Recording with Zoom F3 - best practice regarding to reduce noise?

[alvin18]

I'm a total newbie to audio recording and would like to know some good practices to get the best low noise results.

My goal is to record sounds in nature (wind, rain, birds, etc.). I use a Zoom F3 that records in 32-bit floating point format. Since I can't get very close to the object, I have to amplify the sound, which leads to an increased noise. I mean natural noise from the recorder and microphones, not ambient noise.

In Zoom F3, it is possible to amplify the sound already in the device. I'm currently using x256 times gain to get more or less correct volume level. However, I still have to raise the volume in the DAW anyway even in this case. If I have to use x1 or x64 or similar amplification, the recording would be extremely quiet. If I import such wav sample into my DAW I practically see just a straight line.

So here are some questions.

To get as low noise as possible, is it reasonable to use the Zoom F3's gain/amplify option, or could it be done in a DAW (I use Cubase 12)?

That means, would it be wise to record without using the gain of the Zoom F3 and adjust the sound level to the right level in the DAW? And which process would be the most appropriate to use for this - normalizing, compressor or limiter? Or do you have other suggestions?

I also use Waves X-noise to reduce the noise but the result is not the best and this plugin can introduce some artifacts.

Here are 2 audio examples - https://drive.google.com/drive/folders/ ... sp=sharing

Microphones are Sennheiser MKE 600 and sE Electronics sE8 using phanton power. Yes I use two totally different mics for stereo image but thats another topic

[Hugh Robjohns]

To get as low noise as possible, is it reasonable to use the Zoom F3's gain/amplify option, or could it be done in a DAW (I use Cubase 12)?

It could and should be done in the DAW. If I recall correctly, the gain options in the F3 you're talking about are purely for the internal waveform display and local monitoring.

The file you export for your DAW will be 32 bit floating point and, in all likelihood, will appear pretty darn quiet in the DAW, so you'll need to rescale or normalise the file to be able to use it alongside more conventional or mastered material.

However, I would recommend not using 32bit FP initially. Instead, I think you would find it easier and get a better understanding if you used 24 bit fixed point recording to get a good feel for the capabilities of both the machine and your mics. Maybe revert to 32bFP later if you are recording in situation that warrants it... but I've made countless nature recordings with 24bit fixed point without problem.

32bit FP is useful for people that have no concept of setting gains appropriately, or for situations where unexpectedly loud sounds might happen... but 24 bit fixed point has more than enough dynamic range for most situations.

And 32bFP doesn't make the machine any quieter than in 24bit mode — you're using the same preamps — it just allows the machine to cope with high peak levels with less risk of overload (at the expense of potential noise modulation artefacts under some circumstances).

...And which process would be the most appropriate to use for this - normalizing, compressor or limiter? Or do you have other suggestions?

If working with 32bFP files, then use the Normalise function, and set it to normalise to a peak of -6dBFS. That process will simply raise the gain so you can see and hear it, without changing the actual audio. Compression and limiting will both change the dynamics of the louder material, and while you might want to indulge in that as part of your later post production, you need to start off just working with the raw audio to assess it and decide what to do with it.

Microphones are Sennheiser MKE 600 and sE Electronics sE8 using phantom power.

These aren't the quietest mics on the planet at 15 and 13dBA, respectively, but they're not the noisiest either. I typically use Sennheiser MKH40s which have a self-noise figure of 12dBA, so quieter, but only by a few decibels.

To get much quieter you're going to have to switch to large diaphragm mics — which adds a whole world of wind noise suppression problems — and I don't find that necessary. Ambient noise is usually the dominant factor.

I'd experiment with optimising the recording gain in 24 bit fixed point mode as a first step and see how you get on with that...

[James Perrett]

I also use Waves X-noise to reduce the noise but the result is not the best and this plugin can introduce some artifacts.

From what I've heard, X-Noise is very primitive compared to other alternatives. In my opinion, Izotope's RX has the best noise reduction available at a sensible price and is light years ahead of X-Noise.

[alvin18]

Thank you for the tips.

If I recall correctly, the gain options in the F3 you're talking about are purely for the internal waveform display and local monitoring.

Changing the amplification level while recording will not change the recording volume of the final audio file. But setting the amplification level before starting the recording then I have these levels reflected in the final recording.

F3 records only in 32-bit float so I can't use 24-bit and I can't set gain.

My main concern is maybe if I boost in F3 recorder or normalize in DAW won't the noise get the same level as final result?

From what I've heard, X-Noise is very primitive compared to other alternatives. In my opinion, Izotope's RX has the best noise reduction available at a sensible price and is light years ahead of X-Noise.

Do you know if the "RX Elements" bundle include any plugin for that purpose? There are 3 different bundles and the description of different plugins is not very clear...
https://www.izotope.com/en/products/rx/ ... .html#comp

[Drew Stephenson]

RX elements has the de-noise plugin and it's very good.

