Composers' Forum

Music Composers Unite!

This is a work that I began as a composition exercise, in preparation for writing a fugal variation to Gerd Prengel’s Beethoven sketch. It was originally to be a short, self-contained fugue based on the D-S-C-H motif. But the music didn’t seem to want to cooperate with that plan — it grew into something much larger, that was not a single fugue, but 3 fugal expositions interspersed over several — eventually, 5 — mostly non-fugal sections, each with its own distinct character, separated by shorter interludes and bridge passages (called Episodes in the score). At the moment I’m calling it Fugal Variations, though the title is subject to change.

Unlike my String Quartet, this work is solidly tonal, though it is tonally restless and never stays in one key for very long. There are also fleeting moments of bitonality in Variation 2, and again in Variation 4. The piece is my creative reaction to the COVID-19 pandemic, both in style and in expression — with modern civilization under threat, I’ve felt an urge both to reengage with old styles, and to write something that expresses the search for hope and deliverance in a time of darkness.

Fair warning: it’s rather long, nearly 23 minutes (sorry!) and is brooding and elegiac for much of its length. But Variation 4 develops a faster, flowing motion and builds to a bracing climax. Variation 5 is by far the longest and is the closest to being a fully worked-out fugue. It begins in deep sadness and remains sombre for over half its length. But then, an episode begins in which a song of solace and hope climbs gradually up the circle of fifths and into the highest registers. Fugal textures then return with renewed vigor, and a short coda ends the piece on a note of stoic acceptance and defiant affirmation.

I hope that you find it engaging, and ultimately uplifting despite the sombre moods. As always, constructive criticism is welcome. I think that as of this writing (7 September), I have made all the revisions to the piece that I have felt necessary, though I am open to suggestions for improvement. This is only my second completed work since discovering the wonders of notation software, and I know that I have much to learn.

The piece was written in Sibelius and rendered using NotePerformer, running under the latest version of Sibelius.

Update 2 August 2020: I have revised the ending slightly and the version on SoundCloud is now out of date. A demo of the current version is at the Google Drive link below. The volume level on Drive is somewhat quiet, so I recommend either turning up the volume or downloading the file and listening on your computer.

Update 5 August 2020: I revised the instrumentation and registration in bars 668-680. I have slight reservations about one change that I made, raising the entry of the subject in the 1st violins an octave higher; but other than that one point, this should be the final version.

Update 10 August 2020: I replaced the rendering with one that has nearly as little distortion, and is more faithful to the score. There were a couple of places in the previous rendering where NotePerformer quite noticeably rushed a pair of eighth notes. There was also one tiny change to the phrasing -- in bar 658, for the record.

Update 21 August 2020: No changes to the score, but an improved rendering that eliminates another pesky flaw where a pair of eighth notes was rushed. There is also less distortion in this one.

Update 7 September 2020: Some adjustments to tempo and phrasing, and a bar of silence before the final cadence in Episode 5 (bar 441). Unfortunately, in this rendering the pair of eighth notes that I mentioned on 21 August (cellos, bar 46) is once again rushed, and Sibelius is doing this consistently with this version of the score. There is also a sudden "explosive" swell on the last eighth note of bar 685. NP clearly does strange things sometimes, beyond the composer's control. This is otherwise one of the best renderings I've gotten yet, though.

Update 18 October 2020: Applied many small changes suggested by a local violist who has studied the score. Mostly the changes were to phrasing and tempo; also a couple of brief pauses were lengthened. Variation 5 is now taken a hair more slowly. I've given up on trying to control the number of players in each section because the NotePerformer string samples seem to sound best either solo, or in the default string orchestra section sizes. So this rendering is for full string orchestra, and there is little noticeable distortion or "swelling" anywhere. The downside is slightly less clarity of the various lines in the denser contrapuntal sections.

https://drive.google.com/file/d/1L_Jr6hVhS7DrglCF0Q-EM-9wS_xt7kCF/v...

The current version of the score is attached.

Views: 472

Attachments:

Reply to This

Replies to This Discussion

What happened here is weirder... in Episode 4, I introduced a light duet between the violin sections (bars 289-297) and had no intention of ever using it again. The 5-note figure in the 1st violins then begins to invade the development in Variation 4 (bars 335-6) and plays an increasingly important role in Episode 5 and in the final "Variation". The thing is, until I was 2/3 of the way through Variation 5 I thought the figure first appeared in bars 335-6 - I had *completely* not made the connection with the duet in Episode 4!

