Is it possible to get 120 db dynamic range from recording to listening room?

[Blumlein 88]

We really need big-league recording and mixing and mastering engineers for a valid discussion on this. What many of us imagine goes on in state of the art recording and mixing and mastering may be pretty close to fantasy. We don’t know what we don’t know. Just looking at it from the back end (reproduction in the home) or speculation as to what is or conjecture as to what could be done based on very isolated consideration of what a mic is capable of and what software is capable of is not going to result in a realistic or informed discussion. Nor is simply comparing what goes on at a live performance as compared to what a mic is capable of and a speaker is capable of. Taking for granted what goes on in the middle from mic techniques to pro recording media and equipment to mixing and mastering can result in a largely fictional narrative and set of assumptions, aka, wild speculation, IMHO.

You can't record quieter than the microphones. So this isn't wild speculation. It is intended to put some outer limits on whether these really wide ranges are possible. I think the example of noise on the Skywalker Scoring stage is likely an example of the quieter end of what is usually the case. You'll also notice I specifically opted out of thinking about studio only creations or movies. Those can be created over any range desired.

Now big league recording or mixing people could certainly add to the discussion, but I don't think it is value-less without them. What usually goes on between recording and release of the material is compression, reverb, EQ at a minimum. None of those are going to expand the dynamic range you can get at home.

 

[amirm]

I'll add a final bit about recordings. Do we have any that reach even 100 db range within them?

This is a difficult question to answer.

We can certainly look at this trivially and arrive with a resounding YES. Take a high-res file that hits peaks of 0 dbFS. The compare that to digital silence in a 24-bit file. Now we have 144 dB of dynamic range.

That is not a correct method of course since we want the dynamic range for when nothing is playing but with the recording chain active. There, the lowest level is noise. Problem is, you can't distinguish noise from music mixed with noise.

The answer to this is to perform a statistical analysis and attempt to determine if the quiet parts of the recording are truly random or have music in them.

Once there, we need to take the spectrum and perform the analysis I have shown relative to threshold of hearing.

I don't know if any readily available tools for this. I have seen the work described in a couple of papers (one from Bob Stuart). I have not seen the outcome of such analysis with specific music files.

 

[Blumlein 88]

Even if 120 db were possible I am not sure it would sound good. See link for the perspective of a mastering engineer https://forums.stevehoffman.tv/thre...n-limiting-during-recording-mastering.631903/

I don't disagree with what Hoffman says here, but that is a different question. Usually modest tasteful compression does improve upon no compression in most people's opinion. If you use that however, you've reduced dymamic range of the music. And too much sounds bad.

 

[Blumlein 88]

This of course ignores the noise in the listening room. My hunch is that investing in a quieter room is probably the best strategy for the consumer. So we have sound damping double glazing, brick walls, muffled ventilation openings and even those not at the side of the house facing the road, ultra quiet mechanical ventilation (thus far we do not need airconditioning), extra quiet water pipes etc. All of this is cheap and easy with a new house but quite impractical for an existing one, of course.

I don't disagree. But I've noted elsewhere and confirmed in a few locations that at least in quieter hours many residential areas have noise not more than 10 db SPL in the 3-5 khz range.

I also posted files at one time that had intermittent levels of noise added at different levels, and asked people to listen at a normal volume and say where the noise wasn't heard anymore. With the idea most people listen at average levels of 75 db or so. The most common response was the noise disappeared at -70 db. Using headphones could change the equation of course.

 

[amirm]

This of course ignores the noise in the listening room.

It doesn't. Same paper I have been quoting analyzes both. I wrote an article for WSR magazine a few years back based on this research. See: https://www.audiosciencereview.com/forum/index.php?threads/dynamic-range-how-quiet-is-quiet.14/

As you see this in this survey, the best rooms were truly silent relative to our threshold of hearing ("min" graph). Even the average is not too bad with 5 to 7 dBSPL of noise in critical middle-frequencies.

SPL meters "lie" in this regard because they are dumb and report the loudest component. In the above graph, we can see even the quietest room has about 30 dB of noise at 70 Hz. People read that and say the minimum noise floor is then 30 dB which is not correct. That noise is inaudible.

Here is a case where your ears can do better than a dumb SPL meter. If you hear nothing in your (dedicated) listening room, then whatever noise there is, is inaudible by definition.

Bass travels through many barriers and hence the reason we always have a lot of it in our rooms. But fortunately our threshold of hearing is very high in that region so it is not an impact.

 

[Blumlein 88]

This is a difficult question to answer.

