SoundDevices NoiseAssist (1 Viewer)

[Jon.Bryant]

Probably a bit of a remote chance, but has anyone with a Sound Devices MixPre recorder purchased and used the noise assist plugin?

The bumf of SoundDevices website, states that the plugin monitors background noise such as traffic, HVAC etc, and then reduces this noise. As I live in an area where it is hard to escape traffic noise this could be useful, provided it doesn't compromise recordings.

I have asked the Sound Devices team for advice, but if anyone has hands-on experience of making bird recordings with the plugin it would be great to get your feedback.

 

[Borjam]

I don't own a MixPre Series II but one of the original series (MixPre 3) which I love. I have used both of the available plugins for mine (Music and Mix Assist) and both have been a real delight for their intended applications.

I am not sure Noise Assist is necessary for your use case unless you really need real time noise reduction. I think that plugin is mostly designed for motion picture/documentary production cases in which the sound mixer sends a ready to use (or almost ready to use) mix to camera.

In your use case probably you will get better bang for the buck using some post production tool such as iZotope's RX which can be purchased at a discount now and then.

As for signal degradation, noise removal plugins may cause it and I am not sure how they will work with "unusual" signal sources such as bird songs. Anyway in case you decide to get Noise Assist it can be assigned to a mix bus so that the ISO track will always be clean.

 

[Joop99]

Probably a bit of a remote chance, but has anyone with a Sound Devices MixPre recorder purchased and used the noise assist plugin?

The bumf of SoundDevices website, states that the plugin monitors background noise such as traffic, HVAC etc, and then reduces this noise. As I live in an area where it is hard to escape traffic noise this could be useful, provided it doesn't compromise recordings.

I have asked the Sound Devices team for advice, but if anyone has hands-on experience of making bird recordings with the plugin it would be great to get your feedback.

The next solution greatly depends if you want to be mobile and if you use symmetric connections. Use a parabolic reflector or so called rifle mics for your recording, setup a tripod with a random mic pointing not in the direction of your subject for recording. Set it up for recording phase reversed. Make the recording, you can choose for direct mixing or do it at home. Record the channels. With direct mixing you can mix the phase reversed channel into the object recording, the amount of effect can be adjusted with the record volume knob. Only you can't undo that operation, so record it at a separate track and do the job at home.
What's happening? You record the object as usual with parabolic reflector or riffle mics. Your random mic does record a little of your object, but everything which is around and is unwanted noise. Because its recorded phase reversed it wil cancel out the the same signals in the recording of the object when added in the mix. You have to experiment, because there is a small phase difference due to the distance between the microphones. Its the same technique applied with hum, that's why you find phase reversal switches on the more serious mixing desks.

 

[Jon.Bryant]

Thank you both for the replies.

In your use case probably you will get better bang for the buck using some post production tool such as iZotope's RX which can be purchased at a discount now and then.

I agree that it makes sense to deal with the noise in post, which would give me more control over the end result. I need to lean how to use the software I have - I note that two products I own have noise reduction plugins!

I think where I struggle is with dealing with wide spectrum variable noise. So noise form individual cars, aircraft (I live not too far from Heathrow!) etc. I was hoping that NoiseAssist may be something clever, that monitoring these variable noises and cancelled them out. SoundDevices are yet to come back, but I will confirm their response when they reply. I think most plugin are good at reducing consistent noise, so perhaps the general rumble from a distant road, but if you are aware if iZotope's RX can deal with variable noise levels I would be interested to know.

Your random mic does record a little of your object, but everything which is around and is unwanted noise. Because its recorded phase reversed it wil cancel out the the same signals in the recording of the object when added in the mix.

Never thought about doing this, but have seen phase inversion used to demonstrate the difference in noise floor between a 16 and 24 bit recording - not a lot if you have a reasonable signal to noise ratio in the first place. I have also thought about using phase inversion to check my edits - basically that I have only deleted noise rather than bird vocals.

I think I will give your suggestion a go as an experiment. I think that doing the phase inversion in post may be best, particularly as I think there will be a lot of fussing with levels (after all you are trying to match a sensitive direction mic, which should reject some off axis noise, with the noise picked up from a less sensitive one, but probably picking up noise from a wider area).

I think it would probably work best with a shotgun, so you have two mics with a flat frequency response - with a parabola, I think you would need to add equaliser effects to the noise channel, to counteract the fact that higher frequency sounds have greater gain in the parabola . Perhaps overthinking it - and an experiment wouldn't be that hard to do in any case.

I only have a cardiodal mic to use to record the noise and suppose an omni would be best, but I think it could work - with traffic noise set up the mics as close as possible with the directional mic facing away from the noise and the cardinal facing towards it; for aircraft noise, perhaps face the cardiodal mic upwards.

