ABX Testing Application

I’m linking below to an app from 2017 called QA_ABX. This program will take a WAV file (your choice of music or tones) and it let’s you inject different types of distortions on the fly.

Once you have picked your distortion type and the severity of distortion, you can audition 3 audio streams: The A stream, the B stream or the X stream.

The distortion will only be applied to the A OR B stream, and it’s random on each iteration. Your task is to identify if the A or B stream has been distorted with the distortion you selected. And if you can hear a difference between the A and B streams, you can then decide if the X stream is the A or B stream.

The distortions are very easy to pick up on the audio analyzer (try with a 1 kHz tone), but very difficult to pick up by ear whether it’s music or a tone.

More info and a PDF at the link below. If you are looking for a starting point, load your favorite audition WAV and play through your favorite DAC and pick 8-bit distortion. You’ll easily succeed at that.

I’d be curious what others can consistently hear for amplitude distortion.

1 Like

I look forward to trying it- just remembered I brought my Bose NR headphones with me so can try it sooner.

Hi Guys, It seems to me the Crossover Distortion is only applying it to the left channel, And FYI I measured it..

Great little app though

Hi @Dean_OAV, thanks for the bug report. Can you hear it though? I really have a hard time nearing some of these distortions, but they can be readily measured.

@matt - I have been meaning to ask you what the TANH distortion is ?

Hi @Var, it’s what some of the DSP guys are using for guitar distortion. The function looks like this (courtesy GPT):

And here’s a sine wave distorted via tanh (again courtesy GPT). Note that it gives you a very fluid means for increasing the distortion and so you can dial in distortion based on amplitude, which is very much how guitar amps operate: More quiet playing and the tones are largely untouched. But as you play more aggressively (loudly) the distortion automatically kicks in.

For ABX testing, the TANH distortion will result in a dead give-away beyond a certain point: The distortion increase will manifest as a volume increase. Keep in mind a 1V sine has an RMS of 0.707, but as you distort that to a square wave, the RMS increases to 1.0 as it morphs from a sine to a trapezoid to a square.

And so, if you dial up the TANH distortion, you might be detecting the amplitude difference and not the distortion difference. I really wish someone with good ears and a great system could share the numbers on what they are able to detect…

Thanks for the nice explanation @matt . I decided to do a test with a 440hz Audacity generated tone .wav file to keep it simple, and used my Bose NC headphones connected to my PC’s line output. I was able to hear crossover distortion @.94 pretty consistently. A bit depth at 8 was not problem, 12 most of the time, but not at 13. Amplitude delta of 0.5db sometimes and .8dB most of the time. Clipping distortion at .98 and .92 pretty consistently. Hearing a Tanh of 10 was pretty consistent. My hearing was tested 6months ago and was decent.

I have noticed a click when switching between any of the matched options and the unmatched.. this is now making it quite easy to detect the correct pair.. I have been using a new set of HD560s which are very analytical along with a fiio KA1 micro usb-c dac in a laptop this is on level tests - now easy to detect 0.5dB level difference with 1k sine wave and clipping distortion bit rate is hard