LinuxMusicians

Fife times in a row I had the "white window" when I loaded a Qtractor session with Dragonfly.lv2 but a correct display when I loaded the same session without restarting Qtractor. But I had cases when a 2nd opening of the Draginfly GUI helped as well.

Usually I don't do that since Qtractor eats up more memory with each new load, maybe there is a memory leak with Linuxsampler and Fluidsynth.

Is it possible that there's a race condition in Qtractor so not all data structures are set up for the LV2 plugin when it's loaded? And is the needed data structure reused from the 1st session when I load the 2nd (that should replace the 1st)?

(Qtractor eats up more and more memory when loading sessions again without restarting Qtractor, don't know why but could be related with memory leaks with soundfonts and/or linussampler).

bluebell wrote: Unfortunately Iforgot to ...
- unload/reload the plugin in Qtractor
- unload/reload the Qtractor session without restarting Qtractor

Next time ...

Yep, this was "next time". Loading my Qtractor session leads to VU-meters constantly to the max in one bus where Dragonfly.lv2 was loaded. Note that in this session I have 2 buses with Dragonfly.lv2.

- Deactivating the plugin brought the sound back
- Re-activating destroyed it again.
- Deleting and re-loading Dragonfly.lv2 in the running session helped.

bluebell wrote:

bluebell wrote: Unfortunately Iforgot to ...
- unload/reload the plugin in Qtractor
- unload/reload the Qtractor session without restarting Qtractor

Next time ...

Yep, this was "next time". Loading my Qtractor session leads to VU-meters constantly to the max in one bus where Dragonfly.lv2 was loaded. Note that in this session I have 2 buses with Dragonfly.lv2.

- Deactivating the plugin brought the sound back
- Re-activating destroyed it again.
- Deleting and re-loading Dragonfly.lv2 in the running session helped.

And a second time.

Something fails at load time. Maybe you can make it write out stuff in the initial phase?

And one more manifestation of maybe the same bug: This time Dragonfly.lv2 didn't stop audio but made Qtractor's audio stutter from time to time. I think it ate realtime CPU.

Maybe there's a loop while calculating (fft?) that runs a loop several times because the end condition is not met (exactly).

Jzust some thoughts.

I added a comment on git in the "issues" section.

bluebell wrote:I added a comment on git in the "issues" section.

Thanks for the report. Will you try the VST plugin to see if you experience the same bug? Based on my own experience, I'm starting to wonder if this bug only happens with the LV2 plugin. You can respond either here or on the github issue.

Edit: Also, are you using multiple instances of the plugin? My project that sometimes exhibits this bug has three different reverb busses... I'm wondering if there is some concurrency bug buried deep in the algorithm that only happens when multiple instances are loaded.

Michael Willis wrote:

bluebell wrote:I added a comment on git in the "issues" section.

Thanks for the report. Will you try the VST plugin to see if you experience the same bug? Based on my own experience, I'm starting to wonder if this bug only happens with the LV2 plugin. You can respond either here or on the github issue.

Edit: Also, are you using multiple instances of the plugin? My project that sometimes exhibits this bug has three different reverb busses... I'm wondering if there is some concurrency bug buried deep in the algorithm that only happens when multiple instances are loaded.

I'll try to use the VST version in my current project.

Yes, I have 2 instances loaded with different settings.

bluebell wrote:Yes, I have 2 instances loaded with different settings.

Oooh, I really think we might be onto something here. Really for real this time!

lilith wrote:I used Dragonfly in a recent Renoise project. And somehow the plugin has some issues. There is some kind of "feedback noise" generated by Dragonfly sometimes and it makes the sound of the whole project completely go off. Observed a similar issue once in Ardour.

Are you using the VST or LV2 plugin? Do you have multiple instances of the plugin in a single session?

Michael Willis wrote:

bluebell wrote:Yes, I have 2 instances loaded with different settings.

Oooh, I really think we might be onto something here. Really for real this time!

lilith wrote:I used Dragonfly in a recent Renoise project. And somehow the plugin has some issues. There is some kind of "feedback noise" generated by Dragonfly sometimes and it makes the sound of the whole project completely go off. Observed a similar issue once in Ardour.

Are you using the VST or LV2 plugin? Do you have multiple instances of the plugin in a single session?

I am not sure if it is the same issue or not but: I have a set of plugins saved in Carla that I use to practice and I had one instance of the LV2 plugin. I've used it many time, but once when opening this same set, the Dragonfly plugin produced a huge noise (there was no input), until I removed it. I did not report since it happened only once and I cannot reproduce it... (Maybe it was not even the plugin's fault, but it was the only one that was producing sound when it happened.)

