In late 2016 I released a free application called QA_ABX for some testing we wanted to run on various distortion types, with some follow-on updates. This app lets you take a pristine WAV file, and it will distort the WAV file "on the fly" seamlessly with a type and intensity of distortion you specify. You can then ABX audition the playback streams, and when finished you can "sign" your results cryptographically.
I put a video up on youtube to show the process (link below)
What is hopefully helpful is that it lets you introduce several different types of distortion at various levels. You can look at quantization, crossover, amplitude, hard-clipping, TANH. These all have a slightly different sound to them.
The settigs are such that there's an "easy" setting for each, so that even untrained ears can recognize what each type of distortion sounds like. And then you can reduce the distortion and try again. Personally, I cannot hear the difference between 14 bit and 24 bit audio with any statistical significance. But I played various bands for many years, and my ears aren't what they used to be.
The ABX comparison is seamless and gapless. You can click with the mouse or use the the A or B keys to switch. And you can use your favorite equipment, favorite room, favorite music, etc. The music just needs to be WAV and the playback device needs to be your primary audio device on Windows.
The aim here is to allow someone to demonstrate skill in discerning whether a stream is degraded or not. And since it can be signed, that means others can review your work and verify if what you claim was achieved. Note this doesn't tell you if equipment is "good" or not. But it does tell you if a person listening through a certain piece of equipment can hear a difference. Once it's established a person can sense a difference beyond what others can, then if they tell you equipment A differs from B, then it's likely believable even if it cannot be readily measured for whatever reason. But to me, the more interesting part is that if the data set is large enough, we can learn if some types of distortion matter more than others.
So, if anyone wants to try, I'd propose to start with quantization distortion and see if someone has superpowers (ears + equipment) that can reveal these distortions.
What you need:
1) Youtube link on how the app works: YouTube
2) Download location for application: Releases * QuantAsylum/QA_ABX * GitHub
3) For the audio source, I'd propose to start Mephisto Waltz Excerpt #1 24b/96K: Steinway and Sons grand piano recording : High definition music | Audiophile music recordings | HD tracks by LessLoss
4) The original blog post on ABX from late last year: ABX Testing and Distortion
– QuantAsylum
What is the highest level of quantization distortion you can identify reliably 5 out of 6 times?
I put a video up on youtube to show the process (link below)
What is hopefully helpful is that it lets you introduce several different types of distortion at various levels. You can look at quantization, crossover, amplitude, hard-clipping, TANH. These all have a slightly different sound to them.
The settigs are such that there's an "easy" setting for each, so that even untrained ears can recognize what each type of distortion sounds like. And then you can reduce the distortion and try again. Personally, I cannot hear the difference between 14 bit and 24 bit audio with any statistical significance. But I played various bands for many years, and my ears aren't what they used to be.
The ABX comparison is seamless and gapless. You can click with the mouse or use the the A or B keys to switch. And you can use your favorite equipment, favorite room, favorite music, etc. The music just needs to be WAV and the playback device needs to be your primary audio device on Windows.
The aim here is to allow someone to demonstrate skill in discerning whether a stream is degraded or not. And since it can be signed, that means others can review your work and verify if what you claim was achieved. Note this doesn't tell you if equipment is "good" or not. But it does tell you if a person listening through a certain piece of equipment can hear a difference. Once it's established a person can sense a difference beyond what others can, then if they tell you equipment A differs from B, then it's likely believable even if it cannot be readily measured for whatever reason. But to me, the more interesting part is that if the data set is large enough, we can learn if some types of distortion matter more than others.
So, if anyone wants to try, I'd propose to start with quantization distortion and see if someone has superpowers (ears + equipment) that can reveal these distortions.
What you need:
1) Youtube link on how the app works: YouTube
2) Download location for application: Releases * QuantAsylum/QA_ABX * GitHub
3) For the audio source, I'd propose to start Mephisto Waltz Excerpt #1 24b/96K: Steinway and Sons grand piano recording : High definition music | Audiophile music recordings | HD tracks by LessLoss
4) The original blog post on ABX from late last year: ABX Testing and Distortion
– QuantAsylum
What is the highest level of quantization distortion you can identify reliably 5 out of 6 times?
Hi QAMAtt: Interesting. Is it a Windows only app? Is it limited to use on a computer audio system (laptop feeding DAC, driving preamp/ amp, driving speakers)? Unfortunately, I have an old time CD based 2 channel stereo.
Even if the app could run on my iMac sound card/ Sennheiser headphones, this setup seems unlikely to have enough resolution.
There is a similar thing on-line Free Online Audio Tests, Test Tones and Tone Generators As an aside, he uses electronic music in some of the tests which make it impossible for me to judge bit depth
Even if the app could run on my iMac sound card/ Sennheiser headphones, this setup seems unlikely to have enough resolution.
There is a similar thing on-line Free Online Audio Tests, Test Tones and Tone Generators As an aside, he uses electronic music in some of the tests which make it impossible for me to judge bit depth
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.