Video: How a noise gate works and why it distorts the signal when the release time is too short

Blue_Mangoo

rs2000

While you're at it, please consider adding one parameter: Downward expansion ratio.
Or just use Apple's native "Dynamics" AU effect.
(kAudioUnitSubType_DynamicsProcessor)

audio_DT

@Blue_Mangoo said:

Interesting and informative video, as ever.

Blue_Mangoo

@rs2000 said:
While you're at it, please consider adding one parameter: Downward expansion ratio.
Or just use Apple's native "Dynamics" AU effect.
(kAudioUnitSubType_DynamicsProcessor)

In what situations do you use that downward expansion parameter? I was thinking to leave it out in order to make the app run lighter on the cpu.

Our current design has a closed gain control, so it can reduce gain without shutting off the audio completely.

Replacing that with a ratio control means the circuit has all the components of a compressor.

rs2000

It's not about adding a compressor, not at all.
Have you seen Auria yet?
What I mean is the adjustable downward expansion ratio that is often fixed at 1:infinite in conventional noise gates.
Sometimes you can adjust the "closed gain" (like you apparently have in your code already) but only few good ones offer this kind adjustment which helps to get much more natural-sounding noise suppression without the disadvantages of a typical old-school gate.
Not offering this kind of adjustment is a missed opportunity IMHO.

Blue_Mangoo

@Blue_Mangoo said:

@rs2000 said:
While you're at it, please consider adding one parameter: Downward expansion ratio.
Or just use Apple's native "Dynamics" AU effect.
(kAudioUnitSubType_DynamicsProcessor)

In what situations do you use that downward expansion parameter? I was thinking to leave it out in order to make the app run lighter on the cpu.

Our current design has a closed gain control, so it can reduce gain without shutting off the audio completely.

Replacing that with a ratio control means the circuit has all the components of a compressor.

Although it is not a compressor, but it has all the same components as a compressor. We have those components available and could use them to replace the closed gain control with a ratio control. But since that will significantly increase the complexity of the design, I am wondering how much different the sound would be if we do that replacement. Personally, I use a noise gate to silence the noise from my guitar pickups, and I don’t think the expansion ratio control would be worth the extra processor cycles in that context. Can you tell me about some contexts where downward expansion sounds better than the simpler noise gate design?

rs2000

@Blue_Mangoo said:
… Personally, I use a noise gate to silence the noise from my guitar pickups, and I don’t think the expansion ratio control would be worth the extra processor cycles in that context. Can you tell me about some contexts where downward expansion sounds better than the simpler noise gate design?

Absolutely. I'll record an example for you.

rs2000

@Blue_Mangoo
An even better example:

Blue_Mangoo

Thank you. I’ll look at this tomorrow and look into switching from gate to downward expansion

One good thing about the expander is that it’s less prone to that zippering problem and therefore doesn’t need such long release time. It might also allow us to use a simpler smoothing filter, which would partly make up for the time lost to computing the gain as a ratio.

McD

More cool engineering disclosure... I hope there are engineering students soaking in this
information.

TL:DL The Noise Gate app will ship soon and well before the Saturator app.

rs2000

@Blue_Mangoo said:
Thank you. I’ll look at this tomorrow and look into switching from gate to downward expansion

One good thing about the expander is that it’s less prone to that zippering problem and therefore doesn’t need such long release time. It might also allow us to use a simpler smoothing filter, which would partly make up for the time lost to computing the gain as a ratio.

Yes! I wanted to write exactly that, so I'm even more happy to see your reply now.
A downward expander can be set to act like a gate but not the other way around, so...

Edit: This is how Apple's free Dynamics Processor UI looks on MacOS when set to downward expansion mode (no soft knee though):

[Deleted User]

The user and all related content has been deleted.

Blue_Mangoo

@rs2000 said:

@Blue_Mangoo said:
Thank you. I’ll look at this tomorrow and look into switching from gate to downward expansion

One good thing about the expander is that it’s less prone to that zippering problem and therefore doesn’t need such long release time. It might also allow us to use a simpler smoothing filter, which would partly make up for the time lost to computing the gain as a ratio.

Yes! I wanted to write exactly that, so I'm even more happy to see your reply now.
A downward expander can be set to act like a gate but not the other way around, so...

Edit: This is how Apple's free Dynamics Processor UI looks on MacOS when set to downward expansion mode (no soft knee though):

This morning I replaced the closed gain control with a downward expansion ratio control. It requires the addition of one log operation and one exponentiation per sample. DSP engineers of the past avoided those operations because of fear they would be too slow. But on modern hardware the speed doesn’t seem to be a problem. Especially if the code doesn't have the processor waiting for the output of a previous calculation before going on to the next one then it seems to be just fine. I think my intention to avoid those functions might be nothing more than a bad habit, so I went ahead and implemented the ratio control. Thank you for suggesting that change.

rs2000

@Blue_Mangoo said:
This morning I replaced the closed gain control with a downward expansion ratio control. It requires the addition of one log operation and one exponentiation per sample. DSP engineers of the past avoided those operations because of fear they would be too slow. But on modern hardware the speed doesn’t seem to be a problem. Especially if the code doesn't have the processor waiting for the output of a previous calculation before going on to the next one then it seems to be just fine. I think my intention to avoid those functions might be nothing more than a bad habit, so I went ahead and implemented the ratio control. Thank you for suggesting that change.

You're welcome and thanks for considering!
If you worry about CPU usage anyway (and low CPU is always good!), you could do two things:
1. Disable log/exp operations when the ratio control is at the "gate"/infinite position
2. Approximate the calculations by multiple linear segments, just as many as necessary to make the resulting sound indistinguishable. (Fast circle drawing in bitmap computer graphics has been done this way).

Blue_Mangoo

@rs2000 said:

@Blue_Mangoo said:
This morning I replaced the closed gain control with a downward expansion ratio control. It requires the addition of one log operation and one exponentiation per sample. DSP engineers of the past avoided those operations because of fear they would be too slow. But on modern hardware the speed doesn’t seem to be a problem. Especially if the code doesn't have the processor waiting for the output of a previous calculation before going on to the next one then it seems to be just fine. I think my intention to avoid those functions might be nothing more than a bad habit, so I went ahead and implemented the ratio control. Thank you for suggesting that change.

You're welcome and thanks for considering!
If you worry about CPU usage anyway (and low CPU is always good!), you could do two things:
1. Disable log/exp operations when the ratio control is at the "gate"/infinite position
2. Approximate the calculations by multiple linear segments, just as many as necessary to make the resulting sound indistinguishable. (Fast circle drawing in bitmap computer graphics has been done this way).

Good points. Thanks again.