[Rad Librarian]

Hi, I run a librarianship blog www.radlibrarian.wordpress.com, and I'm working on a feature where I compare the sq of the "playaway" mp3 players that so many libraries loan out, with audiobooks on cd, and with audiobooks ripped to 128kbs, and audiobooks ripped to apple lossless.
 
In order to be accurate in my evaluation, I'd like to make sure that I cover all the necessary ground when desribing the sound of these formats.
 
So far I'll be evaluating
 
background noise
"lifelike" quality of voice
and the sound in general.
 
Its in "in general" part that I'm having problems with.
 
Any help would be greatly appreciated.  This will be a subjective review of the audio (I won't be doing any waveform analysis for example) so otherwise correct criteria and terminology will be very important.
 
Hope you can help.  Thanks!
-Jason

 

[Head Injury]

Best way to go about it would be an ABX test, not a subjective review.
 
Get yourself Foobar and the ABX comparator, rip all the files you need, and compare them two by two.
 
Don't bother testing CD and Apple Lossless together, the only differences will be from what you play them on. Lossless is literally the same data.

 

[Rad Librarian]

Thank you for your suggestions H.I, but this is going to be a subjective review.  So, respectfully, suggestions on  criteria and terminology, not the tech to use for the analysis, is what I need.
 
I appreciate your point about lossless.  I'll bear that in mind when writing up the review.  I'll still do the test, so that I can say I've done it, etc.

 

[nick_charles]

 
Quote:

Thank you for your suggestions H.I, but this is going to be a subjective review.  So, respectfully, suggestions on  criteria and terminology, not the tech to use for the analysis, is what I need.

 
With respect this is not the best subforum to ask those questions, you got the polite answer already. Here comes the rude one. Your evaluations if solely based on your subjective impressions in an uncontrolled sighted manner without carefully accounting for things like levels, normal human bias, output impedance on devices and so on will be misleading at best and worthless at worst. Whereas with careful objective comparisons you can derive some potentially useful information, you could measure the noise levels, you could contact Findaway and ask for data on their players. what you are planning is just another weak anecdotal set of comparisons. There are plenty of subjectivist subforums here to ask those questions...

 

[mikeaj]

The subjective impressions are important to give the target audience a reference they're familiar with (they probably wouldn't understand any waveform analysis anyway), but they're next to meaningless unless they're done blind.  I still think that you could measure noise levels or come up with some technical details that are easy to understand and are relevant, but if you don't want to do that, that's understandable.
 
The exact words used to describe the sound are not that important.  Different people have different interpretations of what different words mean anyway, even if it's something like "bright" or "crackly."  Imprecision and fluidity of language is good for literature but not so much for conveying technical details, in my opinion.  It's more important that the results are more meaningful due to some modicum of experimental control, than for the descriptions to have more fancy audiophile terminology.  You said you wanted to do this "in order to be accurate in [your] evaluation" so this is the way to go.
 
Just be sure to evaluate the samples using the same source recording (and try a couple different audiobooks), without you knowing which is which when you're listening to it.  Take notes on what you hear, and score how well you think it sounds.  Then later you can go back and match up which impressions and scores match up with which audio file (128 kbps or lossless).  ABX can also be considered, but that's not the only thing you can do.
 
For audiobooks I would imagine modern 128 kbps encoding would do pretty well though.

 

[nick_charles]

Quote:

 
For audiobooks I would imagine modern 128 kbps encoding would do pretty well though.

I have a load of audiobooks and 128kbps is subjectively (ha ha) fine, the human voice has almost nothing above 10k and the vast majority of energy in the 1k - 3K region. Many early audiobook rips one could find were done at 22khz sampling and often as little as 32kbps !
 
 

 

[mikeaj]

Most energy is 1kHz - 3kHz?  That's somewhat of a surprise to me.
 
Let's just say that I am acutely aware of the upper limitations of my poor baritone singing voice (outside the falsetto range), from mandatory choir participation back in the day.  The fundamental frequency for speaking tones, even for women, has to be say...under A 440 Hz for sure, right?
 
Is most of the energy in speech from 's', 't', 'p', etc. sounds, or is at least part of the reason the 1 - 3 kHz band contains so much energy, due to high presence of overtones?

 

[Rad Librarian]

Mikeaj, thank you very much for your constructive comments!  They will be very helpful in my analysis of the audiobooks I'll be listening to. 
 
Since you were kind enough to offer your help I'll give you a bit of insight into the project, which I have just begun today.
 
I have 5 audio books as both cds and as playaways, so 10 total.  One of the playaways is a playaway labeled "HD Audio".  Curiously, when I called the media contact for Findaway and asked her to describe the differences between the "regular" and "HD" players, she did so using extremely vague gobbledigook about what "better sound" the HD unit offered.  I politely asked for more information and she promised me an email from their engineering dept. today. She was very helpful when it got down to brass tacks. So that should be interesting.
 
As you wrote, it is true that for my audience, which is mostly librarians, there is no point in getting exremely technical.  They are not sound engineers, but they are content and quality enthusiasts.  Eyes would glaze over and I'd lose my audience for sure if I started posting brickwalled waveforms, etc.
 
Your suggetsed methodolgy for the project is excellent.  Thank you again.
 
