Sine energy experiment (Sound Design / Master)

[RobC]

I earlier mentioned topics alike, but was lacking time to do anything more than theorizing.

This would be about creating an experimental (pre)mastering standard and to see how it works.

It was inspired by the fact that everyone has a different hearing spectrum, thus people find some sounds more annoying than others. That it seems to be not just dependent on the sounds' harmonics (if at all), but that they don't match an individual's hearing spectrum at all, having crazy boosts for some.

I understand, that there's the 'natural sound reproduction' argument, but if there's anything surprising I noticed, then it's: when hearing anything equalized "flat" to my ears, possibly anything sounds wonderful. It opened a whole new world for me, and there's a reason why I'm hooked on researching it more.

The base idea was and still is to split any sound's frequency spectrum into multiple frequency bands, then take static sine waves at the center frequency of the said bands, and set each band's audio level to match the respective center frequency sine wave's level. When I tried a quick test the other day, I noticed that it didn't really damage dynamics - though I feared it might.
Doing something like this, would result hearing an average loudness similar to said sine waves', while still remaining dynamic.
Now, equalizing our system until we hear each sine wave equally loud, would result in hearing the frequency spectrum of the processed sound more or less equally loud, too.

Not only does this create a pleasant sound in my opinion, but also it would open many new possibilities.

My question is, if I would work this idea out in a slight publication-like fashion, with audio experiments / tests, etc. would anybody read and maybe try it?
It would be all open of course.

I'm mostly interested in people who are open for experimentation and are willing to forget for a moment what's right and what's wrong by whatever standards.

[Loque]

I do not get it what you had done. While your experiments sound interesting and i tried something similar not so long ago, i need to ask...You adjusted the gain of several bands to a sine playing in that band? Does not that create a "flat" eq curve in the end? That sounds pretty boring (as my own test told me). What did i miss here? Where is the difference?

In things of frequencies, i actually experiment a lot with disharmonics, frequency/pitch shifting and different wave-form-fundamentals (which have a HUGE impact to the sound). What you have written here, made me think of stuff like, why not limit this to several frequency bands? Got to check that tomorrow...

[RobC]

I do not get it what you had done. While your experiments sound interesting and i tried something similar not so long ago, i need to ask...You adjusted the gain of several bands to a sine playing in that band? Does not that create a "flat" eq curve in the end? That sounds pretty boring (as my own test told me). What did i miss here? Where is the difference?

In things of frequencies, i actually experiment a lot with disharmonics, frequency/pitch shifting and different wave-form-fundamentals (which have a HUGE impact to the sound). What you have written here, made me think of stuff like, why not limit this to several frequency bands? Got to check that tomorrow...

You sort of did the same thing, but it's also important to create a master equalizer, which you set so that you hear each sine wave equally loud - on the given system you listen, with your own ears of course. That's what sounded awesome to me.

A simple flat spectrum on its own is still not enough, cause the system and our hearing still changes the final curve we hear.

In my opinion, the characteristics of a sound are far more important than near transparent effects that only change the spectrum itself. Thus I'm rather for using different modulation types (especially FM), and deforming the sound, too, before even looking at things such as filter effects and alike.

[RobC]

Does not that create a "flat" eq curve in the end? That sounds pretty boring (as my own test told me). What did i miss here? Where is the difference?

I also need to add, that the master equalizer setting that I just mentioned to you, would never be exported applied on a sound or song! That would be up to the listener to set to their own hearing on their given system or systems individually, of course.

[Loque]

Hum...my own loudness feeling instead of a flat EQ curve...got to try this. Since the "hearing" is _not_ flat and a bit individual, i am not sure what i could expect. Should sound similar to, pink noise?

[selig]

A simple flat spectrum on its own is still not enough, cause the system and our hearing still changes the final curve we hear.

I honestly don’t believe that’s true - what evidence can you present to back up that claim? Years of amazing music recording suggest what you’re saying is not true and is not necessary. How do you account for all the great music made without taking what you say into consideration?

My ears have changed over my lifetime, probably greatly (I’m 57 now). But my mixes from my early 20s sound the same as my mixes today, certainly not a different frequency response that would account for the changes in my hearing. How does that fact fit into your theory?

