Dynamic Range and Noise Floor

[Eujene]

I read that:

" In its purest form, dynamic range in music production is simply the decibel (dB) difference between quietest and loudest sounds in a mix or audio file. Therefore, songs with a wide dynamic range will have a larger gap between the loudest sound and quiet sound compared to songs with more consistent sound quality.

Dynamic range can also refer to the loudest and quietest sounds hardware or audio systems can properly represent. The bottom range of this value is called the noise floor."

If you want to listen to the quietest sound like the sound of silence. Do you need high dynamic range? Or Is it possible to have quietest sound only without regards to loudest sound? Meaning is quietest sound always coupled to dynamic range or can the two be decoupled? How?

 

[MaxwellsEq]

ASR is full of clear explanations of this. If you use the search feature you will find some very informative items.

 

[Eujene]

Ok. I have read so much info about it. Dynamic range is related to bit depths of the ADC and occur because of narrow windows capture of real sound. But for real sound like birds whispering in the forest amidst the sound of thunder in the sky. How do you use dynamic range to describe the natural sound without concept of ADC or even analog recording medium like tapes.. Can we say the forest still has dynamic range but only unlimited? I can't find this info about whole night of googling and reading and searching the database here.

 

[unpluggged]

Can we say the forest still has dynamic range but only unlimited?

The dynamic range of mechanical oscillations in a medium is always limited: by the energy of the brownian motion of this medium's particles, at minimum, and by the energy of the shock wave in that medium, at maximum. For standard atmospheric conditions, this is about 191 dB SPL.

For average human hearing, the dynamic range is limited by the threshold of hearing and the threshold of pain, which range is considered to be about 130 dB.

Edit: for human hearing, this is the absolute sensitivity difference. When listening to specific audio signals, we cannot hear all this range due to masking effects. That's why we don't hear, say, background noise such as tape or amplifier hiss when music is playing at normal volume.

Edit 2: after a quick search, the atmospheric brownian noise at standard conditions has about −24 dB SPL, which means that the DR of the atmosphere as a medium is about 215 dB.

 

[Blumlein 88]

Ok. I have read so much info about it. Dynamic range is related to bit depths of the ADC and occur because of narrow windows capture of real sound. But for real sound like birds whispering in the forest amidst the sound of thunder in the sky. How do you use dynamic range to describe the natural sound without concept of ADC or even analog recording medium like tapes.. Can we say the forest still has dynamic range but only unlimited? I can't find this info about whole night of googling and reading and searching the database here.

0 dbSPL (SPL is Sound Pressure Level) for humans is arbitrarily defined at 20 micropascals. Under some conditions some might hear a bit lower sound levels.

So analog or digital in nature SPL is the sound pressure. As @unpluggged has already said, 130 db SPL is the edge of pain and sound cannot be louder than 192 or 191 db SPL or it is clipped by the air. Think of the waveform. A sound at 192 db will reach zero pressure and will reach twice Atmospheric on the peak. If it is beyond this it will have clipped peaks and negative pressure or slight vacuum meaning it is distorted like a clipping amplifier. This level of sound is a pressure of a bit over 100 kilopascals.

So if you have an excellent ADC with 120 db dynamic range, how that relates to natural recorded sounds depends upon the microphone used and what amplification has done to the signal. The current quietest place on earth is -20 db SPL in a Microsoft anechoic chamber (unless someone has recently done better). Most places are going to have at least 20 db SPL of natural ambient noise.

Another interesting tidbit. Some quote the range of human hearing as what is audible at 60 db SPL. This reduces the oft quoted 20hz to 20khz range to 30-18000 hz approximately . As the sound level goes down the range is reduced.

 

[amirm]

But for real sound like birds whispering in the forest amidst the sound of thunder in the sky. How do you use dynamic range to describe the natural sound without concept of ADC or even analog recording medium like tapes..

You do this by first setting the loudest level you want to hear. Then at the other end is threshold of hearing which is around 0 to -5 dBSPL at 3 to 5 kHz. Loudest non-amplified concerts clock at 120 to 125 SPL. In that regard, to reproduce that performance with zero noise during silence, you need 120+ dB of dynamic range.

In your example, if a bird is chirping at 60 dbSPL and there is no noise floor, that would be the full dynamic range you need to reproduce it

 

[Eujene]

The dynamic range of mechanical oscillations in a medium is always limited: by the energy of the brownian motion of this medium's particles, at minimum, and by the energy of the shock wave in that medium, at maximum. For standard atmospheric conditions, this is about 191 dB SPL.

