What am I hearing? Can it be filtered out?

[TyFord]

Hi,

 

I'm working on a music project for a client. I am using a friend who is a composer to build tracks. He has an acoustic grand piano and records into PT with a pair of C414 mics.
The files I have received from him (below) have some odd ringing, like harmonic distortion. Can you tell me what the cause might be and if these files can be filtered to remove the "fizz?"

Thanks,

Ty Ford

 

 

[cjf]

Can you load the file into Audacity to see if it shows anything odd?

 

https://www.audacityteam.org/

[TyFord]

I saw some strange waveform artifacts when I zoomed all the way into Pro Tools. That doesn't tell me how it got that way or if here's a way to fix it.

 

Thanks,

 

Ty Ford

[Axiom05]

What is the format of the recording, bit rate? Is it lossy or lossless? If I had to guess, sounds like it is over processed/data compressed.

[TyFord]

24-bit .wav, 48 kHz.

[sandyk]

 

Quote

The files I have received from him (below) have some odd ringing, like harmonic distortion. Can you tell me what the cause might be and if these files can be filtered to remove the "fizz?"

Thanks,

Ty Ford

[TyFord]

so, that's a "no."

[yamamoto2002]

It seems it is recorded at 48kHz. 4kHz or above frequency component contains non-musical harsh digital noise: The spectrum pattern is repeated above 4kHz,  it is somewhat similar to the aliasing noise when the same digital sample value is repeated 12 times (Fig.3) ...

 

Cutting the signal 4kHz or above (Fig.2) to alleviate this noise. Sound becomes somewhat Lo-Fi ish

Delete4kHzOrAboveResult.mp3

 

Fig.1

 

Fig.2 Cut 4kHz or above frequency components

 

 

Fig.3 Downsampled to 4kHz PCM and zero-order-hold upsample to create 48kHz PCM

[bluesman]

That sounds like it may even be in the piano.  It’s not a common occurrence, but a loose or broken winding on a string can do this.  So can extraneous material sitting on a few strings.  There are many felt parts in there, and any of them could shed a piece.

 

I once had a similar sound that turned out to be the power cord from my iPad lying against the strings.  I leave a tablet on the lyre because so much of my music is in PDFs - and I left the charging cable plugged in, thinking it was secure on the wood.

[TyFord]

Thanks for that! 

 

My vendor, while assuring me that nothing was wrong, sent a file later that didn't have the problem. 

 

Phew!

 

Ty Ford

[yamamoto2002]

I tried to improve the SQ using machine learning, and it was limited-success...

 

Inferred_PianoModelNN_epoch30_8to16_LR_a.flac

 

Network model used is Audio Super Resolution : https://kuleshov.github.io/audio-super-res/

 

 

[TyFord]

Thanks for your effort Mr. Yamamoto. So, mostly over 4 kHz? As the piece builds, your work is more effective!

Any thoughts on what went wrong during recording?

 

[yamamoto2002]

I'm not sure what happened but maybe it is error of clocking something like: the PCM signal is recorded at 48kHz SR, later it is filtered through DAC - analog effectors - ADC chain on editing/mastering and its DAC is set to 48kHz SR and 48kHz signal is played but some clocking is set to 44.1kHz and once in every 12 samples or so are skipped ?

 

I think, when neural network is well trained, solo piano sound will be recovered very well from only DC to 4kHz info. The network is trained using 100 tracks of my piano solo music collection (Clementi, Hanon, Debussy, Ravel, Messiaen, Chopin, Schumann and Beethoven) as its 8kHz PCM input and desired 16kHz PCM output pairs and the learning took only for 3 days with one Nvidia card

[TyFord]

Wow! Very impressive! Thanks so much for the process notes.

I was able to convince the composer/recordist eventually that something was wrong. My client could not hear the problem!

We got new tracks and they sounded much better. I haven't had the heart to pin the composer down on what he found. That'll probably come later after a few beers.