Importance of the reponse above 16 kHz

[daftcombo]

Hi,

I am 34 and can still hear up to 16 kHz and even more (20 kHz it seems, if I pay good attention).

My speakers (Epos K3) have a range of 41Hz - 30kHz. Amplifier and DAC range are even larger.

Frenquency reponse of my system is quite flat after EQ (measured with a calibrated Behringer ECM8000 and REW, and corrected by convolution with an impulse created in RePhase and processed in JRiver), with a descending curve of 0.6 dB per octave from 40Hz to 20kHz.

Recently I've tried a Low Pass filter (Linkwitz-Riley 96dB/octave @ 22kHz). Up to 16kHz, it doesn't change anything. But it gives -2dB at 20 kHz and works plenty afterwards.

I listen to FLAC 44.1kHz/16bits, so there shouldn't be much in the 20kHz - 30kHz range.

But still, I find the sound smoother and less tiring with this filter.

It is theorically less "Hi-Fi" though.

Is it really possible that the huge attenuation after 22kHz make my ears/brain feel more confortable? Is it just because of the -2dB @ 20kHz? Is the trick elsewhere?

Thanks!

EDIT: As you will read below, I can't spot the supposed differences in "smoothness" in an ABX test.
I'm still interested in replies if someone has successfully taken an ABX in that frequency range.

 

[Soniclife]

Have you tested yourself blind with this?

 

[soundwave76]

Not relevant to me anymore, since I can't hear above 11kHz anymore...

 

[daftcombo]

Have you tested yourself blind with this?

No yet. I don't know how to do it. I would need a friend to switch between convolving files in my back.

But I was wondering if anyone else made similar experiments.

 

[Shadrach]

I'm impressed. I've never come across anyone over the age of 25 who can hear 20kHz.

 

[Pluto]

I don't know how to do it.

Get yourself a copy of Foobar and the necessary ancillary tools, all free. You will need the ABX tool and, probably, the WASAPI I/O module to keep the signal sent to your computer's output nice and clean.

The ABX tool will provide you with the means of comparing a signal via your filter with the unfiltered version.

 

[edechamps]

I am 34 and can still hear up to 16 kHz and even more (20 kHz it seems, if I pay good attention).

Are you sure you're actually hearing the 20 kHz tone, as opposed to non-linear distortion artifacts that might be created when the tone is being played?

This is not a theoretical question. For example: I have measured a number of DACs that have a steep reconstruction filter designed in such a way that tones near 20 kHz and above will create imaging artifacts (so, for example, at a 48 kHz sample rate, a 20 kHz tone might also create an an imaging artifact at 28 kHz). This spurious image tone can then interact badly with the original 20 kHz tone in devices downstream of the DAC (amplifier, transducer) to create intermodulation distortion which will definitely land in the audio band, well below 20 kHz. This can then trick the listener into thinking they can hear a 20 kHz tone, while what they're actually hearing are non-linear distortion artefacts that happened to land in the audible band.

 

[daftcombo]

I'm impressed. I've never come across anyone over the age of 25 who can hear 20kHz.

I wrote that because I did a small blind test with a 20 kHz playing (or not) in Foobar and spotted the 20 kHz whenever it was playing.
I am not sure it was a completely clean signal. edechamps probably has a point. My DAC was an Apogee DUET 2 during the test.

 

[daftcombo]

Get yourself a copy of Foobar and the necessary ancillary tools, all free. You will need the ABX tool and, probably, the WASAPI I/O module to keep the signal sent to your computer's output nice and clean.

The ABX tool will provide you with the means of comparing a signal via your filter with the unfiltered version.

This tool can't switch between two convolution / EQ files.

 

[daftcombo]

Are you sure you're actually hearing the 20 kHz tone, as opposed to non-linear distortion artifacts that might be created when the tone is being played?

This is not a theoretical question. For example: I have measured a number of DACs that have a steep reconstruction filter designed in such a way that tones near 20 kHz and above will create imaging artifacts (so, for example, at a 48 kHz sample rate, a 20 kHz tone might also create an an imaging artifact at 28 kHz). This spurious image tone can then interact badly with the original 20 kHz tone in devices downstream of the DAC (amplifier, transducer) to create intermodulation distortion which will definitely land in the audio band, well below 20 kHz. This can then trick the listener into thinking they can hear a 20 kHz tone, while what they're actually hearing are non-linear distortion artefacts that happened to land in the audible band.