How do you convert the 191 db SPL to 0 db and the negative values below it. Is it also 0 to - 191 dB? A 32 bit ADC has 192 dB. So 32 bit ADC is the ultimate bit depth for capturing sound on earth? What would happen to 64 bit audio interface with corresponding 386 dB?

For average human hearing, the dynamic range is limited by the threshold of hearing and the threshold of pain, which range is considered to be about 130 dB.

Edit: for human hearing, this is the absolute sensitivity difference. When listening to specific audio signals, we cannot hear all this range due to masking effects. That's why we don't hear, say, background noise such as tape or amplifier hiss when music is playing at normal volume.

Edit 2: after a quick search, the atmospheric brownian noise at standard conditions has about −24 dB SPL, which means that the DR of the atmosphere as a medium is about 215 dB.

 

[amirm]

Note that you can't determine the dynamic range of recorded music. Simple example: song starts with digital silence. Then any level after that represents infinite dynamic range! (divide by zero). True dynamic range requires statistical analysis of quiet parts of music to separate noise from music -- which is hard. From research MQA has done in encoding millions of tracks, they found best case dynamic range to be about 18 bits.

 

[DVDdoug]

For music (or other program material) I like the term "dynamic contrast". But most people say "dynamic range" for both the performance/recording and the playback system/equipment.

For music, it's often measured as the difference between the peak and average or peak and RMS. (The difference between peak and RMS is the Crest Factor.) This can be misleading because our ears don't perceive short-term peaks as loudly as longer-duration loud parts. Crest factor also a simple calculations that ignores the fact that our ears are more sensitive to mid-frequencies.

There is another standard called LRA (loudness range) which tries to better approximate perception and it takes frequency content and the duration of loud and quiet parts into account.

The quiet parts can also be misleading because you can have a song that fades-out to total digital silence. Digital silence is minus infinity dB and that "calculates" to infinite dynamic range. (But once you have any non-zero value, you get quantization noise.)

Dynamic range in the reproduction chain is limited by noise. It can be acoustic or electrical noise in the recording, digital quantization noise, analog noise on a vinyl record or tape, etc. The only thing limiting the other end... The maximum loudness, is how loud your amplifier/speaker/headphones can go, or how loud you can stand it.

If you want to listen to the quietest sound like the sound of silence. Do you need high dynamic range?

No, but you need low noise. It's difficult to record quiet sounds because of preamp hiss and/or acoustic noise. A bird in the forest doesn't need a lot of dynamic range.

 

[DVDdoug]

How do you convert the 191 db SPL to 0 db

There's no calibration between digital levels and acoustic levels but there is a direct correlation. The acoustic dB SPL loudness depends on the volume control, amplifier power & gain, the efficiency of your speakers, how close you are sitting to your speakers, etc. Except movie theaters have a standard calibration and recording studios often have a defined (but not necessarily standardized) calibration between digital and acoustic levels)

But if you reduce the digital level by -6dB, the acoustic level also drops by 6dB, etc.

The digital 0dBFS digital reference is the "digital maximum". With integer formats, it's the maximum you can "count to" with a given number of bits. So digital dB levels are normally negative. A 24-bit file has bigger numbers than an 8-bit file, but everything is scaled at playback-time to match your DAC. (Floating point uses a different reference. It can go way over 0dB and for all practical purposes there is no upper or lower limit.)

The 0dB SPL reference is approximately the quietest sound that can be heard.

and the negative values below it. Is it also 0 to - 191 dB? A 32 bit ADC has 192 dB. So 32 bit ADC is the ultimate bit depth for capturing sound on earth? What would happen to 64 bit audio interface with corresponding 386 dB?]/quote]
You can only get "extreme" dynamic range on the digital side (or in the digital domain) . ADCs & DACs also have an analog side and that limits the practical/usable range.

 

[Blumlein 88]

Note that you can't determine the dynamic range of recorded music. Simple example: song starts with digital silence. Then any level after that represents infinite dynamic range! (divide by zero). True dynamic range requires statistical analysis of quiet parts of music to separate noise from music -- which is hard. From research MQA has done in encoding millions of tracks, they found best case dynamic range to be about 18 bits.

I doubt there is any recorded music with 18 bit dynamic range unless you include fully electronic music. Do you happen to know a place to read their work on this (I seem to recall seeing it or a summary). Would be nice if they listed the 10 highest dynamic range recordings they uncovered.

