I started this analysis in another thread, but moving to its own thread as I think it's deserving. To recap, I pulled the audio from the leaked youtube video, and analysed the waveform. I also created a comparison audio clip using my phone speaker and a simple podcast level audio recording setup.
I amplified the 500hz sample to look at it a bit closer, and it looks like this
![https://hotcopper.com.au/data/attachments/4891/4891239-007e02256957bd79f99520285290bc3a.jpg](https://hotcopper.com.au/data/attachments/4891/4891239-007e02256957bd79f99520285290bc3a.jpg)
Compare this to the output from my phone, equalised to the same amplitude, with likely a lower quality microphone to the one they used in that video:
![https://hotcopper.com.au/data/attachments/4891/4891240-0bd9944bc963c9ad60754db2a1162ff2.jpg](https://hotcopper.com.au/data/attachments/4891/4891240-0bd9944bc963c9ad60754db2a1162ff2.jpg)
So the first thing you might blame is youtube compression, it's not the original audio right? So I pulled the audio from a 500hz tone sample video, extracted it in the same way I extracted the AP audio, and provide it below
![https://hotcopper.com.au/data/attachments/4891/4891253-ec3a49d9f80ac3d06fba9b4312c21a54.jpg](https://hotcopper.com.au/data/attachments/4891/4891253-ec3a49d9f80ac3d06fba9b4312c21a54.jpg)
It's a (nearly) perfect reproduction of the original signal, meaning the process of uploading a video to youtube and then extracting did not cause the visible distortions.
So what are we seeing here? Something that they are supposed to have solved by now - ultra high frequency artifacts polluting the wave. DSR is supposed to means using ultrasonic waves that combine to recreate the lower frequency source wave. I'd estimate the ultrasonic recreation of the source wave is 10-20% out from the source signal. Here's an overlay of the two waves.
![https://hotcopper.com.au/data/attachments/4891/4891276-8c95bd92351759a37a736ed50b567d3e.jpg](https://hotcopper.com.au/data/attachments/4891/4891276-8c95bd92351759a37a736ed50b567d3e.jpg)
What you're seeing here is not just ultra high frequency noise though, the wave is also shifted in the time domain, which is the distortion most obvious in the low frequency sample. Here's how the 20hz looked
![https://hotcopper.com.au/data/attachments/4891/4891291-a55955d933f95fd53cf296a94be61dcb.jpg](https://hotcopper.com.au/data/attachments/4891/4891291-a55955d933f95fd53cf296a94be61dcb.jpg)
and here's how it should look
![https://hotcopper.com.au/data/attachments/4891/4891288-7341bc048a90eaf66010b8dc8d93576d.jpg](https://hotcopper.com.au/data/attachments/4891/4891288-7341bc048a90eaf66010b8dc8d93576d.jpg)
Here's an overlay
![https://hotcopper.com.au/data/attachments/4891/4891294-4cae8d7ed944e0a4e737058087d9b03b.jpg](https://hotcopper.com.au/data/attachments/4891/4891294-4cae8d7ed944e0a4e737058087d9b03b.jpg)
The high frequency distortion is less relevant here, but the wave is extremely skewed on the downward edge of the wave.
I said before that perhaps this issue could be fixed in software, but they've had 10 years, and this is the level of quality they're producing? I suspect we've found out why no one has released a demo yet, and why this was released in the form of a deniable "leak". If this is the quality level of the prototype, why are they mass producing? Are they assuming that they can increase the quality if they mass produce?
It's not evidence that they can't eventually achieve quality audio output, but it is evidence that at the time that the video was recorded, the AP software/hardware they were using was producing a very poor reproduction of the source signal.