I. The total dynamic range

The total dynamic range available for recording a virtual orchestra is the difference from the maximum, loudest tutti chord where all notes trigger samples with velocities of 127, to the minimum loudness of a single note played with a velocity of 1.


01. Set the track fader
Set all faders with sample playback software to -6.0 dB.

02. Set the output level of the samples
Set the output level of the patches within the sample player also to -6 dB. This is the default value with most patches anyhow.

Now we have a headroom of 12 decibel.

03. Absolute maximum loudness - staccato
Play a tutti chord where all instruments play a staccato ff sample with a velocity of 127. This will produce the maximum loudness possible. The master fader will not clip, but may raise to 0.01 dB Peak.

04. Absolute maximum loudness - sustain
Play a tutti chord with sustaned notes, all ff samples and with a velocity of 127. This sustained chord will show us the maximum level of this instrument combination. The maximum loudness will be circa 12.5 dB RMS when the chord reaches nearly 0 dB Peak.

05. The minimum loudness of the dynamic range
Play a timpani pp single note at velocity of 1. This will produce a level of circa -48 dBFS Peak. This level will be the absolute minimum of our orchestral dynamic range.

.

II. Level, Headroom and Dynamics

06. Set sample patches to a default volume output of -6 dBFS

07. Set virtual track faders at a default value of -6 dBFS

The level reduction to -6 dB with the fader, plus the level reduction to -6 dB withhin the player software, ads up to a total of 12 dB level reduction. This -12 dB is our headroom, and enough reserve to play large ff tutti.

.

III. The measured loudness of an orchestra recording

The decibel values are from a recording who has dynamics from ppp (as soft as possible) to fff (as loud as possible):

ff = -12.5 dB RMS (as loud as possible)*
mf = -18.0 dB RMS (normal loudness, not loud)
mp = -20.0 dB RMS (normal loudness, not soft)
pp = -27.0 dB RMS (soft)
ppp = -48.0 dBFS Peak (as soft as possible)


All decibel values are dB RMS as AES-17 standard.

.

IV. Dynamic Indications - Velocity to dBFS

This PDF visualizes the MIDI velocity and the produced decibel of the four layer string patch VI-14_mV_sus_p-ff. The produced maximum peak is -7.5 dB in the ff layer, and the minimum is -36.7 dB with the pp layer.

download link:
http://vsl.co.at/upload/users/57/Dynamic_Indications_Velocity_to_dBFS.pdf


.

V. Keyswitch_4 Layers_Velocity to dBFS

This PDF shows the same dynamic range analysis of the b]VI-14_mV_sus_p-ff[/b] four layer string patch as in the previous post, but the graph is with lines.

download link:
http://vsl.co.at/upload/users/57/Keyswitch_4_Layers_Velocity_to_dBFS.pdf

Legend:
Red line = ff layer
Yellow line = f layer
Green line = mf layer
Turquoise line = p layer

Every colored line represents the full MIDI velocity range of 1-127 of the particular layer, and the corresponding level in dbFS Peak is indicated in the scale on the left in dark blue letters. As you can see, there is a little level gap between the layers.

The blue line with the largest dynamics (-47dB to 0dB), is a sinus test tone played with MIDI velocities from 1 to 127, and represents the largest possible dynamic range playable with MIDI velocities.

.

VI. The maximum possible dynamic range with MIDI velocity

The maximum dynamic range possible using MIDI velocities from 1 to 127 is 47.0 dB

.

Reference Point

The stereo master fader, we bounce/print the mix thru, stays at 0.0 dB. This master fader is our reference.

.

- For visually controlling the overall loudness at the stereo master, we open a meter. For wide range dynamic orchestra material preferably a VU meter where zero is at -20 dB RMS.

.

After calibrating the virtual mixer in the way described above, the stereo sum will not clip at the master fader, You can add several more instruments and notes without driving into clipping. It seems, that a headroom of 12 dB is an ideal value for working with orchestral music scores.

The obvious reason that all orchestra tracks playing simultaneously does not clip the stereo master is, that you calibrated with the loudest tutti. The not so obvious reason is, that -12 dB headroom is an ideal value. It would need a lot more notes playing velocity of 127 in order to drive the stereo master into 0 dBFS+ clipping. Another reason is the rudiculous level uniformity of VSL thru out the whole library, this uniformity prevents from being surprised by an unexpected higher level.

When adding more and more instruments and/or notes, then at a certain point the headroom is used up, and the audio will clip. That is the moment we have to set a larger headroom. The clipping will occour at the stereo master, and you simply give more headroom by pulling all track faders back by a small amount. Often a reduction of 0.1 dB on all faders to -6.1 dB is enough, and the stereo master fader will not clip again thru the whole composition.

.

I did a quite similar approach to my template. With some deviations.
As I work in Samplitude which relies on a 32 bit floating point mixing engine I don't have to care for clipping *before* the master fader. If the D/A-output clips I just lower the master fader. This is an approach which doesn't work on ProTools, for example. Furthermore it is just an approximation which needs to be adjusted for every individual piece.
In my template I was more concerned about the relative ff levels. So trumpets double loudness than horns. horns double loudness than strings and woodwinds. But I didn't set these levels by numbers but by ear.
On the other hand, since all instruments can play more or less equally soft the midi programming doesn't translate automatically between the instruments. I think about applying input filters to the individual instruments so instrument programming can be moved around freely without adaptation.
Angelo, you don't mention relative levels. How did you set these up?

