Head-Fi.org › Forums › Equipment Forums › Dedicated Source Components › Cary 303/300 - my upsampling blind test
New Posts  All Forums:Forum Nav:

Cary 303/300 - my upsampling blind test

post #1 of 22
6/18/06 at 11:10pm
Thread Starter 
After I upgraded my amp the difference between upsampling levels has become bigger. I know I'm not imagining things anymore, and this blind test confirms that to me. I only did 6 tries in this test because I could have gone on for days while still staying above average even if I had kept guessing all the next answers.

Cary 303/300 CD player has 44.1, 96, 192, 384, 512, 768 kHz upsampling levels. Answering correctly 6 times in a row is 1 chance in 46656. I scored 4 of 6 which is 106 chances in 46656 or 99% accuracy. Was it luck?

_________________________

Blind test #1 - Upsampling 6 levels (Random CD and track)

1) 768 kHz. My answer = 768 kHz CORRECT

2) 96 kHz. My answer = 96 kHz CORRECT

3) 44.1 kHz. My answer = 512 kHz WRONG (this was a very smooth/muddy track unlike the others)

4) 384 kHz. My answer = Could not decide between 192 - 96 - 44.1 kHz WRONG (previous smooth track messed me up)

2 hours listening to 768 kHz, different CD, 'DJ Tiesto - Magik One'

5) 512 kHz. My answer = 512 kHz CORRECT

2 hours listening to 768 kHz, different CD, 'DJ Tiesto - Magik Two', and 'Magik Three'

6) 512 kHz. My answer = 512 kHz CORRECT

/Ended test because of fatigue

Result: 99% accuracy.
_________________________

I have used 768 kHz upsampling for over 8 months, lower levels hurt my ears because it's too harsh. I don't have "golden ears", I can only hear a difference between sources if they have different upsampling levels, other than that, I hear no other differences. However, I do hear a difference if they are both plugged into the wall with stock cords and without isolation feet.

P.S. Cables, power conditioning and vibration isolation make a bigger difference than 44.1 vs 768 kHz upsampling. They are not snake oil.
Reply
post #2 of 22
6/19/06 at 12:28am
At the very least, it wasn't "99.8% accuracy".

The probability of getting 4 correct answers by rolling a die would be:

(6 choose 4) * (5/6)^2 * (1/6)^4 = (5^3) / (2* 6^5) > .008

So, there goes your "99.8% accuracy". It is < 99.2%

Moreover, you probably would have been even more content with 5 or 6 correct answers -- so, we should add in those (smaller) probabilties also.

(Even more so, imagine for the sake of argument the following: you confuse 44.1 kHz with 512 kHz and vice versa every time. Do you think that would still be evidence for showing an audible difference, even though you would technically be wrong in each instance?)
Reply
post #3 of 22
6/20/06 at 9:53pm
Thread Starter 
This can't be luck now.

Blind test #2 - Upsampling 6 levels ('Hallucinogen - Twisted', track4, haven't listened to this CD for weeks)

1) 192 kHz. My answer = 192 kHz CORRECT

2) 44.1 kHz. My answer = 44.1 kHz CORRECT

3) 96 kHz. My answer = 96 kHz CORRECT

4) 768 kHz. My answer = Could not decide between 384 - 512 - final answer = 384 kHz WRONG
(I was too excited and it messed me up, excitement makes it sounds more harsh? It sounds very smooth when I'm relaxed.)

/ Ended test because of fatigue
Reply
post #4 of 22
6/20/06 at 9:57pm
luck, placebo, too few tries, fake.
Reply
post #5 of 22
6/20/06 at 10:08pm
Thread Starter 
Quote:
Originally Posted by Hello
At the very least, it wasn't "99.8% accuracy".

The probability of getting 4 correct answers by rolling a die would be:

(6 choose 4) * (5/6)^2 * (1/6)^4 = (5^3) / (2* 6^5) > .008

So, there goes your "99.8% accuracy". It is < 99.2%

Moreover, you probably would have been even more content with 5 or 6 correct answers -- so, we should add in those (smaller) probabilties also.

(Even more so, imagine for the sake of argument the following: you confuse 44.1 kHz with 512 kHz and vice versa every time. Do you think that would still be evidence for showing an audible difference, even though you would technically be wrong in each instance?)
I must have made it too simple. I don't know formulas because I quit school when I was 14. I did it the "raw style" and calculated all the possibilities of getting 4 or more correct.

.008 times 46656 is 373.248 possibilities, shouldn't it be an even number?
Reply
post #6 of 22
6/21/06 at 1:59am
Quote:
I have used 768 kHz upsampling for over 8 months, lower levels hurt my ears because it's too harsh. I don't have "golden ears", I can only hear a difference between sources if they have different upsampling levels, other than that, I hear no other differences. However, I do hear a difference if they are both plugged into the wall with stock cords and without isolation feet.

P.S. Cables, power conditioning and vibration isolation make a bigger difference than 44.1 vs 768 kHz upsampling. They are not snake oil.
Well... if you can hear differences in power cables and isolation feet (I know I can't), its difficult to imagine that you'll miss differences between sources :/
Reply
post #7 of 22
6/21/06 at 2:39am
Thread Starter 
Quote:
Originally Posted by Tachikoma
Well... if you can hear differences in power cables and isolation feet (I know I can't), its difficult to imagine that you'll miss differences between sources :/
My power, vibration and RFI conditions are really bad where I live, probably the worst in this forum.