Not very portable, but it could be a blast on occasions, if it it works well - we have some Nightingales nearby, but unfortunately the place borders a main road - an ideal test case!

 

[Jon.Bryant]

Response from Sound Devices is as follows

"Thank you for contacting Sound Devices. Noise Assist will work across the entire spectrum, but it is really geared to reduce background noise while dialogue is prominent. My fear is that some of the subtle bird vocalizations would be affected by Noise Assist. You might have to experiment a little there. I would suggest applying NA to one of your mix tracks for comparison. So, for example, if you are sending an equal audio mix to both the left and right channels, you can run NA on the left channel only. Then you would have 2 results:

• Left channel mix of audio (destructive N/A recording).
• Right channel mix of audio (non-destructive recording).

Yes, NA will definitely help with aircraft noise! If it's extremely loud, there is not much that NA can do, but if it's background noise similar to traffic, humming fans, etc., Noise Assist will definitely help reduce that type of airplane noise"

If I really wanted to get technical, I suppose I could phase invert the recordings with and without, to check whether the bird vocals were impacted - if the plugin works well for birds, all I should be able to hear should be noise.

Interesting that it is suggested the plugin will work on aircraft. This is in response to me asking whether the plugin can work on semi transient background noise such as overflying aircraft. I suspect that this would not be that simple - unless you were going to try to turn up the reduction levels during the recording, you would need to guess a reduction level before the plane approached and was at its loudest.

I just tried a DeNoiser on my DAW and I am not that not that impressed! Looking for a VST noise reduction plugin, I almost fell off my perch when I saw an outline price for the Cedar plugin, which seems to be the best in the field. Perhaps an option if I win the lottery! The examples on the web for the Cedar hardware deNoiser do however seem spectacular - at least when dialogue is being recorded.

 

[Jon.Bryant]

Thinking about it again, I think phase inversion using two separate mics will only work for low frequency sounds with a long wave length. My logic is as follows;

Say the speed of sound is 343m/s at 20 degrees, then a 1kHz sound will have a wave length of 0.343m. This would mean that if the mics were only 171.5mm apart (1/2 the wave length), they would already be 180 degrees out of phase. If you then phase inverted the two recordings you would actually be doubling the noise at this frequency! The problem becomes worse at higher frequencies, meaning that the two mic capsules would need to be very close or coincidental. I think that any separation between the mics would probably lead to a kind of banding effect for wide frequency noise, with some frequencies cancelled out and others doubled, depending on whether the separation between mics coincided with X wave lengths or X+0.5 wave lengths.

I think that careful and manual post processing, is probably the only real solution - either that or I need to move somewhere where it is quieter!

 

[Borjam]

I agree that it makes sense to deal with the noise in post, which would give me more control over the end result. I need to lean how to use the software I have - I note that two products I own have noise reduction plugins!

But not all of them are born equal. Although noise reduction software is geared towards removing continuous nois. Which ones do you have?

I think where I struggle is with dealing with wide spectrum variable noise. So noise form individual cars, aircraft (I live not too far from Heathrow!) etc. I was hoping that NoiseAssist may be something clever, that monitoring these variable noises and cancelled them out. SoundDevices are yet to come back, but I will confirm their response when they reply. I think most plugin are good at reducing consistent noise, so perhaps the general rumble from a distant road, but if you are aware if iZotope's RX can deal with variable noise levels I would be interested to know.

I don't think so. Well, actually the noise reduction plugins you can "train" can help reduce the noise by several dB as long as the spectrum of the annoyance does not change. Cars and planes, for example, shift in frequency thanks to Doppler effect. Sigh.

Never thought about doing this, but have seen phase inversion used to demonstrate the difference in noise floor between a 16 and 24 bit recording - not a lot if you have a reasonable signal to noise ratio in the first place. I have also thought about using phase inversion to check my edits - basically that I have only deleted noise rather than bird vocals.

I think I will give your suggestion a go as an experiment. I think that doing the phase inversion in post may be best, particularly as I think there will be a lot of fussing with levels (after all you are trying to match a sensitive direction mic, which should reject some off axis noise, with the noise picked up from a less sensitive one, but probably picking up noise from a wider area).

It can be complicated but worth a shot. Remember that you need actually two processes in order to harness the full power of that approach. Phase inversion and some variable delay plugin. The variable delay will change the effect in a frequency selective way but maybe (I said, maybe!) in a way that removes mid frequency noise and using high pass filtering to remove rumble might help.

I think it would probably work best with a shotgun, so you have two mics with a flat frequency response - with a parabola, I think you would need to add equaliser effects to the noise channel, to counteract the fact that higher frequency sounds have greater gain in the parabola . Perhaps overthinking it - and an experiment wouldn't be that hard to do in any case.