I would first seek to prove any theory I had before seeking solutions to what could turn out to be a non-existent “problem”.


Sent from some crappy device using Tapatalk

[RobC]

Hum...my own loudness feeling instead of a flat EQ curve...got to try this. Since the "hearing" is _not_ flat and a bit individual, i am not sure what i could expect. Should sound similar to, pink noise?

Not exactly, more like your personal gray noise, but more simply, you'll hear everything quite balanced on the spectrum.

[RobC]

A simple flat spectrum on its own is still not enough, cause the system and our hearing still changes the final curve we hear.

I honestly don’t believe that’s true - what evidence can you present to back up that claim? Years of amazing music recording suggest what you’re saying is not true and is not necessary. How do you account for all the great music made without taking what you say into consideration?

My ears have changed over my lifetime, probably greatly (I’m 57 now). But my mixes from my early 20s sound the same as my mixes today, certainly not a different frequency response that would account for the changes in my hearing. How does that fact fit into your theory?

I would first seek to prove any theory I had before seeking solutions to what could turn out to be a non-existent “problem”.


Sent from some crappy device using Tapatalk

Like I said, this is experimentation. In short, we adjust any given sound's frequency spectrum to sine waves, which makes them flat in average. That sound may get affected by our sound system if it doesn't have a flat response when it comes to audio reproduction, and of course our unique hearing system will affect how we hear it in the end, too. Sure, if we sit there for quite some time, our hearing slowly adjusts and equalizes it ~ but if we do that ourselves, for our listening joy, the sound won't hit us as 'god this sounds boring'.

And as I said as well, if interest shows, I'm willing to work out what I mean.

I understand that a mix with varied, weird sounds (considering frequency spectrum) can stand together into something nice, but the individual sounds can still have an annoying-sounding frequency spectrum. - But if every sound is slightly flattened out, and we adjust the equalization of our playback device for our hearing and taste to such song, taking that reference set of sine waves, everything gets set to; then not only can a mix sound great, but every sound on its own can sound more pleasant.

Personally, I had good experiences with this theory so far. It's simply a different approach.

[selig]

Two words: Fletcher and Munson...
Can you tell me how wide the difference is between different listeners? If Fletcher/Munson is the average (you don’t mention F/M in any of these conversations, btw), what is the deviation?

That’s where I would start if I was going to test a theory such as yours. Meaning, I’d see if the difference was even all that great and whether it’s worth testing for in the first place, while also acknowledging there is a LOT of data out there already on this subject.

What I’m thinking is that what you are doing is adjusting your sine waves for the Fletcher/Munson curves.

And the “problem” with that is that they are DIFFERENT for each SPL level of playback. Which means there is no single adjustment curve that accounts for all playback levels.

So another question: what SPL reference level are you using with when aligning the tones to your hearing? How are you accounting for the differences when listening at different levels? What about how the frequency response to our ears changes as a song fades in/out? How do you account for that?


Sent from some crappy device using Tapatalk

[RobC]

Two words: Fletcher and Munson...
Can you tell me how wide the difference is between different listeners? If Fletcher/Munson is the average (you don’t mention F/M in any of these conversations, btw), what is the deviation?

That’s where I would start if I was going to test a theory such as yours. Meaning, I’d see if the difference was even all that great and whether it’s worth testing for in the first place, while also acknowledging there is a LOT of data out there already on this subject.

What I’m thinking is that what you are doing is adjusting your sine waves for the Fletcher/Munson curves.

And the “problem” with that is that they are DIFFERENT for each SPL level of playback. Which means there is no single adjustment curve that accounts for all playback levels.

So another question: what SPL reference level are you using with when aligning the tones to your hearing? How are you accounting for the differences when listening at different levels? What about how the frequency response to our ears changes as a song fades in/out? How do you account for that?