1) Lights sometimes blinking. Dimmer gives audible sound that changes sometimes.
2) Vibration on floor that made table give audible humming sound. Replaced legs of table and humming went away.
3) Radio broadcast towers 3 km from window. My first amp (Corda HA-1 Mk1) picked up radio stations clearly. Projector screen blinking often. Cordless mouse suddenly stops working (gets jittery) at certain times of the day, after 20 mins it works perfectly again.
Reply
post #8 of 22
6/21/06 at 8:31am
I currently have a Cary 303/200 which has selectable up-sampling to 96 kHz.
I can always pick out when the up-sample is engaged. Listen to the cymbals and other high frequency information. With up-sampling on, the decay of the cymbal crashes last longer, and you hear an airier more open sound.

Once you pinpoint what it does it's easy to hear. You may pick up on some other aspect of the sound besides the highs, but that's the one I hear the easiest.
TR
Reply
post #9 of 22
6/21/06 at 9:16am
How was your experiment designed? Was it randomised, double blind or single blind, were you listening to the same track, etc...

In my limited upsampling experience, yes, the individual sounds are better defined, but the ambience and the illusion of soundstage collapses. Upsampling is a digital processing step, thus if done badly will not serve to improve the actual sound fidelity (source signal conversion fidelity), at best it will not damage it.

Remember, this upsampling thing does not add any new information, it verifiably just moves the HF filtering much further up the frequency range (beyond the CD cutoff of 20 kHz) thus giving you more HF information which as far as it is known to the medical world is inaudible beyond 20kHz in the best case.... gets worse with age.

Cheers,

Victor
Reply
post #10 of 22
6/21/06 at 9:40am
Quote:
Originally Posted by vmajor
How was your experiment designed? Was it randomised, double blind or single blind, were you listening to the same track, etc...

In my limited upsampling experience, yes, the individual sounds are better defined, but the ambience and the illusion of soundstage collapses. Upsampling is a digital processing step, thus if done badly will not serve to improve the actual sound fidelity (source signal conversion fidelity), at best it will not damage it.

Remember, this upsampling thing does not add any new information, it verifiably just moves the HF filtering much further up the frequency range (beyond the CD cutoff of 20 kHz) thus giving you more HF information which as far as it is known to the medical world is inaudible beyond 20kHz in the best case.... gets worse with age.

Cheers,

Victor
I've always found up-sampling imroved the sound....with this CD player.
Reply
post #11 of 22
6/21/06 at 11:36am
Thread Starter 
Quote:
Originally Posted by vmajor
How was your experiment designed? Was it randomised, double blind or single blind, were you listening to the same track, etc...
It was a randomized single blind test, it's accurate unless I have a 6th sense or something. I listened to a random CD and track which I don't know very well. When I listened to the same track I scored correctly more often.

Quote:
Originally Posted by vmajor
In my limited upsampling experience, yes, the individual sounds are better defined, but the ambience and the illusion of soundstage collapses. Upsampling is a digital processing step, thus if done badly will not serve to improve the actual sound fidelity (source signal conversion fidelity), at best it will not damage it.
There is more ambience detail and the soundstage becomes bigger when I use higher upsampling. Perhaps no upsampling is a little more transparent because it bypasses the upsampling chip which makes the signal path shorter, but it sounds too harsh and it has less detail. 768 kHz sounds the best, it's barely smooth enough and I can listen for 16 hours without fatigue.
Reply
post #12 of 22
6/21/06 at 12:12pm
Quote:
Originally Posted by Patrick82
There is more ambience detail and the soundstage becomes bigger when I use higher upsampling.
You can't "produce" more details... Upsampling makes it easier to remove clock noise without treble rolloff and sometimes it sounds smoother due to dithering techniques. Nothing wrong in enjoying this "flavor"...
Reply
post #13 of 22
6/22/06 at 7:13pm
Thread Starter 

#3

Blind test #3 - Upsampling 6 levels ('Sylver - Little Things', track2, haven't listened to this CD for weeks)

1) 512 kHz. My answer = 384 kHz WRONG

2) 768 kHz. My answer = 768 kHz CORRECT

/ Track ended before I could try 3rd time


Okay, now I have done 12 trials and got 8 correct, any formula guys who can calculate?
Reply
post #14 of 22
6/22/06 at 7:38pm
EDited out...for now
Reply
post #15 of 22
6/22/06 at 8:20pm
Quote:
Originally Posted by Patrick82
Blind test #3 - Upsampling 6 levels ('Sylver - Little Things', track2, haven't listened to this CD for weeks)

1) 512 kHz. My answer = 384 kHz WRONG

2) 768 kHz. My answer = 768 kHz CORRECT

/ Track ended before I could try 3rd time


Okay, now I have done 12 trials and got 8 correct, any formula guys who can calculate?
Well by my calculations the chances of guessing 8 out of 12 right are:
[Combinations of 12 items taken 4 at a time] * [(5/6)^4] * [(1/6)^8]
- The first factor is the number of ways you can miss 4 and get 8 right, given 12 trials
- The second factor is the chances of getting 4 wrong
- The third factor is the chances of getting 8 right

This reduces to: (12*11*10*9 * 5^4) / (4*3*2 * 6^12) which is about 0.000142 or about 0.014%

Of course, you must add in the chances of getting 9 right, 10 right, 11 right, and 12 right, but these are quite insignificant.

Your results ARE statistically significant.

EDIT: I slightly miscalculated - the first factor should be the combination, NOT the permutation. 2nd poster Hello got that right in his calculation - I just thought through this too hastily the first time. Anyways, this reduced my first number further by a factor of 4!, or 24. Calculations above modified to reflect this.
Reply
New Posts  All Forums:Forum Nav:
  Return Home
  Back to Forum: Dedicated Source Components
Head-Fi.org › Forums › Equipment Forums › Dedicated Source Components › Cary 303/300 - my upsampling blind test