Welche Auswirkungen haben die Einstellungen?

  • Mal in die Runde gefragt, gibt es zu den einzelnen Punkten Beschreibungen was diese bewirken und welche Auswirkungen diese haben? Vielleicht auch als Sammlung für einen extra Erklärungsthread?!

    Zufällige Ansprache ist klar, Sample size und Stereo auch. Attack Loading und Release Loading kann ich mir nur was drunter vorstellen. Bei der Sound Engine hört es aber auf mit meinem Wissen.

    Ich finde auch leider nichts dazu.

    pasted-from-clipboard.png

    Melodeum.de - Wissenswertes zu Harmonium

  • Der Vollständigkeit halber hier noch reinkopiert (ist Englisch):

    • Active polyphony management
      When polyphony reaches 3/4 of the current maximum value, release samples (reverberation tails) are faded out in order to conserve polyphony.
    • Release Sample Scaling
      For those sample sets with long reverberation tails ("wet"), enabling this option will fade out the release sample if the note is played staccato. It will have little to no effect on "dry" sample sets.
    • Randomize Pipe Speaking
      This will apply a very small random detuning before each pipe begins to speak. The detuning is small enough so that the ear can't appreciate the tuning difference and large enough so that it approximates the tiny interferences (due to physical location) that occur when playing multiple real pipes in an organ.
    • Load last file at startup
      This will trigger the automatic load of the last used sample set.
    • Interpolation
      This dropdown controls the method used to interpolate the waveform when resampling the samples.
      • Polyphase
        This method gives better audio results at the expense of higher CPU load.
      • Linear
        This method lowers CPU load at he expense of audio quality. It is the only usable option when lossless compression is enabled, as polyphase interpolation is not yet implemented with lossless compression.

      The assertions about audio quality should be used as guidelines only. As a rule of thumb, polyphase gets better treble, linear gets better bass, and this is most noticeable in sample sets that use interpolation extensively (e.g. retuning a detuned organ). Audio quality is highly subjective, so the user is strongly encouraged to listen for himself and retain the setting that suits him better.