Contrast that with the downward turn in the 2nd violins in the final cadence. I got that idea when I was composing Episode 3 and knew it was a bit of a cliche, so I tried to find a way to make it a basic motif of the piece. So I planted it there and used it in many different contexts throughout the remainder of the piece. If I was really good I'd go back and work it in somewhere near the beginning, but I'm not that ambitious.

Episode 3 was supposed to sound ironic, even a bit sardonic. It's the only place in the piece where you hear pizzicato - I wanted that to underline the irony. It may be a bit too pleasant to come across that way though - and it does end in a melancholy sort of way.

BTW, if you do revisit my piece again, please give me at least a couple of days to get my latest (and likely final) revision up there. After using my sound chip intensively as I did this morning, I need to give it a rest and not even open Sibelius for a while in order to get a distortion-free rendering. I don't understand why this is, since Sibelius doesn't (AFAIK) use the sound chip when rendering to a file. But the pattern is fairly consistent. (It also seems to be affected by ambient temperature - the cooler the better.)



H. S. Teoh said:

I confess to doing the same in Exuberance. :-P My fugue skills were not up to snuff to get through several passages, so I just took the liberty to introduce new material, which I later on then retroactively reused to make it look like it was planned all along (it wasn't). Still, I'm quite pleased with the result, and am not overly concerned about a strict, technical interpretation of fugue.

[...]

Finally, I hope that the major key Episode 3 wasn't TOO subtle - it really WAS supposed to stick out like a sore thumb! Sort of like someone showing up for a funeral in a bright white Sunday Communion dress. It even ties a pretty little IV-V-I bow around itself, not once but twice... the only place in the piece where that happens.

[...]

Well, it did stand out, but I guess it depends on how sore of a thumb it is that you're thinking about... I meant it more in the way that it didn't sound out-of-place or awkwardly inserted into an otherwise dark piece; but certainly within the context of the piece itself it presented quite a contrast.

In any case, I think I'll need to listen to this again in order to get a better idea on its details -- a 20-min piece is far too long to completely grasp with merely 2 listens.

Yeah sardonic is rather hard to pull off. It has to straddle that delicate balance between being overwrought and too convincing, and not going far enough so people missed the point.  One way I might do it is to (ab)use bitonality by having one theme "fight" against another as if the main material is "protesting" the happy material. But again, not easy to pull off in a convincing way that doesn't turn into an epic struggle instead of something sardonic. Or another way is to play a theme in major key but stubbornly harmonize it in minor key simultaneously -- I did that at the end of my Threnody to make it sound bittersweet... but then that might not work for sardony.  I dunno. Just throwing out wild ideas here.

As for sound quality depending on temperature: sounds like Sibelius is using your chip for rendering somehow, maybe routing some instruments through it to take advantage of on-chip sound banks? Alternatively, it could be that too high of a temperature causes something to malfunction and soak up system resources, so Sibelius suddenly has a lot less CPU/memory to work with and falls back to a lower-quality rendition.  Such is the curse of the deeply-rooted real-time design of modern DAWs and VST's.  I've been wanting to write an alternative batch-based renderer that is not constrained by real-time requirements and can take its sweet time to produce as good a quality as I wish, regardless of my actual hardware capabilities. But it's likely too ambitious of a project for one person to pull off...

Sardonic is probably a little too strong a word for what I intended. Ironic is closer to the mark I think. Something a little along the lines of what I was getting at is the 2nd movement of Mahler's 9th, where the music is basically a Landler, but the scoring makes it clear that this isn't a straight country dance but more like a parody of one. I'd hoped that the out-of-place simplicity of the harmony, the symmetry of the phrasing, and the timing of the pizzicati would make my Episode 3 a little too pat to be taken as just a pleasant interlude. I was probably trying too hard to be subtle and so the effect didn't quite come across.

The thing about the distortion is that it happens ONLY with NotePerformer as the playback configuration. So I don't know how much of it is Sibelius and how much is NP. There is definitely something weird in the NP string samples that gives the sound a harsh quality sometimes - but not consistently at the same places in any given score. Sometimes it sounds like actual clipping, but if you play the audio file in Audacity, the bar graph indicators never get even get close to the yellow, much less red. At other times, the end of each note has an unnatural swishing - or even a "breathing" - sound that can be distracting and annoying. I don't know, not everyone hears it in any given rendering. I guess I'm especially sensitive to it - it drives me up a wall sometimes.

