From discussion on FACEBOOK on the library in question:
I'll try get some more details. The premise is that the latency when analyzed in terms of personnel is a defect in sw synth that is felt to a greater or lesser depending on the habits of the player. For percussive sounds and an attack with a minimum (see the plan) is obviously much more noticeable and annoying, and conversely for a pad sound from the attack is not immediate in a few ms + or - are absolutely innavvertibili.
There are people playing quietly with 20-30 ms latency without experiencing discomfort, there are those who exceeded 5-6 ms (I am among them) I strongly feel any discomfort. However, with a any sense of adaptation, you can also play with a latency of 80-100 ms, without experiencing great difficulties rhythms. Organists who know something of classic organ pipes may be more than 30 m, which is precisely 100 ms of latency. For a real piano
latency (the time between the moment when hitting the button and when you hear the sound) is approximately 1 ms, when the sound travels in a ms 34 cm, and the ears are more or less at that distance from the strings (cm longer, less :-)). cm In a synthesizer
HW latency is the time between when you pressed the text and when the sound produced. Ditto for the definition modules Synth MIDI input with which we must add 1 ms of delay due to the fact that the MIDI NOTE ON lasts about 1 ms (the midi works at 31.25 kbits / sec and it takes three bytes of the Notes on). When analyzing the various synth Hardware average latencies are around 3 or 4 ms max (I tried with my old Roland Sy85 or D110) and reading the forums this is confirmed for other synths. Now to
sw synth in a situation where I have a card with hardware MIDI input: here is the sum
latency - the delay of the MIDI Note On message (at least 1 ms Jitter + possible)
- delay introddotto by Bus Audio card (less than 1 ms for the PCI and PCIEX), on average one or two FireWire cards for ms (IEEE-1394) and often something more for the USB that the inevitable need to add jitter to be negligible at best and more unlike heavy for non-optimized hardware configurations
- the latency due to the buffer used by the Audio tab. This is equal to the SR of 48000 BUFFER-coupled delay of 96Sample (2ms) 144Sample (3ms), 240Sample (5ms) 480Sample (10ms), etc, etc
- the time it takes the software create the sound of the synth which depends on the type of synthesis, its implementation and of course the power of the CPU
With the method and then I will mention my hardware configuration (processor at 2.4 GHz Core2Duo + + masterkeyboard EMU1212M PCI Sound Card (a plan) Yamaha P155, ottengo con un Buffer di 144 sample (3 ms) una latenza MISURATA pari a 6 ms per software quali il campionatore native instruments Kontakt, 7 ms per software synth quale Native Instrument FM8 e 8 ms per il "vecchio" Korg M1 software.Provando con il buffer minimo , ovvero 96sample(2ms), ottengo sempre 2 ms in meno di latenza, ovvero 4ms per Kontakt, 5ms per FM8 e 6ms per M1
Per misurare la latenza utilizzo un metodo documentato sul web che è in assoluto il più sicuro, ovvero quello di prelevare dal cavo Midi il segnale "audio" degli impulsi binari digitali(avendo cura ad non alzare di molto il volume per non recare danni all'impianto audio) e ovviamente di campionare (a 48000 Hz) suono prodotto + impulso midi. Il risulato Wave which is a trivial sw audio (wavelab in my case) you measure the delay between the pulse and sound production. For the record, midi pulses are easily recognizable and those noteon measure exactly 1 ms, as predicted by theory.
Now for the library Alicia's Keys Piano: proceeding with the buffer of 144 samples (3 ms) which is with Kontakt that plays a sample of test results in a latency of 6ms, with the preset Flat Alicia's Keys Piano I get a latency measured equal to 26 ms. This is perfectly consistent with the fact that displaying the various libraries of the sample shows that among the sample start and the beginning of the sound, there are about 20 ms, and this is exactly the delay that add to 6ms due to my HW + SW which produces the annoying latency: 6ms +20 ms = 26ms.
not happy I went to browse the various paramente Editing and found the parameter Finger Attack - Latency (0% - 100% Default = 50%). Well this parameter varying from 50 to 100, I measured an increase in latency from 26 ms to 36 ms per Finger Attack = 100% and a decrease from 26 to 16 ms per Finger Attack = 0%. In fact, you can reduce the delay to a value acceptable, although still too high compared to 6ms, which could give my system (as measured for other libraries). I went to browse in the script to try to figure out how to further reduce this figure: I found what I was looking for, or a variable $ Bq2cx, described as "Latency: Select a low number for low latency. Select a high number for a larger part of the original attack of the note to be included." Evidently, this parameter increases the value of Sample Start increasing it to max 10ms, I stopped, unable to find a way to increase this value up to a max of 20 ms, which would solve the problem. The script syntax is complicated and there are well ... 144 KBytes of code!
the way: I enjoyed create me a plan, taking second layer of Alicia Keys's (which is a library of 8 levels) and for correcting hand-eighth start for each sample. For the eighth I was able to reduce the latency to 8ms long for! E a suonarlo si sentiva: eccome!
Ultima cosa: se non erro la versione non piratata di Kontakt non prevede ne l'esportazione dei sample e ne la modifica e/o visualizzazione degli script; in tutti i casi il mio piccolo studio non ha ovviamente alcun fine se non quello del piacere della ricerca di una spiegazione ad un "piccolo" mistero!!!
Benny
0 comments:
Post a Comment