    • Number of CPU cores
      This number states how many threads GrandOrgue creates to spread the load. Cores number is the recommended choice for computers without Hyperthreading. Less doesn't use the whole computer. More wastes resources while managing overhead. With Hyperthreading enabled, the CPU load seems more evenly spread among virtual cores.
    • Workload distribution
      This setting has the same role as Number of CPU cores, and is dedicated to spliting the workload. The same value as the number of CPU cores is recommended.
    • Cores used at loadtime
      This number states how many threads GrandOrgue creates to load samples in memory. It has NO effect when loadind samples from cache. Higher speed-up loading while reducing the available memory for samples. A zero (0) value means classic load.
    • Recorder WAV format
      This selects the format used by the Audio recorder. The supported formats are: 8 bit, 16 bit, 24 bit PCM and 32 bit IEEE float.
    • Record stereo downmix
      This textbox enables GrandOrgue to record multi-channel audio in simple stereo. This option is disabled by default, as it increases CPU usage even if the audio recorder is off. When stereo downmis is disabled, all audio channels are recorded in the WAV file.
    • Paths frame
      Settings store
      This selects the directory where preset files are stored. The Browse button opens a file browser where the desired directory can be chosen.
    • Cache store
      This selects the directory where cache files are stored. The Browse button opens a file browser where the desired directory can be chosen.
    • Sample loading frame
      Lossless compression
      Uses a lossless algorithm (i.e. the original samples can be perfectly reconstructed) so that RAM requirements are lessened. Particularly useful for large sample sets.
    • Load stereo samples
      Selects whether stereo samples are loaded in Mono or Stereo If set to Mono, all samples are loaded in Mono. When compression is disabled, this will save 50%, but when compression is enabled, much less will be saved because the compression is much less effective on monoaural inputs. This setting can be set to don't load, if you want GrandOrgue to avoid loading any samples. Via Organ Settings, it is still possible to load only specific parts of the organ.
    • Sample size
      Selects whether the samples are loaded with 8 or 12 or 16 or 24 bits precision.
    • Loop loading
      Selects which loop(s) are loaded if the samples have multiple loops.
      • First loop: Load only the first loop.
      • Longest loop: Load only the longest loop found in the sample. Longer loops usually feel more lively.
      • All loops: Load all loops found in the sample. While playing, the loops are cycled using a round-robin scheme.
      • Memory: First loop less than Longest loop less than All loops
      • Load time: Increases with the amount of data loaded
      • Polyphony: No impact
    • Attack loading
      The sample set creator can provide multiple separate attack files.This selects which attack(s) are loaded.
      • All: Load all attack files.
      • Single attack: Load only the "best" provided file.
      • Memory: Single attack less than All Polyphony
      • No impact: Load time Increases with the amount of data loaded
    • Release loading
      The sample set creator can provide multiple separate release files.This selects which release(s) are loaded.
      • All: Load all release files. The sound engine selects which release to play according to how much time the sustain sample was held. For example, it usually selects the shortest release sample while playing staccato.
      • Single release: Load only the "best" provided release file.
      • Memory: Same as Attack loading Polyphony
      • No impact: Load time Same as Attack loading
    • Memory limit
      Chooses the size of the memory cache where samples are loaded. This is especially useful when running the 32 bits Windows build on 64 bits platforms to avoid a crash while loading a big sample set. If this parameter is set to zero, it tries to load as much data as possible. Achtung: In this case, GrandOrgue keeps loading until the OS fails to provide memory. It can provide large amounts of swap space with expectable adverse impact on performances.
    • Sound output frame
      Sample rate
      Selects the output sample rate. Allowed values are 44100, 48000 and 96000 Hz. The samplerate should match both the configured samplerate in the audio interface and the samplerate of the recorded samples to avoid resampling. If not, resampling can occur in all layers of the audio stack, and audio quality can suffer.
    • Samples per buffer
      Allowed values range from 16 to 1024 by increment of 16. This parameter sets the output buffer size. Larger values usually reduce sound artifacts at the expense of latency.
    • Cache frame
      Compress cache
      Selects whether the disk cache must be compressed when created or updated.
      • Memory: No impact
      • Polyphony: No impact
      • Load time: Loading from an uncompressed cache is I/O bound, while loading from a compressed cache requires more CPU. With a slow disk + fast CPU, compressed might be better. This feature is closely related to the hardware capacities, so the user is encouraged to test for himself and retain the best setting for his computer.
    • Automatically manage cache
      Selects whether the cache must be automatically created or updated when the sample set is loaded.
    • Perform strict ODF
      Organ Definition File (ODF) syntax has much changed since the days of Hauptwerk™ version 1. Some Hauptwerk™ 1 keywords are not used by GrandOrgue and are reported as warning in the pop-up message window. Introducing ODF warnings to GrandOrgue is a two-step development process (This checkbox enables strict ODF mode.):
      • 1. A new warning is added to the ODF parsing utilities. It is displayed only in strict ODF mode, to give samplesets designers time to remove the new warning from their Organ Definition Files, without disturbing users.
      • 2. After a suitable number of months has elapsed, the warning is moved to normal mode and is always displayed.
    • Anmerkung
      Sampleset designers are strongly encouraged to update their Organ Definition Files in order to remove warnings and improve user experience. Likewise, GrandOrgue users are strongly encouraged to report new warnings to the sampleset designer.
  • So wie ich es verstehe wäre es wohl am besten wenn man ein großes Set verkleinern muss langsam anzunähern. Ich würde also zuerst einmal schauen 24 Bit zu reduzieren auf minimal 16 Bit. Danach die Attack und Release auf Single. Wenn das nicht reicht die Loops.

    Was ich spannend finde ist, dass der Speicherverbrauch scheinbar durch Mono quasi halbiert wird. Nun könnte man ja auf die Idee kommen einfach 24/16 Bit Mono zu wählen. Da stellt sich nur die Frage was akustisch betrachtet besser wäre, 24/16 Bit Stereo mit Loops und allem reduziert oder Mono mit allem aktiviert...

    Was meint ihr zum letzten Punkt Mono Komplett oder Stereo reduziert?

    Melodeum.de - Wissenswertes zu Harmonium

  • Ich kann es oben nicht mehr bearbeiten, wohl zu lange her. Hier nun die überarbeitete und optisch übersichtlichere Version mit weiteren kleinen Details, die ich gestern in der Eile und ohne zweiten Bildschirm ausgespart habe:

    Kapitel 6. Midi & Audio settings

    This tabbed dialog is displayed when the Audio/Midi Settings menu item is chosen in the Audio/Midi menu.

    It manages "system-wide" GrandOrgue options. In other terms, these options apply to any loaded sampleset.

    These options are located at different places according to the platform.

    Windows: registry key HKEY_CURRENT_USER\Software\Our Organ\GrandOrgue and %APPDATA%\GrandOrgueConfig file

    Linux: $HOME/.GrandOrgue and $HOME/GrandOrgueConfig

    Mac: $HOME/Library/Preferences/.GrandOrgue and $HOME/Library/Preferences/GrandOrgue

    Anmerkung

    When launched with an instance name (GrandOrgue -i instanceName), the name GrandOrgue becomes GrandOrgue-instanceName, the name GrandOrgueConfig becomes GrandOrgueConfig-instanceName

    The user can reset settings to default values by deleting the "options files" according to the platform.

