4X even 8X.... hell 16X Oversampling (Upsampling)

[Teegarden]

@LittleStudios, jJust missed your latest post. 

In your new video again a crazy amount of difference...
(Cool tool that Metaplugin, didn't know it). 

My remarks still stand...

Edited October 6, 2021 by Teegarden

[LittleStudios]

12 minutes ago, Teegarden said:

You did not test higher project sampling rates. In your test it seems like just 2x oversampling at 24bit/48kHz is enough to even out the artefacts. So how much more oversampling is still worth the resources (88.2, 96, ...192 kHz with 2x, 4x,...oversampling). How do we know where to stop? (just guessing that not everybody wants to play hours with sinus waves in order to find out...)

Don't be fooled by the icon to enable oversampling, "2X".  Yes by default it is 2X oversampling.  I have manually set my sample rates on a per plugin basis.  My T-Racks plugins are maxed out at a sample rate of 192 KHz.  So the test demonstrated in the video had plugin oversampling at 192 KHz (4X at a project sample rate of 48 KHz).  Be aware that Cakewalk doesn't account for latency introduced by enabling plugin oversampling went running on a parallel track.  I need to test two identical tracks, but one with oversampling and see if Cakewalk manages this scenario correctly.

[Mark Morgon-Shaw]

7 hours ago, LittleStudios said:

Now, that is a single tone sweeping through the frequencies.  Imagine cymbals, vocal air, crunchy guitars, synths all contributing to this effect.  You'll end up with a lot of unwanted noise folding back into the audible range.

I never said you can't prove it exists , what I said was it's moot because 99.9% of the listeners can't hear it - hell ...most of them can't tell 128k MP3 from a CD.

[Starship Krupa]

@LittleStudios, can you try your experiment but turn off upsampling at render and instead set your export rate to 88.2 or 96?

I tried my experiment with playback in my MIDI-only project, and exporting at 88.2 gave pretty much the same psychoacoustic results as upsampling all the plug-ins at 2X. I know that you're trying to avoid the larger rendered file sizes, so just as an experiment.

A possible benefit to rendering at a higher sample rate is that any ProChannel plug-ins will benefit. And those include an 1176 clone, a console emulator and a saturation effect, all possibly able to generate frequencies beyond the Nyquist.

And Chris, thanks again for opening this topic up. As you know, I was skeptical, but I can hear a difference.

[Starship Krupa]

@Teegarden, you were curious as to whether the performance hit results I saw were specific to my vintage laptop.

I repeated my playback experiments with running the project at 88.2 vs. upsampling all the plug-ins, and same thing, switching the rate to 88.2 incurred less of a performance hit than 2X'ing all the plug-ins.

And, when I played the project at 88.2 and all the plug-ins 2X'd, I got the same "bum note" in my bass arp track as I did when rendering that way.

1 hour ago, Mark MoreThan-Shaw said:

I never said you can't prove it exists , what I said was it's moot because 99.9% of the listeners can't hear it

Did you read my post where I tried various forms of higher sample rates (plug-in and project)? I must be in the top.1%, because I can hear a difference, and to me it's not even that subtle.

Yes, I agree that after going through <256 lossy conversion, the differences probably wouldn't be audible, but I buy FLAC's on Bandcamp and purchase uncompressed or FLAC albums and songs from sites like HDTracks (OMG, Radiohead's Moon Shaped Pool at 48/24). This is becoming more of a trend.

If you look at the Cakewalk documentation, one of the benefits of higher sampling rates is said to be "Phase shift is drastically reduced." p. 976 of the Ref Guide.

Now I've speculated about how it can be that two power amps can sound different, how it can be that there's so much of a difference in detail and image. In my studio, I have two power amps for powering passive monitors. One is a Crown D60 amp from the '70's, designed for radio station studio monitoring and other pro audio uses. It's "only" 30 watts per channel. The other is an Alesis RA-100, originally designed and sold as a mate to the Alesis Monitor Ones that make up one of my monitoring systems. It's rated at 70 watts per channel. I tried an experiment where I switched my Monitor Ones and my Boston A70's back and forth between the two, and the difference was stunning. I had another musician friend in the studio at the time and he was also blown away by the difference. He's by no means any kind of audio freak, but he described the speakers being driven by the RA-100 as "squeezed" or "choked." or just "smaller." Even in mono, the Crown had a more vivid image.

I suspect that the Crown was designed with a lot of attention to phase distortion and maybe group delay. I think something similar goes on with lossy compression and plug-ins that may sound optimal at higher rates.

So why would lower rate lossy compression sound "2D" rather than "3D." Theory says that since humans can't hear anything above the frequency range, it's not possible to hear a difference. My (fully admit self-educated) theory is that there are other factors like phase and group delay that get messed up. It's not just about frequency response. @bitflipper suggested that I read Fletcher and Munson to get a better understanding of it, but I haven't yet.