[James Perrett]

As Drew says, RX Elements has Voice Denoise while RX Standard has a general Denoise. I must admit that I've not used Voice Denoise because I assumed that it was optimised for voice and most of the time I'm working with wideband music so I stick with the standard Denoise that can be found in RX Standard.

Do you use Voice Denoise on sources other than voice Drew?

[Hugh Robjohns]

Changing the amplification level while recording will not change the recording volume of the final audio file. But setting the amplification level before starting the recording then I have these levels reflected in the final recording.

Ah... apologies... I didn't recall correctly!

F3 records only in 32-bit float so I can't use 24-bit and I can't set gain.

Darn it... sorry... confusing myself — I have an F8n Pro which allows 24 bit recording.

My main concern is maybe if I boost in F3 recorder or normalize in DAW won't the noise get the same level as final result?

I would imagine so, yes.

[Matt Houghton]

For the mic/recorder noise, I'd suggest you try taking your mics and recorder somewhere very quiet and, with the same gain settings as used on your recording, capture around 30 seconds of this boosted self noise in isolation. Then use a section of that recording to create a fingerprint for noise removal in RX, or whatever DAW plug-in you are using.

That should get rid of lots more of that hiss, without having any 'outdoor sounds' interfere with the noise fingerprint (something that will likely cause more artefacts in the noise processing, or mean you can't push it as far).

If doing this in RX, personally I'd use the spectral noise removal module and do it in two or three separate offline passes.

Also, possibly worth pointing out that if recording more localised sounds (eg a specific bird or birds in a particular tree, rather than say the birdsong in the whole area) you could review your mic options. Eg a parabolic reflector might come in handy for reducing unwanted ambience.

[alvin18]

For the mic/recorder noise, I'd suggest you try taking your mics and recorder somewhere very quiet and, with the same gain settings as used on your recording, capture around 30 seconds of this boosted self noise in isolation. Then use a section of that recording to create a fingerprint for noise removal in RX, or whatever DAW plug-in you are using.

Thats a clever tip, thank you, I will try!

As Drew says, RX Elements has Voice Denoise while RX Standard has a general Denoise. I must admit that I've not used Voice Denoise because I assumed that it was optimised for voice and most of the time I'm working with wideband music so I stick with the standard Denoise that can be found in RX Standard.

Do you use Voice Denoise on sources other than voice Drew?

I'd like to know that too before I buy Elements bundle and compare it with Waves X-noise Other RX bundles are maybe too expensive to buy just to try because I am not doing any professional movies. Just want to use some decent quality nature sounds in my wildlife videos.

[James Perrett]

I'd like to know that too before I buy Elements bundle and compare it with Waves X-noise Other RX bundles are maybe too expensive to buy just to try because I am not doing any professional movies. Just want to use some decent quality nature sounds in my wildlife videos.

If you can put up a short sample somewhere I can run it through Voice Denoise. You can PM me a link if you would rather not post publicly.

It is also worth saying that once you enter the RX ecosystem you will receive some pretty good upgrade deals. I started with Elements and I'm now using RX Advanced.

[alvin18]

If you can put up a short sample somewhere I can run it through Voice Denoise. You can PM me a link if you would rather not post publicly.

Thank you for this kind proposal:)

Here you can find two files: "birds peak normalized.wav" and "birds original.wav" - https://drive.google.com/drive/folders/ ... sp=sharing

In normalized version I used Cubase´s peak normalization @ -6 dB and not loudness normalization. Original wav is without any processing right from the recorder, just converted to 24 bit.

[James Perrett]

So I've had a quick play and put the results up at

https://drive.google.com/drive/folders/ ... sp=sharing

I found a section which appeared to be just noise and used the same section to learn from for each file. I also did multiple passes so that you can hear the effect of just one pass of NR and 2 passes.

The files labelled RXVDN are voice denoise on its voice setting while the files labelled RXVDN-Music use Voice Denoise's music setting. The files labelled RXSDN use Spectral Denoise.

All are set to provide 12dB noise reduction for each pass which is usually a sensible starting point.

[alvin18]

So I've had a quick play and put the results up at

https://drive.google.com/drive/folders/ ... sp=sharing

I found a section which appeared to be just noise and used the same section to learn from for each file. I also did multiple passes so that you can hear the effect of just one pass of NR and 2 passes.

The files labelled RXVDN are voice denoise on its voice setting while the files labelled RXVDN-Music use Voice Denoise's music setting. The files labelled RXSDN use Spectral Denoise.

All are set to provide 12dB noise reduction for each pass which is usually a sensible starting point.

Have to say that RX versions sound really a little bit better than wave's x-noise versions.

[Drew Stephenson]

Sorry, late back to the conversation. Yep, I've used voice de-noise on non-voice sources - acoustic guitar and double bass. I've not tried the automatic function in this context, only using the learn option. As is often the case with these things, a couple of light passes often has fewer side effects than one more heavy-handed one.