Garritan Personal Orchestra - Stuttering

[JoeGBradford]

I've downloaded the Satie midi file from the Spitfire Orchestral walkthrough here https://www.spitfireaudio.com/thepage/ and made a reasonable job of assigning it to three instances of Miroslav Philharmonik in a CbB project - which plays perfectly

However when I created a new project, swapping out Philharmonik for Garritan PO playback stutters.

Can anyone suggest the likely cause of this - could it be to do with the buffer size or is it something to do with RAM or HDD loading times - does Garritan load differently from Philharmonik? I have other Garritan Orchestral projects, including a Sibelius 2nd Symphony, which have much more going on at once and they play fine - why would this be different?

I'm on Win 10 latest update and use onboard soundcard with ASIO4All - I tried swapping to WASAPI but couldn't get any sound when I did! 

Appreciate any suggestions!

[Matthew Sorrels]

ASIO4ALL is weak at best no surprise you can't get that to work well.  But no matter what you are using for an audio driver (or audio interface) if your ASIO buffer size is too small and the synth you are using needs too much CPU time, Cakewalk can't fill the buffer before it goes "live" and that causes pops and clicks.  Among solutions to this is to use a larger ASIO buffer, use a better ASIO driver, use a synth that doesn't need so much CPU power, alter the synth's settings to use less CPU power (turn off effects/reverb), update drivers and other software running your computer that cut into the CPU available for Cakewalk and the synth, buy a faster computer, buy a audio interface with a lower round trip latency/lower CPU overhead/etc.  All of these things serve to make it more possible for your DAW and your soft synths (and effects) to process a buffer of information before the buffer has to be published to your speakers.  If any thing slows that down, even just a little bit, the audio buffer will be played on the speakers but since the data isn't complete you get a glitch.

[JoeGBradford]

Thanks. I checked the CPU load in the meter in CbB and it appears very low.  Asio4All has always worked fine for me. As I say I have other orchestral projects with more going on and have no problems with those just this one. I'll try altering the buffer size though when I went into the audio dialogue there didn't seem to be an option for that. I need to be careful I don't increase latency too much though

[Mánibranðr Studios]

It is possible that you're rubbing up against the limits of the data transfer speeds of your HDD, considering that most sample based instruments use DFD mode streaming. I'd recommend investing in a solid state drive, and migrating the sample collections to that drive.

Edited November 7, 2019 by Freyja Grimaude-Valens

[Matthew Sorrels]

It's not total CPU that is an issue, filling an audio buffer for realtime is a bit different than that.  There are only so many clock cycles available before the next buffer has to go to the speakers (depending on the sample rate you are running at).  If your CPU is busy doing anything other than constantly filling those buffers, like servicing other interrupts or if (as is the case with ASIO4ALL) the driver software isn't all that good and the hardware isn't that good, the amount of time available is going to be a problem.  This is why people buy very expensive audio cards with well written ASIO drivers instead of using the built in motherboard hardware and ASIO4ALL. 

You can use LatencyMon to measure how much time other things are taking on your computer.  Run it while trying to play your project and see what it says it taking all the time.

https://www.resplendence.com/latencymon

You may also be able to alter your soft synth to need less time, perhaps by making it load more data/use the disk less, etc.  It's a very delicate balancing act.  And one soft synth may be very different than another.  And one DAW may be very different than another.  But at the end of the day it's all the same math, the buffer has to go live to the speakers on the clock, if it's not done yet you will get a glitch.  Anything that takes your computer away from filling that buffer can prevent you from getting it finished.  Use LatencyMon to figure out what is taking the most time away, then see if you can alter it.  For lots of things you can't, other than updating Windows/Drivers/OS.  But at least you'll get some idea of what is taking the time.

[JoeGBradford]

Thanks - will see if I can sort it - just seems odd that it is only this project where there is a problem - especially as it is not as demanding as my other GPO projects

At the end of the day I'm quite happy with the Miroslav version of this particular programme but obviously I wouldn't want GPO to create problems on other projects. As I am only a hobbyist it's not really worth me upgrading my hardware until I need to or until SSD prices come down a bit more 

[kitekrazy]

BTW the Aria engine and Player are updated.   The player will have to be updated at Make Music.

[JoeGBradford]

Ah thanks didn't know that!

[David Baay]

Are there tempo changes in the project? Possibly Aria Player is not handling them as well as a Miroslav, though I would expect it should. If the stuttering persists after freezing/bouncing to audio, that would take the audio driver/hardware out of the equation. It could also be that some particular controller in the MIDI is causing problems with Garritan. If you want to share the Garritan version of the project somewhere, I could give it a run  on my good-but-less-than-state-of-the-art system to see how it behaves.