Another factor is what genre we're talking about. I'm in awe of artists like Tipper and Telefon Tel Aviv, who create detailed sonic immersive spaces. There's a higher percentage of their fans who know that their stuff will sound better in lossless and with bit perfect playback, on a system with less phase distortion, etc. I'd like to impress those people too. Even for people who don't listen "critically," I think the extra detail can come through subliminally.

[LittleStudios]

@Starship Krupa

One thing to keep in mind is that for every upsampled plugin,  those plugins must be downsampled back to the project sample rate.  This amounts to twice the work per plugin.  If audio quality is your goal and you want use higher sample rates, your approach is driven by your system limitations.  I see two ways of looking at plugin level upsampling:

1. You don't have a powerful processor.  Only upsample the plugins causing an audible difference.  If you apply upsampling to all your plugins,  your CPU is doing twjce the work it would if you simply set your project sample rate higher.

2. You have limited drive space. By upsampling your plugins instead of your project sample rate, the audio files contributing to your project can be smaller in file size, saving drive space.

Another thing to consider about upsampling is a plugin's max operating sample rate.  If your project sample rate is 96 KHz and you upsample a plugin that tops out at 96 KHz, going to 192 KHz could cause issues.  Luckily you can tailor the individual upsample rates in the AUD.ini file.

[Starship Krupa]

@LittleStudios, The odd thing about it, which @Teegarden points out, is that the Cakewalk documentation specifically says that plug-in oversampling on playback is supposed to be less of a processor hit than running at 88.2 or 96. The results of my testing (albeit on older systems) indicate otherwise.

I'm not so concerned about the size of the audio files; 500G SSD's go for about $50. I have a 3TB backup drive. Of course, reading them is potentially another performance hit, but I have an SSD, so I'll just have to see.

My current strategy is that I'm going to enable 2X on most of my plug-ins (except the Meldaproduction and other ones that have internal sampling) during render only. At this point, with the trailing edge hardware I have, it's more important to me have stutter-free playback while mixing than it is to have the quality bump. At render time, it doesn't matter, I can crank my buffers up to 200 and let it churn.

Then I can be pleasantly surprised after rendering at hearing more detail in the finished mix. ?

[LittleStudios]

1 hour ago, Starship Krupa said:

@LittleStudios, The odd thing about it, which @Teegarden points out, is that the Cakewalk documentation specifically says that plug-in oversampling on playback is supposed to be less of a processor hit than running at 88.2 or 96

I think the Cakewalk documentation may need a refresh.  Could also be that the authors of the documentation  were making the assumption that the user would only enable plugin upsampling on a handful of plugins.  Which I think makes sense.  You would think that if they intended on the user enabling upsampling on a bunch of plugins they would have come up with a more useful user interface for such a task. 

Edited October 7, 2021 by LittleStudios
grammar

[Starship Krupa]

5 minutes ago, LittleStudios said:

Could also be that the authors of the documentation  were making the assumption that the user would only enable plugin upsampling on a handful of plugins.

That must be it. Find by trial and error which ones make a difference and only upsample those. The thing is, in my testing, the ones it made the most difference with (A|A|S Player, Phoenix Stereo Reverb) are ones that really challenge the audio engine. My Plugin Alliance fx seemed happy either way, and didn't really impact the engine either.

[LittleStudios]

I just had an interesting thought.  Using the AUD.ini you can set specific sample rates to oversample at for each plugin.  Knowing that not every plugin has the same maximum operating sample rate, can we define a maximum sample rate for each plugin in the AUD.ini.

For example, let's say I set my project sample rate to 192 KHz.  Well, the Waves J37 tape plugin tops out at 96 KHz.  If I set the plugin upsample rate in the AUD.ini file for the Waves J37 to 96 KHz and enable plugin upsampling, will it, for lack of better words, downsample the Waves J37?  Basically project running at 192 KHz and the Waves J37 running at 96 KHz.

Would this work?  Is this even necessary?  Does Cakewalk recognize the maximum sample rate of a plugin and not exceed it? 

