[aroon]

Hey guys, long time since ive been around.

I had an idea recently to do a study on how well people can actually tell the difference between different bitrates of encoded music [i.e. ordering 64, 128, 192, 256, and 320kbs mp3s of the same song by perceived quality]

I'm trying to configure my source setup. I need a setup that will provide good clean sound [of course] for less than 500 dollars. My current though is an ipod mini [less than 250 thanks to the academic price] and pair of good headphones. I was looking at the Senn HD580's which, when coupled with the ipod mini, maintains my budget.

I'm worried, however, about how well the ipod mini can power the HD580's. at 300ohms, it sounds like the mini could have some issues. anyone have any experience with this specific pair? any other advice such as using a different source, different headphones, or both? I really want to steer clear of using any amps to help power the headphones since a) it introduces all kinds of questions about altering the sound in a subjective way to the listener and b) purchasing amps will almost definiately throw me voilently and ruthelessly out of my budget!

thanks for the feedback in advance!

 

[mjg]

it sounds like ety er4-p's are for you.

 

[Taurui]

Depends what you want from the study. Do you want results that matter for the average consumer or do you want results like "Omg I never knew there was so much difference" ?

Oh, I wouldn't do it with etys. They're much too dependant on a good fit to make a study with lots of people.

 

[bangraman]

I would say the Etymotics too, but if you're doing testing I'd recommend the S cable on the P's. Alternatively, the MDR-CD3000 is a possibility... it'll work, although it'll look really weird hooked up to the Mini.


Although Taurui has a point too... In which case the iPod earphones should be what you shuld use.

 

[aroon]

my initial thought was to go with ety's i have a pair myself and i love them but then...

they go IN your ear. to use them with a large number of people id have to either clean them or replace the tips for every subject which costs money, is a pain, and frankly, could be potentially GROSS. so im leaning away from the in-ear/just normal ear buds.

also, the reason i would like some high quality cans is so that the only thing determing how the users perceive the quality of the song is the bitrate itself. if, for example, i used the stock ipod earplugs [which, for the afformentioned reason, i wouldnt, but i digress] once you get to a high enough bitrate you might not be able to hear the difference not because it doesnt exist or you, as a person/subject, couldn't tell the difference, but because the earphones just arent good enough to portray a such a minute difference.

edit- btw, i know the ipod doesnt output perfectly flat sound but accorinding to many people's experiences it is very good quality and im assuming any defects can be neglected. keep in mind this survey will be targeted to the avg consumer [namely people with ages from 18-40, most of whom will be undergraduate psych students].

i hope you understand what im saying =]

so to summarize, im looking for some high quality cans that can be driven without an amp by an ipod mini for around 250 bux or less.

 

[bangraman]

Ultrasone HFI-650 is an accurate monitor. It's isolated so environment will be less of an issue.

 

[WmAx]

The Sony MDR-V6(Not MDR-V600), also known as the MDR-7506(same but with higher price) is a good feed monitor. It is a resolving headphone as well. However, it does not seem to be favored for it's tonal balance on general commmercial recordings, even though the tonal balance seems to me to be nearly as close as one can get in comparisons of live acoustic vs. the headphone, using linear measurement microphones and standing at the same distance as the microphones from the live sources. The MDR-V6 can be purchased for approx. $70 USD each new or you can find factory refurbished units for as low as $50 USD each. The unit is extremely efficient, thus can be powered adequately from nearly any portable audio device.

-Chris

Quote:

Originally Posted by aroon Hey guys, long time since ive been around. I had an idea recently to do a study on how well people can actually tell the difference between different bitrates of encoded music [i.e. ordering 64, 128, 192, 256, and 320kbs mp3s of the same song by perceived quality] I'm trying to configure my source setup. I need a setup that will provide good clean sound [of course] for less than 500 dollars. My current though is an ipod mini [less than 250 thanks to the academic price] and pair of good headphones. I was looking at the Senn HD580's which, when coupled with the ipod mini, maintains my budget. I'm worried, however, about how well the ipod mini can power the HD580's. at 300ohms, it sounds like the mini could have some issues. anyone have any experience with this specific pair? any other advice such as using a different source, different headphones, or both? I really want to steer clear of using any amps to help power the headphones since a) it introduces all kinds of questions about altering the sound in a subjective way to the listener and b) purchasing amps will almost definiately throw me voilently and ruthelessly out of my budget! thanks for the feedback in advance!