Nick_Charles, when the project is done and the post is up at www.radlibrarian.wordpress.com I'll let you know in this thread, and you can freely savage it in the comment section of my blog should you so desire.  I only ask that you wait until the project is done before you comment.
 
Thanks all!
 
btw, I do tend to encode my personal audiobooks at 128kbps, but I very much prefer lossless.  If I had the space, thats how I would go...

 

[Willakan]

Pah! Any librarian worth their salt can interpret a THD graph!
On a serious note, you probably could test a few of the players with RMAA. (http://audio.rightmark.org/index_new.shtml). It isn't a very accurate or sophisticated tool, but it could be useful for ensuring none of the players you measured had any significant technical flaws. There's a helluva lot of stuff it can't measure, but I don't think your readers will be very interested in hearing how a given player is prone to ultrasonic oscillation into high-capacitance loads...

 

[Rad Librarian]

Thank you Willakan!  All constructive comments, like yours, are greatly appreciated!

 

[nick_charles]

Quote:

 
Nick_Charles, when the project is done and the post is up at www.radlibrarian.wordpress.com I'll let you know in this thread, and you can freely savage it in the comment section of my blog should you so desire.  I only ask that you wait until the project is done before you comment.
 
Thanks all!
 
btw, I do tend to encode my personal audiobooks at 128kbps, but I very much prefer lossless.  If I had the space, thats how I would go...

It is not a matter of savaging anything it is about meaningful comparisons. This subforum is the one place where we can talk openly about meaningfully controlled listening tests. You are already showing a personal bias that would seriously undermine any comparisons done in a casual manner. Mikeaj's suggestions are bang on, unless you control how you do the comparisons they will just not be very useful.
 
As for Findaway - you need to ask the right questions , and ask the right people. Amongst the first questions is why on earth they believe HD is in any way necessary for audio books ? - the 2nd question would be what do they mean by HD ? - since the device uses proprietary codecs they can call it whatever they like, whereas for us here HD has a very specific meaning viz a recording that has more data than a redbook CD - typically this means 24 bits per sample instead of 16 and anywhere from 88.2khz to 192Khz sampling rate - lossy compressed formats are not normally described as HD - if it is not lossy then the HD variant device must have a stonking big chip in it - 40 hours of lossless compressed redbook audio equates to about 12.5GB - a lossless compressed 24/88.2 variation could be 40GB plus !

 

[Willakan]

I would strongly support Nick_Charles et al in that you should attempt, if possible, to do blind comparisons between players. This does, however, require very precise volume matching to be of use, for which you need to precisely measure the volume - matching by ear is grossly inadequate for this.

 

[Rad Librarian]

very helpful comments all.  I thank you again!

 

[Rad Librarian]

Here's a handy chart I found online that breaks down bit rate and the memory sizes of the payaway units.
 
 

[size=x-small]Here you can find a scheme which indicates the possible playing time of the player, depending on the size of the player (in Mb) and the bitrate of the encoding to mp3-format (only to be used as an indication, because this can vary depending on the audio used) :[/size]
[size=smaller]Compressionratio :[/size]	[size=x-small]8 kbps[/size]	[size=x-small]16 kbps[/size]	[size=x-small]32kbps[/size]	[size=x-small]64kbps[/size]	[size=x-small]128kbps[/size]
[size=smaller]Playaway 64 Mb[/size]	[size=x-small]15:50h[/size]	[size=x-small]8:25h[/size]	[size=x-small]4:50h[/size]	[size=x-small]2:00h[/size]	[size=x-small]1:00h[/size]
[size=x-small]Playaway 128 Mb[/size]	[size=x-small]32:20h[/size]	[size=x-small]17:00h[/size]	[size=x-small]8:20h[/size]	[size=x-small]4:10h[/size]	[size=x-small]2:05h[/size]
[size=smaller]Playaway 256 Mb[/size]	[size=x-small]64:50h[/size]	[size=x-small]34:10h[/size]	[size=x-small]16:45h[/size]	[size=x-small]8:25h[/size]	[size=x-small]4:10h[/size]
[size=x-small]Playaway 512 Mb[/size]	[size=x-small]129:40h[/size]	[size=x-small]68:20h[/size]	[size=x-small]33:30h[/size]	[size=x-small]16:50h[/size]	[size=x-small]8:20h[/size]

 
According the email from the rep, there is also a 1 GB unit.  Nick_Charles I did, per your suggestion, ask at what bit rate they assign to "HD Audio" players.
 
 
 
 

 

[Rad Librarian]

Also, it should be noted that these players only have an analog out, there is no possibility of extracting the digital signal.  Looking into RMAA, to see if it will work considering that limitation.  If I can play analog signals into the program, then cool.  If not, I'll have to do my best on my own.
 
Btw, I freely admit a slight bias against playaways.  Otherwise I wouldn;t care and wouldn't bother doing this project.  However, I am approaching this with an open mind as to the results. 
 
Also, clearly I should perform an ABX test. Tricky to do, but I could enlist the help of a friend to switch out the headphones for me.  I'll be using 2 sets of headphones, ksc75 and grado sr225.
 
I do sincerely appreciate these comments, even if my environmental limitations will make this project challenging and imperfect.