Michael,

Do you think the graphics issue has anything to do with OpenGL that various people are using? Someone made a post on the Reaper forum here mentioning something about it. I searched this thread for instances of the word "OpenGL" but it only came up once. No one experiencing the graphics issue (since you'd fixed it previously, for me anyway) has mentioned what version of OpenGL their video adapter uses. It's worth seeing if that's the issue. I keep forgetting that some Linux users stick with older hardware, more (that I've noticed) compared to Windows users anyway.

(PS. it still works fine on my system, in Reaper, with my video adapter.)

JamesPeters wrote:Michael,

Do you think the graphics issue has anything to do with OpenGL that various people are using?
...
(PS. it still works fine on my system, in Reaper, with my video adapter.)

I'm glad it's still working for you. You might be right about some people experiencing problems due to their OpenGL version.

bluebell wrote:I'll try to use the VST version in my current project.

Yes, I have 2 instances loaded with different settings.

Bluebell, do you experience the graphical bug (all white where the spectrogram should be) when you use the VST? Also, do you know what OpenGL version you have?

I wanted to also ask if the "spin" function is supposed to be subtle or more noticeable, since I have had trouble knowing if it does anything.

With "wander" set low I can hear the reverb panning, and set higher it's harder to notice that aspect, so I figure I have some idea how that function works. But for "spin" it seems when I think I have figured out what it does, if I adjust its setting (irrespective of the other parameter settings I've chosen), I can't reliably tell the difference it makes. I might just not be using settings which reveal how noticeable the function is, and maybe it's intended to be subtle overall.

JamesPeters wrote:I wanted to also ask if the "spin" function is supposed to be subtle or more noticeable, since I have had trouble knowing if it does anything.

With "wander" set low I can hear the reverb panning, and set higher it's harder to notice that aspect, so I figure I have some idea how that function works.

From your description, you know as much about these mystery dials as I do I didn't write the algorithm, I just wrapped it in a plugin that I could use.

The reverb algorithm consists of multiple overlapping tails, each of which independently pans around in the stereo space. I think wander is an average number of milliseconds that they are separated from each other, so with wander set to zero, they all end up panning back and forth together. Spin is the frequency that they pan back and forth, so with a spin of 1Hz, the round trip from right to left to right takes one second.

The algorithm has another parameter called "spin factor", I decided to make this a fixed 30% instead of exposing it as yet another dial, since all presets in hibiki reverb had it set to 30%. It might be easier to distinguish what these two dials do if it there were a way to set it at 100%, as right now it mixes 30% of the spin and wander effect and 70% of the reverb without that effect.

If you could reveal that parameter "under the hood", I'd be glad to test it. In Reaper you can forgo the UI of the plugin by clicking the "UI" button in the FX chain window, such that it only shows sliders for the parameters available for adjustment. Sometimes a plugin has a parameter that can be adjusted in this view, which isn't visible in its UI. So if you could take advantage of this--having that parameter exist for adjustment, but not be in the UI specifically--that would be handy.

I notice Dragonfly has some amount of "swirl" to it irrespective of what I do. It's easy to mask with the Diffuse parameter, but it would be nice if it didn't require that. I'm being picky at this point. I only bring it up now because perhaps that Spin Factor parameter would be helpful in this regard.

I had another thought, about the "exploding sound" problem (which I don't experience using Reaper and the Linux VST version): remember how the decay time, if at "0", would cause this? (If I recall correctly, an earlier version of Dragonfly had this parameter able to be adjusted to "0"). With its current minimum setting of "0.1", it's still possible to make the output high (easily exceeding 0dBfs) even without high settings of each "volume" fader (dry/early/wet). I don't mind since when I'm doing something "wrong", I tell myself "well don't do THAT again!" But I wonder if that has anything to do with the current problem some are experiencing. Perhaps the LV2 version "initializes" at the minimum setting or something like that, even briefly? I know nothing of these things but I'm throwing the idea out there, just in case. You could try changing the minimum decay to "0.4", since honestly that's probably perfectly fine for most situations.

Getting ready for the 2.0.0 release, I have started working on publishing manuals for both plugins. Thanks to a lot of help from a technical writer named Chris Share, the current draft of the manual for Dragonfly Hall Reverb is here:

https://michaelwillis.github.io/dragonf ... anual.html

Any feedback is welcome. The manual for Dragonfly Room should be available soon. Hopefully these manuals will be helpful in clarifying how to use some of the features.