I only have a cardiodal mic to use to record the noise and suppose an omni would be best, but I think it could work - with traffic noise set up the mics as close as possible with the directional mic facing away from the noise and the cardinal facing towards it; for aircraft noise, perhaps face the cardiodal mic upwards.

There is a technique used sometimes for radio communication that uses an entirely different approach. Place a microphone as close as possible to the unwanted noise source (nearby road for instance) and play with phase inversion, delay and level. Or at least use also a directional microphone for the noise source. You want the minimum possible desired signal on the noise microphone.

Not very portable, but it could be a blast on occasions, if it it works well - we have some Nightingales nearby, but unfortunately the place borders a main road - an ideal test case!

Maybe a cheap wireless microphone would make the close noise miking approach more portable

Response from Sound Devices is as follows

"Thank you for contacting Sound Devices. Noise Assist will work across the entire spectrum, but it is really geared to reduce background noise while dialogue is prominent. My fear is that some of the subtle bird vocalizations would be affected by Noise Assist. You might have to experiment a little there. I would suggest applying NA to one of your mix tracks for comparison. So, for example, if you are sending an equal audio mix to both the left and right channels, you can run NA on the left channel only. Then you would have 2 results:

• Left channel mix of audio (destructive N/A recording).
• Right channel mix of audio (non-destructive recording).

Yes, NA will definitely help with aircraft noise! If it's extremely loud, there is not much that NA can do, but if it's background noise similar to traffic, humming fans, etc., Noise Assist will definitely help reduce that type of airplane noise"

If I really wanted to get technical, I suppose I could phase invert the recordings with and without, to check whether the bird vocals were impacted - if the plugin works well for birds, all I should be able to hear should be noise.

Interesting that it is suggested the plugin will work on aircraft. This is in response to me asking whether the plugin can work on semi transient background noise such as overflying aircraft. I suspect that this would not be that simple - unless you were going to try to turn up the reduction levels during the recording, you would need to guess a reduction level before the plane approached and was at its loudest.

I just tried a DeNoiser on my DAW and I am not that not that impressed! Looking for a VST noise reduction plugin, I almost fell off my perch when I saw an outline price for the Cedar plugin, which seems to be the best in the field. Perhaps an option if I win the lottery! The examples on the web for the Cedar hardware deNoiser do however seem spectacular - at least when dialogue is being recorded.

And still Cedar is mostly designed for dialogue recording as far as I know.

Thinking about it again, I think phase inversion using two separate mics will only work for low frequency sounds with a long wave length. My logic is as follows;

Say the speed of sound is 343m/s at 20 degrees, then a 1kHz sound will have a wave length of 0.343m. This would mean that if the mics were only 171.5mm apart (1/2 the wave length), they would already be 180 degrees out of phase. If you then phase inverted the two recordings you would actually be doubling the noise at this frequency! The problem becomes worse at higher frequencies, meaning that the two mic capsules would need to be very close or coincidental. I think that any separation between the mics would probably lead to a kind of banding effect for wide frequency noise, with some frequencies cancelled out and others doubled, depending on whether the separation between mics coincided with X wave lengths or X+0.5 wave lengths.

I think that careful and manual post processing, is probably the only real solution - either that or I need to move somewhere where it is quieter!

Again, delay + phase inversion + level. The closer you match the delay to the distance between microphones, the better it will work for higher frequencies.

 

[Jon.Bryant]

As always I seem to have collected more gear and software than I need (or can regularly use). I have
1/ WaveLabs Pro - Bought a long time ago due to the ability to do precision edits to spectrograms (which was quite unique at the time - at least at the software price point). This is the product where I tried noise reduction the other day and thought it was not that good.
2/ Reaper - Bought this as Wavelab were not doing Ambisonics at that time. Haven't tried noise reduction in Reaper yet, but I understand it has the software has the functionality.
3/ Adobe Audition. This is actually part of the Adobe Creative Cloud, which I have for my work. I haven't really tried Audition until this morning, but the noise tools seem quite good.
4/ Spectral Layers - a great product for editing spectograms. I have tried applying a low pass gate to a file to eliminate excess wind noise, then splitting out the remaining noise with this tool and 'painting back in' a low noise wash, to make it sound more natural. I thought this worked quite well. Another cheat I am thinking about is to capture a 'clean' ambient sound without aircraft or traffic noise, then use this software to extract the noise from a subsequent bird recording and replace the noise with the 'clean' ambient recording. I quite like how you can isolate and listen to the noise, tonal and transient layers separately, so you can listen for any artifacts of a bird vocal that has been placed in the wrong layer automatically and then try to rectify.

Again, delay + phase inversion + level.

Great advice. I was about to give up on the concept, but looking forward again to seeing whether I can make the two mic system work.

Thanks

Jon