@mathis said:

In my template I was more concerned about the relative ff levels. So trumpets double loudness than horns. horns double loudness than strings and woodwinds. But I didn't set these levels by numbers but by ear.
On the other hand, since all instruments can play more or less equally soft the midi programming doesn't translate automatically between the instruments. I think about applying input filters to the individual instruments so instrument programming can be moved around freely without adaptation.
Angelo, you don't mention relative levels. How did you set these up?

Yes, this is the big question for me, too. I'm mostly concerned with relative levels - basically, in the sense of making 4 Horns fff distinctly mask out 1 solo violin fff...

Great work, though, Angelo! Thanks for sharing.

J.

@hermitage59 said:

I think Angelo's intent (and considerable reflection and work in this) is to provide a 'baseline' to start from, and give those who wish to build a template somewhere to begin. (Yes Angelo? Was this the idea?)

Yes, that what it is, a baseline, something you setup, and who makes work easier, and better controllable.

The most important thing is that you never change the playback volume, but once the SPL's are calibrated leave the volume knob at the marked 85 dB SPL.

.

The creation of a calibrated environment will give you several advantages:

You will have a loudness reference.

You will know at any time where you are with the dynamics, visually as well aurally.

You can work within the corner data of a particular delivery standard, for example, you can mix a broadcast soundtrack to -9 dB peak maximum by simply observing that the master fader shows maximum -9 dB peak.

.

@jbm said:

Yes, this is the big question for me, too. I'm mostly concerned with relative levels - basically, in the sense of making 4 Horns fff distinctly mask out 1 solo violin fff...

The position of the solo violonist in front of the orchestra, as opposed to the four horns sitting on the back of the ensemble will solve this problem

.

About templates: I don't see how templates with presets of relative levels between instrument groups could be of practical use. I don't know what instrument balance the next composition is callling for, and I never had twice the same instrumentation.

.

Loudness and the available dynamic range

Every composition has a maximum peak. Most often this peak is a tutti chord. A smaller ensemble, for example a chamber orchestra, reaches the maximum loudness with less instruments then a large orchestra. A quartor has a complete other loudness structure then a chamber ensemble or a large philharmonic orchestra. But, no matter how large or small the ensemble is, one thing have all incommon, respectively stays the same, this is the total available dynamic range on the final media.

Calibrating the maximum peak with a "ff tutti chord"

The idea of the ff tutti chord calibration is based on the fact, that you the composer, knows most often in advance where the loudest point is in a composition, and therefor you can set the needed headroom in advance.

The maximum peak in a recording session in a concert hall, is only in so far different, that the recording engineer can check the required headroom during rehearsal, and adjust the recording level and having enough headroom, so it doesn't clip when he records a final master.

A "ff tutti calibration chord" for a classical period composition could be as follow:

1 Piccolo note
2 Flute notes
1 Oboe note
1 Englishhorn note
1 Clarinet in Bb note
1 Bass Clarinet note
1 Bassoon note
1 Contrabassoon note
1 Trumpet note
1 Horn note
1 Pianoforte chord
1 Timpani note
1 Gran Cassa note
1 Piatti note
1 Vln I note
1 Vln II note
1 Vlc note
1 Vlc note
1 CB note

Now play back the full chord with ff samples with as many notes per instrument you think it will have at the end, and all at velocity 127, this will produce the signal with the maximum peak level. This signal you use to calibrate your VI or Kontakt track fadersto an uniform level, for example the suggested -6 dB, or more if the stereo master still clips.

.

Here am example of a "ff tutti calibration chord" preliminary made to set the headroom of a composition.

The chords are voicings as they are in the composition, but for the purpose of calibration played with short sounds, so the reverberators can also be programmed.

Name: ff tutti calibration chord_seven octave_wind+strings.mp3
Size: 3MB

Download Link:
http://vsl.co.at/upload/users/57/ff_tutti_calibration_chord_seven_octave_wind-strings.mp3

.

01. Set the track fader
Set all track faders with a sample playback software to -6.0 dB.

Does this mean for example if I have and aux in my DAW that is bringing in audio from an outboard cpu that I set it to -6.0? Or do I set the output faders in the outboard cpu of host Plogue Bidule to -06? Would you explain this step more and where it applies?

02. Set the output level of the samples
Set the output level of the patches within the sampleplayer also to -6 dB. This is the default value with most patches anyhow.

Where is this done in the VI player?

@dpcon said:

[b]Does this mean for example if I have and aux in my DAW that is bringing in audio from an outboard cpu that I set it to -6.0? Or do I set the output faders in the outboard cpu of host Plogue Bidule to -06? Would you explain this step more and where it applies?

I would leave the source signal loudness unchanged, and balance the effect signal at AUX-Return until it fuses with the mix.

The idea is to set up a listening environment, where we have the loudness under control, aurally as well visually.

The aural side is done by calibrating the monitor to a standard. This standard is that the monitor speakers produce an SPL of 85 dB when a signal of -20 dB RMS is played back. Important is, that you never change this loudness once it is calibrated. When I would change the volume, for example with the volume knob on the amplifier, I loose the loudness reference my ears are calibrated to.

The visual control is done by reading a meter with a peak and a rms scale.

.

Don't get lost in the endless field of audio engineering, Angelo. To my opinion, keeping this topic as close to your initial intentions is most likely the best idea for now.

-----------------

Moderator's Sidenote: The critical discussion on Angelo's approach continues HERE:
-> http://community.vsl.co.at/viewtopic.php?t=10761
Thanks!

Okay Dietz, good point.

.