Running Kemper at 96k

I'm a recording engineer, here's some of what I have observed.

For most sources and with good converters, I can't hear a difference between sample rates on an individual track; an exception would be a solo instrument in a very nice room. I get stems recorded at various sample rates to put in a master session and the original sample rate doesn't seem to make a difference, especially once it's all mixed.

As far as sessions go, I hear a difference. There are many factors - the converters, the DAW's summing, the plug ins, etc. For whatever reason, 44.1k sounds a bit harsh to me, 48k has a sweetness I like, so if I'm doing demos I can happily work at 48k.

I don't use 88.2, probably because of an outboard reverb I used to have that DID NOT LIKE 88.2k. Habits.

I use 96k for most serious sessions, disk space is cheap so I don't worry about it. I asked another engineer friend why he uses 96k, if he could really hear a difference. His answer was that he wasn't sure he could hear a difference but that "he was in a better mood at the end of the day working at 96k." That isn't a joke, I know what he means.

Theoretically, 44.1k should be all we need, but the execution of perfection is elusive!

Best...H

Quote from lightbox

And another reason would be in situations where latency is crucial (live sound e.g. in broadcast or even concerts). You just don't want a bunch of oversampling linear phase "plugins" in your signal chain considerably adding to the overall latency (and potentially even severe phasing issues between "parallel" busses. So you rather use 96kHz (if you can) instead of 48kHz on your desk and keep most of the nasty stuff happening in low latency "plugins" well above the audible spectrum ... and have the DA stage take care of that.

Phasing issues should be solved by the DAWs latency compensation. Doesn't it work?

I wonder why plugins use linear phase filters for that.

We do oversampling in the Profiler as well, as you know.

There is better and latency-free methods than using linear phase filters.

I do my cruical recording on harddisks not larger than 1 GB.

Every bit has more space on these less dense harddisks.

The result is a more airy and relaxed track. You should try it. Difference is obvious.

PS: Seagate harddisks sound the best IMHO. But I wlll do more tests.

Quote from ckemper

I do my cruical recording on harddisks not larger than 1 GB.

Every bit has more space on these less dense harddisks.

The result is a more airy and relaxed track. You should try it. Difference is obvious.

PS: Seagate harddisks sound the best IMHO. But I wlll do more tests.

????

Quote from ckemper

I do my cruical recording on harddisks not larger than 1 GB.

Every bit has more space on these less dense harddisks.

The result is a more airy and relaxed track. You should try it. Difference is obvious.

PS: Seagate harddisks sound the best IMHO. But I wlll do more tests.

Too funny Christoph.

There was a dude over at GearSpace who insisted that if he compared a raw, original WAV file with the same one sent over the internet to him, he could hear a difference. Something bad must happen in the digital aether, he figured.

People eventually managed to convince him that the bits don't lie and to therefore trust the byte-size comparison and null test they talked him into doing. Even after doing the null test and getting a perfect 0dB result, he still didn't want to believe that they were the same... for a while. That's confirmation bias if ever I saw it.

One small thing:

When the mastering engineer said in the video I linked to that peeps couldn't reliably distinguish between a 320kbps MP3 and a WAV / AIFF delivery mix, I'm sure some would instantly think, "Hang on, I can easily tell!" These peeps are surely using sub-par conversion methods, such as iTunes. Even at 320kbps, iTunes has been tweaked to keep people happy and therefore compromises a little on quality in order to be able to render very-quickly.

I could easily hear the HF clarity loss and smearing of transients when I bothered to compare... 2 years after converting my entire iTunes library using the built-in algorithm. I could've kicked myself! I started all over again back in 2011 using L.A.M.E. at 320kbps, CBR and true stereo (not interleaved), and couldn't be happier.

So, if anyone's naturally-sceptical about that M.E.'s claims, I'd predict that he or she hasn't heard the conversion done properly. Consumer-oriented processes such as those found in iTunes are tailored for the masses, something that almost always isn't appropriate for we audio-engineer types, as I'm sure you'd know.

Quote from ckemper

Phasing issues should be solved by the DAWs latency compensation. Doesn't it work?

Read again, Christoph I was talking about live sound (e.g. broadcast or live shows), not studio work. I was just adding another real world application where higher sample rates like 96kHz do actually make sense. Maybe you got confused from me using the term "plugins" (in quotes).

My comment wasn't related to the Kemper Profiler by any means.

Quote from ckemper

PS: Seagate harddisks sound the best IMHO.

Hahaha, that was hilarious. Thanks for the good laugh and good luck with your tests with Seagate. As you certainly know: "Sie Geht, Sie Geht Nicht"

I use solid state hard drives, just back up to spinning discs, as a result of this thread I have a mountain of analysis and a/b listening to do, without having to a/b my disc drives!