    Enhancements frame

    • Active polyphony management
      When polyphony reaches 3/4 of the current maximum value, release samples (reverberation tails) are faded out in order to conserve polyphony.
      • Memory: No impact
      • Polyphony: Conserved
      • Load time: No impact
    • Release Sample Scaling
      For those sample sets with long reverberation tails ("wet"), enabling this option will fade out the release sample if the note is played staccato. It will have little to no effect on "dry" sample sets.
      • Memory: No impact
      • Polyphony: No impact
      • Load time: No impact
    • Randomize Pipe Speaking
      This will apply a very small random detuning before each pipe begins to speak. The detuning is small enough so that the ear can't appreciate the tuning difference and large enough so that it approximates the tiny interferences (due to physical location) that occur when playing multiple real pipes in an organ.
      • Memory: No impact
      • Polyphony: No impact
      • Load time: No impact
    • Load last file at startup
      This will trigger the automatic load of the last used sample set.
      • Memory: No impact
      • Polyphony: No impact
      • Load time: No impact

    Sound engine frame

    • Interpolation
      This dropdown controls the method used to interpolate the waveform when resampling the samples.
      • Polyphase: This method gives better audio results at the expense of higher CPU load.
      • Linear: This method lowers CPU load at he expense of audio quality. It is the only usable option when lossless compression is enabled, as polyphase interpolation is not yet implemented with lossless compression.

      The assertions about audio quality should be used as guidelines only. As a rule of thumb, polyphase gets better treble, linear gets better bass, and this is most noticeable in sample sets that use interpolation extensively (e.g. retuning a detuned organ). Audio quality is highly subjective, so the user is strongly encouraged to listen for himself and retain the setting that suits him better.

      • Memory: No impact

      • Polyphony: Linear gives more polyphony

      • Load time: No impact

    • Number of CPU cores
      This number states how many threads GrandOrgue creates to spread the load. Cores number is the recommended choice for computers without Hyperthreading. Less doesn't use the whole computer. More wastes resources while managing overhead.
      With Hyperthreading enabled, the CPU load seems more evenly spread among virtual cores.
      • Memory: No impact
      • Polyphony: Raising the number of cores raises the polyphony attained before the sound starts to break.
      • Load time: No impact
    • Workload distribution
      This setting has the same role as Number of CPU cores, and is dedicated to spliting the workload. The same value as the number of CPU cores is recommended.
      • Memory: Same as concurrency
      • Polyphony: Same as concurrenc
      • Load time: Same as concurrency
    • Cores used at loadtime
      This number states how many threads GrandOrgue creates to load samples in memory. It has NO effect when loadind samples from cache.
      Higher speed-up loading while reducing the available memory for samples. A zero (0) value means classic load.
    • Recorder WAV format
      This selects the format used by the Audio recorder. The supported formats are: 8 bit, 16 bit, 24 bit PCM and 32 bit IEEE float.
      • Memory: No impact
      • Polyphony: No impact
      • Load time: No impact
    • Record stereo downmix
      This textbox enables GrandOrgue to record multi-channel audio in simple stereo. This option is disabled by default, as it increases CPU usage even if the audio recorder is off.
      When stereo downmis is disabled, all audio channels are recorded in the WAV file.
      • Memory: No impact
      • Polyphony: Slightly - like an additional audio interface
      • Load time: No impact

    Paths frame

    • Settings store
      This selects the directory where preset files are stored. The Browse button opens a file browser where the desired directory can be chosen.
    • Cache store
      This selects the directory where cache files are stored. The Browse button opens a file browser where the desired directory can be chosen.