 

[intlplby]

what is your methodology for the study going to be?

what exactly is the variable you are measuring? how are you measuring it?

will they compare different level bitrates to each other or to a lossless source?

are you going to use music? or are you going to use test sounds?


be aware that a subjects lack of familiarity with a certain genre of music will make them less able to pick out certain details..

i.e. use classical music with a population, such as older adults, where most people are familiar with and listen to classical music regularly will most likely result in higher selected bitrate.... i.e. they may percieve no difference between 256 and 320

in contrast to:

use classical music with a population, such as college students, where most people are unfamiliar with or don't listen to classical music much if at all will most likely result in a lower selected bitrate..... i.e. they may percieve no difference between 128 and 192.......

you will most likely have to have 2 factors:

bitrate - continuous - 5 levels
music type - nominal - N levels (where N is number of different songs , all of which are in different genres of music)

e.g. 5 levels - Jazz, Rock, Punk, Classical, Opera....


make sure you choose high quality recordings.... studio copies... remastered if available....

it may even be good to include a worksheet where the subject rates their familiarity with each genre on a lykert scale (5 or 7 point)..... or a scale where they put the 5 genres in order with 1 being the genre they are most familiar with and 5 being the genre they are least familiar with

another option may be to have them listen to sweeps or pulses are various volumes..... perhaps combine the sweeps or pulses with pink noise and ask them when they can hear the pulse or sweep


you could start a sweep or pulses and give them a hand buzzer and everytime they hear the sweep start they press the button and you record response time from the start of the sound file to when they first heard the sweep or pulses....... pulses would work better because you would have more defined levels and if you left enough space between the pulses you could compensate for different lag times between subjects


are you using DOE (design of experiments)?

if so what kind of design are you using or what model are you using? taguchi array, response surface design or something else?

is the bitrate going to be the only factor?

what statistical software package are you using? SPSS or JMP?

how many subjects do you expect to get?

what is the power you are trying to achieve? 1% or 5% or some other power?


that is everything i can think of right now

 

[intlplby]

ps .... how are you going to maintain the same volume across all subjects?


volume may be another factor to worry about.......

also ambient noise will be environmental noise

ipod battery voltage will be an issue and amp voltage could become part of environmental noise

if you want to be able to adequately measure the difference between these it will be important that you get a pair of phones that are geared towards detail....

coloration isn't so much of an issue........

you also need to be aware that without an amp you won't be able to get full reproduction of the source...... and since most mp3 algorithms are geared towards clipping at the top and the bottom that is where you are really going to be measureing the difference.....

i don't see meauring the highs to be too much of a problem but measuring percieved bass will be without a good setup....

it may be best to see if there is a fellow head-fier in your area that may be willing to lend you his equipment.....

 

[ampgalore]

ER4S and CD3000 are worthwhile choices. Don't use the HD600/650 for monitoring, they color the sound, but they are pleasant colorations.

 

[intlplby]

i dont think sound coloration is an issue.... it appears he is measuring percieved detail differences..... coloration should only be a problem if the phones color the sound differently between bitrates.....


does the soundstage change when you change bitrates?

 

[aroon]

Quote:

Originally Posted by intlplby what is your methodology for the study going to be? what exactly is the variable you are measuring? how are you measuring it? will they compare different level bitrates to each other or to a lossless source? are you going to use music? or are you going to use test sounds?

i dont know all the details of the study yet, im just considering it all and i was bored at work so i thought id start on the fun part, looking at equipemt =]

my thought was to have them compare the same song [as in music] at different bitrates and JUST order them in increasing quality as they perceived it without a lossless version to compare to.

Quote:

Originally Posted by intlplby be aware that a subjects lack of familiarity with a certain genre of music will make them less able to pick out certain details.. i.e. use classical music with a population, such as older adults, where most people are familiar with and listen to classical music regularly will most likely result in higher selected bitrate.... i.e. they may percieve no difference between 256 and 320 in contrast to: use classical music with a population, such as college students, where most people are unfamiliar with or don't listen to classical music much if at all will most likely result in a lower selected bitrate..... i.e. they may percieve no difference between 128 and 192....... you will most likely have to have 2 factors: bitrate - continuous - 5 levels music type - nominal - N levels (where N is number of different songs , all of which are in different genres of music) e.g. 5 levels - Jazz, Rock, Punk, Classical, Opera....

yup. ive thought about that and im not sure what im going to do. im thinking of having them repeat the experiment for multiple generes. using one genre for each age group isnt an experimentally sound imo.

Quote:

make sure you choose high quality recordings.... studio copies... remastered if available.... it may even be good to include a worksheet where the subject rates their familiarity with each genre on a lykert scale (5 or 7 point)..... or a scale where they put the 5 genres in order with 1 being the genre they are most familiar with and 5 being the genre they are least familiar with another option may be to have them listen to sweeps or pulses are various volumes..... perhaps combine the sweeps or pulses with pink noise and ask them when they can hear the pulse or sweep you could start a sweep or pulses and give them a hand buzzer and everytime they hear the sweep start they press the button and you record response time from the start of the sound file to when they first heard the sweep or pulses....... pulses would work better because you would have more defined levels and if you left enough space between the pulses you could compensate for different lag times between subjects

i really just want to see whether the average joe can really tell the difference between different bitrates, and what is the bitrate where they really cant tell a difference anymore. now that i write this, i think it might be logical to include a lossless version of the song in the experiment just to see if people rank it lower than some lossy version.

