15.2 Player music

[SchorschiSchrumpf]

~added~
I should have asked this earlier but does the new system reflect volume for +pppp+ and +ffff+ or are we will stuck at +ppp+ and +fff+?

The game correctly implements the +ppp+, +pppp+ issue already and always did, according to the ABC standart +pppp+ has exactly the same volume as +ppp+.
Citing from the abc-standard 2.1: !pppp! = !ppp! = 30, !pp! = 45, !p! = 60, !mp! = 75, !mf! = 90, !f! = 105, !ff! = 120, !fff! = !ffff! = 127

I think what Turbine is trying to do is not to change the engine of the music system (which would also include adding more octaves or change the ways volumes are calculated). It's just about replacing the samples that are used for the singular tones.

Thats like 11 instruments with 37 tones each ... just put them all in one folder, write a script that picks the loudest 1/10 of a second of each sample (or maybe a bit more), take the modulus of its fourier transform, multiply with one of those spectral weighting functions (A-weighting or something), integrate over the spectrum and normalize all the 11*37 samples by that. That's not magic and doesn't need any human judgement, that's what those spectral weighting functions were build for.
The details of that process are probably going to be the issue, but I'm sure it can be done in less than a day by someone who knows about signal processing and one scripting language like matlab or python.

[Thurallor]

Thats like 11 instruments with 37 tones each ... just put them all in one folder, write a script that picks the loudest 1/10 of a second of each sample (or maybe a bit more), take the modulus of its fourier transform, multiply with one of those spectral weighting functions (A-weighting or something), integrate over the spectrum and normalize all the 11*37 samples by that. That's not magic and doesn't need any human judgement, that's what those spectral weighting functions were build for.
The details of that process are probably going to be the issue, but I'm sure it can be done in less than a day by someone who knows about signal processing and one scripting language like matlab or python.

I think you're oversimplifying the problem of comparing the perceived loudness of e.g. horns vs. drums, but for most of the instruments, this approach should work reasonably well.

Why don't you whip up a quick Matlab script (or SciPy, I guess) and throw it over the wall?

[Emere]

Thank you Bruzo
~hugs~

[Lolorndinfey]

Ok so you should see some volume balancing tomorrow along with a switch back to the live version harp sound.

We do not intend to change the system again unless it is to add a new instrument. This pass was purely to correct the tuning issues and inconsistent note timing that was problematic in the Live version. Were we to change anything in this system again past 15.2, it would be because the musician community was in dire need for us to fix a bug/ tuning issue.

Best news I heard all week. Thank you so much Vyvyanne!

[Siegfriedpf]

Just did a quick run through using one tune (Ashokan Farewell, if you are interested ABC is below) on all instruments. IMO, with the exception of theorbo and harp, balance seems good. Theorbo could probably be dropped one more notch in volume, while harp could go up one. Samples all seem nice.

X: 1
T: Ashokan Farewell Solo lute or harp (2:01)
R: Folk
C: Ashokan Farewell, by Jay Ungar
N: Arranged by Falibrand of Landroval
Z: Transcribed by LotRO MIDI Player: http://lotro.acasylum.com/midi
%  Original file: ashokansolo-lute.mid
%  Transpose: 0
L: 1/4
Q: 100
K: C

