Latency, nudge and midi

Per ZUNGLE:

Audio Loop Back Test(As I do it)

....Open SHS

...go to track view

...right click time ruler see options and set time ruler format to Samples.....not Absolute Time or Musical Time.

...set metronome to Audio/recording.......... no pre-count or pre-roll.

...Arm track #1 and record audio metronome to track #1. About 4 measures. (you use Delta 44 right?)( If so you can record audio metronome by selecting mon mixer for track#1 input...make sure its also selected in Delta control panel.)

...Turn off all metronome features audio, midi... all.

...Arm track #2

...In real time record track#1 onto Track #2....You should now have 2 audio tracks #1 is original #2 is real time recorded copy.

Remember you cannot simply copy track#1 to #2 or paste or do any kind of offline edits, its not the same.

...You may record track #1 to #2 by selecting mon mixer for track#2 input. Hint I leave my monitors off while doing this to avoid nasty feedback loops.

Depending on interface a user may have to send audio out of machine and then back in to record in real time from track#1 to #2. Coming out of machine lengthens audio path and could add more latency.

...Now magnify track heights a little.......pick a wave peak in track#1 and the relative wave peak in track #2.
...They probably look good in placement, with beat and each other....at least to the naked eye.

...Mark that peak with marker or now time.

...Now magnify track width and height... the integers on time ruler should become more accurate as you magnify, take it down to actual bits( integers of 1)

...Find your reference wave peak in track#1 and Look for relative wave peak in track #2

...Count along the time ruler .....the difference in wave peak #1 to wave peak #2 is your exact... to the sample offset(latency).

...It could be a few samples or hundreds, even thousands............. but its there.

Because you love bench marks play with your ASIO and WDM drivers and buffers each time do a test. This will accurately tell you whats happening with each change.

Knowing your exact offset helps you deal with latency accurately and efficiently.

Now you can set up nudge and eliminate offsets with a single key stroke. I don't know if its just me, but it seems like nudge feature isn't a priority for a lot of users.

Did this make sense ?

Midi test would use same principles with different mechanics.

Sorry - my turn to misunderstand the question. the post I gave above is for audio, not for MIDI. Frank's got your answer there about MIDI.

ORIGINAL: fep

Thanks Reese, That is basically what I did to calculate audio latency which I tried to explain in my first few sentences.

Thanks Frank, Lots of good info, food for thought. You mention that midi latency will change as the CPU workload changes. I'd assume then that audio latency will also change as CPU workload changes. Is that correct?

Also, in HS 4 you can nudge midi clips in increments of samples. HS 4 will just move the whole clip by however many samples you nudge it for and the undelying midi events will move by that many samples. I just tested that and at a tempo of 95bpm a nudge of 71 samples = 2 ppq and at 180bpm a nudge of 71 samples = 5 ppq (960 ppq is the resolution in HS 4).

But the problem remains, what is a good practical way to deal with latency.

Latency is a fact of life and you just have to live with it. As for the audio engine that is one big program and it spits everything out at the same time no matter what the CPU load up to the point it starts to drop out. The software synths are seperate processes and can degrade without indication it's happening.

HS4 is trying to offer a consistant user interface so it will let you nudge by samples but in reality I don't think it can output midi at a fraction of a ppq, so it just "appears" to be able to do it. It makes a calculation and gets as close as it can but still moves the data by the ppq resolution, it's not sample accurate just like sheet music doesn't show exactly when the notes will happen in time, midi doesn't either but at least it's finer then sheet music. You don't see many scores with 960th notes.

Most sound cards have direct monitoring or you can use a mixer to get around audio latency while tracking but there is nothing you can do about software synth latency. What I have done before is use a sound from my midi keyboard to monitor to input the performace so I'm in time with the song when I hear it. Then change the output of the midi track to the software synth later. That seems to work well to keep me in time as long as the sound is close to the same ADSR as the one I'll be using in the software synth. To hear the onboard sound and the playback I use a mixer or an extra input on my sound card and use the sound cards DSP mixer to echo it back to the same outputs Sonar is playing on. With that setup I don't even need the software synth loaded when tracking.

Also you can lower CPU load by doing only one software synth at a time, bounce to audio, then move on to the next one. Also, don't add a bunch of effects till you are done tracking and get ready for the mix. I like to bounce all tracks to audio before I start the mix so I can nudge them if needed and have all the envelopes and effects bins available. It's also best for archiving, after all I might not have some of these software synths in the future when I go back to do a remix.

there's still going to be SOME time between when you press a key on the keyboard and when the data arrives in the software. with no softsynths running, that time should be very small.