Quote:

are you using DOE (design of experiments)? if so what kind of design are you using or what model are you using? taguchi array, response surface design or something else? is the bitrate going to be the only factor? what statistical software package are you using? SPSS or JMP? how many subjects do you expect to get? what is the power you are trying to achieve? 1% or 5% or some other power? that is everything i can think of right now

HAHA! im no stats major. in fact, im a sophmore cs major so you're over my head in this area. bitrate [idealy] is going to be the only factor. as far as software im not sure, may be SPSS but only because thats all ive ever heard of. i plan on finding a stats professor to mentor me through this so hopefully he/she can help me out in that area...and all this other stuff! i definately have over 100 subjects, may be over 200. freshman pysch majors are required to participate in at least 4 experiments so ill be getting a lot of them =]

to answer a bunch of other questions:

im on a budget not because thats all the money i have but because i can only get up to 1000 dollars in funding from a grant. borrowing equipment isnt really an option. if i REALLY need the money there are ways i can get more funding [up to 2000] but i really think i can do this in under 1000.

as far as sources, i just wanted to hunt down some well recorded cd's. its not the best but its not expensive, and they're readily available.

as for the environment, ill definately have a room just for the experiment. CMU has lots of small, solid, almost sound proof rooms so i dont think that will be a problem. I do realize sound isolation is a big deal though.

 

[intlplby]

i think the pink noise/pulses is your best bet if you want to get good reliability and accuracy



the pink noise will simulate the music and the pulses will simulate the details that the person is listening for..........


as far as learning the statistics or finding someone to help you with it i recommend reading six sigma quality control information to learn it..... my experience with psychology professors is that most are no where near as proficient in using statistics as well as engineers and quality control people.....

find out the issues for dealing with human subjects with regard to psych after........


i have a psych degree from a fairly decent pscyh program and i will tell you that honestly most psych people really don't know statistics all that well

JMP is a much much better program than SPSS but most people are only familiar with SPSS...... JMP is made by SAS ... you may be able to get a student copy

make a CMoy for about 30-50$

you may even want to consider buying the tpa6120a evaluation module to use as an amp....... everyone says it works real well with high impedance phones like the 580 and the EVM package is 50$

the amount a good amp will colour the sound is far less than the issues brought about by underpowered phones


one option may be to feed the MP3 into one ear and the lossless into the other ear and ask them to tell the different...... make sure that they are listening to a mono recording fed to both channels or flatten the stereo into mono and feed to both channels

when doing the DOE do a balanced design where half the subjects listen to it Right ear lossless and left ear mp3 first and then right ear mp3 and left ear lossless and the other half listens it vice versa



if you got college freshman and softmores most students will be 18-20 years old


if you do go for music go with pop music most everyone will be familiar with....

i hate to suggest some of the following artists but it may be your best bet because it guarantees most people will be familiar

britney spears, puff daddy, beyonce, coldplay, dido, etc

if you don't go with current pop music choose artists like

the rolling stones, pearl jam, jimi hendrix, the who, the beatles, etc


for classical music choose well known pieces like:

beethovens 9th
mozarts requiem
wagners ride of the valyries
etc

bitrate may be the only factor but you need to check for other things that may color your results


something most psych people never look at are the residuals and i've never met a psych person that has even heard of DOE (heck, most engineers and scientists haven't heard of it).... DOE is an extremely valuable tool... if you dont see it in your psych classes make sure to learn it on your own

 

[pnrgi]

The a900s have great sound quailty at 200$ and you dont need an amp.

 

[John Reeves]

Hi,

You are probably aware that as people age their sensitivity to high frequencies deteriorates. In order to ensure that this is not a critical factor, which affects your results, you will need to carry out an audiogram test on each subject. You can probably get a source for this and carry out the tests yourself.

Additionaly, audiogram results will vary for a particular individual if they have a wax build up in their ear or if they are suffering from a cold, sinus problems etc.

You will need to examine each subjects ear to ensure this is not the case. Maybe if this is a college project you can enlist the aid of medical services or something to give each subject a quick check.

I am only aware of this because my son suffers from a slight defficiency in hearing at certain frequencies and has had audigrams taken over the years. Fortunately he seems to have stabilised, he's now 13, and the problem is only minor.

Anyway he's got good taste: Led Zep, Linkin Park, Nirvana etc and I've always wondered if this slight defficiency has affected his musical tastes or is it my extremely excellent influence.

Best of luck and looking forward to "hearing" your experiences....pun, LOL.

Cheers
John