z2 ^d3/8 z/8 g3/8 z/8 [^g5/4^G,3/8] z/8 ^D,3/8 z/8 [C13/8z/2] =g3/8
z/8 f3/8 z/8 ^d3/8 z/8 [c13/8C,3/8] z/8 ^G,3/8 z/8 [^D13/8z] ^A3/8
z/8 c3/8 z/8 [^c5/4^C,3/8] z/8 ^G,3/8 z/8 [F13/8z/2] =c3/8 z/8 ^A3/8
z/8 ^G3/8 z/8 [^A,3/8F3/4] z/8 F,3/8 z/8 [^C13/8F5/4^G5/4] z/4 F3/8
z/8 [^G,3/8^D3/4] z/8 ^D,3/8 z/8 [=C3/4^G3/4] z/4 [=G,3/4c3/4] z/4
[F,3/8^d3/4] z/8 C3/8 z/8 [F13/8^g3/4] z/4 c'3/4 z/4 [^C,3/8c'5/4]
z/8 ^G,3/8 z/8 [F13/8z/2] ^c3/8 z/8 c'3/4 z/4 [^D,3/8^a13/8] z/8
=G,3/8 z/8 [^G3/4^C3/4] z/4 [^C3/4=G3/4^d3/8] z/8 =g3/8 z/8
[^g5/4^G,3/8] z/8 ^D,3/8 z/8 [=C13/8z/2] =g3/8 z/8 f3/8 z/8 ^d3/8 z/8
[^A13/8=c13/8^F13/8^D,3/8] z/8 ^A,3/8 z/8 ^C3/8 z/8 ^D3/8 z/8
[^G,3/4^F3/8^A3/8] z/8 c3/8 z/8 [^c5/4^C,3/8] z/8 ^G,3/8 z/8
[=F13/8z/2] =c3/8 z/8 ^A3/8 z/8 ^G3/8 z/8 [^A,3/8F3/4] z/8 F,3/8 z/8
[^C13/8F5/4^G5/4] z/4 F3/8 z/8 [^G,3/8^D3/4] z/8 ^D,3/8 z/8
[=C3/4^G3/4] z/4 [^D,3/4c3/4] z/4 [=C,3/8^d3/4] z/8 ^G,3/8 z/8
[^D13/8c3/4^g3/4] z/4 [^d3/4c'3/4] z/4 [^D,3/8=G3/4^d3/4] z/8 ^A,3/8
z/8 [^C3/4^A3/4=g3/4] z/4 [^D3/4G3/4^c3/4^a3/4] z/4
[^G,3/8=c13/8^g13/8] z/8 ^D,3/8 z/8 =C3/4 z/4 c3/8 z/8 ^c3/8 z/8
[^G,3/8^d5/4] z/8 ^D,3/8 z/8 [^G,3/4C3/4z/2] =c3/8 z/8
[^G,3/4C3/4^G3/4] z/4 [^g13/8C,3/8] z/8 ^G,3/8 z/8 [^D3/4C3/4] z/4
[^G,3/8^d3/4] z/8 C,3/8 z/8 [f5/4^C,3/4] z/4 [^D,3/4z/2] =g3/8 z/8
[F,3/8^g3/4] z/8 =G,3/8 z/8 [^G,3/8^d3/8] z/8 [^D,3/8c5/4] z/8
[^G,3/4C3/4] z/4 [^D,3/4^A3/4] z/4 [^G,3/8c5/4] z/8 ^D,3/8 z/8
[^G,3/4C3/4z/2] ^A3/8 z/8 [=G,3/4^G3/4] z/4 [F,3/8F13/8] z/8 =C,3/8
z/8 [^G,13/8z] ^C3/4 z/4 [^D,3/8^D3/4] z/8 ^A,3/8 z/8 [^C3/4=G3/4]
z/4 [^C3/4G3/4z/2] ^d3/8 z/8 [^d13/8^C,3/8] z/8 ^A,3/8 z/8
[^D3/4G3/4] z/4 c3/8 z/8 ^A3/8 z/8 [^G3/4^G,3/8] z/8 ^D,3/8 z/8
[c3/4^G,3/4=C3/4] z/4 [^D,3/8^d3/4] z/8 F,3/8 z/8 [^F,3/8^f13/8] z/8
^C3/8 z/8 [^F3/4^A3/4] z/4 [^F3/4^A3/4z/2] ^g3/8 z/8 [=f5/4^C,3/4]
z/4 [^D,3/4z/2] =g3/8 z/8 [=F,3/8^g3/4] z/8 =G,3/8 z/8 [^G,3/8^d3/4]
z/8 ^D,3/8 z/8 [^G,3/4=C3/4c5/4] z/4 [^D,3/4z/2] ^G3/8 z/8
[^G,3/8^D3/4] z/8 ^D,3/8 z/8 [^G,3/4C3/4^G3/4] z/4 [=G,3/4c3/4] z/4
[F,3/8^d3/4] z/8 C3/8 z/8 [=F13/8c3/4^g3/4] z/4 [^G3/4c3/4] z/4
[^D,3/8^A5/4] z/8 ^A,3/8 z/8 [^C3/4z/2] ^G3/8 z/8 [^D,3/4^C3/4=G3/4]
z/4 [^G13/8^G,3/8=C13/8] z/8 ^D,3/8 z/8 ^G,3/4 z/4 [=g3/8^d3/8] z/8
[^c3/8g3/8] z/8 [^G,3/8c'5/4^g5/4] z/8 ^D,3/8 z/8 C3/8 z/8
[^D,3/8=g3/8^a3/8] z/8 [^G,3/8^g3/8f3/8] z/8 [^A,3/8=g3/8^d3/8] z/8
[^d13/8c'13/8=C,3/8] z/8 ^G,3/8 z/8 ^D3/8 z/8 ^G,3/8 z/8
[^A,3/8^c3/8^a3/8] z/8 [C,3/8^d3/8c'3/8] z/8 [^C,3/8f5/4^c5/4] z/8
