Phil Leigh;610116 Wrote: > . > 5) ADM tests for "nulls" - i.e. similarities. The higher the score, the > deeper the null, meaning that the two files are more similar. > So, where the null score is high, no measurable change has occurred. A > lower score indicates a bigger change. A high null score CANNOT happen > by accident. > > Test 1 (the control) > > Separate recordings made a few minutes apart. Nothing was changed. > > Null score=96.96 (as a guide - that's quite a high score/deep null) > > Test 2 - base level mods (kernel & watchdog - not jive2) > > Null score = 96.20 > > Test 3 - as Test 2 + disabling the digital and USB outputs > > Null score = 90.02 > > Test 4 - as Test 3 + disabled wlan > > Null score = 86.99 > > Test 5 - as Test 4 + disabled screen > > Null Score = 84.49 > > Remember: bigger number = smaller difference > > As you can see, the biggest changes were induced by disabling the > unused audio outputs and the wlan. > > Analysing the difference file from test 5 shows that the difference is > not "musical" in content - it is broadband noise (obviously at a very > low level). > This means that there is no change in actual frequency response as a > result of the mods. Nor is there a change in the timing of the > samples. > > Might this be audible to someone? I can't say it will or won't. All I > can say is "it might be audible". > > > I did another test 2 days ago on the new ttvol100 mod...(on top of the > base mods as per test 2 above) > > The null score was 96.23 (so it didn't make any difference - zip, > zilch, nada). My non-scientific assessment of the margin of error in > these test is +/- 0.5 by the way. It's very hard to be precise about > this! > > > Conclusions: > > 1) Some people may be able to hear some changes via the analogue outs, > as a result of some of these mods, depending on their ears and > equipment. > > 2) the vol100 mod does precisely nothing (but only IF you already have > volume set to maximum) - also remember I am testing SBS 7.6 with a fix > for bug 14807. > > Phil I have a couple of questions about your tests and ADM, if you have a moment (and by the way hope you get better soon). Am I right in assuming that the maximum null (ie no change whatsoever)=100? If so I am a bit confused about the control difference being 96.96 and the margin of error being .5 -shouldn't the control difference be related to the margin of error. What else accounts for the control measurement being less than 99.5? Am I right in assuming that the significance of the case 2 difference is marginal? It seems quite close to the control figure, and I am intrigued by how much variation there would be if you repeated the control test at the end. I am not a scientist or statistician obviously, but had a vague idea that in scientific test a significant variation was one of which you are 95% confident that it isn't the result of random variation or other factors- how do we apply that to this data? NB As ever I promise that these are genuine questions however daft or misguided they may appear.
-- adamdea ------------------------------------------------------------------------ adamdea's Profile: http://forums.slimdevices.com/member.php?userid=37603 View this thread: http://forums.slimdevices.com/showthread.php?t=84742 _______________________________________________ audiophiles mailing list [email protected] http://lists.slimdevices.com/mailman/listinfo/audiophiles
