So the first order of business was to see how flat the response of each cable was.
First I generated a 40 second clip of white noise , I saved this as a Wav file and ran it through a spectrum analyser,
this is the reference point.
Then I burned the sample to CD and tested each of my CD players by playing it back and recording the analog
output to find the one most like the reference in shape and overall level. I settled on a Denon DCD910 analog
output which had the lowest deviation from the reference from 20 - 20K using stock cables.
Now I set about more critically measuring the response for the 4 cables I bought plus my own 77c specials.
To do this I played back the white noise sample and recorded it to my PC. To cater for random variations
I did each recording with each cable **10** times.
Each resulting wav file (70 of them, I tested the AQ cables aligned by the arrows and unaligned to test for the claimed directionality) was clipped to exactly 24.003628 seconds, each file is precisely
4.03 MB (4,234,284 bytes) all white noise. Due to careful editting ( the same protocol used for each file)
no wav file can differ from any other wav file in alignment by more than 1/1000th of a second.
Each file was then run through a spectrium analyser with 1024 sample points. Each spectrum was exported as a text file.
Then I started up Excel and I loaded each set of ten trials into a separate worksheet. This allowed me to see the Min, Max and Average levels for any frequency. Also it allowed me to see how much variation there was in any set of trials.
For instance the biggest difference between any two trials for any frequency with the Stock cables was 0.056db, the average difference between the max and min value for all frequencies for stock cables was 0.032 db.
Tartan cable 0.032 and 0.014
BlueJeans 0.047 and 0.016
AQ G-Snake Not Aligned 0.061 and 0.016
AQ G-Snake Aligned 0.038 and 0.013
AQ Sidewinder Not 0.050 and 0.014
AQ Sidewinder Aligned 0.044 and 0.014
The way that the recording was set up meant that each recoding was always at a slightly lower absolute level than the reference. Thus the best cable by definition at any frequency would have the highest value or the lowest attenuation from the reference. This made life easy.
In no case did any cable boost any frequency above the reference level.
Each set of 10 trials for each cable was averaged and the results fed into the worksheet with the reference values.
The maximum difference found between any two cables at any frequency point was 0.029db. The average difference between the best and worst cable across all frequencies was 0.012db. Note that the identity of best and worst cable did differ between frequencies.
Two of the cables I purchased are supposed to be directional. I recorded samples using the Audioquest G-Snake and Audioquest Sidewinder both correctly aligned and incorrectly aligned. I recorded 10 samples with each of the 4 combinations.
The G-Snake correctly aligned was in fact measurably different from the G-Snake incorrectly aligned. When I fed the results into SPSS the result was significant.
the maximum difference at any frequency was 0.022db and the average difference was 0.001db. For the Sidewinder the maximum difference was 0.011db and the average difference at all frequencies was 0.002db.
Do any of these cables roll off any more than any others. I tested the differences in level between 1K and 20K, 2K and 20K, 3K and 20K, 4K and 20K, 5K and 20K, 6K and 20K, 7K and 20K, 8K and 20K, 9K and 20K, 10K and 20K, 11K and 20K, 12K and 20K, 13K and 20K, 14K and 20K , 15K and 20K and finally between 16K and 20K.
The maximum difference in roll-offs between any two cables for any two frequencies was 0.016db
Stock(77c) vs Sidewinder ($60) , max diff 0.020db, ave difference 0.009 db
Stock(77c) vs G-Snake ($35) , max diff 0.013db, ave difference 0.003 db
Stock(77c) vs Blue Jeans($26.75), max diff 0.014db, ave difference 0.004 db
Stock(77c) vs Tartan ($5) , max diff 0.009db, ave difference 0.003 db
Tartan ($5) vs Sidewinder ($60) , max diff 0.014 , ave difference 0.006db
Tartan ($5) vs G-Snake ($35) , max diff 0.009 , ave difference 0.003db
Tartan ($5) vs BJC ($26.75) , max diff 0.007 , ave difference 0.002db
BJC($26.75) vs Sidewinder($60), max diff 0.011 , ave difference 0.005db
BJC($26.75) vs G-Snake ($35), max diff 0.012 , ave difference 0.004db
Sidewinder vs G-Snake max diff 0.018 , ave difference 0.009db
I just loaded my dataset up into SPSS and ran a repeated measures ANOVA with cable type as a covariate and some ANOVAs between cables.
The results show that for these cables there was no significant effect on changing cable, no cable was significantly different from any other. Frequency also did not alter the result, no cable was significantly different at any frequency.
The other result was that there was much more variation between trials than between cables, this is entirely predictable as a cable is a passive device and and ADC is not, so my decision to average 10 trials was justified. Amusingly the repeated measures variation was significant. Remember this was a variation of no more than 0.056db between any two trials for any given cable at any frequency, but when you have 4600+ data points little differences become statistically significant.
I just tested the Sidewinder vs the stock cable using a cymbal crash with lots of harmonics and transients.
Max difference 0.102340db average difference 0.003539db
Average overall signal
Stock------------ Sidey ------------ difference
-37.743888 ******-37.742651 ****-0.001237
This graph shows the stock cable with the sidewinder dropped by 20db so you can see the patterns
For the Anoraks - the distribution of differences against frequency
And finally difference vs signal level
The max variation was 0.08db at 129hz and -54db, the average difference is 0.001408db.
These were done using a five second sample of digital silence and recorded back via the ADC. I ran the tests ( 10 trials) for both stocky and the Zu Oxyfuel.
Average Noise levels Stock -95.7975db.........Zu Oxyfuel -95.8598db
This is quite close to the theoretical noise levels for 16 bit systems
Low frequency noise pattern
I also did a few tests on my $2.11 Monoprice and the $60 Sidewinder. Interestingly each cable has a distinct noise pattern with specific low level spikes at different frequencies.
To place these noise figures in context, here is the noise juxtaposed against the cymbals crash.
In absolute terms the stock cable is the worst for noise with three low frequency spikes between -78 and -76db ,- Monoprice is as good as the most expensive cables for noise, but I will need to run 10 trials.
This text is between a DH labs BL-1 Series II Silver Plated Copper cable and the solid copper Sidewinder with the cymbals sample.
Maximum difference is at 19488hz where the difference is -75.3254db to
-75.3695db i.e a difference of 0.0441db, average difference from 20 to 20K is 0.0014 db. See graph below
Maximum difference at any frequency = -0.00436 db , the Monoprice cable has less attenuation at all frequencies, the average difference = - 0.00266 db.
No graph posted, no point , impossible to see the deviations.
For these tests I made a slight change to my protocol I used a 16K FFT analysis which translates to 7K sample points. Since I have all the original wav files I may redo them with a smaller FFT later if it is a bust.
Noise differences ranged from
Silver Average Noise level = -97.8339 db
Copper Average Noise level = -99.5085 db
Average difference = 1.6746 db
Range of differences from
Freq (Hz) silvercopperdiff
Differences vs frequency
Silver copper diff
Range of differences
Differences vs frequency
Pattern of Frs against Frequency
Edited by nick_charles - 3/1/11 at 3:56pm