^G,3/8 z/8 F3/8 z/8 [^G,3/8c'3/8^d3/8] z/8 [^C,3/8^c3/8^a3/8] z/8
[=C,3/8=c3/8^g3/8] z/8 [^A,3/8^G3/4f3/4] z/8 F,3/8 z/8
[^A,13/8^c5/4^g5/4] z/4 f3/8 z/8 [^d3/4^G,3/8] z/8 ^D,3/8 z/8
[^G3/4C3/8] z/8 ^D,3/8 z/8 [^G,3/8^D3/4=c3/4] z/8 =G,3/8 z/8
[F,3/8^G3/4^d3/4] z/8 C3/8 z/8 [c3/4F3/8^g3/4] z/8 C3/8 z/8
[F,3/8^d3/4c'3/4] z/8 ^D,3/8 z/8 [^C,3/8^d5/4c'5/4] z/8 ^G,3/8 z/8
^C3/8 z/8 [f3/8^G,3/8^c3/8] z/8 [=C,3/8^d3/4c'3/4] z/8 ^C,3/8 z/8
[^a13/8^c13/8^D,3/8] z/8 ^A,3/8 z/8 [^C3/4=G3/4] z/4 [=g3/8^d3/8] z/8
[^c3/8g3/8] z/8 [c'5/4^G,3/8^g5/4] z/8 ^D,3/8 z/8 =C3/8 z/8
[^a3/8=g3/8^D,3/8] z/8 [^g3/8f3/8^G,3/8] z/8 [=g3/8^d3/8^A,3/8] z/8
[^D,3/8^f13/8^a13/8c'13/8] z/8 ^A,3/8 z/8 ^C3/8 z/8 ^D3/8 z/8
[^c3/8^a3/8^G,3/4] z/8 [^d3/8c'3/8] z/8 [^c5/4^C,3/8=f5/4] z/8 ^G,3/8
z/8 F3/8 z/8 [^d3/8c'3/8^G,3/8] z/8 [^c3/8^a3/8^C,3/8] z/8
[=c3/8^g3/8=C,3/8] z/8 [^G3/4f3/4^A,3/8] z/8 F,3/8 z/8
[^c5/4^g5/4^A,13/8] z/4 f3/8 z/8 [^d3/4^G,3/8] z/8 ^D,3/8 z/8
[=C3/8^G3/4] z/8 ^D,3/8 z/8 [^D3/4=c3/4^G,3/8] z/8 =G,3/8 z/8
[^d3/4F,3/8^G3/4] z/8 ^G,3/8 z/8 [^D3/8^g3/4c3/4] z/8 ^G,3/8 z/8
[^d3/4c'3/4C,3/8] z/8 ^C,3/8 z/8 [=G3/4^D,3/8^d3/4] z/8 ^A,3/8 z/8
[^A3/4=g3/4^C3/8] z/8 ^A,3/8 z/8 [^c3/4^a3/4^D3/4] z/4
[=c13/8^g13/8^G,3/8] z/8 ^D,3/8 z/8 [^G,13/8z] [^G3/8c3/8] z/8
[^A3/8^c3/8] z/8 [^d5/4^G,3/8=c5/4] z/8 ^D,3/8 z/8 ^G,3/8 z/8
[^G3/8^D,3/8c3/8] z/8 [=C3/4^G3/4^G,3/8] z/8 ^A,3/8 z/8
[=C,3/8^g13/8^d13/8] z/8 ^G,3/8 z/8 ^D3/8 z/8 ^G,3/8 z/8
[C,3/8^G3/4^d3/4] z/8 ^D,3/8 z/8 [^C,3/8f5/4^c5/4] z/8 ^G,3/8 z/8
^C3/8 z/8 [=g3/8^G,3/8^d3/8] z/8 [^C,3/8f3/4^g3/4] z/8 ^D,3/8 z/8
[^G,3/8^d3/8=c3/8] z/8 [^D,3/8c5/4^G5/4] z/8 ^G,3/8 z/8 ^D,3/8 z/8
[=C3/8^A3/4] z/8 ^D,3/8 z/8 [c5/4^G,3/8] z/8 ^D,3/8 z/8 ^G,3/8 z/8
[^A3/8^D,3/8] z/8 [^G,3/8^G3/4] z/8 =G,3/8 z/8 [F,3/8F13/8] z/8
=C,3/8 z/8 [^G,13/8z] ^C3/4 z/4 [^D,3/8^D13/8] z/8 ^A,3/8 z/8
[^C13/8z] [=G3/4^D3/4] z/4 [^C,3/8^d13/8G13/8] z/8 ^A,3/8 z/8 ^D3/4
z/4 [^D3/8c3/8] z/8 [^C3/8^A3/8] z/8 [=C3/4^G3/4^G,3/8] z/8 ^D,3/8
z/8 [^D7/8c7/8^G,3/8] z/8 ^D,/2 z/8 [C3/8^G7/8^d7/8] z/8 ^D,/4 F,/4
z/8 [^F,/2^A2^f2] z/8 ^C/2 z/8 [^F2z5/4] ^g z/4 [=f5/4^C,3/8] z/8
^G,/2 z/8 [^C7/4z/2] =g/2 g/2 z/8 ^g3/8 z/8 [^G,/2^d7/8] z/8 ^D,3/8
z/8 [^G,7/4c11/8] z3/8 ^G3/8 z/8 [^D/2^G,7/8] ^G/2 z/8 [^G,7/8c3/8]
z/8 ^d/2 z/8 [=G,7/8^g3/8] z/8 c'/2 z/8 [=F,7/8^d3/8] z/8
[^g11/8z5/8] =C7/8 z/8 [^Gc'] z/8 [^c11/8^a11/8^D,/2] z/8 ^A,3/8 z/8
[^D9/8z5/8] [=G5/8^d5/8] z/8 [^D,5/8^C5/8^A5/4=g5/4] z/8 ^D5/8 z/8
[^G,6=c6^g6]

