How much dynamic range is too much

[Miguelón]

Minnesota, the record label is BIS. The Beethoven Symphony cycle of Osmo Vanska/ Minnesota Orchestra is only available as SACDs.

sorry, I remember it was a double consonant elsewhere, but leaving an 's' to put an 'n' preserve the number of bits

I will not try to write Mississippi without before reading Mark Twain, hopefully I did.

I try to listen other recoedings from the conductor, this one is really forced to histrionism IMHO

EDITED: aha, Sibelius, this is another kind of beast. Let's try, first minutes and hadn't a heart attack, dynamic is under control

 

[rdenney]

That's because you're listening to FM. That always happens on FM, at least these days.

Yeah, since about 1979.

Rick "but it's fine for the car and for music in the background of other activities" Denney

 

[krabapple]

I suppose is both the conductor and the edition process, never listened Karajan in live concert but had the reputation of a very special sound.

Nothing against, he was one of the greats, and revitalized a lot the classical orchestra for the big public.

if you have some time to listen to Vikingur Olafsson (this time under Deutsche Gramophon) I'm curious to know more about his technics of recordings, I have the impression that places the mic inside the piano.

very special pianist, on the oppsite side of actual performances that exhibit a machine-gun like ultra fast version more on the athletic side.

I'm thinking about Yuja Wang and Lang Lang, more concretely, the public is more attracted by their speed than the expression

Arcady Volodos is even faster, but it belongs to other generation

This all seems to be wandering rather far from the question of when dynamic range is 'too much'.

 

[Miguelón]

This all seems to be wandering rather far from the question of when dynamic range is 'too much'.

Yes, you're right, I'm very enthusiastic to classical and I got so far, but for that I can open a thread about the recordings in classical interpretes

 

[ZestClub]

I have a 16Kw Hi-end PA system, the pro-ribbons can go to 130db.
Its in a large studio room - a normal hi-fi would sound like a tin can in the corner in comparison.
But I only use it once a month on average otherwise its likely to cause hearing damage.

I have a set of test tracks used by audiophiles that range from classical through to 32 bit electronic music.
There seems to be few really good recordings that can take advantage of a system like that.

 

[Newman]

I have measured peaks from my seat in a large performing ensemble, sitting in front of the timpani, at 108 dB SPL A-weighted. Assuming the above scales, the quiet bits would be at about 40 or 45 dB SPLA, which seems to me about right and validates the range in my mind.

I doubt that such ranges are ever recorded as such, which is, in my view, why recordings never sound quite realistic.

Single-sample peaks are of no interest in the quest for realistic sound. The intervals over which sampling is averaged would need to match human perception, and no faster.

 

[Miguelón]

Single-sample peaks are of no interest in the quest for realistic sound. The intervals over which sampling is averaged would need to match human perception, and no faster.

According to a mic company, they recorded 130 dB peaks (I suppose audible) inside a grand concert piano, obviously none of the musicians will listen that because it was directly over the strings.

Before I mentioned Vikingur Olafsson particular recordings (studio), the mechanical noises are very audible and in the Organ Sonata BWV 528 II movement arrangement recording gradual crescendo goes pretty loud.

Personally I like some hyper close recordings as I can remember when I played myself with the front of the piano leaved.

Glenn Gould recorded his latest version of Goldberg Variations with mics placed in face of the nude piano, but it said it liked the naked sound because the precision of the feedback, sound traveling from strings to head with reflections can cause some delay specially in grands.

Another extraordinary recording in dynamics is Sviatoslav Richter's Pictures at an Exibition (Mussorsky ), I don't know if is a myth or not but he broke 3 strings of the piano during the concert. The quietest parts of Promeade variations and "cum mortuus in lingua morta" passages are really low also.

 

[rdenney]

Single-sample peaks are of no interest in the quest for realistic sound. The intervals over which sampling is averaged would need to match human perception, and no faster.

Agreed in general, and the 100ms sample in the paper I cited is perhaps the shortest tone burst that a person can hear without needing to be amplified above its current level. But the histogram I cited isn't an observation of single-sample peaks, but rather amplitudes aggregated into 100ms bins from continuous music. The usual graph I recall from Alton Everest's Master Handbook of Acoustics reported the shortest tone burst (presumably in a field of relative silence--a single sample) that can be heard without needing to be amplified, meaning that shorter standalone tone bursts will not be perceived as being as loud. I'm sure there are all sorts of qualifications and limitations on that measure, but I don't recall them and I don't have the book handy.

In this picture of snare-drum waveforms...

(source: https://www.soundsandwords.io/drum-sound-classification/)

...initial impulses seem to be quite short--as little as 20 or 30 ms--plus a much longer reverberant tail. If that impulse clips, the waveform for the hit won't be as tall with respect to the reverberant tail, and that seems to me to affect the sound significantly, despite that the peak is a much shorter duration than 100ms. The difference between the hit and the reverberant tail in the third column, for example, is greater than how much the hit would need to be amplified to be perceived as being as loud.

Note that snare drum hits are one example of short peaks that are distinct from the wall of sound--potentially a single sample that does stand alone--and particularly in orchestral music, but also in quite a lot of jazz and old minimally processed rock.

(The spectral waterfalls don't seem to show the difference in amplitude very well, and sometimes I wonder if we think spectrally at the expense of looking at amplitude variations. The color variations that are supposed to reveal dynamics are not nearly as precise an indicator of amplitude as the waveform itself.)

In any case, any difference in the recording or playback and the initial signal is a distortion, and the question is how much of that dynamic distortion is needed to affect the perceived sound. So, the required dynamic range may be greater than the 75 dB range measured in the first paper I cited, because the peaks that affect it may be shorter than the 100ms bins into which they aggregated their samples. I think of it as a floor, because clearly the time period of interest isn't any longer than 100 ms. And, again, this is validated by the minimum dynamic range (as usually measured with audio--broadbanded) that I perceive as sounding "realistic".

Rick "thinking that clipping/compression happens somewhere in the recording/mastering/distribution/playback process much more often than we suspect" Denney