Quote from lightbox

Read again, Christoph I was talking about live sound (e.g. broadcast or live shows), not studio work. I was just adding another real world application where higher sample rates like 96kHz do actually make sense. Maybe you got confused from me using the term "plugins" (in quotes).

My comment wasn't related to the Kemper Profiler by any means.

Granted!

My point was, that the plugin industry doesn't make their homework and forces the users to go for higher sample rates.

Making non optimal plugins will feed the myth that high sample rates are advantageous in general.

The Profiler is an example how it would work.

The homework for them is:

- Use oversampling where it's necessary for technical reasons (as described in the Fabfilter video). If optional, tell the user the reasons.

- Don't use linear phase but minimum phase filters for up- and downsampling.

There is also a myth that linear phase filters are better for this purpose, but again the difference is not noticable.

Linear phase filters are usefull for audio signal editing, however. (But not even there in every case)

Quote from ckemper

I do my cruical recording on harddisks not larger than 1 GB.

Every bit has more space on these less dense harddisks.

The result is a more airy and relaxed track. You should try it. Difference is obvious.

PS: Seagate harddisks sound the best IMHO. But I wlll do more tests.

???

Quote from ckemper

I do my cruical recording on harddisks not larger than 1 GB.

Every bit has more space on these less dense harddisks.

The result is a more airy and relaxed track. You should try it. Difference is obvious.

I'm surprised that such an educated person would overlook the obvious value of storing your bits in the cloud. Infinite room per bit, and the very definition of airy.

Also, since everyone listens to music on the Internet these days your bits would be properly conditioned for streaming, and this reduced friction would help to minimize latency.

Quote from wwittman

and I’m saying YOU can’t hear the difference.

I wish we could get together in a nice studio and play some 44.1K tracks and some 96K tracks and have you tell me which one is which. I'd believe it when I witnessed it. So far I have never seen anybody that could do it, but hey, maybe you have super advanced ears. Agreed, I can't do it, and I would bet a good sum money if I could that you couldn't either. Even small differences in the quality in the recordings would be way more obvious.

I've recently come to the conclusion that digital audio quality is a moving target. This week I had to provide a remix of a song originally recorded back in 2006.

The song was recorded on a motu 896 with spl preamps. I used a handful of waves plugins at that time and minimal eq, L2, ren vox, ren comp, ren verb plate.

I was a cubase user then and I still am now.

I loaded the project up, swapped in a few plugins that were no longer supported and mixed down to see how it compared.

The original premaster sounded like it was playing inside a cardboard box vs this one.

I then worked through it track by track using all the new tools I have and remixed, again a massive step up in quality.

I was always aware of the woolly sound back then, that's why I invested in a lynx aurora and all new pre amps and mics. These improvements were all noticeable. The motu had definitely sounded like a step backwards from the akai I used before it and the tape before that.

But now I listen and think it must be the recording engine that has taken significant steps forward, the source was fine, even the mix, no eq changes, like for like limiters, compressors and verb, it's all digital, but sounds infinitely better now. We have already established that the original digital stereo file couldn't have gone off like tape, but the export process and handling of the audio files must have improved massively over the last 15 years.

Is there a point to this observation, I suppose its the consideration about recording things and archiving them in the best quality we can, maybe it's not sample rate but bit depth, maybe both, because we don't know what technology will be available to us in 15 years time.

I can confidently say that the technology for us to hear well above 20kHz won't be there.

Even if it were, via, say, hearing aids that boosted said HF component of the signal, why would anyone bother putting up with the expense and discomfort for so little return?

The 16kHz area has the psychoacoustic effect of lightening the mood / bringing about positivity, or so I was taught back in the '90s by a professor of 3 disciplines. I conducted an informal experiment at a gathering back then that seemed to confirm this and learned later on that some local nightclubs boosted that area of the spectrum by-default.

Why am I saying this? Well, that's the only gratifying HF tweak I've heard of that "the masses" / consumers might benefit from. Sure, there may be others, but somehow I doubt they'd be above 20kHz.

So, whatever improvements in digital audio come along down the track, the ultrasonic stuff simply won't matter. It never did and never will because we can't hear it and don't care... because we can't hear it. As you speculated, mixing engines and plugins have been where the real improvements in the last 10 years have occurred (IMHO), but I think we're already beyond the point of diminishing returns.

I therefore suggest we all try to appreciate what we have now 'cause, especially to the layperson and to a great extent us, future improvements are going to be very-tough to discern. All IMHO of course, YMMV.

I’ve followed a good portion of what’s been written so far. People making their living and people scientifically studying the capture and reproduction of sound have weighed in.

I am not part of that group.

As a spectator (and a comparative layperson) in all of this, one question came to mind when looking at audio resolution and the reproduction of recorded music. Perhaps I’m missing something, so point it out if I am.

If 192 is ‘better’ than 96 and 96 ‘better’ than 48.....why then has the lowly vinyl record been resurgent with listeners?

