Voice edit: what does Cutoff filter do

Started by BogdanH, May 30, 2023, 10:11:30 AM

Previous topic - Next topic

BogdanH

Recently there was a discussion in this forum about voice editing, where Cutoff filter was mentioned. We all know that Cutoff setting influences the sound of voice, but because Yamaha documentation isn't that specific, we can't really know how big the changes are at certain setting. Means, we make setting so the voice sounds to our liking, without knowing what actually changes (and by how much). That moved me to make some further investigation.

For that purpose I created a voice where a single "white noise" sample is used. In short: white noise is a sound that contains all frequencies from 20Hz to 20kHz and all frequencies have the same loudness level. Means, white noise is linear in whole audio range and we need that so we can recognize changes after applying settings.
After importing white noise voice into keyboard, I started to record audio onto USB stick at various Cutoff filter settings. Then I loaded resulting wav files into audio editing software on PC, so I could see the changes.

I have recorded at the following Cutoff settings: 0, -20, -30, -40, -50 and -60 (btw. there's no visible change at -10, so I skipped that).
And here you can see resulting screenshots.





-I have no idea for how long images will be hosted at that site (I'm sorry if you missed them)

As we can see from screenshots, depending on Cutoff setting, frequency starts to cut at:
  0: no cut (same as original)
-20: ~10kHz
-30: ~4,5kHz
-40: ~3,0kHz
-50: ~1,7kHz
-60: ~1,2kHz

-from above table we can approximately interpolate results for other settings (i.e, Cutoff value -25 will start cutting frequency at about 7kHz).
If we observe falloff, we can see that it's about 10dB/octave at all settings (which is quite steep).

That's it. Now we all know what Cutoff does  :)

Bogdan
PSR-SX700 on K&M-18820 stand
Playing for myself on Youtube

Misu

Hi Bogdan,

To see what cut of make you can go in YEM - Common Sound and when the value is 0 you can see no difference for sound, if you choose the -64 value for example the sound become stiffed(hidden) without high frequency.(at cymbalom the players put a handkerchief between the strings)
Since common parameters can be changed in the keyboard, these changes can also be seen directly in the keyboard.
Unfortunately, this does not help to create a usable voice in style.
To see this, cut-off settings must be made in Wave assign General LPF Filter and then each Element - Velocity can have a different sound depending on the velocity.

Best regards
Mihai
PSR SX-900; PA 1000; AKX 10

BogdanH

hello Mihai,
Changing Cutoff in YEM (Common section) does the same as changing Cutoff on keyboard. And yes, we can hear that voice is being changed, but we (majority, I guess) don't know what causes that change and by how much. The only thing that manual says is: Determines the cutoff frequency or effective frequency range of the filter. Higher values result in a brighter sound -which is very basic info (we can hear that by trying anyway). Btw. nowhere in manual is mentioned that values above zero (0...+64) have no influence and I think my article explains why's that.
In short, my article is pure academic. It was not my intention to describe how to do this or that, because it would be out of scope.

Greetings,
Bogdan
PSR-SX700 on K&M-18820 stand
Playing for myself on Youtube

DerekA

To get the classic synth type sounds, pick a bass or pad, whap up the resonance and play with the filter. Lower filter values make the resonance more prominent.
Genos

BogdanH

For sake of completeness, I would like to add some further info related to Cutoff filter...

In first post of this thread I described Cutoff filter behavior on keyboard (which we can change by editing voice). The same Cutoff (and Resonance) setting can also be found in YEM voice editing: Common section, Sound panel -where both values are set to 0 by default.
When creating new voice (in YEM), then we usually define settings in Common section to our liking, although after installing the voice, we can change those values on keyboard anytime later. In short, it doesn't matter if we define Cutoff in YEM Common section or we do that on keyboard later.

More important to know is, there's another Cutoff setting in YEM, which we can find in each Element section, General panel. Here it is simply called Filter, which is by default set to LPF (Low Pass Filter). Other Filter settings are:
Gain: -50...+50 (default 0),
Cutoff: 0...100 (default 100)
Resonance: 0...100 (default 13)

Let's forget about Gain (is kinda self explanatory) and concentrate on Cutoff and Resonance values. As we can see, value range for each setting is totally different from those on keyboard (and in YEM Common section). As we know, settings in Element(s) are permanent and we can not change them later on keyboard! Because we can't listen to result made by these settings (we need to install the voice first), changes we make here are more or less a matter of lucky guess. Or should we leave them at default settings? Resonance is set to 13 by default... but I don't wish artificial resonance, because my wav samples are perfect as they are. Should I set Resonance to 0? No need to open YEM manual, because answers aren't there.

The answer to all questions is: setting values are just numbers that have no logical meaning and so it's impossible to predict the result. From testings I have done, Cutoff frequencies are as follows:

Cutoff Freq
100=   20.0 kHz -no cut
90 =  ~20.0 kHz -no cut
88 =  ~10.0 kHz
86 =   ~8.0 kHz
84 =   ~7.0 kHz
82 =   ~5.0 kHz
80 =   ~4.5 kHz
78 =   ~4.0 kHz
76 =   ~3.5 kHz
74 =   ~3.0 kHz
72 =   ~2.5 kHz
70 =   ~2.2 kHz
68 =   ~2.0 kHz
66 =   ~1.8 kHz
64 =   ~1.4 kHz
62 =   ~1.2 kHz


-and falloff is ~10dB/Octave (I didn't try, but I assume Filter LPF6 is maybe 6dB/Octave?).
I didn't check lower (than 62) settings, because it's pointless. Also keep in mind, that my results are not 100% exact, but they're accurate enough for practical use.

I didn't make some deep analysis about Resonance settings, because my goal was only to find the setting that has no impact on original sample sound (no resonance). And according to my findings, that is at Resonance=6 (although the difference between 6 and 8 is very small).
And what happens at Resonance 0 and 13? At Resonance=0 there's about 3dB/oct decrease and at Resonance=13 about 3dB/oct increase at the Cutoff setting. 3dB seems not to be a lot, but difference can be clearly heard at higher frequencies.

Btw. I didn't come to all those numbers by listening to the sound (would be impossible), but by observing resulting spectrogram (see first post in this thread).
I know all this is quite boring for majority... I'm doing that for remaining 2% and for myself, of course  ;D

Bogdan
PSR-SX700 on K&M-18820 stand
Playing for myself on Youtube

Joe H

FYI... the COMMON parameters in YEM are converted to the Voice Set file. That means all those parameters are available in Voice Set.  Voice Set only applies to the Left and Right Voices.  There is no Voice set file for Voices used in the style.

Joe H
Music is the Universal Language!

My Article: Using Multi Pads in registrations. Download Regs, Styles & MPs:  http://psrtutorial.com/music/articles/dancemusic.html

BogdanH

hi Joe,
Thank you for commenting. Yes, that's correct and is something we need to keep in mind: voice settings that we apply on keyboard (by editing voice) or in YEM Common section, are ignored when voice is used in style. When voice is used in style, only voice settings in YEM Element(s) apply.
Or if we look from another angle, settings that are made in Element(s) are permanent -we can't override them by editing voice on keyboard -and so, needless to say how important settings in Element(s) are at voice creation.

Bogdan
PSR-SX700 on K&M-18820 stand
Playing for myself on Youtube