Could you git pull, and see if there is a difference with the recent version?
Ideally compile with make OPTIMIZATIONS="-DDEBUG" that will print additional information while the synth plays.
You will see messages like
fluidsynth: noteon 2 42 34 00089 8.160 7.733 0.000 24
fluidsynth: noteoff 2 42 0 00089 7.798 25
The last number in each line is the current polyphony.
Hi, thanks so far!
I built the update and crashes are good but ride is still truncated, curiously If I redraw the ride notes as 1/4 notes they sound better… What data would you like from the terminal output or do you prefer a screen capture with the whole process?
Interesting, the duration of the notes does not matter at all, so I wonder what the difference is.
I tested this by bouncing a simple MIDI region twice. First using 1/4 notes for Ride Tip, then shortening the notes to minimum duration. The two bounces pass a NULL test.
I’ve used Blonde Bop Hot Rod LV2.
Do you use a pattern that uses both Ride Cymbal Tip, and Ride Cymbal Bell which triggers an exclusive stop?
No, not in the I Agree file, they are straight ride hits but Ride bell needs to certainly be released from the exclusive ride group across all AVL Drumkit versions, its not useful as it is…
When I redrew the Ride notes I deleted the existing notes and drew in new ones within Ardour so I didn’t simply resize the existing notes, would that make a difference if the notes were newly drawn in Ardour’s editor?
No. I also just checked than potential overlap of notes make no difference in the sound of AVLDrums.
Could you send me the new file? I’m curious to do a NULL test (assuming you have placed the ride hits at the same location).
OK,
Here is the actual Ardour session, it has the MIDI file imported and the tempo set and I have put relevant location markers to show the song parts. I have deleted and redrawn the Ride tip notes in ‘Chorus1’ and assigned them an 1/8th note value (I was mistaken before it was 1/8 notes not 1/4 notes). The ride tip notes in Chorus2 and the Bridge have been untouched for comparison. To me the newly drawn notes still sound a bit odd but not nearly as obviously cut off as the originals.
Lol I have an album project of about 12 songs that are depending on getting this figured out… no pressure… 
The difference between Chorus 1 and Chorus 2 are velocities.
When I add a plugin to force all velocities to the same value, both Chorus 1 and Chorus 2 sound identical. Try it yourself:
To test I have removed all notes except for the Ride: http://robin.linuxaudio.org/tmp/IAgree%20-%20Drums-1.mid
In Chorus 1 the Ride alters between velocity 64 and 74. They are the same velocity layer and hence and same sample.
Chorus 2 has a wider velocity range and some randomness between 59 to 100. It spans two different velocity layers and hence sounds different.
Checking the samples “SabianAAX-20-Ri-2” (vel. 52…77) and “SabianAAX-20-Ri-3” (vel 78…102) they do indeed sound rather different.
–
PS.
The first two velocity layers (Ri-1 and Ri-2) sound similar, as do the last two (Ri-3 and Ri-4).
Also while “Ri-4” has a higher peak, it is less loud than “Ri-3” and also shorter.
Indeed. Done now in avldrums.lv2 v0.6.1-4 (upcoming 0.6.2 or 0.7.0)
Hmmm, The conversion from SFZ to SF2 really mangles the sample filenames which are organized by ascending number in the SFZ file, it’s quite annoying… Are we certain that this truncation of the filenames doesn’t shuffle the sample (and thus velocity) order?
Filenames in SFZ:
// 51-SabianAAX-20-RideTip-HR-
<group> group=3 key=51 pan=36 loop_mode=one_shot lochan=1 hichan=16 off_mode=fast off_by=3 offby=3
<region> lovel=1 hivel=26 sample=51-SabianAAX-20-RideTip-HR-1.wav
<region> lovel=27 hivel=52 sample=51-SabianAAX-20-RideTip-HR-2.wav
<region> lovel=53 hivel=77 sample=51-SabianAAX-20-RideTip-HR-3.wav
<region> lovel=78 hivel=102 sample=51-SabianAAX-20-RideTip-HR-4.wav
<region> lovel=103 hivel=127 sample=51-SabianAAX-20-RideTip-HR-5.wav
Filenames in SF2:
It appears we have a secondary problem and this why the velocities between sample 4 and 5 are so pronounced and sound weird… This has probably affected all of the kits for years thanks to the archaic filename handling of the SF2 spec… 
… or due to you picking rather verbose names in the first place, and CamelCase no less 
but yeah, 16 chars is somewhat limiting.
