Freesound Forums

Hi,

I need advice how to record sounds better that have a high dynamic range.

For example thunderstorms: How can I record the non thunder-part
in a recording louder ?

I lose so much sound in between the thunderroars -
because its clear, I HAVE to level the mic for the loudest sound..

What is the solution here ? Do I need a better recorder with auto level ?
My current recorder is a tascam dr-05 (Im still really happy with it)

Is it done in post-processing ?

Thanks for your help

As long as the quiet part is recorded well enough quality-wise once you set the recording not to clip at the loudest peaks, you have quite a few available options in post: gain riding/staging, compression of the loudest part (even limiting if it suits you), upward compression of the quiet part, full compression above the "silence" level, and eventually a gate/expander below it to compensate the makeup gain or upward compression...

I'd say, try increasing the gain of such quiet parts in your recordings (pure mathematical gain, no amp simulation) and check how they sound.
if they're ok, you can leave it to post; if the safety settings put the quiet parts too close to the background noise you'd like to remove (or too lo-fi anyway), you'll need some compression/limiting/auto-leveling at recording stage.

Hey there,
Another thing you can check is the bitrate of your recording, 16 24 or 32. The higher the better.
32 bit is the hot thing right now of "no need to set gain" but you can get away with 24bit and leaving a lot of headroom.

Otherwise, indeed set the gain for the loudest thing, it can be hard to judge sometimes.
But dont be too scared of recording quiet and adding gain in post. You can probably get away with adding 30db in post, if not more!
If in doubt leave more than 12 db headroom

Good luck ^^

When I record a coming thunderstorm I start with a noise-floor -40dB.
That means that there is a lot of space for loud thunder.

In post I use Audacity for post compression from a VST, so I can experiment around the loud parts of the recording finding a nice balance between quiet and loud parts. And yes, there is a limit on that bet very acceptable.

These are my recordings using a handheld recorder from Sony.

https://freesound.org/search/?q=Thunderstorm+klankbeeld+sony&f=&s=Duration+%28longest+first%29&si_tags=0&si_name=0&si_description=0&si_packname=0&si_sound_id=0&si_username=0&d0=0&d1=*&ig=0&r=0&g=0&dp=0&cm=0&mm=0

Good luck experimenting the coming summer. It took me a lot of experimenting, but great lessons.

Regards

Hello RandomRecord19! I am the owner of the same recorder. And he's fine with me, too. But I do a variety of post-processing of almost all records. It is unlikely that auto-adjusting the levels on the recorder can give adequate results. I've never even turned on the limiter on the recorder. And as Klankbeeld has already said, postprocessing is an excellent lesson for gaining experience and training skills in working with sound and mastering sound software, implementing the theory studied, etc. But first you need to study the theoretical foundations of working with sound. Figure out what it is and why it exists, for example, compression, equalization, and many, many other things. The main point is not to use "preset thinking" here. I mean, we need to try to ensure that each soundtrack, if it requires post-processing, is processed in its own way. I will share my thoughts and experiences. Probably, using the example of this recording of a thunderstorm, which I recorded in 2021 https://freesound.org/people/newlocknew/sounds/580440 / . Previously, there was a version of this post without postprocessing on a separate page, but I deleted it. I remember for sure that I set the recording level to 5 units (on Tascam DR-05x, the recording level is in conventional units). It was written in 16 bit/48 kHz format. And I was lucky that I guessed right here, and there was never a clipping anywhere. Although the storm dynamics were just extreme. The following post-processing was applied here: a cut-off low-frequency filter (at 20 Hz / 18 Db per octave) to remove low-frequency "debris". This often helps to slightly increase the average level in the editor. The level of the Side component has been slightly increased for minimal stereo expansion (now I'm doing it in a more efficient way, but I'll tell you about it separately if anyone is interested). And then I cut the peaks with a limiter, in three stages (there was no compressor in the circuit, but only a limiter, which of course is quite rude). That's all.
I recently made a fresh version of the processing of this record, where I applied a more fresh look. Maybe I'll post it soon. In this version, in addition to using a compressor and limiter, a dynamic equalizer was also used. It was used for two reasons. 1. Very fine suppression of low frequencies of thunderclaps by a shelf filter, with a set trigger threshold. Thus, due to a slight attenuation of the bass component of the thunderclaps, I slightly compressed the dynamics of the recording. The main thing here is to choose the right filter frequency, Q-factor and suppression level, i.e. do not overdo it, and do not deprive the foundation rolls. 2. This made it possible to use a more gentle compression, which was highly desirable, since when the compressor was triggered, the "sinking" of rain noise was clearly audible on some rolls. I also tried to use a multiband compressor here, but I liked the work of the dynamic equalizer and compressor more. I also "cleaned up" the resonant hump at 250 Hz. In general, I often slightly suppress various resonances on rain recordings and not only, which may appear in different acoustic recording conditions. These may be resonances from the sound of rain on a metal roof, for example. And again, the main thing here is not to overdo it. After the compressor, and stereo expansion (a bundle of stereo imager and mid/side exciter+ compensating equalizer) nevertheless, we had to apply a light limitation. Unfortunately, it could not do without this. Anyway, it was the most difficult thunderstorm recording I've ever made. It was difficult to achieve a balance: to preserve the natural dynamics of thunderclaps, and at the same time bring the rain noise level to approximately the level at which I perceived it when recording. And of course, it is not necessary that you will need exactly the same processing scheme on another record.
Someone will find all this too excessive and will be right in their own way. But as already noted, all this trains your skills.

@Newlocknew

What a great lesson.
Thanks

Klankbeeld! I'm always happy to share!))) Thanks!

So it seems post processing is the solution.

I know this feature(s) in audacity (what I use only)
but I was always scared to use it to make my recordings louder..
scared of decrease of quality or noise increase.

So I could use amplify or loudness normalization or a other "normalize"
and theres a compressor..

Which one would be a good "no brainer" for me ?
(something that give a good boost in volume without quality decrease)

I record mostly nature sounds.

Thanks alot for all these answers and help

Well, I fear we're not yet at "no brainer" stage; AI will get us there eventually, but for now it's better to practice with the tools we have until we make them do what we want.

The very nature of each recording might require a different approach from the previous one.

Quality decrease is not really an issue, it can only happen if you misuse the tools in post, e.g. causing distortion by pushing the signal above 0dB; an eye on the level meters will help you avoid it.

As for the noise, as suggested above you two main ways of taming it before the general volume increase: a high pass filter is perfect if the noise frequency range doesn't overlap with the signal range; a typical case is your gear's self noise, often a low frequency hum. If there's overlap, you'd have to resort to noise removal/reduction.
In both cases, I'd suggest to do it as first step, and do the boosting part later.

Of all the mentioned tools, compression is the easiest to misuse and cause more harm than good or, at the very least, get unsatisfying results; e.g. you listen and you can tell exactly when the compressor kicks in.

Imagine rain and thunder, if you only want to compress the thunder and you don't set the parameters well, you'll notice a volume change in the rain; in this case you might be better off with a dynamic equalizer focused on the thunder frequency range, leaving the rain untouched.

An easier way to use the compressor in a case like this would be to have it always active, not just on the loud parts, so you'd reduce the overall dynamic range without ups and downs; just set the threshold right above the noise level and you're done.

If you take some time to practice, you might find a good general workflow and either create macros in Audacity or, in a DAW, save a session template with the chain of plugins already set in the ballpark to speed up your next processing.