[Lolorndinfey]

Jan 23: 17:09 server time

Just completed a test of the current volume levels per today's build. For your consideration;

(OTM = on the money, NC = no change) where OTM has a tolerance of +/- 3-5%, close enough to previous levels.

Recommended volume adjustments compared to LIVE levels in volume.

-20% Lute *
NC Harp
-40% Theorbo
-10% Flute
OTM Clarinet
+5% Horn **
OTM Bagpipe
OTM Pibgorn ***
OTM Drum
+20% Cowbell


Bug when stopped bagpipe mid song the last note kept repeating.

*Lute has a higher sample quality with built in reverb? less bass. better mid range, overall shift to upper registers. Qualities of mandolin. Better tone and acoustic qualities

** Horn loss in bass tones, and louder upper octave, with a shorter decay - different sound

*** Pibgorn actually about 12-15% softer than before, but tones in upper octave way better. This is a solo-able instrument now.

Overall with the exception of Harp, the better sound samples either has built in reverb, broader range of frequency in the sample, or general shift away from bass or lower octaves. It's not just the shorter decay in some samples, but the higher fidelity of the note samples, such as the mandolin qualities of lute now.

[Emere]

Jan 23: 17:09 server time

Just completed a test of the current volume levels per today's build. For your consideration;

(OTM = on the money, NC = no change) where OTM has a tolerance of +/- 3-5%, close enough to previous levels.

Recommended volume adjustments compared to LIVE levels in volume.

-20% Lute *
NC Harp
-40% Theorbo
-10% Flute
OTM Clarinet
+5% Horn **
OTM Bagpipe
OTM Pibgorn ***
OTM Drum
+20% Cowbell


Bug when stopped bagpipe mid song the last note kept repeating.

*Lute has a higher sample quality with built in reverb? less bass. better mid range, overall shift to upper registers. Qualities of mandolin. Better tone and acoustic qualities

** Horn loss in bass tones, and louder upper octave, with a shorter decay - different sound

*** Pibgorn actually about 12-15% softer than before, but tones in upper octave way better. This is a solo-able instrument now.

Overall with the exception of Harp, the better sound samples either has built in reverb, broader range of frequency in the sample, or general shift away from bass or lower octaves. It's not just the shorter decay in some samples, but the higher fidelity of the note samples, such as the mandolin qualities of lute now.

After taking my band onto BR today, I would have to say your recommendations for volumes is pretty spot on. The only exception I would make is on the Theorbo. I would suggest maybe not so much of a reduction since many of us have wanted just a hair more volume out of it but that is just me nitpicking. I would say -30 to -35% however that would mean adjusting files later and since I have been a huge advocate of not having to do that unless completely necessary, I will digress.
My observations are that everything sounds almost delightful now. I can actually recognize my songs today and they sound great for the most part.
The exceptions to this are the flute and clarinet. The banjo/mandolin sounding notes ruin the instrument. It makes it sounds as if someone is filling in odd notes with a banjo and it doesn't flow well with any of my solos. The clarinet still has two notes that sound completely odd compared to the rest and those should be addressed as well however, the overall quality of it seems much nicer save for the couple of odd sounding ones. My biggest concern however is the loss of notes while playing all six characters on songs that are lute heavy. It seems as if there is a lower note cap on Bullroarer and when I play for example, my John Henry song, which consists of three lutes, theorbo, bagpipe, and harp. I am hearing a loss of bagpipe notes and lute notes, however, individually, they all appear to play well. As for some of my other songs, they sound marvelous and I'm very excited to see what else is done before this goes live. To the developers I have to say this....Awesome job with this last update on Bullroarer, you guys are without a doubt going in the right directions and the difference between my early tests yesterday and todays tests are like night and day. Keep up the wonderful work and there will be many bunny hugs from Fincin :P

P.S. Bunny Pets for hobbits!!!!!

[Lolorndinfey]

Yes. the Theorbo is an odd beast. tone is richer, yet shifted higher in register. That means less bass on the lower octave, more richness to the bottom end of the upper octave. According to the equalizer graph, it's actually double, but the shift in the mean average of frequencies means it's quieter.
By pure numbers, -50% is the graph value. by ear, adjusting the player music meter nets a result of -37 to -40. I toggled playback of LIVE vs BR side by side, and got that value.

Yes, I've heard the dropped notes on some of the new samples in the first 8 bars also, but subsequent playback of the same song results in proper playback.
It's like the new samples are memory intensive, had to load first, then is cached in memory. A pet peeve.

Work around this by playing a tuning song to get it cached, and it should be golden.

Play the clarinet by itself. you hear the error in the bad notes. It's like a mechanical breath or a rub on the microphone. I called it an error during recording, like a microphone pole rub. Extraneous noise not generated by the instrument. Not a showstopper, but a pet peeve.

I'm really excited and hope they can squeeze in another set of volume tweaks before release, and please. More contradictory or corroborating reports. More the better. Let Turbine get a feel of the trend, not just one or two players.

Is there any possibility of manipulating the harp samples to have a longer sustain, bring back the bass quality to it in the lower octaves, and do whateer is necessary to enhance the polyphonic nature of it, which is usually mistaken as tone? When a harp string vibrates, the adjacent strings are also vibrating slightly , resulting in a more bell curve, as opposed to a spike on the equalizer. I think that's what's missing to make it better. Not echo, not mid range, but the polyphonic nature of it.

Personally, if volume tweaks are within 10% of each instrument's recommendation, I'll be a happy camper.

[Emere]

I cannot agree more Lolorndinfey. I lack the fancy know how to test as you have done and I lack any musical abilities other than my transcribing for Lotro. But it does sound like you pretty much hit the nail on the head. I would be very put out to see this go live quite yet as the lute and clarinet for me, would be show stoppers. We (the bands) have held ourselves to the highest standard of quality that Lotro and our own personal limits have allowed for and going live at this point without a bit more work would be a step backwards for us. I think this probably applies to most everyone judging by the performances at Winterstock this year.
Its going to be a whole new ball game once this does go live as some of the tricks and methods we've learned over the last several years to create a certain sound or feel to songs may not work anymore but it will be grand fun to learn new tricks.

