release
dev builds
extras
themes manual
wiki
device status forums
mailing lists
IRC bugs
patches
dev guide
translations

# Rockbox mail archive

Subject: Re: Optimum Audio Levels - Vote Results

# Re: Optimum Audio Levels - Vote Results

From: BlueChip <cs_bluechip_at_webtribe.net>
Date: Thu, 03 Jun 2004 05:04:11 +0100

Okay, peeps,

Quicker than expected the results are already in on todays spot
survey ;)

This is where the Jury Stands on listening volume.

It goes down a little further than my deaf ears and cheap headphones
had expected, but in the graph at the end of this email we have a
small block of high-rise council flats at 65 & 70; between 80 and 92
we have the bulk of our audience with the sears snack-bar at 85 and
the empire state foodery at 90.

Beyond my attempt at humour (see above) the graph kinda speaks for
itself really.

So I propose the following as the volume scaling algorithm for
NON-techies.

Before you read on, understand that the new Audio control system offers
a dB scale for the techie anyway (most controls have a toggle'able
user/geek mode available)

NOBODY reported listening at volumes lower than 50%/Rockbox == -52dB/MAS
So anything that happens below there is close to irrelevant. That is,
we can afford to be sloppy with our scaling in the 0..49% bracket in
favour of MAXIMUM accuracy for 50..100%

I suppose it could be argued that we could lose accuracy at 96..100%, but
as I said, tinnitus and cheap headphones mean that is where I listen a lot
...so as I am writing the code, accuracy stays at the top ;)

The first thing to consider, is that whatever happens, the Rockbox% numbers
will be scrapped for the new (dare I say "improved") scaling system.

Rockbox = 50% will produce MAS = -52dB
This means that all values from -52..0..+12 need to be in 1dB steps
There are 52+1+12 = 65 steps which are commonly used (by the voters)
This means we would need to use up NewScale = 36..100% with 1dB accuracy
Leaving 1..35% for the remaining 62 steps (0% is 0% is mute)
That's almost 2dB per step... hmmmm...

The solution is obvious :)

NewScale: 0=Mute; 1..27% are each 2dB steps; 28..100% are each 1dB steps
The upshot is 1dB accuracy down to -60dB (a further 8dB of accuracy beyond
that which was voted for)

These are not the results I expected at all, so another immense thank you
to everybody who has helped create this conclusion. Should anyone feel
I have missed a consideration, please feel free to offer your thoughts here
at this mailing list.

Cheers,

BlueCh:p

The following is an ascii-art graph and not suitable for screen-readers:

12_|
11_| #
10_| #
9_| # #
8_| # #
7_| # # ######
6_| # # # #########
5_| # # # ########### #
4_| # # ################ #
3_| ###############################
2_| ####################################
1_| ############################################
0_| ###################################################
"""""""""""""""""""""""""""""""""""""""""""""""""""""""""
| ... | | | | | | | | | | |
00 50 55 60 65 70 75 80 85 90 95 100%
-52 -46 -39 -33 -27 -20 -14 -8 -1 +5 +12dB

>Hi guys,
>
>First of all a great big thank you to everybody who replied. You help is
>much appreciated :)
>
>I have read every single response and catalogued them; and will make some
>analysis of the results and tweak the algorithm accordingly
>
>
>>I would REALLY like a setting that produces "line level"
>>
>>Don't wan' no stinkin' karry okie.
>
>"LINE OUT" is there :)
>
>The Empty-Orchestra option is gonna be cool - I'm already dreaming about
>it (the interface, not the singing), but fear my time may be about to be
>torn away from such severe Rockbox development - so it may be a while.
>
>
>Either you have incredible headphones or I am going deaf - lol
>
>>If you do line level what are you going to define "line" as?
>>-10 is about normal for consumer stuff
>>+ 4 is line for pro stuff.
>>Thats dBu
>
>and
>
>>As someone who has tried to use an Archos in professional applications, I
>>would strongly suggest -10dBu. It's consumer grade, at best.
>
>scope tests:
># Chris Muth reports 90% is the closest match to the line level from his
>car CD player
># I recall a more detailed test (but cannot find the email now) where the
>result was 93%
>...The manual claims that 0dB (on it's scale) will produce line-out
>
>So as there have been _two_ tests which I know of, and the average of the
>two is almost exactly what the MAS Engineers claim - I am going to reap
>the experience of the chip designers and the two Rockboxers who got out
>their scopes and call 0dB (on the MAS3587 "VOLUME" scale) LINE-OUT
>
>Hope that satisfies all the points which were raised. If not ...ask more ;)
>
>Cheers again for all your time people,
>
>BlueChip

_______________________________________________
http://cool.haxx.se/mailman/listinfo/rockbox