Sampler issues latest build

[Dan Korneff]

Looks like the samples cutting out and not exporting properly are a buffer issue. Setting preload to -1 works great... but eats up a huge amount of RAM. Time to tinker with some settings :)

[Christoph Hart]

Are you using an external HDD?

[Dan Korneff]

@Christoph-Hart I'm not. I have an SSD and SATA in this rig. It does it on both.
Setting preload to 16000 made it work well with a has a small RAM footprint. I'm pretty happy with that.
Still can't export a monolith though ;)

[Christoph Hart]

Yes, that export issue has something to do with your weird AMD CPU not supporting a SSE3 instruction :) I am using Intel CPUs on all my test machines, but I'll figure this out when I find the time.

The streaming buffer drops are worrying though. How many voices are you playing before you get the drop outs? You should be able to play at least 400-500 voices before the streaming engine starts getting busy.

[Dan Korneff]

@Christoph-Hart That's probably why my Mac doesn't crash when making monoliths... It should be noted that the intel Mac does get the same "file is corrupt" error in Console.

When I do a test with a bunch of hits, such as a drum roll, it can use up to 100 voices. I've haven't seen it get above 200 with a full multi mic drum kit going.

[Christoph Hart]

Hmm, but the example screenshots you showed are just playing a few voices and still drop out, so it can't be a performance issue actually.

How many mic channels are you using? Ah, and does changing the preload size to -1 fix the overlap not being retriggered?

[Dan Korneff]

@Christoph-Hart said in Sampler issues latest build:

How many mic channels are you using?

This example has 4 mic positions.

does changing the preload size to -1 fix the overlap not being retriggered?

HOLY CRAP! It does!

[Christoph Hart]

Alright, then we just know that something is funky with your streaming thread. Getting close...

[Dan Korneff]

@Christoph-Hart Maybe I spoke too soon. It worked on one quick test, but now it seems to be random if it triggers with overlap

[Dan Korneff]

@Christoph-Hart After looking into this further, I think it comes down to basic NoteOn / NoteOff commands.

In this first pic, I have 4 overlapping notes. Only 3 of them are triggered.

If I arrange the midi to display where the overlaps occur, I can see that the sampler wont trigger a 2nd midi note (of the same note) without first receiving a NoteOff command.

[Dan Korneff]

It's actually more apparent when done with the Sine Generator using the same midi.

[Dan Korneff]

Looks like this is just normal behavior of midi in general. I guess it's up to the designer on how to deal with this?

[Cyberdyne387]

@Christoph-Hart said in Sampler issues latest build:

FYI, that's the code that handles samplerates:

if (isOldMonolith)
{
    return 44100.0; // doesn't matter anyway (the sample rate is stored externally)
}
else
{
    const static double sampleRates[4] = { 44100.0, 48000.0, 88200.0, 96000.0 };
    const uint8 srIndex = (sampleDataByte & 0xC0) >> 6;
    return sampleRates[srIndex];
}

Now it also makes sense why @staiff is struggling with his 27,4kHz sounds :)

Thanks for this notice. I too have needed to repitch my own samples after compression due to unusual sample rates (frequent ones for me being 32 and 50 kHz).

[d.healey]

Have you converted your samples to HLAC monoliths?