Stock VSTi Fails Timing Test Using Host Seq

[Perfect Human Interface]

There are plenty more tests that you and I need to do to prove things conclusively, but my gut feeling is that FL is doing several things in non-standard ways - quite possibly with good reason (low CPU for example), but in terms of VST timing, it has already proved itself the 'odd-man-out' by not dealing with delay compensation correctly for the PPQ signal.

If you guys want me to pester them on the forums for answers again I can, but you'll have to explain your observations a little more so I understand what's going on. Or of course you could go ask yourself if you're registered on the forums.

Also, just in case it matters, there is a PPQ setting in FL that may effect things.

The explanation about 'visual quantising' of the display from GoI has been bothering me too...
Getting audio timing bang on is hard to do - it has to be done in real time. Drawing a wave display? That's a piece of cake, the CPU has all the time in the world to do it - the individual cycles of waves are displayed without any distortion - it shows the bar/beat lines in the right place - so why show the wave start points on screen in the 'wrong' location?

Because it's almost completely insignificant, you can just use Edison, and they have many other things to work on. Trust me, that's the answer. Really this is imperceptible unless you have long audio (5 mins), zoom in as far as possible, and lean in close to your monitor, so in my opinion not worth much concern.

[billv]

There are plenty more tests that you and I need to do to prove things conclusively, but my gut feeling is that FL is doing several things in non-standard ways - quite possibly with good reason (low CPU for example), but in terms of VST timing, it has already proved itself the 'odd-man-out' by not dealing with delay compensation correctly for the PPQ signal.

Will do another test in reaper to re-confirm this.
I think you might be right here trog.
My real problem here is lack of "expertise" in FL settings. there's heaps of them.
In fruity wrapper, that holds plug-in, there's even more options.
To get to the bottom of this properly, we need to have our test enviorment set up
with all options set the way they should be for "precision timing testing".
Otherwise, we are just going to go round in circle's for months.

Or of course you could go ask yourself if you're registered on the forums.

Yeh I'm registered, good idea, might be time to get FL set up by the owners.

PPQ setting in FL that may effect things

again...going in circles...lots of things may have an effect....what's the right combination... .
so yeah, might be time to get FL set up properly for what we want to achieve.
Thanks guys....
Also,
Note on PPQ:
Check my last post in VST Timing thread.(might be of interest)

[trogluddite]

Because it's almost completely insignificant,

Possibly the most dangerous excuse in software development - " it's only a little bug, no-one will notice." - a very slippery slope, that one (I should know, I've done it often enough!!). But, yes, in real terms not exactly a deal breaker - I doubt many folks are using FL to do the kind of detailed analysis that billy is up to.

But the lack of PDC on the PPQ signal 'out of the box' IS significant. It is a signal whose sole purpose is to act as a standard timing reference - without compensating for the audio buffers, it will always arrive late, guaranteeing a timing error for any plugin that assumes the de-facto standards have been observed.

Of course, this cuts both ways - likely we would not be debating this if SM had not had such lousy MIDI timing for so many years - and lack of 64bit and being a few versions behind with the VST spec' don't exactly help FS's reputation as a plugin development tool.

If you guys want me to pester them on the forums for answers again I can

I'm quite happy for you to pass on whatever observations I've made - but I've no interest in FL as a regular tool, and don't sell my plugins, so it's purely an academic curiosity for me.

[MyCo]

I'm just wondering, how all of you testers are generating "Waves" from a plugin in Fruity? The reason for this question is: For Rendering the Audio buffer doesn't matter at all. Infact you can render even without having an audio device. Although the rendering settings in FL are important, because interpolation introduces a huge amount of delay. Also in the rendering dialog is an option for delay compensation, that has to be checked. And when you render to mp3, the codec adds extra latency.

There are also other options that introduce latency, eg. in the Audio Settings when "Mix in buffer switch" is unchecked, or when "Tripple buffer" is set, or Interpolation is enabled.

Regarding PPQ, in the audio settings dialog make sure "Align tick length" is unchecked, else the PPQ timing is processed in steps and not linear.

[billv]

re-tested the stock vsti about 8 times in a row, everytime it was perfect..

it's only a little bug, no-one will notice

So searching from a different angle, i got this.
See if you can get your heads around this one.

1: Vst exported from FSM via FS
2: Same FSM loaded into FLFS
This is the exported VST

ScreenShot199.png (5.26 KiB) Viewed 21109 times

This is Host running FSM in FLFS

ScreenShot200.png (6.02 KiB) Viewed 21109 times

At least my exported plug is still consistant.
Seems a bit late compared to the host version, but at least it's still the same.

Don't know why the host is suddenly nailing it either.
Only downloaded v11 a hours back, and trying to leave all settings as they in the demo
as a tempory reference point

[billv]

There are also other options that introduce latency, eg. in the Audio Settings when "Mix in buffer switch" is unchecked, or when "Tripple buffer" is set, or Interpolation is enabled.
Regarding PPQ, in the audio settings dialog make sure "Align tick length" is unchecked, else the PPQ timing is processed in steps and not linear.

Thanks for the tips Myco.
There's till too many un-known variables for my liking. i just can't be sure......

[Perfect Human Interface]

I really think you're better off using a recorded sample/wavetable instead of an oscillator because that will (I think?) give you identical waveforms every time.

Last night I was trying to put together something that would play an impulse on MIDI input. Found out I didn't know how to play back a wavetable or sample "just once" (I guess you have to calculate the length and then manually cut the signal level at the end). After dicking around for a while I dug up a drum pad example someone else made and started gutting it, but then my computer bluescreened in the process.

Damn shame they couldn't just let you control the damn impulse component!

If anyone knows their way around this stuff and can whip something up (maybe in Ruby?) that simply spits out one sample of value 1 at note trigger and then all 0's, I think that would make this process easier.

[trogluddite]

one sample of value 1 at note trigger and then all 0's, I think that would make this process easier.

Yes, that's a good idea - it ought to at least lest os see whether any errors are on the MIDI or audio side. Should be possible Ruby using the 'frame' to mono - but needs a bit of reading up first because some of the 'frame' components also add a buffer's worth of latency (I think just the input ones, but best to double check!)

[billv]

I really think you're better off using a recorded sample/wavetable instead of an oscillator

Yeh, I'm set up for that. Can use the The Drum machine I uploaded a while back. Will have a go.....

But in general, I'm repeatedly seeing the result......

It would have been nice to see this everytime


So I'm almost ready to push this onto my backburners.
Waiting for feedback from Image Line.

[billv]

using a recorded sample/wavetable instead of an oscillator

Tried it three times using different waves. Fail. Not a single cancellation.

If i keep working on the polarity issue, will open new thread for it.

I'm putting this thread down to user error.
Obviously the whole time didn't have host settings correct.
Initializing everything by downloading the FL11 update, seems to have worked.

Stock VSTi nails it everytime now. All good.