What you write is exactly the reason why I wanted to use that 96dB/octave filter above 22kHz in the first place: reduce intermodulation distortion caused by 20kHz+ tones.

By the way, I wonder if such a filter improves/deteriorates imaging or do nothing at all.

 

[Alex-D]

This tool can't switch between two convolution / EQ files.

You can use "convert" function in Foobar transcoding from FLAC to FLAC with any supported DSP plug-ins applied.
Then it's possible to use ABX tool to compare a filtered file with a non-filtered one.

 

[Pluto]

This tool can't switch between two convolution / EQ files

True, but you could pre-render the material under test as WAV files and switch between them to compare.

 

[daftcombo]

True, but you could pre-render the material under test as WAV files and switch between them to compare.

Yes, how can I do that? Is there a way to do it in Audacity or in iZotopeRX?

EDIT: Just saw Alex-D post. Will try that!

 

[Soniclife]

No yet. I don't know how to do it. I would need a friend to switch between convolving files in my back.

But I was wondering if anyone else made similar experiments.

As others have explained above the trick is to pre-render the changed file and compare that to the original. This will definitely show if you can hear the difference, and if so then the question becomes why.

I'm not suggesting you cannot hear the difference, but it's always good to start with good data on an audible difference before getting into speculation on why you hear what you do.

 

[Pluto]

Is there a way to do it in Audacity or in iZotopeRX

Just apply the process(es) under test and save the result as a WAV file! This is what iZotopeRX is for – you could even use the iZotope plug-ins via Audacity if you wanted to.

Forgive me saying this, but the foregoing is so obvious that I fear that I am somehow missing the point of your question.

 

[daftcombo]

Get yourself a copy of Foobar and the necessary ancillary tools, all free. You will need the ABX tool and, probably, the WASAPI I/O module to keep the signal sent to your computer's output nice and clean.

The ABX tool will provide you with the means of comparing a signal via your filter with the unfiltered version.

Alright, I installed ABX and tried it. Funny moments announced!
Now I need to convert a few tracks and let's go

 

[daftcombo]

To begin with, I did a mp3 320 (V0) vs 128 kbps on 16 trials, on AKG K701. I was lost. 50%. Bye bye Hi-res.

 

[Pluto]

Bye bye Hi-res.

Yes, a properly controlled test is just so revealing and, quite often, the truth can hurt.

But you'll get over it

 

[Soniclife]

To begin with, I did a mp3 320 (V0) vs 128 kbps on 16 trials, on AKG K701. I was lost. 50%. Bye bye Hi-res.

If this is your first try at proper blind testing it might take some time to learn, but it's very valuable, and you are in an elite group of audiophiles who have ever tried one.

 

[wiggum]

Are you sure you're actually hearing the 20 kHz tone, as opposed to non-linear distortion artifacts that might be created when the tone is being played?

This is not a theoretical question. For example: I have measured a number of DACs that have a steep reconstruction filter designed in such a way that tones near 20 kHz and above will create imaging artifacts (so, for example, at a 48 kHz sample rate, a 20 kHz tone might also create an an imaging artifact at 28 kHz). This spurious image tone can then interact badly with the original 20 kHz tone in devices downstream of the DAC (amplifier, transducer) to create intermodulation distortion which will definitely land in the audio band, well below 20 kHz. This can then trick the listener into thinking they can hear a 20 kHz tone, while what they're actually hearing are non-linear distortion artefacts that happened to land in the audible band.

Nah, modern DACs are not that bad. THD stays below 0.01% and IMD is anywhere from 0.4 to 0.01% This is for the most popular DACs not some boutique ones.

OP is unnecessarily worried. There are 3 things that tells that it is most likely placebo effect.

1. Harmonics of musical instruments fall to single digit % by 15kHz.
2. Critical band of the human ear at 15kHz is ~ 1.5kHz and at 20kHz is ~ 2kHz.
3. The 800 lb gorilla in the room is distortion of speakers. It ranges anywhere from 0.1 to 1%. There is no way either the THD or IMD of anything else in the audio path to be audible above speaker distortion.