[bigshot]

This test vividly demonstrated to me how super audible frequencies tend to be bunched up in a fairly narrow band just above 20kHz, but they are at such a low volume level, I can't see how they would ever have any kind of presence above the volume of the audible frequencies. A lot of times I see people talking about recordings done at 192 or the necessity of distortion levels below -70dB. Tests like this are great to demonstrate just how small small actually is and how high high is.

 

[RRod]

Here's another example.

 

[bigshot]

That one is interesting. The level is so low and when it's normalized up, it sounds like jingle bells. Is it possible that is some sort of harmonic distortion? Or perhaps the ajtifacting on the pitch up has made it sound like that? It doesn't resemble the sound of massed strings at all.

 

[Arpiben]

@RRod
Nice Job. Lol.

 

[RRod]

That one is interesting. The level is so low and when it's normalized up, it sounds like jingle bells. Is it possible that is some sort of harmonic distortion? Or perhaps the ajtifacting on the pitch up has made it sound like that? It doesn't resemble the sound of massed strings at all.

The form of the HR content makes no sense, being low from 20k-32k then rising again to 40k with what looks like inverse harmonics. I'd wager it's some kinda of aliasing artifact that no one 'saw' (because no one hears this stuff to fix it). More examples to come now that I've got some stats run. There was one track that had super-loud HR content; turned out the DL was bad and half of the R channel was white noise ^_^

Here's another example where the content actually makes sense.
And another for good measure.

 

[bigshot]

Do you think they might be putting in random HF content just so it looks like there's something there?

 

[Arpiben]

Do you think they might be putting in random HF content just so it looks like there's something there?

From a spectral point of view it doesn't appear to be the case with the 4 files we have until now.
Hp1 -> beginner mistake and clear aliasing. At first I thought @RRod did it on purpose.
Hp2 -> Ultrasonic 'boost' due to close mics & mixing changing the natural power density per frequency.
Hp3 -> strong spurious at 32.7(68) kHz.Probable clock pilot or residual from other equipment in the surroundings (HF mics or other stuff)

 

[analogsurviver]

Do you think they might be putting in random HF content just so it looks like there's something there?

I can't speak or vouch for anybody else; I don't.

But it is hard to avoid the noise from the venue; below 20 Hz, there is likely to be ventilation/temperature air conditioning gear noise, Then there is what I have ultimately termed Tesla's Greeting from the Grave : mains frequency (and the whole myriad of its harmonics) hum and buzz - the easiest way to tell if it is an european or an american recording...

Above 20 kHz ( and slightly below, beginning around 19 kHz ) , there may be movement detectors, various lighting "accompaniement " , an ocassional insect ( ah, those charming moments in the quietest passage usurped by the close flyby of a mosquito/fly/bee/wasp/whatever ... ) - and who knows what else is lying in ambush, waiting to jump me in the future.

As I record in DSD - and almost exclusively on location, preferably live - I can only check for this LATER - as it requires the DSD to be converted to PCM in order to be able to use spectrum analyzer such as Voxengo SPAN. I would love to have the time available to set up an additional PCM recording rig to see what is going on right away - but given the real world time constraints, I will always put microphone placement as a priority - and not > 20 kHz " environmental pollution ". And when recording acoustical music, it is relatively easy; anything that requires electricity for the musicians to be heard, may well be booby trapped with various switching power supplies - which, once assembled in the actually used system, can have nasty ultrasonic output.

There is at least one soundcard that takes the "juice" for its phantom mic supply off the main switching PS - producing, as you might have guessed - a small, but distinly defined spike > 20 kHz; in whatever it tries to record, from the earliest possible stage...

Now, imagine a multichannel multimiked recording ... using say at least three mic preamps from at least two manufacturers. Unless the equipment is not strictly vintage ( with decent and solid linear power vsupplies ), all those switching power supply frequencies are bound to find their way trough all the cabling, one way or another. Even in balanced XLR connection, there is something called common mode rejection ratio - CMMR - and it definitely is NOT infinite; some of the noise ( including > 20 kHz ) WILL come trough .

So, above 20 kHz spikes and other noise not related to music itself most likely it is not intentional, but probably next to unavoidable in real world situations. And is something to be added to the "watch for and avoid" list for the future recordings.

 

[Arpiben]

@RRod

It was a real pleasure looking a bit deeper at your file Hrex3. Adding to the pilot tone there are lot of distortions and chirp for bats.
Amazing.

Adaptive Spectrogram (divide frequency scale by 2):

A different presentation with spectrum at different timing:

 

[bigshot]

Man. At its best, it's a small sliver of inaudible frequencies at inaudible volume levels. At its worst, it's a whole mess of noise and distortion. I'm beginning to think that the guy who downsamples all of his HD tracks to 16/44.1 is on the right track.

 

[Arpiben]

Yes and No.
Even if not audible keeping high ultrasonic digital content like the above is a potential danger for IMD or aliasing further down in the audio chain.
IMO better LP filter such 24/96 files or keep your 16/44.1.

Dealing with the existing distortions in the audible range starting from DC up to 24 kHz or more nothing much you can do.
You should listen to the couple of chirps range 14kHz-22kHz between 04s and 16s.

Dealing with levels, they are incredibly high (yellow colour in the spectrogram picture) or dB level in spectrum view.

 

[bigshot]

I was saying that without any distortion, there isn't much there... content wise or volume wise... and with it, it's can be a loud mess of noise. Best to just filter everything off and save file size by making it 16/44.1. I think you're just saying the same thing I did.

 

[RRod]

Put up examples 4-6 here; perhaps more expected material.

 

[Arpiben]

@bigshot

Now we have samples with enough ultrasonic.
Now you have sensed the ultrasonic content with frequency translation or pitch shift.

Did you try comparing both versions in order to distinguish them?

By both versions I mean: original and LP filtered (24kHz or 22 kHz) ones (files not provided).
Depending on ultrasonic power content you will need to adjust levels up to 0.4 dB .

 

[bigshot]

I've compared "HD" audio to redbook before. I know that I can't hear anything above 18kHz or so. Ultrasonic content doesn't improve music for me at all. I was interested here in finding out what exactly was up there. I think I've discovered that super audible frequencies in commercial music are like a dusty old attic. Either they're pretty much empty or they have junk stacked up. I already knew that, but now I have a better idea of how empty "empty" is and what the junk sounds like.