Jemusic wroteI can do some tests. Can you tell us what your setup was or how you recorded these figures. For example recording the output from your sound card back into the program would include the input latency of the system. Compared to just hearing it.

I did the following to test:
At each host I've set up 2 MIDI tracks and 1 audio track. One MIDI track is playing a trigger at each beat (120 bpm) for 8 beats. The output of that track goes to the input of an Instrument track containing a sample player (Battery, Halion, Kontakt, GA etc.) playing a click sound on each trigger (panned right). The audio track contains the same click sound (panned left) at each beat and is just there for reference.

Now when I record 8 beats in the host and simultaneously record the audio output on an alternate device or with an alternate piece of software I can measure the difference between the sound in the left channel (reference) and the right channel (monitored instrument sound) and finally deduct the amount of latency introduced by the MIDI triggers that are recorded on the track containing the VST Instrument. I know how much time it took in real time between the moment the MIDI trigger was received and the sound was generated.

So let's say the difference between the reference click and the real time sound is +470 samples and the MIDI trigger was received at +46 samples, the time between the real time generated sound and MIDI trigger is 424 samples (470-64). Which boils down to 9,6 ms at 44.1kHz.

A less accurate but simpler test is loading two different hosts at the same time (if you ASIO device supports multi-client mixing) with the same VST Instrument simultaneously playing the same mono sound. Pan one host left and one host right. Trigger the both Instruments with an external device or by sending MIDI data from one of the hosts or another host to both hosts. If the sound becomes stereo, you know there's a deviation. Downside of this test is, you don't know if the deviation is within the MIDI treatment, audio treatment or both.

Try it, you'll be amazed how differently hosts handle MIDI and VST Instruments.

Below is a video showing Studio One is responding slower than Cubase under the exact same circumstances.

I am confused. What is creating the reference track. You have not explained it that well. What is creating the track that would not be experiencing any latency. The metronome?

niles wrote

Jemusic wroteI can do some tests. Can you tell us what your setup was or how you recorded these figures. For example recording the output from your sound card back into the program would include the input latency of the system. Compared to just hearing it.

I did the following to test:
At each host I've set up 2 MIDI tracks and 1 audio track. One MIDI track is playing a trigger at each beat (120 bpm) for 8 beats. The output of that track goes to the input of an Instrument track containing a sample player (Battery, Halion, Kontakt, GA etc.) playing a click sound on each trigger (panned right). The audio track contains the same click sound (panned left) at each beat and is just there for reference.

Now when I record 8 beats in the host and simultaneously record the audio output on an alternate device or with an alternate piece of software I can measure the difference between the sound in the left channel (reference) and the right channel (monitored instrument sound) and finally deduct the amount of latency introduced by the MIDI triggers that are recorded on the track containing the VST Instrument. I know how much time it took in real time between the moment the MIDI trigger was received and the sound was generated.

So let's say the difference between the reference click and the real time sound is +470 samples and the MIDI trigger was received at +46 samples, the time between the real time generated sound and MIDI trigger is 424 samples (470-64). Which boils down to 9,6 ms at 44.1kHz.

A less accurate but simpler test is loading two different hosts at the same time (if you ASIO device supports multi-client mixing) with the same VST Instrument simultaneously playing the same mono sound. Pan one host left and one host right. Trigger the both Instruments with an external device or by sending MIDI data from one of the hosts or another host to both hosts. If the sound becomes stereo, you know there's a deviation. Downside of this test is, you don't know if the deviation is within the MIDI treatment, audio treatment or both.

Try it, you'll be amazed how differently hosts handle MIDI and VST Instruments.

Below is a video showing Studio One is responding slower than Cubase under the exact same circumstances.

phpBB [video]

Wow, this is very interesting. Thanks for testing this, niles. The jitter (variation in latency) makes it extra disturbing.

I don't know much about MIDI drivers and how they distribute signals, but do you think this variation could be caused by the MIDI driver or does it have to be inside the DAWs? Some people say you can't sense jitter like this, but sure you can. And if you layer f.i. two drums, you'll clearly hear it phase throughout the song.

Either way, jitter or not, S1 is apparently slower to respond than Cubase using the very same settings and hardware. I really hope the next update repairs this.

Jemusic wroteI am confused. What is creating the reference track. You have not explained it that well. What is creating the track that would not be experiencing any latency. The metronome?

You can use a bounce of the VST you are testing with. Only make sure the VST is as simple as possible and just playing the same sample. Obviously you have to use the same files and VST's in all hosts you are testing.

Alternatively I attached a click reference file, it's at 44.1kHz and has 8 click triggers at each 22050 samples (120 BPM). I also attached a single shot to load up in a sample player that is triggered each 22050 samples. The one shot starts exactly the same as each click in the reference and also is 44.1kHz.

https://daw.one/shared/files/Monitoring ... _Files.zip

Skaperverket wroteEither way, jitter or not, S1 is apparently slower to respond than Cubase using the very same settings and hardware. I really hope the next update repairs this.