Companies like Waves are inconsistent with the maximum sample rates their plugins run at: 

Codex Wavetable Synth     96 kHz

dbx® 160 Compressor / Limiter     192 kHz

DeBreath     96 kHz

DeEsser     192 kHz

Doppler     96 kHz

Dorrough Stereo     192 kHz

Dorrough Surround     192 kHz

Doubler     192 kHz

DTS Neural™ Mono2Stereo     48 kHz

DTS Neural™ Surround DownMix     48 kHz

DTS Neural™ Surround UpMix     48 kHz

Dugan Automixer     96 kHz

Dugan Speech     96 kHz

Eddie Kramer Bass Channel     96 kHz

Eddie Kramer Drum Channel     96 kHz

Eddie Kramer Effects Channel     96 kHz

Eddie Kramer Guitar Channel     96 kHz

Eddie Kramer Vocal Channel     96 kHz

Electric 200 Piano     96 kHz

Electric 88 Piano     96 kHz

Electric Grand 80 Piano     96 kHz

Element 2.0 Virtual Analog Synth     96 kHz

EMI TG12345 Channel Strip     192 kHz

eMo D5 Dynamics     96 kHz

eMo F2 Filter     192 kHz

eMo Generator     96 kHz

eMo Q4 Equalizer     192 kHz

Enigma     192 kHz

F6 Floating-Band Dynamic EQ     192 kHz

Flow Motion FM Synth     48 kHz

[Mark Morgon-Shaw]

22 hours ago, Starship Krupa said:

Did you read my post where I tried various forms of higher sample rates (plug-in and project)? I must be in the top.1%, because I can hear a difference, and to me it's not even that subtle.

It's not that subtle to me either, but I spend hours every day listening to audio critically. It's what I do.

The rest of the family wouldn't tell the difference, they are casual listeners who consume music on their phones, laptops or smart speakers or possibly ICE system. I don't even think my teenagers have ever heard a lossless format.  They are typical listeners and they will never ever hear the difference in a mix.

However if you give my wife a photo which looks perfectly good to me she will point out if the focus is soft, pick up on chromatic abberations, converging verticals and blown highlights.  

As experts we focus on tiny details which are insignificant to the uncaring masses. Hence..moot. 

 

 

[Starship Krupa]

4 minutes ago, Mark MoreThan-Shaw said:

As experts we focus on tiny details which are insignificant to the uncaring masses. Hence..moot.

I suppose I'm interested in producing sounds that other critical listeners can enjoy. And the others can get the gist, or maybe even experience it subliminally. Being able to pick out detail and soundstage in a recording is an exhilarating experience, even an emotional one. I've read people on hi-fi listening forums use the term "tearjerker" to describe the experience of listening to a favorite recording for the first time on a really good system with bit perfect playback. I've experienced it myself. It has an emotional impact, realizing how much is in there that I had missed.

Just because most casual listeners won't notice it doesn't mean that nobody will hear it. And even for them, I do believe that aliasing and compression artifacts can lead to ear fatigue.

Radiohead's "Everything In Its Right Place" is the song I use to test sample rates and compression, and it has so many little hidden things in it that I can only hear with a good setup. Tiny little reverse reverb on the lead vocal. Little sound snippets hard panned. When I first got a bit perfect setup going, I called my housemate down to listen to the song, and he did literally get tears in his eyes. It's one of his favorite songs.

I'd like to create music that can help people have that experience, that rewards critical listening. It creates joy in me, and I'd like to do that for other listeners.

My ambitions are humble, I guess. I'd just like to get my songs up on Bandcamp. Bandcamp has a really good sounding CODEC for their website, and they featured FLAC downloads. I spend money there because I can get my music in the format I prefer.

So, to me, not moot. I'm not producing music aimed at teenyboppers (although I love it when my friends' kids listen to my stuff and like it). People don't stay teenagers forever (in numbers, anyway), and perhaps when they get older and we don't need to compress music to fit a thousand songs on a phone, they'll learn to listen closely. Whatever I put out there, I don't want it to sound crappy in 20 years when (I hope) playback systems will have advanced. My stuff may be forever buried by the sands of time, but who knows.

[Mark Morgon-Shaw]

12 hours ago, Starship Krupa said:

I've read people on hi-fi listening forums use the term "tearjerker" to describe the experience of listening to a favorite recording for the first time on a really good system with bit perfect playback. I've experienced it myself. It has an emotional impact, realizing how much is in there that I had missed.

Yes, I've listened to HiFi setups that cost in excess of £100k  and it's a completely different level of experience. But the demo recordings they use are not by people making music in their spare rooms. They're generally top class musicians recorded in world class faciilites with high end equipement playing great music.

Not Jon from Basingstoke in his shed making sub par Prog drivel on his laptop so he can get 7 likes on Facebook.

So it's all relative.  Songwriting - Performance - Production - MIxing - Mastering all trump recording media and  delivery format. I can get more enjoyment from  a great song that was recorded 70 years ago in mono , over an AM radio  than virtually every home produced self released track I've ever heard. And I've heard a lot.

I agree that N'th degree improvements can wring every last ounce of detail if you have the ears and the equipment top hear it , but if it's not top notch material to begin with then nobody to whom it might matter will be listening.