Distortion in Euphonium

[tubatimberinger]

I posted something about this a while back, but didn't have time to add any examples. I'm using the stock library and stock reverb in N4. I keep experiencing distortion. After a while I have been able to isolate it to a specific pitch "G" on top space (bass clef).

I attached a couple sounds and a page from one of the pieces.

Even at a soft dynamic

Firm Land 2013-distortion.jpg (161.54 KiB) Viewed 7470 times

Thanks for any ideas.

-t

[Surfwhammy]

I keep experiencing distortion. After a while I have been able to isolate it to a specific pitch "G" on top space (bass clef).

I did a few experiments, and as best as I can determine at present this does not involve the NOTION 4 native reverb . . .

BACKGROUND

(1) Musical pitch and frequency are logarithmic, which also is the case with loudness and panning, although there are linear aspects to panning . . .

One of the things I discovered during the three or so months when I temporarily went Country Western and decided to start singing bass is that it is not so easy to sing bass, because the difference from one note to the next is very small in terms of pitch . . .

Using string bass at standard tuning where "Concert A" is 440-Hz as an example, the low-pitch "E" string is 41.204-Hz, and the low-pitch "A" string is 55-Hz . . .

The notes in this range are {E1, F1, F#1, G1, G#1, A1}, and the difference from the lowest to the highest is approximately 13.8 cycles per second . . .

Scientific Pitch Notation (wikipedia)

For singing, this requires very precise accuracy, since each note differs from the next by only 2 or 3 cycles per second (or "Hertz [Hz]"), which is a tiny difference at best . . .

If the pitch is off by 2 or 3 cycles per second, then you entirely miss the note and are singing another note . . .

In contrast, in higher registers, it is possible to miss a note by a few cycles per second and not be noticed . . .

(2) Instrument Sampling Techniques . . .

There are two general ways to sample instruments, chromatic and non-chromatic, and the way which samples every note is called "chromatic"; and since chromatic sampling requires more samples than non-chromatic sampling, chromatically sampled instrument libraries typically cost more and are larger than non-chromatically sampled instrument libraries . . .

Non-chromatically sampled instrument libraries have samples for some but not all of the notes, where as an example, C4 ("Middle C") might be sampled but C#4 might not be sampled, while D4 is sampled, with the result that when a C#4 is specified in the music notation the software "engine" for the sampled instrument library computes the C#4 using an algorithm which among other things does a bit of logarithmic extrapolation and other activities, although this probably is done by multiplying, and the base note will be either C4 or D4, which are the two nearest actually sampled notes to C#4; and if C4 is used as the base note, then the C#4 note is computed upward from C4, but if D4 is used, then the C#4 is computed downward . . .

For notes that differ only by a half-step or perhaps no more than three half-steps, the computed "missing" notes are very realistic and typically sound the same as actually sampled notes, but it depends on the playing style, dynamics, articulation, pitch, and other characteristics, where as an example, if the instrument is an electric guitar played through a Fender amplifier with tremolo at a fixed rate, then the tremolo rate for computed notes will be either slower or faster than the tremolo rate for actually sampled notes, and whether the tremolo rate is slower or faster depends on whether the nearest actually sampled note is higher or lower, respectively . . .

[NOTE: I think there is a way to keep the tremolo rate constant for computed notes, but doing this requires knowing that tremolo is used, and probably requires a lot more complex set of algorithms, which probably would take too long to run. Anything is possible, but whether something is practical is another matter . . . ]

In the same way that notes can be sampled chromatically or non-chromatically, this also happens with dynamics, articulations, and other playing styles and techniques, where yet another set of variables includes the type of microphones, room acoustics, digitizing techniques, and so forth . . .

Some instruments are sampled at various dynamic levels (pianissimo, forte, and so forth), and some instruments are sampled in every possible articulation, but this is nearly never the case, because it takes a lot of time and requires a lot of resources (musicians, instruments, studios, recording engineers, digitizers, and so forth), where the typical reality is that dynamic levels are computed, but primarily by raising or lowering the MIDI volume level, which ranges from 0 to 127 . . .

Doing dynamics this way can work nicely, but it s not the same as having a musician play the instrument at the specific volume level, because when trained musicians play at different dynamic or volume levels, they usually change the way they play, and this in turn affects pitch, attack and release times, texture, and so forth . . .

(3) Volume level affects loudness, pitch, texture, and other stuff . . .

As shown in the updated Fletcher-Munson Curves (see below), the perception of pitch and texture depends on the volume level . . .



Equal Loudness Curves (wikipedia)

(4) You need a calibrated full-range studio monitor system running from 10-Hz to 20,000-Hz at 85 db SPL to hear everything accurately, and you need a sonically neutral sound isolation studio . . .