Capturing is one thing, but when placed on vinyl the highest resolution music recordings seem become nearly pointless. Not to mention fly in the face of this discussion from a listeners point of view.

Well, vinyl sounds great in frequencies that actually matter, Brother Ruefus - ones we can hear. You're not going to get a 20kHz response out of any vinyl LP on the planet.

So that puts the frequency debate in some perspective I guess.

It's instructive to also point out that the dynamic range of records is terrible (somewhere in the 60->65dB range IIRC); even good-quality cassette tapes can beat it. 16 bits (ye olde CD) delivers a theoretical dynamic range of 96 dB and yet people love listening to vinyl, many even for classical music, with only two thirds of that.

Anyway, these days you're lucky if you get 15dB dynamic range out of a commercial release thanks to loudness maximising.

This video explains why sample rates above 16 bit 44.1khz is a waste of time and energy. Nyquist is already dealt with using this sample rate so any higher than that is not going to improve things and most likely will result in diminished returns.

Quote from Ruefus

If 192 is ‘better’ than 96 and 96 ‘better’ than 48.....why then has the lowly vinyl record been resurgent with listeners?

Just for fun, let me try to give you a layman's example/explanation:

Digital recording adds a rigid wall somewhere behind the end of your playing field. At 48kHz the wall is "24 meters" from the opposite baseline. Say you're right in the middle of the playing field and you throw a ball at the wall. Obviously it will bounce back from the wall towards you.

In analog audio, there is no wall. Throw your ball the same you did before ... the ball won't bounce back, it will just keep going away from you until its energy is gone. It will not come back and disturb what's going on on the playing field

Now back to digital audio ... At 192kHz the wall is 96 meters from the opposite baseline. Throw the ball towards the distant wall with the same energy like before. There's a good chance the ball doesn't have enough energy to bounce (or roll) all the way back to you.

Now finally imagine all players on the field constantly throwing balls at the wall. You can certainly imagine that the "reflected" balls from the "48kHz playing field" coming back from the wall will have a negative impact on what's going on on the field. They add unwanted chaos, especially in the upper part of the playing field.

Hope it's at least funny to read this but I think it explains the issues with the sample rate creating a rigid wall inevitably reflecting everything that wants to go beyond the wall.

Quote from lightbox

Just for fun, let me try to give you a layman's example/explanation:

Digital recording adds a rigid wall somewhere behind the end of your playing field. At 48kHz the wall is "24 meters" from the opposite baseline. Say you're right in the middle of the playing field and you throw a ball at the wall. Obviously it will bounce back from the wall towards you.

In analog audio, there is no wall. Throw your ball the same you did before ... the ball won't bounce back, it will just keep going away from you until its energy is gone. It will not come back and disturb what's going on on the playing field

Now back to digital audio ... At 192kHz the wall is 96 meters from the opposite baseline. Throw the ball towards the distant wall with the same energy like before. There's a good chance the ball doesn't have enough energy to bounce (or roll) all the way back to you.

Now finally imagine all players on the field constantly throwing balls at the wall. You can certainly imagine that the "reflected" balls from the "48kHz playing field" coming back from the wall will have a negative impact on what's going on on the field. They add unwanted chaos, especially in the upper part of the playing field.

Hope it's at least funny to read this but I think it explains the issues with the sample rate creating a rigid wall inevitably reflecting everything that wants to go beyond the wall.

Display More

Ok. This helps explain at a high level, how digital differs from analog.

If I understand, the key difference is analog doesn’t ‘reflect’ what it can’t capture, whereas digital does - and in doing so creates artifacts when that data has nowhere to go (or can’t be accurately captured)?

Am I close?

Quote from Ruefus

Am I close?

Yes, you are very very close

Now when you think a little further from that point:

It's not so much about the "capturing" only. It's also about audio processing once you're in the digital domain.

To use my previous layman's example:

Even if all players agree to stay inside the playing field at all times ... once the game starts and the players want to make the game interesting, they will likely ignore the rule and e.g. hammer the ball towards the goal. Damn, that shot went over the goal to the stands behind the goal. But hey, it looked great and made the game exciting.

In audio terms, you often want to create e.g. "saturation" to color the sound, make it more exciting. This automatically creates frequencies beyond the "agreed playing field" ... and you need to take measures to prevent that from happening.

One way to do that is to enlargen the entire venue dramatically (aka 96 or 192kHz) and only take care of the remaining unwanted effects at the very end of the digital processing chain .... or .... make sure that each individual "saturation stage" plays by the rules and stays within the agreed limits. The latter requires time and calculation power which results in added latency.

This added latency doesn't matter in post production but can be a serious issue in live production where latency must be kept to a minimum.

Was this helpful as "the second stage" of explanation?

In short. If you use plugins that create saturation (harmonics) which many people do, and those plugins are not coded optimally (which some aren't - but you might still like them for features etc.) then higher sample rates can be very relevant for you. And audibly superior.