Just please, please, Dev's......fix the few items left before you kick this out the door. All the years and work people have put into their music so far here is worth that little more effort to put this out right. The music society is very strong right now and I would hate to see them get a kick in the face. We do thank you so very much for the great effort so far, the listening to us, the strive for perfection you have shown so far. You guys rock!

[SpiritusLOTRO]

My understanding was that the musician community wanted these updates enough to make adjusting ABC files worth it. If that is not the case then we need to know that so that we can leave the system as it is on Live.

This was certainly my position. I don't mind having to do volume adjustments to past work for a music system that going forward has proper dynamic balance and tuning between the instruments. Doing so would create a strong base for the addition of new instruments and features such as expanded octave ranges.

Regarding the current beta build, it's coming along nicely. My comments:

-------------------------
Inter-Dynamic Balance
-------------------------
Going by ear, I feel the harp needs a ~15% increase in volume. The lute could come down by ~15%. The theorbo could come down by ~15%. The flute could stand maybe a ~5-10% increase. Obviously these are spitballs, but give a general range of what I'm hearing. The rest of the instruments seem to be well balanced to my ear.

--------------------------------------------------
Tuning, Timbre, and Intra-Dynamic Balance
--------------------------------------------------
All of the instruments individually seem well tuned. I cannot speak to how they blend together.

The biggest area of work is in timbre. Many individual notes have an off timbre that make them stand out too much from the rest of the tones on the instrument. I am detailing those ones that I feel deserve attention below:

Clarinet:
--------
F3 -- Has a "loose reed" sound that makes it stand out from the other notes. It is identified by a noticeable buzzing vibration on the attack of the note.

Flute:
-------
C2 -- Has a late attack compared to other notes. Meaning, the breath pushed across the lip plate is more gradual in reaching the desired force. This can either cause a note to sound "late" in the rhythm, or depending on the note length, not sound at all, or sound very soft/cut off compared to the other notes.
F3 -- Has a very hard attack compared to other notes. Meaning, the breath is pushed out very forcefully across the lip plate before settling into the desired force. This note plays very much like a note with a permanent "accent" and can make it undesirably stand out in a piece.
C#4 -- Has a late attack. See Flute C2.
D#4 -- Seems to be about ~15% softer dynamically compared to the other notes.
E4 -- Has a late attack. See Flute C2.

Horn:
-------
C#2 -- Has a noticeably "buzzy" timbre compared to the other notes. This timbre aberration is sustained throughout the duration of the note sample.
C#3 -- A slight, but noticeable "buzzy" timbre as if someone played a note on a saxophone with a slightly hard attack. Unlike Horn C#2, this is very short and does not sustain throughout the note.
F3 -- It either needs about at ~10% dynamic increase or more depth in the timbre. The note is noticeably more mellow than the others.
D4 -- A slight, but noticable "buzzy" timbre. Similar to Horn C#3.

Bagpipe:
-------
Honestly, I could nitpick here, but I feel a little variation in timbre is a good thing on the pipes as it gives them a more rustic sound. And the dynamic balance, while could use some micro adjusting, is good enough for my ear across the board. Good Job.

Lute:
-------
D#2 -- Has a noticeably "metallic" sound compared to the other notes, as if it was picked by a fingernail or pick instead of the finger.
D#3 -- Same as Lute D#2. Slightly less pronounced.
B3 -- Same as Lute D#2. Very pronounced.
B4 -- Same as Lute D#2, but also has an odd, off-tune reverb. Difficult to describe, but immediately recognizable once heard.

Harp:
-------
*EDIT: If I understand you correctly, this may be useless information as I believe you said that the old samples can't be adjusted? And the harp has been reverted. If so, then disregard.
All notes -- This might be an aberration on my end, but it seems like all the notes on the harp have a slightly delayed attack compared to other instruments. Meaning, there is a small delay between prompting a note and the actual beginning of the tone; as if the sample itself has a tiny pause built in. I would need more confirmation from others before considering this an actual issue.
D#2 -- Has a slight "metallic" sound compared to the other notes, as if it was picked by a fingernail instead of the finger.
C#3 -- Same as Harp D#2.
F3 -- Same as Harp D#2.
G3 - Same as Harp D#2.
G#3 -- Same as Harp D#2, but more pronounced.

Theorbo:
-------
E2 -- This note alone is causing my audio output to peak on the bass, which leaves me to believe that the bass is turned up too much on this note in the sample itself, rather than an issue with my card or output devices.
B3 -- Extremely heavy "metallic" sound on this note compared to others. Very out of place and sounds more like a low lute note than a mid-range theorbo note. A must fix.
C#4 -- Too much pitch shift on the reverb. Meaning the "wavy" sound a plucked string gives off after played is unnaturally broad -- almost as if this specific note has a finger vibrato applied, rather than just its natural vibrato.
D#4 -- Same as Theorbo C#4, less pronounced.
G4 -- Same as Theorbo C#4, less pronounced.
G#4 -- Same as Theorbo C#4, less pronounced.
A4 -- Has the pitch shift reverb issue as in Theorbo C#4, but also has a "hard attack" sound on the pluck that is noticeably different than the timbre on the attack of the other notes.
A#4 -- Has both of the issues of Theorbo A4.
C5 -- Has both of the issues of Theorbo A4.

