View Single Post
  #2  
Old 11-15-2000, 08:56 AM
evandaum evandaum is offline
Member
 
Join Date: Jun 2000
Location: Los Angeles, CA, USA
Posts: 395
Default Re: extensive test results of 24bit flying calf a/d converter

Did you make SURE the levels were EXACTLY the same on all these transfers? I'd bet what they were hearing were subtle differences in gain, not actual differences between the converters. For this test to work, you have to be VERY anal about the level matching or all is lost.

-Evan


<BLOCKQUOTE><font size="1" face="Arial,Helvetica,sans-serif">quote:<HR>Originally posted by coaster:
yes, its 24bit. they make them that way now.

i found one for DIRT CHEEP on ebay and bought it, new in the shrinkwrapped box.

hooked up to my audiomedia 3 card, i have done some extensive testing of actual data, and subjective response.

my conclusions are scary!

hard testing data suggests a response curve that is poorer than the converters in the am3. this can be measured against various sources running through both converters (midiman and am3)

at this point i was convinced the am3 converters were better.

but i noticed that when i ran a MIX through both sets, the midiman sounded "better".

so i designed the following test to see what my response would be, as well as 10 other people in the audio world, and 5 people NOT in the audio world.

obviuosly there is some innacuracy in my test but bear with me till you hear the results.

the test:

burned a cd of three tracks for playback

track one: bounce to disc stereo file of a large session.

track two: the same bounce, played out the am3 card, recorded into the am3 card analog inputs- no mixer- just a direct connection.

track three: the same bounce, played out the am3 card, routed into the midiman, then into the spdif input of the am3. for this recording the session setup clock was changed to spdif.

listeners were NOT told what they would be listening to, or what the test was about, or even what type of results i was looking for at all. i figure this is the most accurate way to extract non-biased info from them.

the listener is played track 2 (am3conv)

then track three(midiman)

they can then switch back and forth from tracks 2 and 3 for a few listens.

14 people said track three was a better mix.

"why" i ask

it's cleaner, the guitar is louder, it sounds more like a record; were some responses.

1 person said track two was better. he offered no reason why.

at this point in the test i told the listener that they had listened to the same mix twice.

this confused everybody, and they swore they could hear a difference between the two (you can).

i then explained what the test was, that they were comparing two different converters.

they then wanted to know what track one contained.

i did not tell them, but then played all three tracks in a row.

all fifteen people agreed that track one was by far the worst of the three(this is where i get confused- this is the original bounce, and is what the other two tracks are generated from)

14 people said track three was the best, one person said track 2 was the best (the same guy as before)

THEN the final part of the test.

keep in mind that noone knows yet what tracks are what at all so far.

i asked them to identify which track (2 or 3) is more accurate to the sound of track 1.

every one tested could easily tell that track 2 was the most like track one.

so the results are so far:

14 people like the midiman track
1 person likes the am3 track
15 people dislike the original bounce
15 people could tell that the am3 track sounded the most like the original bounce

WOW!

this is not what i expected from this test!

what do you all think of this??????!!!!!

<HR></BLOCKQUOTE>

Reply With Quote