• WANTED: Happy members who like to discuss audio and other topics related to our interest. Desire to learn and share knowledge of science required as is 20 years of participation in forums (not all true). Come here to have fun, be ready to be teased and not take online life too seriously. We now measure and review equipment for free! Click here for details.

"Low Frequency Interpolation", another BS software from the author of "Minority Clean"

KSTR

Major Contributor
Joined
Sep 6, 2018
Messages
1,824
Likes
3,892
Location
Berlin, Germany
Another funny software,"Low Frequency Interpolation", from http://www.mics.ne.jp/~cdorya/

Discussion of Minority Clean is here.

Translated with deepl:
Because the sound is too accurate, it causes the following problems

Why can't I hear the bass?
When editing the music, I ran a SINC 512 point resample on the 96kHz edit, so anything below 94Hz is fragmented. As a countermeasure, you can use low frequency interpolation or use speakers smaller than 8cm which will make anything below 100Hz blurred.

Headache and nausea!
Speakers that feature a flat balanced output, such as the BRAVIA LCD TVs, can cause headaches and nausea when playing music that has been edited for resampling, as only the high frequencies are strong. The movie Cats Cats is the worst example, where the headache and nausea become increasingly unbearable to watch.

The cause of smartphone hearing loss is the studio's editing process.
SINC resample processing is used in all waveform processing. Also, if you do 16-point processing with your phone's equalizer, anything below 1378 Hz will be fragmented. If you want to edit at 96 kHz, you need to change the frequency to 23 Hz with 2048 points.
Sinc resample points
The SINC resampling software is designed for 44.1kHz/48.0kHz. Fragmentation below 43Hz/47Hz can be resurrected with an analog buffer, but it will overlap with the resampled bass, resulting in unbalanced and cheap audio.

What is the low frequency interpolation process?
When processing SINC resampling, we resample from fragmented bass information by skipping seven points in the referenced music data. Initially, we used a single point skipping method for odd and even columns, but now we use a seven-point skipping method because it is more effective to skip large reference values. Some distortion and noise may be felt. Low frequency interpolation processing The sound quality can be used regularly in Low mode.

Ok, he's blathering about resampling during musisc production / editing with a sinc filter of finite length which is alleged to corrupt bass frequencies below sample-rate / filter-length as these frequencies "are not processed". Far out. No real understanding of what's going on in the various topics he touches.
He shows this table with the "corner frequencies":
1622234706999.png




I did a test run of this "LowFreqResample65536x16x8.exe" software (using various settings) and guess what, it does nothing to the bass frequencies, as checked by DeltaWave:
1622236259874.png

Rather, we see the filter ripple (this was using the "128" size, with higher values the ripple is smaller/nonexistent)

1622235637784.png

But, hey, something is happening at high frequencies, there is a transition very close to fs/4 (11025Hz)?


Looking at the phase plot we find this funny stuff:
1622235553816.png

As suspected, exactly at fs/4 the phase changes from 0degrees to a constant -90degrees and continues this way up to fs/2. The slope of this step change is governed by the selected "sample size" in the software. A very strange allpass filter this is. No idea if this any intentional but if it is, what has it to do with the bass?


The impulse response of the filter looks like this
1622235761111.png



The software, besides doing silly things, is also buggy. The first few seconds have a sample offset of 0.5 samples but after a while that step-changes to 1.5 samples, as if there was a single missed sample?
 
Last edited:
OP
KSTR

KSTR

Major Contributor
Joined
Sep 6, 2018
Messages
1,824
Likes
3,892
Location
Berlin, Germany
I don't know if people are aware of the implications of the impulse response.... besides the "deformed" center section it has massive pre- and pos-ringing, and that happens at very audible 11kHz. Every transient with content in that region will ring the 11kHz bell. It would be no wonder if that is perceived as more "airy" -- and because we've been told so, also "better bass".
 
Top Bottom