This is a continuation of the post here. I hope some others can give a try and share results here as I think it's very important to find those that can demonstrate unique skill in this area.
I'm using source material from LessLoss--the 24-bit/96k version of Mephisto Waltz Excerpt #1. You might find other source material is better/worse at deducing limits.
On this source material and a Roland Tri-capture DAC, I am able to reliably determine 12 bit truncation (shockingly unimpressive, I know), 0.97 crossover distortion and - 2 dB of amplitude imbalance (also unimpressive, with caveats however). This DAC has some noise floor issues, and I'd hope with another DAC I could do better on the truncation.
For more background, you can read here.
With all ABX testing, don't bother listening to the X stream if you cannot hear the difference between the A and B stream. That is really they key: Listen carefully until you can tell the A stream from the B stream. And then listen to X.
For the bit depth test, the QA_ABX is truncating the source material (24-bit) down to 12-bit on the fly. The key to hearing bit-depth truncation is the noise floor. More volume will help. Listen to the A stream, then listen to the B stream. And make sure you can really hear the differences (and if you are having trouble, go to 8 or 10 bits and work towards 16 bits). And then listen to the X stream and decide if it is A or B. On a spectrum analyzer, truncation appears as a rise in noise floor.
For the crossover distortion, you want to listen for buzziness . Cross over distortion appears as a picket fence of harmonics and thus you'll hear high frequency energy where there was none. Again, listen to A and B streams until you can hear the difference, and then listen to X. The algorithm to introduce crossover distortion is very simple and subtle: Find a crossover, and the multiply the first point after the crossover by a value slightly less than 1. That is it. When I indicate I can hear 0.97 crossover distortion, that means that on all zero crossings the first sample after the zero crossing is multiplied by 0.97.
For amplitude distortion, one of the streams will have its gain changed by an amount you select. In my case, I picked -2 dB and so either the A or B stream was 2 dB quieter. I think I could do better here with source material that had a lower peak-to-average ratio. But piano music with lots of staccato notes amidst a huge dynamic range (playing style) is very hard for me to hear the 2 dB imbalance. With something highly compressed, then 1 dB would be expected.
My results are below. You can take any of these results and paste it in the app's verify window and it should tell you the signature is correct. To copy a test (only do a single test, not the entire group) just grab the text from and including the Q of the QA_ABX all the way down the final pipe ('|') symbol just after the signature that appears after the Finished message.
-- TO VERIFY, CUT STARTING HERE (BUT NOT INCLUDING THIS LINE) --
QA_ABX v0.95
Loaded file: LessLoss-Mephisto-Listz-I-24_bit_96_kHz.wav @ 7/28/2019 1:56:41 PM
File MD5: f2eaff89f99881cebf357b4b6c68ee8b
Distortion: BitDepth Value: 12
Sample Rate: 96000 Bit Depth: 32 Channels: 2 Format: IeeeFloat
Fail @ 7/28/2019 1:57:02 PM
Pass @ 7/28/2019 1:57:19 PM
Pass @ 7/28/2019 1:57:29 PM
Pass @ 7/28/2019 1:57:55 PM
Pass @ 7/28/2019 1:58:01 PM
Pass @ 7/28/2019 1:58:16 PM
Pass @ 7/28/2019 1:58:25 PM
Pass @ 7/28/2019 1:58:35 PM
Pass @ 7/28/2019 1:59:02 PM
Pass @ 7/28/2019 1:59:18 PM
Pass @ 7/28/2019 1:59:34 PM
Finished @ 7/28/2019 1:59:37 PM
|e27725382124b934542c914917273a6c|
-- TO VERIFY, CUT FINISHING HERE (BUT NOT INCLUDING THIS LINE) --
QA_ABX v0.95
Loaded file: LessLoss-Mephisto-Listz-I-24_bit_96_kHz.wav @ 7/28/2019 2:08:33 PM
File MD5: f2eaff89f99881cebf357b4b6c68ee8b
Distortion: Crossover Level: 0.97
Sample Rate: 96000 Bit Depth: 32 Channels: 2 Format: IeeeFloat
Pass @ 7/28/2019 2:09:12 PM
Pass @ 7/28/2019 2:09:39 PM
Fail @ 7/28/2019 2:09:54 PM
Pass @ 7/28/2019 2:10:12 PM
Pass @ 7/28/2019 2:10:25 PM
Pass @ 7/28/2019 2:10:36 PM
Fail @ 7/28/2019 2:11:10 PM
Finished @ 7/28/2019 2:11:58 PM
|20aa354abd41f602cf5b7200c5b8796f|
QA_ABX v0.95
Loaded file: LessLoss-Mephisto-Listz-I-24_bit_96_kHz.wav @ 7/28/2019 2:12:25 PM
File MD5: f2eaff89f99881cebf357b4b6c68ee8b
Distortion: Amplitude Value: -2.00
Sample Rate: 96000 Bit Depth: 32 Channels: 2 Format: IeeeFloat
Pass @ 7/28/2019 2:13:10 PM
Pass @ 7/28/2019 2:13:17 PM
Pass @ 7/28/2019 2:13:32 PM
Pass @ 7/28/2019 2:13:54 PM
Pass @ 7/28/2019 2:14:28 PM
Pass @ 7/28/2019 2:14:41 PM
Pass @ 7/28/2019 2:15:10 PM
Finished @ 7/28/2019 2:15:15 PM
|21c34135be275b697d1b6f55b73801c7|
I'm using source material from LessLoss--the 24-bit/96k version of Mephisto Waltz Excerpt #1. You might find other source material is better/worse at deducing limits.
