a good cable, it’s a combination of factors including the way that, “materials affect ranges of frequencies in dissimilar amounts”.

 

[micv]

14 hours ago, msmcleod said:

Using the "x2" button within Cakewalk also produced a significant improvement.
 

Never knew where the 2x really apply until now.

If I could sum up from a usage perspective:

- No real benefit if export a project recorded at a lower rate to a higher resolution, since the project is processed at the low rate.

- For plug-in that produces harmonic (distortion or saturation type plugs), higher resolution will reduce the anti-alias or artifacts that got 'bounced back' into the hearing range.  To improve this, use Cakewalk "2x" button AND the plug-in options "Upsample on Playback - Upsample on Render".

- To mix an entire project at a higher resolution, have to create a new project at the higher rate and import the individual tracks in.

 

[Gswitz]

I think that's it in a nutshell, @micv.

[Noel Borthwick]

On 1/21/2020 at 11:15 AM, micv said:

This is zero'ing in on what I don't yet understand.  The project is at 48k, when I export to96k, my soundcard got set to 96K, what rate is the project now?  I thought that the entire project would be up-converted to 96k, but if what it does is process at 48k and just simply up-convert the 'bounced' audio to 96k, then I can see that there is no real value.  If that's the case how would you actually mix that the higher rate?

Picking a  higher sample rate in the export dialog will not help at all if the original project is at a lower sample rate. The project isn't converted to 96k.
All the export is doing is upsampling the FINAL output of the render, essentially processing and mixing all tracks at the lower sample rate and then converting the final result. This isn't very useful other than producing a file at 96K. 

If you want to take advantage of upsampling in plugins you should use the the plug-in upsampling feature which will add a upsample -> process -> downsample stage to all plugins you choose to upsample.  Or have all the tracks in the projects at 96k.

[mettelus]

^^^^ circling back to the OP, that would be the "answer"

[micv]

You guys are the best.  This forum is the main reason I stay with Cakewalk for so long.  thanks

[Adrian Bartholomew]

Sample rate has nothing to do with headroom.  Bitrate does.

Sample rate defines the highest frequency possible to be represented, which is theoretically half the sample rate.  For e.g., 44.1kHz can only reproduce as high as 22.05kHz.  In reality, the source is filtered way below that to avoid aliasing, the same phenomenon found in video where the wheels of a car appear to be going backward.  In audio, this is represented by ultra low nasty frequencies.

Bitrate defines the number of volume levels represented.  For e.g., 1 bit audio would contain zero dynamics.  Either full volume or silence.  A 2bit converter produce 3 volume levels or silence.  The noise floor would be very nearly as loud as the material.

Compare those to 16bit - 65,536 levels. 
24bit - 16,777,216 levels, which is 256 times more granular (headroom) than 16bit!

Though 16bit audio is quite sufficient for human psychoacoustics (the smallest difference in volume that the human brain can discern), mixing 16bit audio files, where most are attenuated, thus not achieving their full dynamic range, results in a mixed audio file way below 16bit.  No one mixes a project where all tracks are placed at 0db nominal level, the only way to maintain a 16bit mixed audio file from 16bit audio tracks.
Hence the value of 24bit audio files.

Sorry for some of the irrelevant info - just wanted to give a fuller picture of the subject.

 

Edited February 25, 2021 by Adrian Bartholomew

[John Vere]

You realize this thread is over a year old.  ?⏰

[sjoens]

Age matters not. ?

[Noel Borthwick]

On 1/21/2020 at 9:11 AM, Gswitz said:

This is true of the bounce happened at a higher sample rate, but i don't think that is what the op is getting.

Project at 48...

Saturation plug

Export audio to 96...

...

This will result in the saturation being processed at 48 then the result converted to 96 with badness printed. Yes?

....

On the other hand, using cakewalk up-sampling to 96 for the plug and exporting audio to 48 avoids the badness of the plug.

...

No harm is done exporting to 96, but unless i misunderstand, i don't think there it's much potential benefit. Perhaps if he has an interface from the 2000s the higher rate might reduce issues related to the interface.

This is exactly right. If you want to benefit from higher sample rates for plugins or synths you have to use the upsampling feature.
You can set the plugins to Upsample on Render and they will automatically be converted only when exporting/bouncing.

Increasing the export sample rate won't do much other than upsampling the final output of your mix. So no sonic advantage.

 

[Craig Anderton]

Having spent a LOT of time on researching this subject (and props to Noel for not blowing me off and figuring out how to upsample), I've found out certain things that I'm going to think are true until shown otherwise...

• Improvements from recording at 96 kHz are due to  minimizing foldover distortion, which can be an issue with soft synths and amp sims. Limiter performance can also improve at 96 kHz.
• A synth that oversamples in real time is at a disadvantage compared to doing it offline.
• I've recorded many classical (i.e., acoustic-only) sessions at 96 kHz. No one could tell the difference between the original multitrack masters, and a  44.1 kHz CD. 
• If foldover distortion is in recorded audio, there's no way to get rid of it.

Interestingly, when the Panasonic DA7 digital mixer appeared, I was sure the EQs sounded better than other EQs I was using in various DAWs and digital mixers.  The high end sounded "sweeter." Now, I'm not a wine-tasting kinda audio guy ("I find the Blurfle EQ pert, yet unassuming"), so a difference has to hit me between the eyes for me to think it's real, let alone important. But I couldn't shake the feeling something about the DA7 EQs was better.

At a tradeshow, I cornered one of the Japanese engineers who worked on the mixer He said they doubled the sample rates for the EQs, so while the mixer was running at 44.1, the EQs were running at 88.2. It made to sense to me that's why the high end sounded better.

 

 

 

 

[sjoens]

So if this is something not published in a spec sheet or feature list how would one know if his unit's EQ's sample rate is doubled?

Edited February 26, 2021 by sjoens