    Sample loading frame

    • Lossless compression
      Uses a lossless algorithm (i.e. the original samples can be perfectly reconstructed) so that RAM requirements are lessened. Particularly useful for large sample sets.
      • Memory: Variable savings up to 40%
      • Polyphony: Approx. 10% lower
      • Load time: Approx. 20% slower. Only affects load without cache - as the compressed data is cached.
    • Load stereo samples
      Selects whether stereo samples are loaded in Mono or Stereo.
      If set to Mono, all samples are loaded in Mono. When compression is disabled, this will save 50%, but when compression is enabled, much less will be saved because the compression is much less effective on monoaural inputs.
      This setting can be set to don't load, if you want GrandOrgue to avoid loading any samples. Via Organ Settings, it is still possible to load only specific parts of the organ.
      • Memory: Variable savings up to 50%
      • Polyphony: Slight impact, as loading in Mono saves a few calculations per sample.
      • Load time: No impact
    • Sample size
      Selects whether the samples are loaded with 8 or 12 or 16 or 24 bits precision.
      • Memory: High impact
      • Polyphony: The polyphony will increase with lower bit sizes and decrease with higher bit sizes.
      • Load time: Slight impact
    • Loop loading
      Selects which loop(s) are loaded if the samples have multiple loops.
      • First loop: Load only the first loop.
      • Longest loop: Load only the longest loop found in the sample. Longer loops usually feel more lively.
      • All loops: Load all loops found in the sample. While playing, the loops are cycled using a round-robin scheme.
      • Memory: First loop less than Longest loop less than All loops
      • Load time: Increases with the amount of data loaded
      • Polyphony: No impact
    • Attack loading
      The sample set creator can provide multiple separate attack files.This selects which attack(s) are loaded.
      • All: Load all attack files.
      • Single attack: Load only the "best" provided file.
      • Memory: Single attack less than All
      • Polyphony: No impact
      • Load time: Increases with the amount of data loaded
    • Release loading
      The sample set creator can provide multiple separate release files.This selects which release(s) are loaded.
      • All: Load all release files. The sound engine selects which release to play according to how much time the sustain sample was held. For example, it usually selects the shortest release sample while playing staccato.
      • Single release: Load only the "best" provided release file.
      • Memory: Same as Attack loading
      • Polyphony: No impact
      • Load time: Same as Attack loading
    • Memory limit
      Chooses the size of the memory cache where samples are loaded. This is especially useful when running the 32 bits Windows build on 64 bits platforms to avoid a crash while loading a big sample set.
      If this parameter is set to zero, it tries to load as much data as possible.
      Achtung: In this case, GrandOrgue keeps loading until the OS fails to provide memory. It can provide large amounts of swap space with expectable adverse impact on performances.

    Sound output frame

    • Sample rate
      Selects the output sample rate. Allowed values are 44100, 48000 and 96000 Hz.
      The samplerate should match both the configured samplerate in the audio interface and the samplerate of the recorded samples to avoid resampling. If not, resampling can occur in all layers of the audio stack, and audio quality can suffer.
    • Samples per buffer
      Allowed values range from 16 to 1024 by increment of 16.
      This parameter sets the output buffer size. Larger values usually reduce sound artifacts at the expense of latency.

    Cache frame

    • Compress cache
      Selects whether the disk cache must be compressed when created or updated.
      • Memory: No impact
      • Polyphony: No impact
      • Load time: Loading from an uncompressed cache is I/O bound, while loading from a compressed cache requires more CPU. With a slow disk + fast CPU, compressed might be better. This feature is closely related to the hardware capacities, so the user is encouraged to test for himself and retain the best setting for his computer.
    • Automatically manage cache
      Selects whether the cache must be automatically created or updated when the sample set is loaded.

    Perform strict ODF

    Organ Definition File (ODF) syntax has much changed since the days of Hauptwerk™ version 1. Some Hauptwerk™ 1 keywords are not used by GrandOrgue and are reported as warning in the pop-up message window.

    Introducing ODF warnings to GrandOrgue is a two-step development process:

    • 1. A new warning is added to the ODF parsing utilities. It is displayed only in strict ODF mode, to give samplesets designers time to remove the new warning from their Organ Definition Files, without disturbing users.
    • 2. After a suitable number of months has elapsed, the warning is moved to normal mode and is always displayed.


    This checkbox enables strict ODF mode.

    Anmerkung

    Sampleset designers are strongly encouraged to update their Organ Definition Files in order to remove warnings and improve user experience.

    Likewise, GrandOrgue users are strongly encouraged to report new warnings to the sampleset designer.

  • Ich habe jetzt vieles getestet, mein Ergebnis ist dass wenn man ein großes Set mit wenig Speicher laufen lassen möchte Mono die beste Option ist. Dann hat man alle Loops und so weiter und trotzdem bleibt die Ausgabe ja Stereo da die Position der Pfeife ja im ODF festgelegt wird. Ob es nun wirklich einen Unterschied macht im Klang aus ich nicht, ich zumindest höre mit Kopfhörer oder Stereo Lautsprecher keinen.

    Melodeum.de - Wissenswertes zu Harmonium

  • Eigentlich sollte GO unter Linux mit dem X virtual frame buffer ohne tatsächliche grafische Ausgabe oder WM funktionieren. xvfb installieren und dann am einfachsten mit dem Wrapper xvfb-run GrandOrgue auf der Kommandozeile starten. Habe ich noch nicht mit GO getestet, hatte aber bisher keine Probleme, andere Software (Browser, Spotify, usw.) so zu starten