 

[Eujene]

For music (or other program material) I like the term "dynamic contrast". But most people say "dynamic range" for both the performance/recording and the playback system/equipment.

For music, it's often measured as the difference between the peak and average or peak and RMS. (The difference between peak and RMS is the Crest Factor.) This can be misleading because our ears don't perceive short-term peaks as loudly as longer-duration loud parts. Crest factor also a simple calculations that ignores the fact that our ears are more sensitive to mid-frequencies.

There is another standard called LRA (loudness range) which tries to better approximate perception and it takes frequency content and the duration of loud and quiet parts into account.

The quiet parts can also be misleading because you can have a song that fades-out to total digital silence. Digital silence is minus infinity dB and that "calculates" to infinite dynamic range. (But once you have any non-zero value, you get quantization noise.)

Dynamic range in the reproduction chain is limited by noise. It can be acoustic or electrical noise in the recording, digital quantization noise, analog noise on a vinyl record or tape, etc. The only thing limiting the other end... The maximum loudness, is how loud your amplifier/speaker/headphones can go, or how loud you can stand it.


No, but you need low noise. It's difficult to record quiet sounds because of preamp hiss and/or acoustic noise. A bird in the forest doesn't need a lot of dynamic range.

What do you make of the E1DA Cosmos ADC. It has huge dynamic range and low noise. Can you give example of other interface with low noise (as low as the E1DA) but poor dynamic range?

 

[Eujene]

I bought a E1DA Cosmos ADC to record the lowest nature sound in the forests (like birds singing or mate whispers). I chose it because it has very high dynamic range. SINAD of -128 dB. How is SINAD related to dynamic range anyway? That was the reason I asked if you need very high dynamic range to record the quietest sound. And Dvddoug answer was you didn't Hence. I could have chosen other gear to record the quietest sound and not the E1DA since dynamic range was not important. Now I want to know what other audio gear as example that has the same low noise as the E1DA or even lower noise but poor dynamic range. Is there such?

 

[Blumlein 88]

I think you have misunderstood DVDdoug a bit. Dynamic range is the noise level vs maximum signal of a device. Skipping some minor details about how this is measured. SINAD is the addition of noise level with THD (THD+N) vs maximum signal. A device could have a modest SINAD and still be low noise because distortion is the main component and the noise part is very low. Or it could have modest SINAD because of fairly large amounts of noise.

Now very low SINADs -120 db or so means both distortion and noise are very low otherwise that number could not be achieved. The E1DA certainly has very low SINAD and very low noise. Now there are devices which have maybe -96 db SINAD, but the noise part of SINAD is very low, much lower than -96 db. A dynamic range measurement would show this. So SINAD is only partly related to dynamic range.

The issue with recording is the self noise and sensitivity of the microphone. Typically specs will either include self noise or Signal to Noise ratio. These are a little different than such measures for electronic gear. Self noise is the equivalent SPL in noise. Something rated with self noise of 10 db SPL generates noise equal to a sound level of 10 db SPL even in a quiet anechoic chamber. Very few places have ambient noise low enough for self noise of 10 db SPL to be a problem. Large condenser mics have generally lower self noise than other mics. There are a couple that are rated around 0 db SPL self noise A-wtd. (Lewitt makes one I'm aware of).

Sensitivity is also important. You will have to amplify the mic signal with the mic preamp. When you have to turn up the gain, you also are amplifying the noise level. So a sensitive mic will need less gain leaving less noise in the signal. Large condenser microphones also tend to be more sensitive. Typically the microphone will be the bottleneck that limits possible dynamic range in a recording.

 

[Eujene]

I think you have misunderstood DVDdoug a bit. Dynamic range is the noise level vs maximum signal of a device. Skipping some minor details about how this is measured. SINAD is the addition of noise level with THD (THD+N) vs maximum signal. A device could have a modest SINAD and still be low noise because distortion is the main component and the noise part is very low. Or it could have modest SINAD because of fairly large amounts of noise.

Now very low SINADs -120 db or so means both distortion and noise are very low otherwise that number could not be achieved. The E1DA certainly has very low SINAD and very low noise. Now there are devices which have maybe -96 db SINAD, but the noise part of SINAD is very low, much lower than -96 db. A dynamic range measurement would show this. So SINAD is only partly related to dynamic range.

