[WindowsX]

Lately sound science is circulating around rmaa and classical analog metrics with measurement methodlogy from past decades. It's becoming harder to discover noticeable measurements in digital audio domain.

Today, I'd like to introduce Diffmaker software. It can detect changes in audio signal using digital differencing method. It's recommended for the following applications as below:

Changing interconnect cables (compensation for cable capacitance may be required)
Different types of basic components (resistors, capacitors, inductors)
Special power cords
Changing loudspeaker cables (cable inductance may need to be matched or compensated)
Treatments to audio CDs (pens, demagnetizers, lathes, dampers, coatings...)
Vibration control devices
EMI control devices
Paints and lacquers used on cables, etc.
Premium audio connectors
Devices said to modify electrons or their travel, such as certain treated "clocks"
Different kinds of operational amplifiers, transistors, or vacuum tubes
Different kinds of CD players
Changing between power amplifiers
General audio "tweaks" said to affect audio signals (rather than to affect the listener directly)
Anything else where the ability to change an audio signal is questioned

Here's Diffmaker homepage that you can download and read useful information like aes paper:

http://www.libinst.com/Audio%20DiffMaker.htm
http://www.libinst.com/AES%20Audio%20Differencing%20Paper.pdf
http://www.libinst.com/Detecting%20Differences%20(slides).pdf
https://www.gearslutz.com/board/gear-shoot-outs-sound-file-comparisons-audio-tests/607481-evaluating-ad-da-loops-means-audio-diffmaker.html

Maybe this can help us observing digital audio better for years to come.

 

[spruce music]

I have used this a fair amount. You have to be careful using it.  The software can be flaky. Sometimes it reports results that make no sense, and sometimes it shows things you have trouble obtaining any other way.  When it works it is almost like magic.
 
What you are likely to find is the old ways of measuring do a pretty good job of fleshing out performance parameters and this software won't disagree.  Doing those measurements is much less tricky or error prone than doing this.  But for some purposes differencing is an excellent idea. 

So, what are you going to compare or test for us to discuss?  What do you think is important that this can show us?

 

[castleofargh]

 it's a very cool little tool and free, so what's not to love. but it's not perfect(RMAA is also far from perfect).
sometimes aligning 2 recordings will fail and result in way more differences than there really are. so just assuming diffmaker results are always factual would be wrong. Spruce M is right about that.
 
all in all with some practice and by trying to use alternative tools as reference to notice when something goes wrong, it can be useful.

 

[pinnahertz]

Cool, flakey, all true.  The real problem is there's no good way to scale a difference against audibility. It's a very hyper-sensitive metric.  If you're not careful it'll make an alarmist out of you.

 

[limpidglitch]

  Cool, flakey, all true.  The real problem is there's no good way to scale a difference against audibility. It's a very hyper-sensitive metric.  If you're not careful it'll make an alarmist out of you.

 
The developers seem aware of this.

Taken from the convention PDF:

Complications
● An audible Difference track doesn’t mean the differences are audible while still part of the full audio signals. It’s a necessary but not sufficient requirement.
● Nearly anything not identical will leave an audible Difference track. And a lot of uninteresting effects can be different.
● There is always noise in an analog stage -- noise is always different. Sound left in Difference track might be real. Silence almost certainly is real.

 

[RRod]

   
The developers seem aware of this.

Taken from the convention PDF:

 
Yes, it's important to note that the whole edifice of lossy codecs rests on having an audible difference file that is inaudible when put back into the file. I've listened to a lot of difference files between hi-res and Redbook, and absolutely none of them have anything audible at my maximal listening volumes in my listening room. That would mean the only way I could possibly hear anything different in the hi-res file would be due to non-linearities (assuming a non-sketchy resample to the higher rate).

 

[pinnahertz]

   
Yes, it's important to note that the whole edifice of lossy codecs rests on having an audible difference file that is inaudible when put back into the file. 

Yes, that's correct.  A lossy codec depends on masking to hide it's losses.  Differencing a lossy file against a lossless one removes all masking.  No fair!

 

[spruce music]

If you get difference tracks that are silent upon playback, you can be sure the difference is inaudible. If your difference track is down by 60 or 70 dB it will be faintly audible, yet the two tracks are probably not audibly different. Above that it gets tricky. Just a for instance, you can difference analog interconnects and you'll get silence. Same for digital or USB cables.

 

[pinnahertz]

If you get difference tracks that are silent upon playback, you can be sure the difference is inaudible. If your difference track is down by 60 or 70 dB it will be faintly audible, yet the two tracks are probably not audibly different. Above that it gets tricky.

This is precisely the problem with difference testing. You get a result where the difference is audible, is that an actual audible difference? The test is hypersensitive, interpretation gravitates towards alarmist. The answer lies in determining what the difference is spectrally, and if the changes to the original are masked by the primary signal. That's such a huge question, and differencing alone skates right past it. There is exactly one condition that we can be sure of: no difference result. We could also assume from known masking curves that difference products in the 70dB and below range that are below the masking curve of the primary audio spectrum are inaudible.

Just a for instance, you can difference analog interconnects and you'll get silence. Same for digital or USB cables.

That's the condition we can be sure of: silence/no difference.