[JoeGBradford]

Thanks David - there are tempo changes - I did wonder whether it might be that - increased buffer size but still a problem

Here's the file - appreciate the help ?

 

TROIS GNOSSIENNES Orchestra - Garritan.cwp

[David Baay]

So I got the same stuttering effect here, and noted it was happening particularly on strings and in places where the tempo wasn't changing. 

I looked at the controllers in the PRV, and found both Modulation and Expression controllers following the same pattern but at different average levels.

In Garritan, Modulation (CC1) and Expression (CC11) both control the Expression parameter, so the controllers are competing with each other, and rapidly switching the level up an down since the curves are at different levels.

Deleting either set of controllers eliminates the stuttering.

 

[JoeGBradford]

Ah thanks - you are a star! I think you (or possibly someone else) mentioned previously that Garritan PO has some unusual midi implementation. I did wonder if that was a potential source of the problem given that the file was intended for Spitfire BBC SO

I'll work on deleting one of those tomorrow - I was interested to see how Garritan and Miroslav compare in terms of sound quality and realism - Miroslav Philharmonik  CE2 is sounding quite good but I thought Garritan may give me more flexibility and range of instruments

Thanks again for your help

[David Baay]

Glad to help. Garritan is somewhat unique in using Modulation to control volume. IIRC, they added support for CC11 at some point to be more compatible with GM conventions. CC2 (Breath) also does the same thing.

You might want to find out what characteristic of the sound is controlled by Modulation in Spitfire, and see if it's possible to emulate that with a different controller.

Here's a summary of controller implementation in Garritan:

https://usermanuals.garritan.com/GPO5/Content/controls.htm

EDIT: I should add, if the parameter controlled by Modulation in Spitfire has a counterpart in Garritan, you can copy the Mod controllers to the  appropriate controller number in the PRV, and then delete the Mod controllers.

Edited November 8, 2019 by David Baay

[JoeGBradford]

Thanks again - I really need to RTFM ?

[JoeGBradford]

They really do seem have made it complicated though!

[haydn12]

The CC1 programming in GPO was caused by a limitations in Kontakt 2 at the time GPO was created.  CC1 is more than an expression controller.  It also does filtering to change the brightness of the instruments.

Jim 

[JoeGBradford]

Thanks Jim - I didn't know that

[JoeGBradford]

Reading around 'modulation only controls dynamic level in Spitfire, while expression controls the actual volume' - '"dynamics" does mean "timbre and level", and not just timbre'. i don't think there is an equivalent to expression in the Garritan CC list that David posted earlier? 

So I have tried to delete the expression CC in my tracks but if I erase the controller data the notes disappear as well - I assume there is a way of deleting the controller data without deleting the notes - or just turning the controller off? I've had a look in the online help and although it says that you can disable an automation parameter by right clicking on the automation read button I think this may just be for audio track automation? I have looked at both my midi and VSTI tracks - I am sure I must be missing something obvious but after a good half hour I am stumped! Hopefully someone will rescue me!  

[David Baay]

To work one track at a time, graphically, set the PRV to  Show Controller Pane in the View menu. Double-clicking the clip to open it in the PRV should then show all controllers as separate lanes in the Controler pane where you can swipe across controllers with the Erase tool to delete them without affecting notes.

https://cakewalk.com/Documentation?product=Cakewalk&language=3&help=Views.05.html

Alternatively, you can select tracks/clips, go to Edit > Select by Filter, and check boxes to select only the controller number you want to delete, then hit Delete on your keyboard. You could also use this method to delete controllers from all tracks that have competing Modulation and Expression controllers in one go, but take care that you don't include a track that has only the one controller type.

The page I linked earlier does not explicit state that CC1/2/11 affects brightness as Jim suggested, but that might be true, I haven't confirmed one way or the other., but have not really noticed that in the past.

I would probably just keep whichever controller is at the highest average level so that you get a good, strong level out of every track as a starting point for mixing.

Edited November 8, 2019 by David Baay

[Variorum]

Since GPO instruments are actually SFZ's,  individual patches can be configured to respond to CC1 in pretty much any way the sound designer wants. Some do change multiple parameters (i.e. volume and filter). The fun part is that you can edit the SFZ files yourself to add that, and other, functionality to the patches. They are very advanced SFZ files, though, so it's not a job for the faint of heart ?