[gorman]

I promised this some time ago and, finally, I'm able to share this information with all fellow Head-Fiers.
First I will describe the experiment and the rationale behind it, then I'll provide the results, along with some comments.

As, hopefully, pretty much everybody in this forum is aware of, listening to headphones and canalphones at high volume is dangerous for the well being of our hearing. Hearing damage, once it happens, it's there to stay. So we should take great care of our one and only listening instrument.

With this in mind, I got in touch with Jerry Harvey, asking if he had the instruments necessary to conduct an experiment of the Sound Pressure Levels that UE-10 achieved inside one's ears, combined with the Rio Karma I usually use to listen with them.

He was ready to do that and confirmed that he had all the instruments necessary for this test, with the notable exception of the Rio Karma. Enter cmascatello who, very very generously, accepted to ship his own DAP to Jerry to test (I live in Italy, for me it'd have been really unpractical).

Now everybody would think "yes, but all songs have different recording volumes, the test would be pretty much useless unless you were listening only to the songs tested". True and I thought about that too.

Enter Replaygain. Replaygain is this nice technology whom many of you surely will be familiar with. For more detailed info, check this link. It boils down to a psychoacoustical analysis of a given music file to check how loud it sounds to our ears, with the target being to bring it under check, make all your music sound relatively at the same volume, avoid clipping, and so on and so forth. The volume target is usually set at 89dB for every song, with an "audiophile" setting available if you want to preserve the relative differences in volume of different tracks on the same album.

With this in mind, I reasoned that analyzing with Replaygain the tracks used for the test would help in coming up with results that could keep valid across all music. And the nice thing is that, upon examination of the results, it appears to work very nicely, give or take a couple of dB.

On with the results, now.

First, the values Jerry measured on cmascatello's Karma:

On Klark Teknic dn6000 A weight

Track 1 / No Doubt - I'm Just A Girl
Volume 10 - 80db average/ 84db peak
Volume15 - 88db average / 92db peak
Volume20 - 98db average / 102db peak
Volume25 - 108db average / 112db peak

Track 2 / Pink - Get The Party Started
Volume 10 - 77db average / 82db peak
Volume15 - 87db average / 92db peak
Volume 20 - 97db average / 102db peak
Volume 25 - 106db average / 111db peak

Track 3 / Alice In Chains - Man In A Box
Volume 10 - 74db average / 79db peak
Volume 15 - 83db average / 88db peak
Volume 20 - 94db average / 99db peak
Volume 25 - 103db average / 108db peak

These first results, without analyzing the recording levels of the different songs, already told me there was sense in this experiment. The Rio Karma, in fact, has been confirmed personally to me by a Rio engineer (UK Rio Engineer on Riovolution forums) to increase volume by 2dB per point of volume setting. We found this to be pretty much true for all these three songs (give or take one or, in one instance, two dB). And I did not share this information with Jerry Harvey at all (it came to me through an MP on Riovolution).

Now, on with the Replaygain analysis.

The three songs resulted with these average recording levels:

Track 1 / No Doubt - I'm Just A Girl: 99.7dB
Track 2 / Pink - Get The Party Started: 98.4dB
Track 3 / Alice In Chains - Man In A Box: 97dB

Again, these results are pretty much consistent with Jerry's findings, give or take a couple of dB. The difference is probably related to Replaygain psychoacoustic model, more than anything else. It's accurate, but to a certain extent, and obviously not as much as real measuring of real music reproduced through real speakers (or headphones in our case).

If we apply Realgain to achieve the 89 dB average replay level, we achieve these results with the three songs:

Track 1 / No Doubt - I'm Just A Girl
Volume 10 - 69,3dB
Volume 15 - 77,3dB
Volume 20 - 87,3dB
Volume 25 - 97,3dB

Track 2 / Pink - Get The Party Started
Volume 10 - 67,6dB
Volume 15 - 77,6dB
Volume 20 - 87,6dB
Volume 25 - 96,6dB

Track 3 / Alice In Chains - Man In A Box
Volume 10 - 66dB
Volume 15 - 75dB
Volume 20 - 86dB
Volume 25 - 95dB

As you can see, we still have this couple of dB of difference among the three songs. Let's average them, to come up to more "real" average SPLs at the different volume settings:

Volume 10 - 67,6dB
Volume 15 - 76,6dB
Volume 20 - 86,9dB
Volume 25 - 96,3dB

Even in a worst case scenario, taking into account the peaks for all the songs, we would get this:

Volume 10 - 72,3dB
Volume 15 - 81,3dB
Volume 20 - 91,6dB
Volume 25 - 100,9dB

Here is the suggested time of daily exposure for different SPLs. You can find several versions of this table. Some more strict, some more forgiving in their suggestions. This is kind of strict and comes from the Nova Scotia Department of Labour:

16 hours for 80 dBA sound
8 hours for 85 dBA sound
4 hours for 90 dBA sound
2 hours for 95 dBA sound
1 hour for 100 dBA sound
30 min for 105 dBA sound
15 min for 110 dBA sound
7.5 min for 115 dBA sound
0 min for above 115 dBA sound (there should be no exposure at this level!)

This is more forgiving and comes from the O.S.H.A (Occupational Safety & Health Administration of the U.S.A.):

85dB 16hr
90dB 8hr
95dB 4hr
100dB 2hr
105dB 1hr
110dB 30min
115dB 15min
120dB 7.5min

I hope that my findings will help fellow UE-10 owners to listen to more safe levels. They will certainly guide me in the future.

Jerry tested UE5c too. They were consistently 1dB quieter than UE-10 Pro.