Pibgorn:
-------
There are still many odd notes on the pibgorn; a tribute to its troubled past. Instead of identifying specifically the issue with each note, I'm just going to use notes C2 - G#2 as my timbre base and list the notes that are off from that:
A2, A#2, B2, C3, G#3, A3, A#3, B3, C4, A4, A#4, B4

Drums, Cowbell, Moor Cowbell:
-------
Nothing to report here. Wasn't much wrong before. In fact, I'm not certain if this has even been re-sampled. Which is fine by me. I have no pressing issues with these instruments. Good Job.

==========================

I hope this helps you fine-tune these instruments. Thank you very much for your hard work and top-notch feedback and engagement. I have no doubt that, working together, we can create a superior music system of which everyone can be proud.

With regards,

Lifimun of Landroval
Bandmember & Kinmember of Les Beaux Chapeaux

[Emere]

Did a second round of testing tonight with some of Bright Stars band members.
We confirmed that three lutes that are note heavy seem to cap the note limit and it was cutting the bass and bagpipe out of one song, and on another song which was only three lutes and theorbo, both the lutes and theorbo started losing notes. That alone is going to kill many songs from being able to be played and should be addressed.
I encountered another problem after playing the lute heavy song, I switched up to a song with two lutes, drum, theorbo and two horns and my sound in game crashed. I had to reboot the game to get it back although game sounds and voice still seemed to be working fine. I have a pretty fair sound card (sound blaster recon 3-d) and have never had that happen before with it.
As eager as I am about the new music samples, I will say again, this is not ready to go live yet.

[Orlor]

Ok, I found a couple of new bugs.

First of all, there is no more drone on the bagpipe. I have a song where about a quarter way in, the drone starts and it carries through to the end of the song. In the new build, the note continues until the end of the sample and then stops until the next drone note appears. Not a huge deal as not a lot of people actually use it but it is something that is missing from the old system.

Second, I have confirmed with Fincin that there is a problem with band pieces with three or more lutes in them. They cut off the Theorbo and Bagpipes and the other lutes. This in my opinion is a deal breaker. It will be unacceptable if it is released with this bug.

[Siegfriedpf]

What do you mean they "cut off"? The other instruments stop playing completely?

[SchorschiSchrumpf]

I think you're oversimplifying the problem of comparing the perceived loudness of e.g. horns vs. drums, but for most of the instruments, this approach should work reasonably well.

Why don't you whip up a quick Matlab script (or SciPy, I guess) and throw it over the wall?

Well you maybe right, but we definitly need something to make those observations objective - we can't have votes on volume levels and looking at the thread here shows that we have lots of maybe similar, but not quantativly agreeing observations here.

So I sat down, whipped together that python script ( https://sites.google.com/site/brutem...lumecompare.py ) and applied it to 4 randomly picked samples, so first is the state of volumes as they are played on bullroarer right now:
Before (current state) : https://sites.google.com/site/brutem...arer_01_24.wav

And this is the result of me de-amplifying 3 of the samples to align them by their peak volume to the clarinet - the differences the script found ranged were around 3db (which is the difference of +f+ to +fff+!!)
After : https://sites.google.com/site/brutem...eakaligned.wav

That one flute tone sounds still too loud (it was reduced by 3dB compared to the clarinet tone). If you open the 2 wav's in an audio-editor you will also see quite clearly that the amplitudes are more similar than before. The big question now is:

Would it help if we made a huge list of all the 11*37 tones and their relative volume differences? Because it would mean that 11*37-1 tones would have to be readjusted in volume - that's something I wouldn't wish anyone to do manually!

I have to add: I didn't properly test that script on reference tones and also the A-weighting is not the optimal function. It's just the result of an hour of coding and testing. Feel free to modify/use the script for whatever purpose you want!

[SpiritusLOTRO]

What do you mean they "cut off"? The other instruments stop playing completely?

It was as if the note cap had been lowered. So, the pipes and theo would sound sometimes, but mostly would not sound at all. Example: The current default sound cap is 64 simultaneous sounds. The behavior I observed was as if the number of simultaneous sounds was hovering around or over 64 tones, causing some tones -- which in this case were identifiable on pipes and theo -- to simply not play by being pushed off the 64-max sound table. The problem is this particular song was no where near creating 64 simultaneous tones and does not have this issue on the live server.

Further testing would be needed to confirm exactly /what/ is causing the behavior, but the above is the best way to describe it. I would seriously doubt it is something as specific as "three lutes = problems," but the issue is there and needs a closer look.

[Lolorndinfey]

Well you maybe right, but we definitly need something to make those observations objective...

Would it help if we made a huge list of all the 11*37 tones and their relative volume differences? Because it would mean that 11*37-1 tones would have to be readjusted in volume - that's something I wouldn't wish anyone to do manually!

I have to add: I didn't properly test that script on reference tones and also the A-weighting is not the optimal function. It's just the result of an hour of coding and testing. Feel free to modify/use the script for whatever purpose you want!

I see what you're doing. Yes. Objective is always good. But when it comes to music. It's a marriage of technical and artform I find.

Analyzing by peak values alone is inherently, scientific. Take the Theorbo for example.