[NOTE: This is explained in great detail in my ongoing topic in the IK Multimedia FORUM . . . ]

The Fabulous Affordable Studio Monitor System Project (IK Multimedia FORUM)

THOUGHTS

I hear the problem, but it is not caused by any single factor . . .

Instead, it is caused by a combination of factors, and finding a solution will be difficult, but I think there is a solution, and the primary reason is that I think trained musicians playing real instruments can play the pieces you are doing correctly, but part of their training includes listening to each other and making what probably are very subtle adjustments in the way they are playing their instruments in real-time on the fly . . .

As explained (see above), the differences in the various notes, dynamics, playing styles, and so forth are very small in the bass clef register, and it is unlikely that the NOTION 4 native virtual instruments are sampled in the required detail to do what you want to do; and intuitively I am not certain that there is any commercially available sampled sound library that has all the required samples . . .

However, if you play Harmonium and Tuba, then you can create your own sampled sounds libraries and use them with a suitable VSTi virtual instrument engine like Kontakt 5 (Native Instruments) or MachFive 3 (MOTU), but it will take a while, since you will need to sample every note in every dynamic level per every articulation, and so forth and so on, which certainly can map to hundreds of thousands of samples, all of which need to be recorded professionally and then digitized . . .

You need a calibrated full-range studio monitor system and a sonically neutral sound isolation studio . . .

In one of the experiments, I created a NOTION 4 score and entered the first six or so measures of Euphonium, and it was obvious what was happening, so I did a few more experiments to isolate the primary problem, and this is the reason for providing the detailed background (see above), because there are several things happening, each of which affects something . . .

Among other things, I hear hiss and noise, as well as comb-filtering, and these are made more prominent by specifying lower dynamics ("pp", "p", and so forth), since in another experiment I did not use any dynamics or articulations, and the problem was less prominent, although obviously present . . .

If I do a few more experiments, I can determine whether NOTION 4 is changing the dynamics by varying the MIDI volume level (0 to 127), but I am reasonably certain that this is what NOTION 4 is doing, and when this is done, it changes other characteristics of the specifically computed sounds for each note, which is one of the reasons that I generally avoid all dynamics and articulations whenever possible, since I do not want any computed notes, and in most instances the only way to get the actual sample for each note is to specify only the note with no other music notation, although there are sampled sound libraries which, for example, have a French horn sampled being played pianissimo and have another French Horn sampled being played forte, so if you use the pianissimo French horn, then you are using the sample rather than something which is computed using an arbitrary algorithm, and the same thing happens when you specify the forte French Horn, but doing this generally requires switching channels rather than using "p" or "f", unless there are corresponding rules that do the same thing as switching channels, as contrasted to simply lowering or raising the MIDI volume level or whatever . . .

Explaining the more relevant aspects of some of what I am hearing might require 10 more posts as long as this post, but one way to understand some of the more subtle acoustic physics is to watch this YouTube video of a presentation made by Ethan Winer at the Audio Engineering Society (AES) convention in 2009, where the most relevant information is in the last half of the video, which is fabulous . . .

Audio Myths Workshop (Ethan Winer et al.) -- AES Show 2009 - YouTube video

Fabulous!

~ ~ ~ Continued in the next post ~ ~ ~

[Surfwhammy]

~ ~ ~ Continued from the previous post ~ ~ ~

Explained another way, one might suggest that general wisdom indicates that if someone is not sufficiently smart to be a human medical doctor, then they might be sufficiently smart to be a veterinary medical doctor, but the reality actually is the exact opposite, at least with respect to meeting the various requirements for successful enrollment in school, where it is considerably more difficult to be admitted to veterinary medical school than to be admitted to human medical school . . .

The similar analogy, metaphor, or simile in music is that if one is not sufficiently smart to play violin or trumpet, then there are viola, cello, string bass, French horn, harmonium, tuba, and sousaphone, but this also is not the case, although it can take a bit of information gathering and pondering to understand, which is the same with the different registers for singing . . .

Everything is complex, but the ways in which each thing are complex differs considerably, and as I discovered when I temporarily went Country Western for a few months and decided to start singing bass, it is not easy to sing bass, because the difference in pitch between notes is smaller, which in turn requires considerably more precision and in turn requires more training and practicing, as is the case with playing string bass or electric bass as contrasted to rhythm guitar and lead guitar . . .

Generally when playing lead guitar in some genres, the overall goal is to play as many notes as rapidly as possible, but in great contrast the overall goal for string bass or electric bass is to provide a typically pulsating deep bass foundation or platform upon which everything else rides as if it were a magic carpet, where the primary focus is on the singer and melody, except during lead guitar solos when the the primary focus is on the lead guitar player and melody; and the fact of the matter is that a skilled bassist can make nearly any singer or lead guitar player sound good, especially when the drummer is vastly skilled and works cooperatively with the bassist . . .

