In our study we adjusted the levels of the distorted signals to have the same loudness. Not doing so gives a que which confounds the results.
gedlee said:
I used to believe that few people would buy sight unseen. I never thought I would do it. In fact, it was the building of my first few own designs that I began doing so, because I began to learn what to look for in measurements to correlate with what I liked in sound (As well as understand what characteristics were important for good sound). I also have noticed huge trends in buying equipment by mail order, it's being considered one of the number one reasons for the failure of brick and mortar stores, so apparently tons of people, more than not, are buying without hearing first. Certain companies have gone to mail order only and hearing their speakers before buying is either impossible, or only possible by finding someone else who has them.
So in short, I did, and I am a happy customer.
Oh and I now realize my comment earlier was wrong, I spoke without thinking it through, you can get 9 out of 10 right and be better than chance, and you can be 8 out of 10 and be awfully close. With 20 trials, it's 15 out of 20 that is better than chance. None the less, if we were to conclude from my results, it would be that with averaging of all results, it was not detectable as better than chance, but that in one longer attempt, one subject could. It would not be conclusive, since I didn't run everyone through identical controlled conditions (once being successful with 20, I would really need to go back and run the others through the 20 as well).
Also sounds like we might have the results confounded though, as the tracks do have a 1db average gain difference between them (apparently do to the distortion processing), which may account for peoples ability to reliably discern the tracks better than chance.
Ok I now finished compiling the trials of 10 or so people using these tracks and calculated a p value corrected for the shear number of trials. The correct P was .67 and uncorrected was .69.
pjpoes said:
Hi!
I really don't know what's the weak point in your set up but as I have mentioned earlier I would no start out with monitors that have distortion in the same levels as the signal. It's hard to say how good the test equipment have to be but experience has shown that high performance often is beneficial when doing research work. Possibly the shure headphones that you have is mediocre distortion wise?
After Earls claim about the in ear phones he used in his test, saying those are "research tools with very low distortion" (or something like that) I checked this site http://www.headphone.com/technical/product-measurements/build-a-graph/ and all shure and Etymotic research in ear phones had relatively poor performance.
I'm close to 40 and I do have perfect (and in some tests better than average for youngsters) hearing without signs of aging.
I guess so. The rms value is different between the files and so is peak value. the original file has lower rms but higher peak (dynamics intact = higher crest factor) just as one can expect from gross compression.
/Peter
MartinQ said:
Martin, I have downloaded the longer files but not done the test yet. Maybe later today.
Would that be like white/pink noise or crackling or clicking?
If the latter look at ASIO drivers and buffer settings. Also you may need to switch of wireless networks of your computer and make sure there's nothing else on teh same bus and also check for IRQ conflicts.
/Peter
gedlee said:
Why would a dipole line source not be a good aproximation to CD? if you slap some small/medium size pistons to a narrow baffle from floor to ceiling you can easily have CD from the lowest frequencies up to 2-3kHz which is way better than your solution and also in a more important frequency range. Add a full length ribbon to that.
Becasue you claim things that is plain wrong, claim having evidence for some things that is merely your opinion and not facts. Becasue you refer to flawed tests as proof for your erroneous claims. Also because you often do this with a bad attitude and chest tumping. Also because several times instead of meeting good arguments you back out of the discussion seeminlgy as an attempt to "safe face" ofte with some less than friendly coment to your "oponent" about lack of respect for your superiority and expertise.
Yea, your room is built the way a roomshould be, your speakers are the best, you are so unique because you have looked into things that no one else have..
Or maybe not!
/Peter
One thing about testing that I do not remember if it has been adressed in this thread allready..
In the FooABX, when you have loaded the files you can listen to the full length of the files or you can chose a short sequence. If you have a gut feeling that there is some difference but have a hard time pointing it out, mark a short sequence and listen to that.
Also try to zoom in on different parts of the sound. Listen to attacks, then listen to decay, listen to ambience and so on.
When a small group of people take tests together one can start out with open tests and help eachother pointing out and finding the artefacts and that will up the sensitivity of the test.
Hifi to me is about reducing all non musical arterfacts from the signal chain. Sure one must put things into perspective but if I can get rid of some coloration, I'd like to do that. Also things add up so if you improve every link in the chain the summed improvement will be bigger than what you get from just improving or "perfecting" on link in the chain.
/Peter
In the FooABX, when you have loaded the files you can listen to the full length of the files or you can chose a short sequence. If you have a gut feeling that there is some difference but have a hard time pointing it out, mark a short sequence and listen to that.
Also try to zoom in on different parts of the sound. Listen to attacks, then listen to decay, listen to ambience and so on.
When a small group of people take tests together one can start out with open tests and help eachother pointing out and finding the artefacts and that will up the sensitivity of the test.
Hifi to me is about reducing all non musical arterfacts from the signal chain. Sure one must put things into perspective but if I can get rid of some coloration, I'd like to do that. Also things add up so if you improve every link in the chain the summed improvement will be bigger than what you get from just improving or "perfecting" on link in the chain.
/Peter
gedlee said:
I think you didn't understand what I was trying to say. I have read Toole's book and I concur with practically all of what he says. In designing loudspeakers, to me it's the measurements that count. Get those right, and the speaker will sound right. So, I'm not in that 5% crowd you're talking about. That 5% mainly consists of people with impaired hearing and my hearing is still fine at this moment.