The issue with recording is the self noise and sensitivity of the microphone. Typically specs will either include self noise or Signal to Noise ratio. These are a little different than such measures for electronic gear. Self noise is the equivalent SPL in noise. Something rated with self noise of 10 db SPL generates noise equal to a sound level of 10 db SPL even in a quiet anechoic chamber. Very few places have ambient noise low enough for self noise of 10 db SPL to be a problem. Large condenser mics have generally lower self noise than other mics. There are a couple that are rated around 0 db SPL self noise A-wtd. (Lewitt makes one I'm aware of).

Sensitivity is also important. You will have to amplify the mic signal with the mic preamp. When you have to turn up the gain, you also are amplifying the noise level. So a sensitive mic will need less gain leaving less noise in the signal. Large condenser microphones also tend to be more sensitive. Typically the microphone will be the bottleneck that limits possible dynamic range in a recording.

Thanks for the explanations. Maybe the reason why low noise gear also has high dynamic range is because since it is expensive to make low noise, may as well raise the dynamic range by having the maximum volume raised before clipping. Correct?

What chip dictate how high the volume can rise before clipping, is it the power amplifier chip? If so, an audio interface can claim to have dynamic range of -130 dB by increasing the maximum volume before it clips even if the noise floor is not low, is this correct? What specific gear does this as example? I think any power amplifier to loudspeaker even cheap brands with hundreds of watts can claim to have dynamic range of even -200dB with very poor noise floor?

 

[Blumlein 88]

Thanks for the explanations. Maybe the reason why low noise gear also has high dynamic range is because since it is expensive to make low noise, may as well raise the dynamic range by having the maximum volume raised before clipping. Correct?

What chip dictate how high the volume can rise before clipping, is it the power amplifier chip? If so, an audio interface can claim to have dynamic range of -130 dB by increasing the maximum volume before it clips even if the noise floor is not low, is this correct? What specific gear does this as example? I think any power amplifier to loudspeaker even cheap brands with hundreds of watts can claim to have dynamic range of even -200dB with very poor noise floor?

No this is not correct. I'll try and put forth some explanations later or link to some informative articles. As to what chip causes this etc., devil is in the details. Good gear can be made with chips, discrete components, even vacuum tubes. Comes down to how well it is designed.

 

[Blumlein 88]

Let me try this approach. Some fine details are not quite accurate, but the overall ideas are.

Let us assume we have a device with a power supply that puts out +10 volts and - 10 volts. I can send it a signal that is a sine wave varying between plus and minus 5 volts. It will be good it can be low distortion. I can even send a signal at or near plus and minus 10 volts. Distortion probably goes up, but might not be horrid depending upon design. If I send a signal which requires 12 volts plus and minus it will hit the rails on peaks and can do no more than 10 volts. The rest is clipped off. Horrid distortion. Now this device with no signal has some noise. The dynamic range is the ratio between that noise and max signal. db always is a ratio. There are limits to how low one can make noise. Yes I could go to plus and minus 20 volt supplies, but noise would go up too.

The physical limits of noise versus signal is maybe 130 db, and not many devices can manage that. Just pumping up voltage and power cannot increase dynamic range beyond those physical limits. The physical limits are related to thermal noise in resistors in devices. So no I cannot make a huge amp with 200 db of dynamic range. Even if I did somehow, the signal is coming from devices that cannot have that much dynamic range so I would be amplifying noise in the signal.

Does this make any sense?

In answer to your question, recording nature sounds, the limit is usually ambient noise, followed by microphone noise and/or microphone preamp noise.

 

[dlaloum]

A relevant but a little tangential comment...

In the digital world, the noise floor is absolute - there is not signal below the noise floor - hence dynamic range is noise floor to max....

In the analogue world, there can (and usually is!) a noise floor well above the bottom of the resolvable audio space - the dynamic range is in fact greater than the max to noise floor range.

You can have tape noise, or vinyl roar, which is at a signal level, above the perceptible quietest audible signal.

Naturally this is of limited relevance when we are all using digital devices - however it may still have relevance when discussing analogue parts of the audio chain - power amps, buffers, analogue preamp circuits - these can add their noise to the signal, and the signal can be percieved below the noise level. (there's noise... and then there's noise...)

With digital, the "noise" level is often conflated with the low end limit of digital resolution... I'm not sure that it is in fact the right term for it.

 

[Deleted member 48726]

This is a nice explanation-->
DR, CF AND POWER

 

[danadam]

In the digital world, the noise floor is absolute - there is not signal below the noise floor

With dither there definitely is.