On BR, playing a song yielded a peak of -8db, mean avg of -28 db
On LIVE, the same song yields a peak of -28 db, mean avg of -41db

Both are slightly different shaped bell curves from 40hz to 12KHz

So, scientifically, it is a complex set of noise, that has different properties in peaks, frequency range of playback, and different mean averages. One cannot take one set of noise and multiply it to match the other. This is where the artform comes in. Close your eyes, listen. Within a 10% or less window, adjust the playback on BR to match what you hear on the LIVE playback. Double check with the frequency equalizer display.

In Houdini, there is a ChOP's editor. It is a visual channel editor than can accept MIDI input, and display them all as a mess of channels on graph, or as a bar. Both playbacks above would be diplayed as if 7 channels (number of instruments) of noise were displayed as 7 lines across a graph. A visual editor of noise. Looking at the graph, you would see visually what was going on. Different properties of the spikes, and the body of the mass of avg noise in the sample.

By your method, just take the peaks and multiply to match. Do you begin to see(hear) the danger in that type of analysis? The perceived volume in the end result will be the BR playback is softer than the LIVE playback. Why? because the peak behavior in the BR playback is different in frequency and ratio vs LIVE. The picture (visually displayed in the ChOPs editor, gives you a visual cue as to what exactly is going on) is more complex than a simple multiply. Just to be clear, you can record both the digital (MIDI instruments) and the playback (audio to digital) as input channels.

Don't get me wrong. Objective analysis is very important. But it is only part of the picture. Bottom line, there's an art to this, and when it comes to debugging, it runs contrary to logic when it comes to the solution. 60% technical, all the rest 30%. Perspiration makes up the last 10. heh

[SchorschiSchrumpf]

I'll try not to stray away from the original topic, I'll focus on the stuff that actually matters for our game and the patch:

*** By your method, just take the peaks and multiply to match. ***

Obviously it is much more complex than just comparing peak volumes. Did you look at the example I provided?

This is not the only method, but do you want to process 11*37 samples all one by one? Given 10 seconds for every pair you'd be working on that for hundreds of hours.
That's just insane - without a technical basis where the tones will at least be in a good approximation to the right volume it is an impossible task.
The state of the music system on Bullroarer confirms my claim there.

*** Don't get me wrong. Objective analysis is very important. But it is only part of the picture. Bottom line, there's an art to this, and when it comes to debugging, it runs contrary to logic when it comes to the solution. 60% technical, all the rest 30%. Perspiration makes up the last 10. heh ***

As soon as they are at 80% of those technical 60% I'll shut up and be happy - so far they are still far away from that.



( had to edit the way I was quoting sorry)

[Lolorndinfey]

Jan 24: 13:34 server time

Completed a test of the current volume levels per today's setting. For your consideration;

(OTM = on the money, NC = no change) where OTM has a tolerance of +/- 3-5%, close enough to previous levels.

Different set of songs tested. Was my previous recommendations misread? Lute, Flute got louder. Need to go more quiet.

I'm scratching my head here. Different songs yielding different results?

Recommended volume adjustments compared to LIVE levels in volume.

-45% Lute
NC Harp
-50% Theorbo
-25% Flute
OTM Clarinet
OTM Horn
+10% Bagpipe
+10% Pibgorn *
OTM Drum
+20% Cowbell

Can I get more to confirm these results? Going nuts here.

* Though more quiet than the LIVE version, pibgorn has a different sound now, and I can stand the more quiet version with the richer tone.

What the...

[Lolorndinfey]

Obviously it is much more complex than just comparing peak volumes. Did you look at the example I provided?

This is not the only method, but do you want to process 11*37 samples all one by one? Given 10 seconds for every pair you'd be working on that for hundreds of hours.

That's just insane - without a technical basis where the tones will at least be in a good approximation to the right volume it is an impossible task.

As soon as they are at 80% of those technical 60% I'll shut up and be happy - so far they are still far away from that.

Yup. originally I was going to play thru each instrument with a test file running thru the octaves at different volume settings. But reality is that you're playing ensemble pieces or solo pieces of accapella, to multi-chord riffs.

But then I thought, play a song. analyze it as best you could. LIVE simulation if you will. The tests were fast, and how effective remains to be seen. Today's test was a head scratcher for eg. I didn't think there was an overnight build, and yet I'm getting different results. But I reported it nevertheless remaining objective about findings. Let Turbine sort it out.

I think that's the key here. How can I test everything? You probably can't in the time allocated, but you can certainly cover as much ground as possible with what you know. Scientific? Probably not. But does it serve the purpose of beta testing? You bet. Because the masses will each report a piece of the puzzle.

Another example. The bad clarinet and lute notes. The clarinet bad notes could be mistaken for a breath or an instrument knock, the mandolin like notes on the lute by itself could be identified as bad notes. But when playing in a song, it...well.. kinda works. Even in real life a perfect performance is reserved for studio recordings. Like Lens flare in cinematography. It's an error, but it adds to the realism. Just not JJ levels. This is what I meant by the artform remark earlier. It's not all scientific and objective. Those bad notes are not show stoppers. Well, that's IMO anyways as an artist/musician.

[SchorschiSchrumpf]

**** let turbine sort this out ****

This is exactly the issue, but I don't see that happening. Not without some proper information on how to fix it - if the ppl doing the modifications would understand how to do it they'd never have put up such messed up mixture of volumes on the test-server in the first place.