He did test with iPod too. But I still have to ask him the replaygain value for the songs he used (which are different from these three). I will publish iPod results ASAP.

 

[Welly Wu]

According to this data, the ReplayGain feature lowered the SPL thresholds considerably at the lower volume levels while maintaining a peak SPL of slightly over 100.9 dB at volume level 25. Therefore, I think it is prudent to utilize this technology for all compressed digital audio files especially when used in conjunction with any canalphone (custom IEM phones too).

 

[John Reeves]

Hi,

Just so I've got this correct. Rip using say EAC, apply replay gain, transfer in FLAC to my Karma. I had to get this in 'cause I just bought a Karma today.

Cheers

John

 

[gorman]

Quote:

Originally Posted by Welly Wu According to this data, the ReplayGain feature lowered the SPL thresholds considerably at the lower volume levels while maintaining a peak SPL of slightly over 100.9 dB at volume level 25. Therefore, I think it is prudent to utilize this technology for all compressed digital audio files especially when used in conjunction with any canalphone (custom IEM phones too).

Replaygain is not a suggestion on my part. I just used it to analyze the recording levels of the different songs. Using it might be a nice way to say: "Ok, if I listen to the music I have on my DAP, whichever music, I can safely listen at volume XX". Otherwise you should be adjusting the volume according to the different recording levels of every album, which is not really practical.

 

[gorman]

Quote:

Originally Posted by John Reeves Hi, Just so I've got this correct. Rip using say EAC, apply replay gain, transfer in FLAC to my Karma. I had to get this in 'cause I just bought a Karma today.

Hi John! I do this:

1) Rip using EAC.
2) Apply replaygain using Foobar.
3) Convert to Vorbis Q6 using OggdropXPd, making use of its feature that allows you to take the replaygain values and incorporate them in the waveform of the Vorbis file.

Because, you see, Karma (and no other player on the market) does not read and apply replaygain tags. If you were to transfer the replaygained FLACs to your Karma, they would sound at their original sound.

I believe there is a way to apply the replaygain to the FLAC waveform as well, but I'm not familiar with it. Plus, you would lose the "perfect archive" thing. Because the FLACs, at that point, won't be perfect replicas of your CDs.

I hope I have given you a clearer picture of things.

 

[john_jcb]

Quote:

Originally Posted by gorman ...........The reason time halfs for every 5dB is because every 5dB sound doubles in intensity................

Very interesting and useful information.

Just a minor point. I always thought that the perceived doubling of sound is every 10dB not 5dB. The calculated value is 6dB but the perceived value is actually 10dB IIRC.

 

[gorman]

Quote:

Originally Posted by john_jcb Very interesting and useful information. Just a minor point. I always thought that the perceived doubling of sound is every 10dB not 5dB. The calculated value is 6dB but the perceived value is actually 10dB IIRC.

I deleted that passage until i can double check this. I was going by memory. And... getting old, I suppose.

 

[gorman]

Yeah, sound pressure doubles every 3dB. Sound perception doubles every 10dB. Which in and by itself tells why it's easy to underestimate the levels we listen at, I guess.

 

[john_jcb]

What always amazes me is how loud 85dB is. Except for brief periods I cannot enjoy listening to my speaker setup for any period of time at anything approaching 85dB. Headphones are even worse. What always worried me with my Ety's is just how loud I was listening. Your data will give people a better idea of where they are at and hopefully save some ears.

 

[loveheadphones]

does anyone know an mp3gain type app which will normalize both mp3's AND wma's? or just one for wma's?

 

[gorman]

Quote:

Originally Posted by john_jcb What always amazes me is how loud 85dB is. Except for brief periods I cannot enjoy listening to my speaker setup for any period of time at anything approaching 85dB. Headphones are even worse. What always worried me with my Ety's is just how loud I was listening. Your data will give people a better idea of where they are at and hopefully save some ears.

You are right. The good thing is that now that I have replaygained all of the songs I listen with the Karma, I know that staying at volume 15 exposes me to an average of 77dB, more or less. Which is almost half the volume as 85dB.

 

[wolfen68]

So around volume 15 on a Karma may be an optimal setting (depending upon gains).....what's the volume scale on a karma, 0-25? I'm curious if a very general relative comparison can be made across all DAP's.

It's probably not possible to generalize since output levels vary so widely from model to model.

 

[gorman]

Quote:

Originally Posted by wolfen68 So around volume 15 on a Karma may be an optimal setting (depending upon gains).....what's the volume scale on a karma, 0-25? I'm curious if a very general relative comparison can be made across all DAP's.

For properly replaygained material (to the waveform level, I want to stress this, because Karma and no other DAP reads the replaygain tags) I would say that yes, more or less 15 is right. In very quiet environment I think you can get quite good results with 12 or 13 too. Bear in mind that each point on the volume scale equals 2dB and you can do all intermediate levels calculations.

 

[random person]

Gorman -- Thank you so much for this analysis! Fascinating stuff!

Naturally I am intensely curious about the Ipod -- when will you have those results? Any qualitative input about them yet? What I'm really after is the answer to this question: Am I reasonably safe confining my listening to 50-60% of the volume slider on the Ipod with the UE5c?

Any info that you have on this will be much appreciated!

 

[gorman]

Quote:

Originally Posted by random person Am I reasonably safe confining my listening to 50-60% of the volume slider on the Ipod with the UE5c?

Ok, bear in mind that this is not the result of the full analysis, just preliminary data. But if you do replaygain to 89 your files (again, the waveform, tags would be useless for iPod), then yes, I believe that 50% ought to be ok.