The distortion happens mostly at high velocities - and so I wrote all the dynamics in my score one level lower than intended. That makes the rendering sound softer than it should - and it's why I tell listeners to turn up the volume a little. The ending is written f but should really be ff!

In that case, it's probably bad samples / clipping inside NP itself. Though that doesn't explain why it's dependent on temperature.  The only logical explanation I can think of for the temperature dependence is that something in your hardware kicks in and throttles CPU speed, etc., in order to keep temperature in check. As a result, the usual CPU bandwidth that should be available to you isn't, and in order to compensate, NP scales back the quality so that it can still manage to crank out that much audio within the given timeframe.

As I said, this is a consequence of the real-time design that most DAWs are designed with these days. I personally prefer working with batch-based processes, which have no such constraints and can therefore guarantee the same quality no matter what your CPU/memory load is.  Of course, that comes with the big caveat that most modern VST's are not compatible with such a mode of operation, so I only have poor-quality samples to work with.   One of those lose-lose situations. :-(

That makes sense... but the confounding factor here is the design of NP itself, which does NOT play through a score the same way every time, but introduces small variations to simulate the effect of a human performance. So the attacks are slightly different, possibly the velocities as well, and that might determine whether the samples containing distortion are used at any given time. How it interacts with Sibelius, and how the two are affected by CPU speed or other hardware factors, I have not the remotest idea. I don't think Arne Wallander would reveal that info even if I asked him, and so far I haven't, not straightaway, anyway.

I tried doing a rendering this evening immediately after rebooting my machine, since the battery temp was relatively cool (29.5*) and the last really good rendering I got was under those conditions. After the rendering was finished the battery temp had risen to 31.8* and the audio was TERRIBLE - lots of distortion that got worse as the piece progressed. I think I will go with this revision though, at least provisionally - the greater variety in instrumentation and the dialogue between the violin sections in bars 688-676 are definite pluses. I have some qualms about raising the subject entry in bar 676 an octave higher as the effect is somewhat jarring, but it definitely contrasts better with the turbulent pseudo-stretto that follows. I'll put a rendering up here when I can get a decent one and see if Gregorio or someone who hasn't heard the piece yet finds it problematic.

This is very strange indeed, that quality will degrade as temperature rises. I can think of any number of weird circumstances that might give rise to this strange effect, but it's really hard to tell for sure without investigating your machine myself.

Playing through differently each time is not necessarily a bad thing; in fact, one might argue it's essential to make an electronic performance even remotely believable. I was in fact planning to do the same in my ambitious dream project of making a batch-oriented DAW/VST host. But yeah, it can make for very annoying troubleshooting if you have bad samples that only show up unpredictably.  The usual way of dealing with this in techie circles is to let the user customize a "random seed": a fixed initial number that the performer software uses to generate its random numbers so that each time round the same sequence of random numbers are generated.  This lets you reliably reproduce problematic spots. And if a particular performance isn't good enough for whatever reason, you can enter a different number to get a different series of random numbers, i.e., get a different performance. But I don't know if NP has such a feature.

Another possibility, if this problem continues to trouble you, is to render each track separately, then use audio postprocessing software to merge it together at the end, after each track has been fully rendered. This reduces the resources required for the performance, which hopefully will avoid triggering the problem samples.  But it does mean more work on your end.

The "random seed" idea is one I actually broached with Arne at one point, though over a different NP issue: one aspect of their "humanisation" algorithms is to introduce small timing shifts at random places in the strings, particularly during legato passages. This was supposed to make the sound of a string section more believable, since the sound of massed strings is actually built up from solo string samples. But they didn't suppress the random timing shifts in settings where you really hear individual instruments, like a string quartet. The result is that the instruments get out of sync with each other at unpredictable times. In some music (like my unfinished - actually barely started - Beethoven sketch variations), there are enough pauses that it is easy enough to get a decent rendering by splicing together excerpts of different playbacks. But if you try to splice other than at a pause, you get very audible clicks that no software that I have can do a good job of eliminating. So with Fugal Variations that was not an option - the reason I switched to a string orchestra setting in a piece I originally intended for string quartet.