Sent from some crappy device using Tapatalk

I mean, after the sounds are set to this sine-standard I described, then the final equalization is up to people's taste - nobody is obliged, I just don't like hearing those boosts and drops when hearing a sine wave sweep. See, there would be no such thing like somebody hears a bassline, wants to hear it more sharply, so they boost the highs a bit on their EQ; then hats start playing, and they immediately take it back, cause the song normally would sound way too sharp. With this method, any change to the sound is far less likely to cause surprises.
Many people use equalizers on their playback device. It's just for adding a little flavor.
The fact that -say- I set my phone to a comfortable audio level, then I would do the equalization either to taste, or the song's reference sine waves, is just a personal touch.
It really is a suggestion of mine (which probably results in something similar to my personal F/M curve then), that we equalize the sound until we hear the used sine waves evenly. It just sounds nice and full. Best equalization I've ever heard cause it fit perfectly in every frequency band to my hearing. Yes, such equalization would only work on 1 set audio level, and indeed there would be changes with fades. However, thing is, I never changed the equalizer setting on my playback devices, no matter how quiet or loud I listened, cause it didn't matter to me that much. It still was pleasant sounding. After a while, I started setting it to a comfortable level, and left it there - especially after learning how our ears can get used to different loudness, where turning it up won't have a long lasting joy, but in return, damages hearing.

All in all, if songs are made with this experimental standard, then if we set our playback device's EQ and audio level to taste and leave it there, then every such song will sound equally loud and equally balanced out.

When it comes to the music industry's standards however, differences between an orchestral, a heavy metal, and a hip hop song will be drastic. It's something that I've always hated.

Most likely, I will work this one out, people can test it, and hopefully realize the benefits it can have.

[Loque]

Can you provide an example? So i can really see/hear what you are doing.

[RobC]

Can you provide an example? So i can really see/hear what you are doing.

I'll get to work with it tomorrow, after that: yes. Gladly it's weekend, just the right time for an experiment.

[RobC]

So, I had a bit more difficulties, since I earlier had some problems to solve with Linear Phase frequency band splitting. When it comes to super accuracy, it can pre-echo and ring really badly. However, with some gating or cutting and general engineering tricks, those artifacts can be overcome. Not everyone has to use FIR filters for splitting, of course. Won't sound as good, though definitely less of a hassle. While in case of FIR, my tricks might work for one-shot sounds, anything else might be quite a challenge.

Anyways, working on it all, unexpected stuff happens now and then.

[selig]

Like "room EQ", this sort of thing (however it's accomplished) colors the audio path you add it to. That's why many studios got rid of room EQ totally over the years, opting for better component choices and better room design/treatment in the first place. And this entire theory assumes a perfectly flat room to start with, right? How else can you precisely judge how loud every frequency should be if the room you're judging it in is not reproducing all frequencies at the same level to begin with?

Beyond that, so many choices would need to be made, assuming research proves this is a valuable approach in the first place (which is not yet the case as far as I've heard).
For example, how many bands are necessary to get the best results?
How best to NOT color the sound when adjusting the frequency response?
How to best take highly accurate measurements without necessarily relying on the user to make manual adjustments?
How to get the adjustment to only affect each specific pair of monitors (you'll need a different curve for each monitor you use, including phones)?
How to measure and account for actual SPL playback variations?

If you approach this haphazardly, again assuming it's even a good idea to begin with, you'll just as likely make things worse as better. Even it was somehow perfectly accomplished, would anyone like it/use it? Anyone who's been around for a number of years would likely NOT enjoy suddenly hearing things differently for the first time in their life - it would mean you'd have to learning how to mix all over again, no? And anyone just starting out could only mix on their own system, as they would be totally flying blind in any other environment not properly calibrated to their ears (see the reasons above why you can't just copy the EQ from one room/speakers to another).

And at risk of repeating myself, so many GREAT sounding records have been made ignoring this concept entirely, that one has to really make the point (and provide proof) that this is at all necessary in the first place! Getting the horse before the cart is unfortunately common in this arena in my experience…or as my brother and I like to call it, this may be a SISOP (Solution In Search Of Problem).

Not trying to take away your fun here, just trying to point you in the right direction to do the groundwork that may end up saving you a lot of time. I'm assuming here (possibly wrongly) that you ultimately want to make music, but I could be wrong - you could be an inventor that prefers to experiment and explore rather than make music (and apologies for either assumption if it's inaccurate).