We can certainly look at this trivially and arrive with a resounding YES. Take a high-res file that hits peaks of 0 dbFS. The compare that to digital silence in a 24-bit file. Now we have 144 dB of dynamic range.

That is not a correct method of course since we want the dynamic range for when nothing is playing but with the recording chain active. There, the lowest level is noise. Problem is, you can't distinguish noise from music mixed with noise.

The answer to this is to perform a statistical analysis and attempt to determine if the quiet parts of the recording are truly random or have music in them.

Once there, we need to take the spectrum and perform the analysis I have shown relative to threshold of hearing.

I don't know if any readily available tools for this. I have seen the work described in a couple of papers (one from Bob Stuart). I have not seen the outcome of such analysis with specific music files.

Here was my simple minded approach. This is from the beginning of a 2L recording. This is the db waveform scale so low levels are more obvious. This ramps up from being inaudible the first 2 seconds to you hearing the hall finally reaching a consistent level in the area highlighted.

There is rising ultrasonic noise from 60 khz and upward. Recorded in DSD perhaps. So I then filtered out everything above 20 khz, and the RMS level of that short segment of hall and other noise is -59 dbFS. Here is the spectrum of it.

Here it is in a linear scale. Fairly like brownian noise.

For comparison here is -59 dbFS of brownian noise.

 

[RayDunzl]

Room noise last night, captured by a UMIK-1 at the listening position.

Floor fan (on a rheostat) about 25 feet away, and washing machine at 36 feet (using a Bosch GLM80 lase distance meter), along with the PC fans and whatever electrical noise polluting the measure.

Red - 1/48 Octave RTA
Green - Spectrum unsmoothed
Blue - Spectrum 1/48 Octave smoothing

 

[Hugo9000]

You can't record quieter than the microphones. So this isn't wild speculation. It is intended to put some outer limits on whether these really wide ranges are possible. I think the example of noise on the Skywalker Scoring stage is likely an example of the quieter end of what is usually the case. You'll also notice I specifically opted out of thinking about studio only creations or movies. Those can be created over any range desired.

Now big league recording or mixing people could certainly add to the discussion, but I don't think it is value-less without them. What usually goes on between recording and release of the material is compression, reverb, EQ at a minimum. None of those are going to expand the dynamic range you can get at home.

This is an astonishing orchestral recording, made in a real concert venue, De Doelen Hall in Rotterdam. Producer Robert Suff, Engineer Thore Brinkmann. This is probably the most impressive recording in my collection of approximately 3000 CDs and SACDs, nearly all classical (operas, symphonies, piano solo, all manner of chamber works, masses, etc., from late Renaissance to Neo-/Post-Romantic). It's not just impressive technically, I also feel it's the best performance of this work I've ever heard (technical excellence is useless to me without musical merit as well, I don't buy things just to have an "audiophile spectacular" lol).

Maurice Ravel
Daphnis et Chloé (complete ballet music)
Yannick Nézet-Séguin/Rotterdam Philharmonic Orchestra
BIS catalogue No. BIS-1850

Original recording format 24/96, released as a Hybrid SACD (Stereo and Surround SACD, and Stereo Redbook CD layer). It's also available as a 24/96 download.

Regarding their recording equipment, I quote from the liner notes for this disc:

BIS's recording teams use microphones from Neumann and Schoeps, audio electronics from RME, Lake People and DirectOut, MADI optical cabling technology, monitoring equipment from B&W, Stax and Sennheiser, and Sequoia and Pyramix digital audio workstations.

The music rises from "hall silence" at around the 3 second mark. With my closed-back headphones (KEF Space One wired version, 32 ohm), JDS Labs Atom on high gain and max volume, I can hear a small amount of hall noise and other sounds. With my open-back Sennheiser HD 58X, I cannot really distinguish that hall noise from my very quiet room (edited to note that I also can't achieve the same SPL, as the HD 58X is 150 ohm vs 32 ohm for the KEF headphones). From there, the recording reaches full peaks at the climaxes (I use low gain and a more reasonable volume when listening to the full recording haha! as I value my hearing).

BIS founder and owner Robert von Bahr comments and answers questions frequently on forums regarding his company's recordings, here is his reply to someone regarding this Ravel recording:

Post by bissie April 8, 2015 (4 of 29)

​

current93 said:​

​

Hi! Please can you comment on sound quality? Previous SACD (Strauss Ein Heldenleben) from these sources was terrible, with very narrow soundstage and low recording level. What about Daphnis?​

Whereas I will not get drawn into a discussion about the quality of the sound, I react as vehemently as ever to the caption "low recording level".