What I was trying to say is: You can have a well measuring conventional box speaker (take for example a Revel Studio, which shows some of the best measured curves I have seen), a well measuring dipole (I believe the Linkwitz Orion measures quite favorably) and you could have the Summa (which I have now seen, also has some nice curves). Now for the points' sake, assume none of these three speakers had any serious resonances. How could anyone reading the measurements, not ever having listened to a dipole or a CD-speaker like the Summa, know how these two would hold up against the Revel?
When I first listened to a well measuring dipole, I had already heard plenty well designed box speakers. Still it sounded vastly different. There was no way of knowing exactly how it would sound, just by reading the measurements. Whether one of both concepts is better than the other or not, might be a matter of taste. So the point is, because the Summa is so different from almost everything else on the market, you DO have to listen before you know how it sounds.
Sometimes it is a matter of finding the right sample. Listen to these Bach samples with a very pure soprano voice. The distorted one (same 5% again) just doesn't sound pure anymore.
http://home.kabelfoon.nl/~dezaire/Bach (1).wav
http://home.kabelfoon.nl/~dezaire/Bach (2).wav
http://home.kabelfoon.nl/~dezaire/Bach (1).wav
http://home.kabelfoon.nl/~dezaire/Bach (2).wav
MartinQ said:
I loadeed the Lizt files but there are some ticking sounds in them which sounds like clocking problems or something.
Did you notice that jeroen_d?
Thanks for your work here but could you please use stereo files? Or perhaps the software you use is unable to process stereo?
Absolutely and the right sample for one person may not be the right for another, we have different sensitivities to different types of artefacts. Just as with visual phenomena and distortion.
I find it a little harder with the piano and voice compared to the strings in the first posted sample files. More on that later.
/Peter
Anyone having problems with Foobar ABX and ASIO with mono files?
I must chose another driver than ASIO because on my system it doesn't work with mono and I prefer to use ASIO to make sure the windows mixer is not involved.
As it is now the DS drivers I have to use instead of ASIO with Foobar results in cliks/interupts but the artefacts I reported about in the Lizt files are in the files though (at the same place in both files when you repeat).
/Peter
I must chose another driver than ASIO because on my system it doesn't work with mono and I prefer to use ASIO to make sure the windows mixer is not involved.
As it is now the DS drivers I have to use instead of ASIO with Foobar results in cliks/interupts but the artefacts I reported about in the Lizt files are in the files though (at the same place in both files when you repeat).
/Peter
Ha, those ticking sounds are really from the recording! Could be the pedal, I don't know.
AddDistortion can only process mono.
For me also the strings worked better than the piano, but the soprano voice was the best sample for me. Also when it concerns preference, the distorted soprano just doesn't sound right.
AddDistortion can only process mono.
For me also the strings worked better than the piano, but the soprano voice was the best sample for me. Also when it concerns preference, the distorted soprano just doesn't sound right.
Another reason to use stero material is that our ability to hear some sounds and signals depends on how the sounds reach our both ears.
There are certain masking phenomena going on and I'm not sure how that would influence a test like this but since we never have identical sounds to both ears in real life it does not makes much sense IMO to use mono material for a test like this, especially when using headphones (which most of us seems to do at the moment).
It's also more pleasing and easier on the ear with stereo.
/Peter
There are certain masking phenomena going on and I'm not sure how that would influence a test like this but since we never have identical sounds to both ears in real life it does not makes much sense IMO to use mono material for a test like this, especially when using headphones (which most of us seems to do at the moment).
It's also more pleasing and easier on the ear with stereo.
/Peter
Hehe, how can such distorted sounds get thru mixing/mastering?
It's not the pedal, it's some artefact from a damaged tape or digital clocking problems. Probably the former.
I have a wonderful CD with Loreena McKennitt where one song is "ruined" by similar clicks. :-(
/Peter
It's not the pedal, it's some artefact from a damaged tape or digital clocking problems. Probably the former.
I have a wonderful CD with Loreena McKennitt where one song is "ruined" by similar clicks. :-(
/Peter
jeroen_d said:
One solution is to import a stereo file into Audacity and split it into two mono files, process in AddDistortion and then go back to Audacity to put them back into stereo.
/Peter
Anyone using foobar's abx tool and having concerns about loudness differences may run replaygain on the files and enable replaygain processing during ABX.
Replaygain analyzes audio based on some model of perceived loudness and generates a gain value to match different audio files. Offline tools are also available so anyone supplying distortion comparison could preprocess the files to have same perceived loudness.
Foobar comes with mono to stereo converter DSP, which just copies the track to the other channel (not a phony pseudostereo algorithm). Foobar DSP chain can be used with the ABX tool too.
Replaygain analyzes audio based on some model of perceived loudness and generates a gain value to match different audio files. Offline tools are also available so anyone supplying distortion comparison could preprocess the files to have same perceived loudness.
Pan said:
Foobar comes with mono to stereo converter DSP, which just copies the track to the other channel (not a phony pseudostereo algorithm). Foobar DSP chain can be used with the ABX tool too.
- Status
- Not open for further replies.