[RobC]

Not trying to take away your fun here, just trying to point you in the right direction to do the groundwork that may end up saving you a lot of time. I'm assuming here (possibly wrongly) that you ultimately want to make music, but I could be wrong - you could be an inventor that prefers to experiment and explore rather than make music (and apologies for either assumption if it's inaccurate).

"this sort of thing colors the audio path you add it to"

Once artifacts are taken care of, it's pretty transparent. Unless you mean the processing, sure, it changes the frequency spectrum of the given sound, once applied, but it usefully does so. That's what I want to achieve.

"this entire theory assumes a perfectly flat room to start with"

No, it's actually turning pretty technical, even a deaf person will be able to do it, so the sound system won't really matter.

"How else can you precisely judge how loud every frequency should be"

This weekend, I figured out that there's not much need to do any manual leveling at all. See, a split frequency band of a simple noise sample, set to -12 dB average ~ compared to a sine wave sample (again -12 dB avg), with respective center frequency to that band, sounded quite similarly loud. When setting multiple bands, and comparing to one-other, the differences weren't that drastic or bothersome - even though as we know, at least our hearing does affect how we hear that flattened frequency spectrum.

"assuming research proves this is a valuable approach in the first place (which is not yet the case as far as I've heard)"

Once I will arrive to the point that I can provide examples to my experiment, then it can be judged how it sounds with my method.

"how many bands are necessary to get the best results'

To be honest, with Linear Phase splitting, in past early tests, when I just messed around, I already got really good results with just about 10 bands. So 1 octave starting from 20 Hz (20-40, 40-80, etc.). When I did the pure sine tone and noise band comparison (above), that one sounded just right, with a 1/3rd octave band. That would be about 30 or so? Depends on processing time, and if there's an easy way to automate the whole process including removing artifacts. Then, if we want, we could do a band per note.
When the listener would set up their equalizer on a song made with this method? 5-10 bands do the trick (or more, if they don't get sick of evening out reference sine tones for themselves with equalization).

"How best to NOT color the sound when adjusting the frequency response?"

Again, if we're talking artifacts, I think it can be made pretty transparent. There might be a trade-off if it can't be cleaned perfectly, but if this works out, then it's worth it in my opinion.

"How to best take highly accurate measurements without necessarily relying on the user to make manual adjustments? "

I think I just answered this question - and with an automated process that takes care of artifacts, too, I think not much can go wrong.

"How to get the adjustment to only affect each specific pair of monitors"

There's no way for that. It's a technically flat sound, that can afterwards be very flexibly equalized by and for 1 user for themselves, on 1 system at a time. Or if they like the flat sound as is, they can listen to it that way.

"How to measure and account for actual SPL playback variations?"

This rather matters at the playback of a finished sound/song. Like I said, there's only 1 way here too. Set 1 comfy level, EQ to taste.

"Even it was somehow perfectly accomplished, would anyone like it/use it?"

I'm preparing this little experiment, so it's not organized just yet, and of course can be cloudy at this stage.
Personally, I believe it could be put to great use, once I'm not the only one who understands it.
It's getting very simple and time saving, too, so why not?

"Anyone who's been around for a number of years would likely NOT enjoy suddenly hearing things differently for the first time in their life"

The flat sound probably would only sound nice to people who have a flat hearing curve. As for starting the EQ-ing to taste - even people who are classical music enthusiasts, tend to equalize the sound. People like adding flavor to it. Hell, even you said, once your studio sound was flattened out, you still did a few changes to the equalization.

"you'd have to learning how to mix all over again, no?"

Not really, or it would be much more simple. If somebody just mixes music so sounds are rather evenly loud, then this pre-flattening does all the work for them. Afterwards, they could just mix around with simple values like taking something back to 75, 50, 25 % etc. After all, everyone's hearing leaves a mark on the sound (tons of static and dynamic factors). Round values change something exact and noticeable.
But this is getting to a bit bigger topic that needs examples and proper explanation.
What I'm working on, is an engineering standard at this point (including mixing, mastering, design, etc.). This might take a month, not just a weekend, but I'm willing to research this. And as I say, when truly done, people can test and judge for themselves. It simplifies a lot of things. And I wouldn't say it causes damage to the sound, otherwise I wouldn't spend time on it.
If this would work, during sound design people really could focus more on creating interesting harmonics, unique sounds, while not worrying what's gonna happen with the frequency spectrum. But same goes for song writing / production, etc.