But I digress. It was in the context of timing flubs that I broached the subject of random seeds with Arne, because the timing problem occurs only rarely in NP under Finale. I wondered if the difference could be in the way a random number generator is seeded in the two notation programs. He basically said there is no difference, and did not suggest a way to alter the randomness, so I took it from that exchange that there was no such way in NP.

As to rendering each track separately - again, because of slight timing differences in each playback - slight rubatos that are coordinated between instrument sections in a single playback, but would occur randomly and to random degrees in each track if the tracks were rendered separately - I don't think this is a viable option in NP.

The score and a link to a rendering of the latest revision is now in the OP. The rendering has some of the "breathing" effect in a few places, but little to no "clipping"-type distortion. (Hopefully no one will notice the "breathing" effect.)

If you've listened to the piece already and would like to hear just the changed passage, best to start listening around 19:00. The changes start at 19:41.

Hi Liz.  This is a beautiful piece!  Ethereal, and very soulful.  I found a certain 'haunting' sense about it (not foreboding, so much). You have a wonderful way of shifting harmony, with many false cadences - which is very effective here.  I was reminded a bit of Monteverdi - (his madrigals)- and a recurring phrase  that referenced a little part Barber Adagio, but probably noticed that due to the mood of your piece. 

My one little quibble might be that you may have introduced a more contrasting section - with tempo ( a faster section)- or further rhythmical subdivisions.. I was wondering (while listening) if you might play the theme, or some variation at double-time, or some such.  But I just went with it, and found it quite satisfying despite that.  On another listening, I may feel differently, and appreciate the Long musical line - contained (or without a big shift), as you have done.

I just want to mention a word about the phrasing.  In your score, it seems the phrasing relates to the utterance of compositional sentences, so theoretically, that is good and necessary to see. But because many of the phrases are quite long, (especially at this tempo) as Im sure you are aware, that there will have to be changes in bow direction midstream. One way of offering a distinction between the 2 ideas is to make dotted slurs for the longer phrases, while also having non dotted slurs indicate where you would have the bow change directions.  (Or you could write actual bowing marks).  

Also, just a little thing, but I noticed 32nd and 64th rests after a phrase.  This is quite precise, but maybe cumbersome to read. Wouldn't the end of a phrase also imply just that kind of short rest? Just a question.

I really enjoyed your piece!  Thanks for posting.

Hi Gregorio,

Thank you so much for the nice comments! I'm glad you enjoyed my piece.

I'm not sure which phrase evokes the Barber Adagio for you - there is certainly no conscious quotation of it or any other work here. Could it possibly be the descending turn that is introduced in Episode 3? If so, I made that a recurring motif so that it could appear at the very end without being heard as just a cliche (which it sort of is).

I'm happy that you picked up on the early Baroque or late Renaissance influence - I think most of that actually comes to me through Holmboe as I have not listened too much in recent years to music from that period. But the harmonic language is certainly not modern except in a few spots.

The 32nd and 64th rests surely won't be present in the final prepared score, if I ever get around to making one - they were designed strictly so that Sibelius would play back those phrases in the way I wanted. I would have used a breath mark, except that in Sibelius, a breath mark halves the length of the preceding note. And without the notated rest, the break would be barely noticeable.

As to the slurs, they are intended to indicate legato. You are right that in some places they are too long and should be broken to accommodate a change of bow direction. I've been doing that slowly, but it is a huge task because of the length of the piece. In places I think there is just too much legato and it will likely be scaled back eventually. I have never used dotted slurs though and am not sure how they work...

As to a faster section - I made the decision early on that the note values would never be shorter than eighth notes, and I wanted the tempo to be pretty much the same throughout, except for slight fluctuations. As a challenge, I limited myself to textural, dynamic, and harmonic means of introducing variety. So Episode 4 and Variation 4 together form the only "faster" section in the piece, a compact and non-fugal development section, building toward the climax that begins in bar 382. (I hope that climax, with its momentary bitonality, came across as bracing and not threatening-sounding; those two sections were meant to give relief from the predominantly dark mood.)

Once again, many thanks for listening and for your feedback!

Liz Atems said:

[...]

The 32nd and 64th rests surely won't be present in the final prepared score, if I ever get around to making one - they were designed strictly so that Sibelius would play back those phrases in the way I wanted. I would have used a breath mark, except that in Sibelius, a breath mark halves the length of the preceding note. And without the notated rest, the break would be barely noticeable.