Yet another way to explain my perspective on what I think you are are trying to do is that you want to make essentially electron microscope level adjustments, but the problem is that the available sound samples, engine, and so forth at best allow only macroscopic control . . .

For example, consider the volume level or dynamics of a single bass or baritone instrument which was sampled in a playing style and dynamic level which has a good signal-to-noise ratio that results in noise and other undesired stuff not being heard . . .

Everything is fine so long as you avoid using dynamics like pianissimo, forte, and so forth, but when you use pianissimo and this maps to the MIDI volume level being lowered, then this affects the signal-to-noise ratio and typically reveals more of the background noise, but it also changes the perception of tone, harmonics, overtones, and so forth, and since the NOTION native instruments are recorded and digitized with generous headroom, which is fine with me, this actually amplifies the problem . . .

As explained in my previous post, one solution is to use samples which specifically were played by skilled musicians in the specific playing style, dynamic level, and so forth, but if you allow (a) 10 dynamic levels ranging from "pppp" to "ffff", including "mp" and "mf"; (b) 100 articulations: and (c) 46 notes for Euphonium, then this maps to 46,000 samples; and if it takes 1 hour for each sample to do the recording and digitizing, this maps to 46,000 hours or approximately 16 years of work for one person who works 8 hours each day or 1 year of work for 16 people . . .

Some of the individual instrument sampled sound libraries from Vienna Symphonic Library (VSL) are chromatically sampled in a virtual festival of dynamics and articulations, but they are expensive relative to the budget here in the sound isolation studio, although I am intrigued by them, and the prices are not so expensive when you only need the vast detail for a few specific instruments, where for example the extensively sampled single instrument Euphonium library costs approximately $220 (US), and it has 7,621 samples . . .

Euphonium (Vienna Symphonic Library)

Vienna Symphonic Library also has four individual Tubas and a Cimbasso, and each extensively sampled single instrument library costs approximately $135 (US) to $165 (US), where for example the Bass Tuba has 12,035 samples and costs approximately $165 (US) . . .

Bass Tuba (Vienna Symphonic Library)

MORE THOUGHTS

I played string bass for a long time and then switched to electric bass, followed sometime later by switching to rhythm guitar and then to lead guitar, and this tends to give me an appreciation of the intricacies involved in playing bass instruments as contrasted to higher register instruments, and you should be able to do what you want to do, but expecting to be able to fine-tune instrument samples might not be so realistic when the particular sampled sounds are more general than specific and exhaustive . . .

I think the Vienna Symphonic Library single instrument collections will make this easier, but they are a bit expensive, so other possibilities include Kontakt 5, which has a nice collection of orchestral samples done by Vienna Symphonic Library, and sometime over the next two months SampleTank 3, where you can use the Miroslav Philharmonic sampled sound libraries and the new sampled sound libraries for SampleTank 3. MOTU MachFive 3 and Symphonic Instrument (MOTU) also are possibilities . . .

However, Kontakt 5 and MachFive 3 do not have Harmonium, and the MachFive 3 Tuba is not good. The Kontakt 4 Tuba is pretty good, though . . .

Miroslav Philharmonik has Tuba but not Euphonium, so not much help there, either . . .

[NOTE: Technically an Euphonium is a baritone instrument, but I make an effort to keep everything as simple as possible here in the sound isolation studio, and instruments that appear on any type of bass staff, including cello, are "bass" in my universe, and an Euphonium plays bass notes, which is sufficient for my definition . . . ]

Summarizing, I think the best current strategy is to get the Vienna Symphonic Library highly sampled individual Euphonium instrument collection--since Vienna Symphonic Library appears to be the only company that has a highly sampled single instrument Euphonium collection--which might also be the case for Tuba, depending on what how extensively you need to be able to control the various playing parameters, styles, and techniques . . .

There is a lot of stuff happening, and all of it interacts in primarily subtle but important ways, which makes fine-tuning everything a bit of a challenge, for sure . . .

For sure!

[Admin]

It sounds like the "distortion" is recorded into the sample. I think that it is spit in the horn itself. If it is just in that sample, it should be covered up by "realism" in the entirety of the score.

-Admin

[tubatimberinger]

after a few more listens; It's what we [brass players] call a double-buzz. It's most likely from fatigue. Mr. Harrild is a beast to be sure, but sampling on tuba all day then switching to Euph is enough to make the best tired. It's not a huge problem as I use Notion for mock ups not a final recording usually. But Hopefully it can be addressed in the future.

thanks

-t

[Admin]

This almost certainly will not be addressed within the Notion sounds. Notion provides the capability for much more expensive VST's if users are looking for more extraordinary playback. We offer a great sounding library for the money, but much more effort has been put into software that is samples only. It is also reflected in the cost.

I am sorry that the G sounds bad in that Dynamic range. If Notion ever goes into making its own new sound libraries, users will be among the first to know. I am sorry that is the best answer that we are able to give at this time.

-Admin