should ReplayGain be off for ABX?
Reply #5 – 2013-11-09 11:09:39
tl;dr: scroll down to the PS, PSS, etc. I would think that if you were to use RG, that "Track" would be more efficient over "Album" if the sources are not from the same album (as intended to be heard). For generic ABX testing, if they are from different sources, use "Track" mode, or not at all. Heck, you can actually set up to three scenarios with ReplayGain: "Track", "Album" and "None". Using "Apply gain and prevent clipping according to peak" is quite reasonable if you want semi-clear results of the ABX test (that is, without any clipping whatsoever, the biggest factor involved). But to ABX against the source in its truest form, without RG, "none" would be more accurate. But then, you and the intended testers would have to go in, fully knowing, that the ABX test is done under the controlled circumstance that no RG was applied (in any manner) and the original mix was provided. If RG was applied, this information would be valuable to the testers as to what they are actually testing. It could influence the testers on a psychological level, but with the "sane" mind, they will understand the circumstances and adjust accordingly. For instance, 3 songs from 3 different artists (say, Rock, Classical, and Jazz), if "Album" applied, I would not participate on the principle that the material to test was not handled appropriately. However, if, for some reason, all three sources (say, Rock, Classical, and Jazz) were of the same nature in musical construction and design, then, perhaps, "album" was an adequate choice and, therefor, provide, somewhat "adequate" feedback from the intended audience of the ABX test. Aiming for ABX means knowing the audience, and, definitely, aiming for the source(s) truest representation, for whatever reason you are ABX'ing. Are you aiming for "codec preference"? Or "transparency"? "ReplayGain"? I'll be more precise: If you intend to include ReplayGain in your ABX testing, make the goal relevant to ABX testing. If you want to find out if more people prefer "Track" over "Album" (given aforementioned knowledge that the test intended for ReplayGain settings), make sure it's mentioned in the... I guess, "introduction" or "description" of the intended ABX testing environment. Testing involves two factors that the most successful feedback: Controlled VS. Uncontrolled. Worrying about "ReplayGain" sounds like an uncontrolled environment. The feedback can only be interpreted as "vague" and inconclusive. You can get a general idea of the intended results. But if you do a controlled environment, you can compare the two (controlled VS uncontrolled) and get much more accurate results. In the case of ABX, controlled is much more likely to yield satisfactory data for your intended purposes. If you want to present criteria without ReplayGain and not notify the listeners of this information, then you are aiming for uncontrolled environments and people will only give you feedback based on what they thought to be, relatively , equal sources for testing. Applying "track" to each source, will, more or less, make it equal on all fields and give decent results for your intended purpose -- which has yet to be declared. The sole purpose of your ABX testing is very unclear. If you could specify your exact intention, then answering the notion of using ReplayGain (at all or even in one method or another) would come into a much more clear answer. As it would seem to me, if I were to be an ABX tester to provide viable feedback to use in a summary of results, based on the intended listening environment. PS: As a tester, I'm not apposed to utilizing both Controlled and Uncontrolled environments to provide the most accurate data for you, the tester, to interpret. PSS: Are you testing for ReplayGain attributes? (Track VS Album?) PSSS: What exactly is your intended purpose of ABX testing? (bitrate, codec, normalization, etc?)