No Sir or Madame, IT IS NOT!!!!! However, it IS - like all SACD:s from my company, a true and unaltered version of what the composer, the conductor and the orchestra actually did produce on stage. Unlike almost all other companies, we don't think we're God and sit and overrule what the composer and musicians do by more or less heavyhandedly compress the music to fit the needs of someone that wants to listen to the music while showering or whatever.

I feel it is most irking that, rather than being hailed to the sky that we present the truth of what was going on, we're over and over again being accused of recording at too low a level.

Our top level is invariably 0 dB. That's the highest level allowed. All the rest is what it was. Other companies take down the loudest parts and up the softest, so that you can listen in the car, in the aforementioned shower or whatever. That seems louder - and it is - on an average. But it isn't the truth.

Now that we have a system that allows for this huge dynamic, you still want us to destroy what the musicians were doing, so as to get the soft parts loud enough for a noisy listening environment?

Sorry, but not BIS, not ever before and not in the future. We do our music for people that really can appreciate what the musicians rather than the balancing engineer did. Having said that, we employ some of the best recording engineers and producers in the business.

Robert

He understandably rejects the complaints on forums and in the occasional music review regarding supposed "low level" recordings as people often call them, that in reality take advantage of as much of the dynamic range as the equipment and processes and formats will allow.

Perhaps I'll try contacting him to see if he would be interested in joining this forum.

 

[Ron Texas]

There is noise in real life. Many months of AC for us southerners, fans, dreaded gas powered leaf blowers, and whatever. One morning this week an 18 wheeler was idling in front of my driveway waiting to drop off a load of bricks. That thing was loud.

 

[Ron Party]

While my memory isn't what it used to be, I seem to remember Arny K. posting that the highest DR he ever measured was around 86. It was a BIS SACD.

 

[Daverz]

Very few people have a quiet enough room to enjoy those super wide dynamic range recordings. I usually end up riding the volume control.

 

[RayDunzl]

There is noise in real life. Many months of AC for us southerners, fans, dreaded gas powered leaf blowers, and whatever.

Spectrum of Noise floor again...

5:36pm on a Saturday afternoon with nothing noticeable going on in the house (black), vs 3:43am with floor fan and washing machine (red)...

And (green), peaks of Steely Dan Gaucho, playing at 70~75dB average SPL.

Music - unweighted (Z) - 6:00pm

Ambient Noise - A and Z-weighted at 6:00pm

Maybe I hear a little distant traffic.

 

[Daverz]

I also live in a second floor condo, so I can't open up the volume on the Shostakovich and Bruckner as much as I might like to unless the neighbors are at work. Luckily pretty thick walls, though.

 

[Ron Texas]

@RayDunzl it could be worse, at least you are in Tampa. Maybe I should play some Steely Dan to get in the Saturday evening mood.

 

[JJB70]

I do think this thread indicates that judicious and careful use of compression is useful.

If recording without compromise so as to maximise DR results in having to listen to music with a hand on the volume control and puts demands on audio systems that few can properly fulfil then it becomes a bit of an empty result. Yes, it is probably a wonderful technical achievement in isolation but from the perspective of most users the result will be sub-optimal.

Compression has gotten a bad name from the excesses we've seen in rock and pop and some of the dire re-masters we have experienced but if used sensitively it has its place.

And that is before even considering that if most music nowadays is played on small BT and multi-room speakers or car audio then inevitably that's what will set the demands for recording. What I do wonder is whether any recording companies would do parallel masters, a compressed one for the mainstream market and an alternative one aimed at people who want a wider dynamic range version?

 

[RayDunzl]

it could be worse, at least you are in Tampa.

It's getting worse...

Where are you?

(visited Corsicana over the 4th of July)

High point was a skunk (possibly poisoned) dying in the garage.

 

[Blumlein 88]

This is an astonishing orchestral recording, made in a real concert venue, De Doelen Hall in Rotterdam. Producer Robert Suff, Engineer Thore Brinkmann. This is probably the most impressive recording in my collection of approximately 3000 CDs and SACDs, nearly all classical (operas, symphonies, piano solo, all manner of chamber works, masses, etc., from late Renaissance to Neo-/Post-Romantic). It's not just impressive technically, I also feel it's the best performance of this work I've ever heard (technical excellence is useless to me without musical merit as well, I don't buy things just to have an "audiophile spectacular" lol).

View attachment 30186

View attachment 30187
Original recording format 24/96, released as a Hybrid SACD (Stereo and Surround SACD, and Stereo Redbook CD layer). It's also available as a 24/96 download.