I saw a lot of ppl posting here stuff like: the lute needs to be tuned 20% down. How exactly is that going to help the person fixing the sound system? Is it 20% reduction of amplitudes or 20% in volume (which is by convention 20*log_10(amplitude)) or is it 20% on the ingame volume slider?
We need to use some technical language here - else you can just say: the lute is too loud or a bit too loud.

[Keli]

It's probably too late now to check any ramifications but I wonder, Turbine, if we can get MaxPlayingSounds=128 set as a default? Currently it reverts to 64 every client build.

Odd little notes etc aside, it is really beginning to sound good. Bagpipe now sounds terrible in comparison to the others but I'm happy to cut down on it's use.

Only really issue I've had in testing is with the really note heavy pieces which is why I wondered about the above. Might solve Fincin's problem; if not then that is a serious bug that needs squashing.

I respect the work of those who have been quantifying the dynamics, I just work by ear!

[Lolorndinfey]

I saw a lot of ppl posting here stuff like: the lute needs to be tuned 20% down. How exactly is that going to help the person fixing the sound system? Is it 20% reduction of amplitudes or 20% in volume (which is by convention 20*log_10(amplitude)) or is it 20% on the ingame volume slider?

If someone told me in game can you bring your volume down 20%, I'm pretty certain they're not using the log scale. But you are right. there's a chance they might, albeit a slim chance. You must've missed my explanation for testing. I was using the in game scale for player music. A simple oversight on my part to detail all explanations. But tbh, if Turbine needed clarification, my e-mail topic link to QA would've had an email requesting clarification.

[Lavs]

have done roughly an hours testing on Bullroarer the last two nights, with two different builds there. I repeated 5 songs the second night that I had done the first but also tried some new things with another player with more alts. I'm sorry I don't have great musical knowledge or tech speak so I am going to have to leave that to others.

Lute volume is better now than the previous build, and I would say just a smidge louder than it is currently on live. There's still some bad notes, but I didn't go through it thoroughly to identify what those were. The most obvious I think was the D#2 that Lifimun identified earlier. Not just in playing a scale, these notes can stand out in a song and throw you. They really do need correcting for an enjoyable player music experience. I also had some wierd choppiness whenever the lute played, particuarly if playing chords quickly. The other person wasn't getting it anywhere near as much as me, but did still experience it occasionally and I think those times did correspond to when it was worst for me. I was hearing notes, but they seemed to be echoed by this choppiness... really not sure how else to describe it and sadly I am not set up to record to share it. Will hopefully get to look into this more in the next few days if BR stays up. I will note I do experience some wierd audio things in live, so this might just be another sign of that, but I really do wonder when it wasn't just me it was happening to.

With flutes we tended to focus on songs that went high and in the past verged on being shrill and unpleasant. This was much improved, so looking forward to that aspect of the flutes.

The pibgorn still has problematic notes. I know many people don't bother with this instrument, but we do find it useful so are hoping it can be improved further.

Theorbo volume is better now than when first tested on BR, not sure if the distortion has been fixed as it may have just been disguised with the lower volume.

Thanks for putting the live harp in, though like others would dearly like some of it's notes to get some improvements (eg those muddly lower octave ones). Of the understanding that the old samples can't be fixed in that way, so I guess that's just something that we will have to continue to live with.

I would also put in a vote for an increase in max sounds. Did experience some notes seeming to be AWOL, but also hard to tell in some songs with all the lute choppiness going on. We weren't playing songs that experience that in live, even with less characters playing than our full band size, so it does seem like there's some bug in there causing this that needs fixing. Increasing the max sounds on top of that would be grand

[Orlor]

I would seriously doubt it is something as specific as "three lutes = problems," but the issue is there and needs a closer look.

Maybe not but the problem didn't appear until after three lutes were involved. We played several 8 part songs with no problems. Fincin played a 8 part song with 3 lutes and you couldn't hear the theorbo or the bagpipes. I played a four part song with 3 lutes and a theorbo and you couldn't hear the theorbo and some of the lutes.

Is the problem as specific as 3 lutes = Problems? I don't know. Maybe, maybe not. However, it does seem that the problem only crops up when playing 3 lutes.

[SpiritusLOTRO]

Maybe not but the problem didn't appear until after three lutes were involved. We played several 8 part songs with no problems. Fincin played a 8 part song with 3 lutes and you couldn't hear the theorbo or the bagpipes. I played a four part song with 3 lutes and a theorbo and you couldn't hear the theorbo and some of the lutes.

Is the problem as specific as 3 lutes = Problems? I don't know. Maybe, maybe not. However, it does seem that the problem only crops up when playing 3 lutes.

It would seem silly that the problem would be "3 lutes = problems." It is more likely that lute parts generally have more tones attributed, so if you have three, it would work more towards perhaps a lowered note cap compared to live.

My suggestion would be to play the songs that are causing the issue, but instead sub harps for the lutes and see if you still have the cut-off. If so, then we know it isn't specific to 3 lutes. And if you really want to get froggy, then put 3 horns on the 3 lute parts, and if the other parts are still cut off, then you can be certain that it isn't so much "3X" as it is an issue of too many tones... which needs to be fixed asap since, for that particular song, there is no way it's capping at 64 tones.