On this source material and a Roland Tri-capture DAC, I am able to reliably determine 12 bit truncation (shockingly unimpressive, I know), 0.97 crossover distortion and - 2 dB of amplitude imbalance (also unimpressive, with caveats however). This DAC has some noise floor issues, and I'd hope with another DAC I could do better on the truncation.
For more background, you can read here.
With all ABX testing, don't bother listening to the X stream if you cannot hear the difference between the A and B stream. That is really they key: Listen carefully until you can tell the A stream from the B stream. And then listen to X.
For the bit depth test, the QA_ABX is truncating the source material (24-bit) down to 12-bit on the fly. The key to hearing bit-depth truncation is the noise floor. More volume will help. Listen to the A stream, then listen to the B stream. And make sure you can really hear the differences (and if you are having trouble, go to 8 or 10 bits and work towards 16 bits). And then listen to the X stream and decide if it is A or B. On a spectrum analyzer, truncation appears as a rise in noise floor.
For the crossover distortion, you want to listen for buzziness . Cross over distortion appears as a picket fence of harmonics and thus you'll hear high frequency energy where there was none. Again, listen to A and B streams until you can hear the difference, and then listen to X. The algorithm to introduce crossover distortion is very simple and subtle: Find a crossover, and the multiply the first point after the crossover by a value slightly less than 1. That is it. When I indicate I can hear 0.97 crossover distortion, that means that on all zero crossings the first sample after the zero crossing is multiplied by 0.97.
For amplitude distortion, one of the streams will have its gain changed by an amount you select. In my case, I picked -2 dB and so either the A or B stream was 2 dB quieter. I think I could do better here with source material that had a lower peak-to-average ratio. But piano music with lots of staccato notes amidst a huge dynamic range (playing style) is very hard for me to hear the 2 dB imbalance. With something highly compressed, then 1 dB would be expected.
My results are below. You can take any of these results and paste it in the app's verify window and it should tell you the signature is correct. To copy a test (only do a single test, not the entire group) just grab the text from and including the Q of the QA_ABX all the way down the final pipe ('|') symbol just after the signature that appears after the Finished message.
-- TO VERIFY, CUT STARTING HERE (BUT NOT INCLUDING THIS LINE) --
QA_ABX v0.95
Loaded file: LessLoss-Mephisto-Listz-I-24_bit_96_kHz.wav @ 7/28/2019 1:56:41 PM
File MD5: f2eaff89f99881cebf357b4b6c68ee8b
Distortion: BitDepth Value: 12
Sample Rate: 96000 Bit Depth: 32 Channels: 2 Format: IeeeFloat
Fail @ 7/28/2019 1:57:02 PM
Pass @ 7/28/2019 1:57:19 PM
Pass @ 7/28/2019 1:57:29 PM
Pass @ 7/28/2019 1:57:55 PM
Pass @ 7/28/2019 1:58:01 PM
Pass @ 7/28/2019 1:58:16 PM
Pass @ 7/28/2019 1:58:25 PM
Pass @ 7/28/2019 1:58:35 PM
Pass @ 7/28/2019 1:59:02 PM
Pass @ 7/28/2019 1:59:18 PM
Pass @ 7/28/2019 1:59:34 PM
Finished @ 7/28/2019 1:59:37 PM
|e27725382124b934542c914917273a6c|
-- TO VERIFY, CUT FINISHING HERE (BUT NOT INCLUDING THIS LINE) --
QA_ABX v0.95
Loaded file: LessLoss-Mephisto-Listz-I-24_bit_96_kHz.wav @ 7/28/2019 2:08:33 PM
File MD5: f2eaff89f99881cebf357b4b6c68ee8b
Distortion: Crossover Level: 0.97
Sample Rate: 96000 Bit Depth: 32 Channels: 2 Format: IeeeFloat
Pass @ 7/28/2019 2:09:12 PM
Pass @ 7/28/2019 2:09:39 PM
Fail @ 7/28/2019 2:09:54 PM
Pass @ 7/28/2019 2:10:12 PM
Pass @ 7/28/2019 2:10:25 PM
Pass @ 7/28/2019 2:10:36 PM
Fail @ 7/28/2019 2:11:10 PM
Finished @ 7/28/2019 2:11:58 PM
|20aa354abd41f602cf5b7200c5b8796f|
QA_ABX v0.95
Loaded file: LessLoss-Mephisto-Listz-I-24_bit_96_kHz.wav @ 7/28/2019 2:12:25 PM
File MD5: f2eaff89f99881cebf357b4b6c68ee8b
Distortion: Amplitude Value: -2.00
Sample Rate: 96000 Bit Depth: 32 Channels: 2 Format: IeeeFloat
Pass @ 7/28/2019 2:13:10 PM
Pass @ 7/28/2019 2:13:17 PM
Pass @ 7/28/2019 2:13:32 PM
Pass @ 7/28/2019 2:13:54 PM
Pass @ 7/28/2019 2:14:28 PM
Pass @ 7/28/2019 2:14:41 PM
Pass @ 7/28/2019 2:15:10 PM
Finished @ 7/28/2019 2:15:15 PM
|21c34135be275b697d1b6f55b73801c7|