Regarding their recording equipment, I quote from the liner notes for this disc:


The music rises from "hall silence" at around the 3 second mark. With my closed-back headphones (KEF Space One wired version, 32 ohm), JDS Labs Atom on high gain and max volume, I can hear a small amount of hall noise and other sounds. With my open-back Sennheiser HD 58X, I cannot really distinguish that hall noise from my very quiet room (edited to note that I also can't achieve the same SPL, as the HD 58X is 150 ohm vs 32 ohm for the KEF headphones). From there, the recording reaches full peaks at the climaxes (I use low gain and a more reasonable volume when listening to the full recording haha! as I value my hearing).

BIS founder and owner Robert von Bahr comments and answers questions frequently on forums regarding his company's recordings, here is his reply to someone regarding this Ravel recording:

He understandably rejects the complaints on forums and in the occasional music review regarding supposed "low level" recordings as people often call them, that in reality take advantage of as much of the dynamic range as the equipment and processes and formats will allow.

Perhaps I'll try contacting him to see if he would be interested in joining this forum.

Well I certainly like Ravel and particularly Daphnis et Chloe. So I downloaded it after listening to some of the streaming samples. I don't have a way to rip SACD so the 96/24 being available in stereo and surround made it attractive to download. When you download you can get both plus 16 bit and MP3.

Thank you for bringing this up. I've only listened to a couple tracks and it is an excellent recording and performance.

In the silent portions gave it a quick cursory look, the noise floor is -65 to -68 db. This is enough at my normal listening levels to be more or less silent, and the 2L being just slightly louder isn't. I don't doubt they may have gotten better noise in some of their other recordings, and you'll not hear this and think "oh this is a bit noisy". BTW, the noise profile of this was also brownian. If I used a 2nd order roll off at 50 hz, the remaining signal was about 10-11 db quieter.

Also how should we account for our uneven thresholds? This BIS recording is below the 95 dbFS level in that 3-5 khz band we are most sensitive using an ERB curve. The 2L recording is in the 80's. This is probably more or less a better way to judge perceived dynamic range of the recording. Probably an area worthy of more discussion in this thread.

Neumann and Schoeps are some of the finest microphones, but they aren't the quietest in terms of self noise.

 

[StevenEleven]

Well I certainly like Ravel and particularly Daphnis et Chloe. So I downloaded it after listening to some of the streaming samples. I don't have a way to rip SACD so the 96/24 being available in stereo and surround made it attractive to download. When you download you can get both plus 16 bit and MP3.

Thank you for bringing this up. I've only listened to a couple tracks and it is an excellent recording and performance.

In the silent portions gave it a quick cursory look, the noise floor is -65 to -68 db. This is enough at my normal listening levels to be more or less silent, and the 2L being just slightly louder isn't. I don't doubt they may have gotten better noise in some of their other recordings, and you'll not hear this and think "oh this is a bit noisy". BTW, the noise profile of this was also brownian. If I used a 2nd order roll off at 50 hz, the remaining signal was about 10-11 db quieter.

Also how should we account for our uneven thresholds? This BIS recording is below the 95 dbFS level in that 3-5 khz band we are most sensitive using an ERB curve. The 2L recording is in the 80's. This is probably more or less a better way to judge perceived dynamic range of the recording. Probably an area worthy of more discussion in this thread.

Neumann and Schoeps are some of the finest microphones, but they aren't the quietest in terms of self noise.

I did something much less rigorous. I pulled up the Ravel / Yannick recording on Spotify and turned the receiver up until I was pretty sure I heard the ambient noise floor at the beginning. It was definitely a higher position on the receiver than normal. I've tried to figure out my room noise in the past and came up with something like 32 dB, A weighted. I then set my dB meter to C weighted, fast peak, at my listening position. Peaks came out at 84 dB a few times, no more. I'm a lot less techy than you guys. I'd guess those are about where the recording was hitting the highest peaks they would allow.

I will say it was a great recording (even on my modest system streaming through Spotify) and I actually found it to be a very comfortable listen in terms of hearing the softest and loudest parts of the music without hitting volume control.

Added it to my Spotify favorites.

 

[Ron Texas]

It's getting worse...

Where are you?

(visited Corsicana over the 4th of July)

High point was a skunk (possibly poisoned) dying in the garage.

Houston, close in where there is lots of noise.

 

[RayDunzl]

I then set my dB meter to C weighted, fast peak, at my listening position. Peaks came out at 84 dB a few times, no more.

Using REW and a UMIK-1, music playing, with Fast C weighting, you can see here a difference of about 16dB between the C-fast (LCFmax) reported maximum and the unweighted "instantaneous (LZpeak) with no time averaging.