LinuxMusicians

Guitarix release 0.35.5

Guitarix is a tube amplifier simulation for
jack (Linux), with an additional mono and a stereo effect rack.
Guitarix includes a large list of LV2 and LADSPA plugins, and support LADSPA / LV2 plugs as well in it's racks.

The guitarix engine is designed for LIVE usage, and feature ultra fast, glitch and click free preset switching and is full Midi and remote controllable (the Web UI is not included in the distributed tar ball).

This release fix a issue within the new online preset downloader.
If you already use V0.35.4, a upgrade is strongly recommended.


Refer to our project page for more information:
http://guitarix.org

Download Site:
http://sourceforge.net/projects/guitarix/

Forum:
http://guitarix.sourceforge.net/forum/

Consider visiting our forum or leaving a message on
guitarix-developer@lists.sourceforge.net

regards
hermann

Thank you very much as always Mr Tramp, I love this FX processor and it's come a really long way! \m/

Thanks for your feedback, asbak.

If you ever run into a issue with guitarix, let me know.

Wow, the sound of my presets changed quite a lot since 0.035.4, at least at 96k.
Did you do something to the Presence? It sounds way flatter now.
Not necessarily a bad thing though, just sayin...

[edit]: Nothing in Guitarix commit log, but I'm not deaf (and I do keep old builds to compare with for such cases).
I think I have a track: there were several library updates on Arch Linux since 0.35.4 was released, exactly 10 days ago.
Could it be this guy? https://www.archlinux.org/packages/extra/x86_64/boost/

Hi

Well, indeed nothing changed in the engine part of guitarix.

But I don't thing that boost could have a influence on the sound, as it is used merely for string formatting in the UI. In the engine, we use some mutex in the init of the convolver engine, but I cant really see that boost could influence the sound.

Are there any updates of gcc or the kernel?

regards
hermann

checking some related packages on arch, I suspect ffmpeg to be the culprit. Updated on 2017-07-15

You could check if it is, by building guitarix with the --includeconvolver flag instead the --convolver-ffmpeg

If it is, I need to investigate.

Good catch Tramp, and thanks again for the first class support.
Building with the --includeconvolver switch, and the sound is back to what it was with my 0.35.4 build from 10 days ago, whereas a stock build with the latest ffmpeg from Arch sounds kind of mushy.

Looking at the ffmpeg package changelog, the only change that occurred in that time frame was just a rebuild with a few dependency updates:
https://git.archlinux.org/svntogit/pack ... e4a6ae46cf

So that could maybe point to gcc which was updated on the 13/07:
https://www.archlinux.org/packages/core/x86_64/gcc/

I'll run more tests when time permits

gimmeapill wrote:Good catch Tramp, and thanks again for the first class support.

You are welcome.

gimmeapill wrote:Building with the --includeconvolver switch, and the sound is back to what it was with my 0.35.4 build from 10 days ago, whereas a stock build with the latest ffmpeg from Arch sounds kind of mushy.

When we developed the ffmpeg convolver engine we've to add a adjustment for the norming difference between ffmpeg and fftw3. This factor may have changed again ( maybe by some underlying library's, or by gcc calculations), I guess replace the lines 464 and 517 in /trunk/src/zita-convolver-ffmpeg/zita-convolver.cc
with
will bring the sound back to normal with the ffmpeg convolver.

You may check that, if so, I could add a version check and set the value accordingly to the need.

regards
hermann

Uptate: Today was rainy so I found some time for more A/B tests.
I didn't try yet to tweak that ratio between ffmpeg and fftw3 (I need to apply a proper patch in the pkgbuild otherwise makepkg will overwrite the file). But I tried to play with build flags and found also that just removing the --includeconvolver and --includeresampler flags (to force the use of the Arch provided zita convolver and zita resampler) produces binaries that have a small tone difference in some patches, altough it's far less noticeable than fftw3 vs ffmpeg. This seems to be audible mostly when using the Bassman cab and Presence at max - although this would probably deserve a proper measurement. Next I'll spend some time on the ffmpeg tuning and confirm back if it works.

BR,

LX

gimmeapill wrote:I need to apply a proper patch in the pkgbuild otherwise makepkg will overwrite the file

You could create a patch (after editing) with and apply it in your pkgbuild with regards
hermann

Ok, I finally managed to build it and have now 3 variants of the latest guitarix git (0.35.5.r2.g07625ed7 as of writing): fftw3, ffmpeg, ffmpeg with tweak.
- ffmpeg sounds not good (thin and midrangey).
- tweaked ffmpeg sounds ok-ish (better but still a tad mushy).
- fftw3 built with external zita libraries sounds the best to my ears.
Bear in mind also that I tested only at 96k on a single preset.
I'll try to record some audio examples out of the same sample to give you a better idea.

Do you really get a difference in sound when using internal versus external zita-resampler and zita-convolver library's ( not the ffmpeg version)?

Well. I built another one with the bundled zita libraries (--includeresampler & --includeconvolver), so I have now 4 different flavors to compare, but the more I test and the less I'm sure
When I play live, the difference between ffmpeg and fftw3 is quite obvious, yet after recording one dry riff in Renoise, pass it through each of the 4 guitarix builds and record the results, I could hardly hear any difference between any. I'll give it a try tomorrow with Ardour see if I can get something more tangible...

He yea, I know this situation very well, when I work to long on algorithm optimisations.
May a Spectral Analysis Plug could make it visible?

However, ffmpeg 3.3.2 arrive today in debian/sid, and I must say that I cant determinate a difference at all.
Gcc7 is still in experimental in debian,

tramp wrote:He yea, I know this situation very well, when I work to long on algorithm optimisations.

Now I feel your pain
But nevermind - the difference is actualy audible.
Not sure why, but I couldn't spot a clear difference between recordings in Renoise, & neither did the builtin spectrum analyzer show anything.
In Ardour however, this is a different story: the 4 versions do sound slightly different, although you probably need some above average headphones or studio monitors to hear the muddy mediums.

So if you want to investigte further, there you go:
https://drive.google.com/open?id=0B2beU ... U1hZmRhLXM

The archive contains:
- the original dry guitar sample - no particular interest in itself
- the guitarix preset
- 4 recordings of that dry signal processed with the same preset & recorded with the four different builds of guitarix (all at 0.35.5.r2.g07625ed7): ffmpeg, ffmpeg with tweak, zita internal, zita external
- an Ardour archive with all of them loaded for easy comparison.

Conclusion: fftw3/zita sounds good (there might be a tiny difference between the bundled zita libs and the external ones but I'm not too sure anymore). ffmpeg doesn't sound as good, even with the tweaking, altought this is probably also a matter of taste.
The only real issue so far is that my AUR pkgbuild included the optional ffmpeg build flag (probably a legacy from the past, as I cannot remember ever adding it), and the sound of that thing changed significantly over the last weeks, apparently due to changes in Arch LInux rather than Guitarix itself.
The pkgbuild is now corrected, so hopefully I won't pollute the next Guitarix annoucement