Published by PitchChanger · Updated · How these guides are made
Hear the difference
Original synthesized diagnostic material, authored at 48 kHz, 16-bit PCM. These are controls for the worksheet, not measurements of this tool or real instruments.
Pause your workspace preview before comparing these clips.
Start quietly. Download the WAV files for external comparison; no listening outcome is asserted.
How to use this workflow
Prepare aligned A and B
Use the same passage, match levels and conceal identifying file names. Practice with known examples before scored trials.
Randomize each X
Use a dedicated comparator or a helper who records independently randomized assignments. Do not disclose answers during the scored block.
Finish the planned block
Record one choice per trial and evaluate the entire planned block, including misses. Do not stop when a favorable streak appears.
An illustrative ten-trial score
Under independent fair guessing with two choices, the probability of at least 9 correct in 10 fixed trials is (10 + 1) / 1024 = 1.07421875%. At least 8 correct is (45 + 10 + 1) / 1024 = 5.46875%. These are exact binomial calculations, not results from a listener.
Multiple passages, repeated attempts and choosing a favorable result after the fact change the interpretation. A small guessing probability does not tell you the audible difference is large or desirable. Failure to identify does not establish universal transparency.
Swipe or scroll sideways to read all columns.
| Checkpoint | Action or value | Interpretation |
|---|---|---|
| 10 of 10 | 1 / 1024 | 0.09765625% guessing tail |
| 9 or more | 11 / 1024 | 1.07421875% |
| 8 or more | 56 / 1024 | 5.46875% |
Keep identification separate from preference
The Steady and Quieter files intentionally differ in level. They are practice controls, not a valid matched test of codec quality; correct matching removes the intended difference. Use unrelated training material so familiar filenames do not leak the scored identity.
PitchChanger exports versions but has no built-in ABX randomizer, hidden labels or trial scorer. Run formal trials in an external comparator. ITU-R BS.1116 describes controlled impairment assessment; this simple ABX worksheet is not a claim of compliance with that standard.