Using DeltaWave to explore BlueTooth and DAC contributions

[celstark]

Given that @pkane is here and wrote the software that made this possible, I thought it good to post up here too. I'm a zero-view "youtuber" who actually only posts up videos there as an outlet for talking about fun technical topics that few actually care about (which is all to pointed an analogy to my day job as a research professor...).

I've wanted to use the Null Test as a way to help people get their heads around what may or may not be audible and DeltaWave is just an incredible tool to be able to do that. Massive props to @pkane for writing it. I've played enough to start to get a handle on its limitations (doing negative controls by testing a device against itself typically) and therefore to put a pair of videos up. Part 1 goes over the concept of the Null Test (I go back to the Carver Challenge days -- I'm old) and uses DeltaWave to calculate the difference in USB-USB vs. USB-BlueTooth or USB-Airplay. What I love about being able to have a video on it is being able to play the original clip and, without changing levels, play the delta. Masking will have us hear less than the delta shows, but when that delta is so, so quiet, it really drives home what the difference really is.

Part 2 then takes it to DACs, comparing a cheap dongle (JCally JM6 Pro) vs. an "entry level" desktop DAC (Modi+). My rig is going to be a limiting factor here, but even with the limit of maybe 105 dB or so of SNR, it, to me, drives home the point. If we're nulling out at 100 dB, you're not hearing that. Listening to the difference signal and hearing nothing -- boosting it like crazy and having the delta be still dominated by pure noise -- to me, really drives home the point in a way that a plot can't (again, I'm a scientist by day -- I love plots). The start of Part 1 there, though, also hits an old informal experiment on the audibility of distortion which, again, is humbling for those that chase SNR levels in the triple digits.

 

[Palladium]

What's the exact BT codec used for the null in Part 1?
I presume it's AAC since you are testing from an iPhone.

 

[celstark]

Yes, it was AAC. Of course, the AAC decoder was on the Topping DX 5 II.

 

[AndewCrstn]

Using a null test with DeltaWave and actually letting people hear the delta instead of just showing graphs sounds way more intuitive, especially for folks who aren’t super into measurements. That part about boosting the difference signal and still basically hearing noise really hits the point home about how tiny those differences are.

Also love that you brought in the DAC comparison (like the JCally vs Modi+ angle). That’s one of those debates where people can get pretty deep into specs, so having a practical demo helps ground it a bit.

Out of curiosity, did you notice any cases where the delta was actually more noticeable than you expected, or has it mostly reinforced the idea that differences are way smaller than people think?

 

[celstark]

One thing I didn't fully incorporate into my thinking when making that video is the fact that just b/c it seems silent when playing back the delta doesn't necessarily mean it won't color or adjust the sound at all. A modulation on top of something at a reasonable level may be easier than that vs. silence. That said, when it's clearly just modest modulation of noise -- yeah, we're not picking that out.

What surprised me, but shouldn't have really, is just how sensitive it can be. Sure, some of that can come from a lack of perfect sync'ing. I've done a decent number of blind tests and they're mighty humbling. Play with his distortion generator a bit and -- wow, it's tough to hear. Yet, DeltaWave picks it out.

 

[celstark]

Thanks! I was a bit shocked when I had the **************** article sent to me. They did a lot of coverage that was nicely done. I don't have a $20k DAC to put through the testing though. But, I, too, was really impressed at how well the null can happen and how little there can be left over.

 

[RandomEar]

Cool little experiment! I must say that I'm a bit surprised how clear the music in the null test still came through (when boosted 40 dB...). You talk about possible small errors in syncing and that is one possibility. If you have three devices like this (DAC 1/2 + ADC) the combination of clock drifts for each pair will probably not be perfectly linear. DeltaWave does offer a non-linear drift correction in the settings (default is just linear) - it would be cool if you could test if that makes a difference.

However, I suspect that the main culprit will be a difference in the reconstruction filter between the DACs. Maybe one even uses minimum phase and the other is linear phase. If you have the option and the time, it would be very cool to see the impulse responses for both of them!

 

[celstark]

Both good ideas. I'd cut down to fairly short (10-30s IIRC) clips which should mitigate the non-linear aspects of the phase, but that's good to know for sure as I think about using this more. The DX5-II, I believe, has selectable reconstruction filters which would allow for a direct test of that.