"anyone just starting out could only mix on their own system"

Just wrote this down, too, but to sum it up, it would have nothing to do with the system's frequency response. I really need to work this out, so it's less confusing.
If every sound is flattened out for a mix and they are on level with each-other, then even if another system sounds differently, the differences between the channels in the mix will be heard as the same this way.

"so many GREAT sounding records have been made ignoring this concept entirely"

I understand, but if there's some chance to get a better sound with less work, then why not?

And to the last part: I'm interested in both the science and artistic part of music/sound, actually. Right now, I'm looking for ways to productively simplify each step of making music, and so that 1 person will be capable of doing everything quickly and with the best possible quality.

[O1B]

I'll get to work with it tomorrow, after that: yes. Gladly it's weekend, just the right time for an experiment.

... so, still no sound example... again....?.... then, so much for...

Can you provide an example? So i can really see/hear what you are doing.

[RobC]

Can you provide an example? So i can really see/hear what you are doing.

Okay, here's a flat noise example. I used 10 bands, each having equal amount of energy. The hard splits are quite audible to me, so rather 1/3rd octave splits will be smooth enough.

In the past, I decided to set a loudness standard according to a noise sample. The reason for that was that it has random information across the frequency spectrum. Now, when we talk about a song, it similarly gets crowded, even if some extreme peaks happen here and there. A noise sample with its fullness seems to give a fair room, a fair amount of limitation ~ guide so to say; where if we keep each sound between those limits, then after we finish a balanced mix with a bit of everything, there should be not much surprises, not much to saturate or clip, in order to reach the target loudness.

For this example, and for one sound, the results seem to be:

Flattening out the average volume of each band to about -25 dB

(The full sample has an average volume of about -15 dB, individual bands -25 dB)

If a band of a sound exceeds 0 dB, then it should be clipped/saturated/limited. I'm more for saturation/soft clipping.
Then I would join the bands and take care of anything above 0 dB.

As for a mix, when loading such flattened sounds and summing them together, there may or may not be a few overshoots when sounds hit together. I would likewise split the mixdown into multiple bands, first take care of any clipping, then rejoin them, and again take care of master clipping.

These are the plans for now.
Like I said before, since there's equal energy in each band, in a mix, such sounds would sound pretty much equally loud and would need no further equalization/shaping.

Then, in the end, since every sound would have a flat spectrum, we can just take a reference center-frequency sine waves and either balance their volumes to equalness with an equalizer (creating a really full sound for everything in the mix), or just to taste. Clearly, 1 system, 1 pair of ears, 1 equalizer. If songs would be made this way, then the same nifty frequency response could be heard.

Everything is very-very raw at this point, though, but will go much faster. Things work according to plan so far.

Oh, by the way, the raw noise sample, which is quite loud, sounds like this - which won't say much yet, cause it's just a flat digital standard. So again, adding the flavor will be in the user's hands.

Now, it's not like one couldn't flatten out even a master, but obviously, if we have access down to individual sounds, it will be more accurate. Master > Mix Channel > Individual sound hits (for example rex slices)

[RobC]

Welp, I guess I'll eventually do a simple short beat, wrap the most important info up, and then who wants, tries it.

[selig]

Welp, I guess I'll eventually do a simple short beat, wrap the most important info up, and then who wants, tries it.

Can I ask a simple question, because perhaps I've misunderstood the intent of this concept.

Is this something you apply during mixing or mastering, something that you put in your master insert (or similar) as a master effect (and becomes a part of the exported mix)

OR, is it something you use ONLY on your speakers/monitors, as a form of "room EQ", and is not a part of the mix/export?

[RobC]

Welp, I guess I'll eventually do a simple short beat, wrap the most important info up, and then who wants, tries it.

Can I ask a simple question, because perhaps I've misunderstood the intent of this concept.

Is this something you apply during mixing or mastering, something that you put in your master insert (or similar) as a master effect (and becomes a part of the exported mix)

OR, is it something you use ONLY on your speakers/monitors, as a form of "room EQ", and is not a part of the mix/export?