Yes, the names are detailed (SFZ and Hydrogen aren’t complaining…) but for people using them in samplers and not within libraries I think it’s important to know what you’re getting. Once you scour the internet and download hundreds of Drum samples and suddenly have dozens of samples named ‘kick’ it’s not very helpful years later.
Soooo, I guess I have some work to do. Whether that will just be done in Polyphone for the SF2 or whether I should do this across the board will remain to be seen but obviously this needs to be fixed ASAP. I haven’t done any detailed work in Polyphone so I will start by seeing if I can rename and reorder from there.
I will create a new Blonde Bop HR SF2 with the sample order fixed and send it to you, or can I just forklift that into the AVL-Drumkits-LV2 source and do a rebuild for it to get picked up so I can test?
Please make notes of the changes you are doing so that I can re-do them, or start by using the .sf2 from the plugin: https://github.com/x42/avldrums.lv2/tree/master/sf2
I have also trimmed many samples in the Black Pearl and Red Zep soundfonts. Some of them had excessive silence at the start. The panning is different as are the exclusive groups.
Yes, you can. The file name needs to be Blonde_Bop_HR_LV2.sf2 in the avldrums.lv2 folder. Reload the session or the plugin for it it be picked up.
I will edit the SF2 from your source url, that way all your changes are retained and we’re all working from the same set. It appears that Polyphone has Bulk rename so this won’t be as tedious as I feared. I will have to pick away at it between jobs so it will likely be a day or two before they are all done.
Thank you and x42 for fixing this!
Of course, shouldn’t have happened in the first place, lucky shot in the dark it showed up, it kind of reminds how long Soundfont2 has been around, lol 16 character filenames are so Windows 3.1!.. Should have something up later tonight…
The irony is even though I sampled these kits and tested them I haven’t actually done a lot of production with them myself until recently. Kinda embarrassing to drink your own Kool-Aid and find a fly in it so quickly… 
fun fact. VST2 parameter labels are only up to 8 chars (really just 7 + NULL termination).
Given that you have the corresponding real drum-kits that’s not too surprising.
OK,
I’m posting the fixed and updated Soundfonts here in case anyone is interested also in the Soundfonts themselves aside from the LV2… (but why…?)
Changelog:
http://bandshed.net/files/FIXED_AVL-SF2.zip
*NOTE!!
The bug that was found only really affected the loudest velocity layer of only some kit pieces, in my experience people rarely use the loudest layer in a sustained way, if you think all your hard work over the years has been ruined it’s probably not even noticeable what changes were made.
In my experience with the MIDI file I linked above the Ride Cymbal still sounds a bit strange as it was originally programmed in another Sequencer. If you are flying MIDI tracks in to Ardour that were created with the AVL-Drumkits Hydrogen Kits or SFZ’s you may have to make some fine adjustments here and there with the LV2 Plugin, you’ve been warned.
My apologies to Robin and anyone else this SF2 snafu may have affected… If you find or hear new bugs or issues please let us know.
Silver lining: I learned a lot more about using Polyphone! 
Thanks a lot!
I have just merged this into avldrums.lv2 v0.6.1-5-ge5e27a1 – I’ll test and let the dust settle before tagging a release.
Meanwhile feedback and testing is very welcome.
No need to apologize. On the contrary, I am very grateful you provide the samples in the first place. And once the issue was found you fixed it on short notice, including a public disclosure. A+
I’m very grateful, too, for offering such valuable tools to the bedroom musician and wanna-be producer (like myself) - and for the pros of course as well