[...]

After having done this sort of thing many, many times, I have come to the conclusion that with the current state of technology, the only sane approach is to have two scores for every project: one for the polished score to be seen by players, conductors, and whoever wishes to study the score (which I call the "real" score), and the other purely for the computer output, a no-holds-barred copy of the "real" score where you pull every last trick there is to pull in order to get the computer to do what you want, no matter if it entails doing stuff that would make the score unreadable, illogical, or otherwise objectionable.  These two copies would, of course, need to be kept in sync, which is a lot of work (that technically computers should be able to automate, but alas, we're not there yet), but they are to be kept strictly separate. As in, they are two completely independent files, so that they don't "pollute" each other.  I find that this is pretty much the only way to produce both a score that doesn't make your eyes bleed and a computer-generated audio file that doesn't make your ears bleed.

Actually, for the second "score", I'd even say that if you're comfortable with working directly with a piano roll, go straight for that and don't even bother writing a "score" for it. Just work directly with your DAW / VST's / whatever, and do whatever it takes to make the audio really good.  For the "real" score, or printed score, throw all audio quality considerations out the window, use notation software and strictly focus on notational concerns.

I've been grappling with how to unify these two sides of the same work, but so far I have not found a good solution yet.  In my own workflow I do primarily work from score, using Lilypond to produce really good-looking scores, and its configurability to insert "invisible" directions that only influence the audio but have no visible effect (and conversely, tag certain score marks as only applying to the visible score but having no effect on audio -- to prevent the computer from misinterpreting something and screwing up the audio). Sometimes I go even as far as writing two versions of the same passage, one purely for the printed score, and one purely for the audio. Nevertheless, in spite of all that, Lilypond's audio output is anything but satisfactory, and I've been looking to writing my own tools for extracting the notes and post-processing them into something more human-like before I feed it to the audio output modules. But with what limited time I have, this hasn't gotten off the drawing board yet. :-(

Actually, an experience today convinced me of what you say here, which others have suggested as well. I had an hour long Zoom meeting with a local violist about my string quartet that I posted here (sometime in February I think). There too, there were a few "oddities" in my score that she couldn't understand, that were hacks I used to make Sibelius play the thing back as I wanted, but which would make no sense to a human reader. Unfortunately I would have to maintain the two scores manually using Sibelius, I don't have a DAW or any other way to produce the audio other than Sibelius playback. I did start investigating a couple of DAWs earlier in the summer, Logic Pro X and Reaper, but didn't get very far with them because of self-imposed pressure to finish Fugal Variations. But the learning curve with any DAW is apparently steep, so I won't be composing that way any time soon. I would certainly love to be able to make music with a good, professional-quality sample library as I HATE the harsh NP massed string sound and the timing idiocy that they carry over to solo string instruments for no good reason that I can think of. It's just not practical right now.

Interesting about Lilypond though, that's a program I'd never heard of. Seems versatile, maybe I should look into it...

BTW, the violist said that the Quartet was her favorite of the three works I showed her, which included Fugal Variations, the opposite reaction to people here it seems, and to my own appraisal of the works. We have another Zoom meeting scheduled for next week, maybe I'll find out why then.

H. S. Teoh said:

After having done this sort of thing many, many times, I have come to the conclusion that with the current state of technology, the only sane approach is to have two scores for every project: one for the polished score to be seen by players, conductors, and whoever wishes to study the score (which I call the "real" score), and the other purely for the computer output, a no-holds-barred copy of the "real" score where you pull every last trick there is to pull in order to get the computer to do what you want, no matter if it entails doing stuff that would make the score unreadable, illogical, or otherwise objectionable.  These two copies would, of course, need to be kept in sync, which is a lot of work (that technically computers should be able to automate, but alas, we're not there yet), but they are to be kept strictly separate. As in, they are two completely independent files, so that they don't "pollute" each other.  I find that this is pretty much the only way to produce both a score that doesn't make your eyes bleed and a computer-generated audio file that doesn't make your ears bleed.

[...]

Reply to Discussion

RSS

Sign up info

Read before you sign up to find out what the requirements are!

Store

© 2020   Created by Gav Brown.   Powered by

Badges  |  Report an Issue  |  Terms of Service