The first one. Can be applied to any recording, whether 1 channel, (maybe each drum hit, or note individually), or a whole mix. The noise sample above has even-averages / RMS in 10 bands.

System itself is untouched during work.



Assume the song is done now. You take it out of the studio and give it to a friend. If we go by the FM curve and they listen at 80 dB SPL on a flat response system, then if the FM curve is their hearing response, they might say it lacks a bit of sub bass and upper treble, but has a boost in the highs, while the rest seems fine. Either they would listen to it as is, or they can apply an EQ if they want. The post-help would be the set of sine waves, which are more easy to equalize than a dynamic, changing song. Since the song has sounds with the same average energy for every frequency band as the sine waves, the applied equalizer would have the same effect. Thus once they set their EQ with the help of the sine waves, then when they use that EQ on the song, they will no longer hear any boosts or drops, but a perfectly balanced frequency spectrum.

[selig]

System itself is untouched during work.

So you do nothing on your end when mixing, this is strictly for playback on the user end?



Sent from some crappy device using Tapatalk

[RobC]

System itself is untouched during work.

So you do nothing on your end when mixing, this is strictly for playback on the user end?



Sent from some crappy device using Tapatalk

Pretty much. Heck, not even the user has to apply the said post-equalization if they are still okay with the flattened sound itself.

[selig]

So you do nothing on your end when mixing, this is strictly for playback on the user end?



Sent from some crappy device using Tapatalk

Pretty much. Heck, not even the user has to apply the said post-equalization if they are still okay with the flattened sound itself.

Wait (and sorry for so many questions), but how does the sound get “flattened” in the first place if what you just said is true - that the system is untouched during work (nothing is added at the mix stage)?

And what exactly does “flattened” mean?


Sent from some crappy device using Tapatalk

[RobC]

Pretty much. Heck, not even the user has to apply the said post-equalization if they are still okay with the flattened sound itself.

Wait (and sorry for so many questions), but how does the sound get “flattened” in the first place if what you just said is true - that the system is untouched during work (nothing is added at the mix stage)?

And what exactly does “flattened” mean?


Sent from some crappy device using Tapatalk

I thought you meant by 'untouched system', that we generally don't equalize our device to our hearing during work, but try to keep its response flat.
The recordings themselves, what we can load as a waveform, do get processed. They get split into frequency bands, and the average levels of those bands are brought to the same level. That's what I dubbed as 'flattened' for now.

But I think it will be best to wait for a beat, made with this method, so you can hear what it will be like. This might be a bit complicated to just talk about.

[selig]

Wait (and sorry for so many questions), but how does the sound get “flattened” in the first place if what you just said is true - that the system is untouched during work (nothing is added at the mix stage)?

And what exactly does “flattened” mean?


Sent from some crappy device using Tapatalk

I thought you meant by 'untouched system', that we generally don't equalize our device to our hearing during work, but try to keep its response flat.
The recordings themselves, what we can load as a waveform, do get processed. They get split into frequency bands, and the average levels of those bands are brought to the same level. That's what I dubbed as 'flattened' for now.

But I think it will be best to wait for a beat, made with this method, so you can hear what it will be like. This might be a bit complicated to just talk about.

Ouch- so the entire mix is run through all these filters, just to make it “flat” (even though flat probably sounds bad)? Have you analyzed good sounding mixes? If so, you’ll notice none of them are spectrally flat, and most of them don’t look alike - indicating there isn’t “one” setting that works for all types of music.

And when you say flat, you mean like white noise on a spectrum, or pink noise (more preferable) or some other arbitrary standard?

Are you trying to take an artistic process, and make it conform to some “standard” response curve (e.g., “flat”)? It sounds like you’re taking the “ears” out of the equation, and just making every frequency range hit some sort of target?

Apologies if I’m missing something here - still sounds like a solution in search of a problem as you’ve described above. But I’m still interested in making sure I’m understanding the basic concept, and trying to understand what “problem” this solution is addressing. Maybe I’ve totally missed your intention here, and if so it is not at all intentional!


Sent from some crappy device using Tapatalk