Exactly, Studio One is responding slower than the rest! That's my point, nothing more nothing less.

Cubase has a lot of MIDI jitter and ASIO jitter, Studio One practically has none.



When you look at the image above you can see how consistent Studio One's MIDI and Audio (and their relation) are (same goes for Reaper) and how Cubase's traditional MIDI tech is jittering but also the ASIO is jittering. I suspect Studio One's own high resolution MIDI equivalent is less subject to jitter than the old MIDI protocol.

However the ASIO part at low latency is the one I'm aiming at. I suspect (speculation) Studio One is using an extra buffer to prevent the ASIO (and maybe also the MIDI) from jittering. ASIO jitter can be noticed at high latency e.g. 1024 samples. The timing of the VST instrument's sound a little irregular then because. At low buffer sizes an extra buffer cycle would be noticeable though, especially when it almost doubles the latency.

What you see in the image are the different hosts. When you look closely you'll see each host responds entirely differently.
Here are the average figures per host, the milliseconds mentioned are the milliseconds between the NoteOn message and the monitored sound. The test was done with an ASIO buffer size of 128 and a sample rate of 44.1kHz

MIDI trigger receive (MIDI In) --> Bitwig + VSTi --> Sound (RT VSTi Audio) = 4.7 ms (average)
MIDI trigger receive (MIDI In) --> Cubase + VSTi --> Sound (RT VSTi Audio) = 5.5 ms (average)
MIDI trigger receive (MIDI In) --> Digital Performer + VSTi --> Sound (RT VSTi Audio) = 6.5 ms (average)
MIDI trigger receive (MIDI In) --> Reaper + VSTi --> Sound (RT VSTi Audio) = 7.6 ms (average)
MIDI trigger receive (MIDI In) --> Studio One + VSTi --> Sound (RT VSTi Audio) = 9.7 ms (average)

Edit: Missing test settings

scottmoncrieff wroteInteresting read...gives some real world example's of what kind of time measurements these figures represent. Blinking for example probably has the same reaction to response time of playing a key on a keyboard ( at the fastest rate ).

1 millisecond (1 ms) – cycle time for frequency 1 kHz; duration of light for typical photo flash strobe; time taken for sound wave to travel ca. 34 cm; repetition interval of GPS C/A PN code

Why not make it clear in musical terms?
1 ms ~1 foot in distance.
So your ear 1 foot away from a marshal amp.

3 ms an arm's-length away from the piano strings on a grand piano.

10 ms The distance between the lead singer and the guitarist.

niles wrote

Skaperverket wroteEither way, jitter or not, S1 is apparently slower to respond than Cubase using the very same settings and hardware. I really hope the next update repairs this.

Exactly, Studio One is responding slower than the rest! That's my point, nothing more nothing less.

Cubase has a lot of MIDI jitter and ASIO jitter, Studio One practically has none.



When you look at the image above you can see how consistent Studio One's MIDI and Audio (and their relation) are (same goes for Reaper) and how Cubase's traditional MIDI tech is jittering but also the ASIO is jittering. I suspect Studio One's own high resolution MIDI equivalent is less subject to jitter than the old MIDI protocol.

However the ASIO part at low latency is the one I'm aiming at. I suspect (speculation) Studio One is using an extra buffer to prevent the ASIO (and maybe also the MIDI) from jittering. ASIO jitter can be noticed at high latency e.g. 1024 samples. The timing of the VST instrument's sound a little irregular then because. At low buffer sizes an extra buffer cycle would be noticeable though, especially when it almost doubles the latency.

What you see in the image are the different hosts. When you look closely you'll see each host responds entirely differently.
Here are the average figures per host, the milliseconds mentioned are the milliseconds between the NoteOn message and the monitored sound. The test was done with an ASIO buffer size of 128 and a sample rate of 44.1kHz

MIDI trigger receive (MIDI In) --> Bitwig + VSTi --> Sound (RT VSTi Audio) = 4.7 ms (average)
MIDI trigger receive (MIDI In) --> Cubase + VSTi --> Sound (RT VSTi Audio) = 5.5 ms (average)
MIDI trigger receive (MIDI In) --> Digital Performer + VSTi --> Sound (RT VSTi Audio) = 6.5 ms (average)
MIDI trigger receive (MIDI In) --> Reaper + VSTi --> Sound (RT VSTi Audio) = 7.6 ms (average)
MIDI trigger receive (MIDI In) --> Studio One + VSTi --> Sound (RT VSTi Audio) = 9.7 ms (average)

Edit: Missing test settings

Very interesting.

Personally, if I had to choose between either higher latency or more jitter, I'd choose higher latency, as that could be compensated for. Jitter is just chaos and I am surprised to see that a couple of the DAWs are all over the place. Reaper appears to be behaving quite similarly to Studio One, except that Reaper's got less latency, so apparently a more powerful implementation can be done. Either way, good thing to be aware of. Thanks for posting your findings, niles. Can't help but to wonder results Live, Logic, Sonar and Pro Tools would produce doing the same test.

so is it gonna be 3.5 or 3.3.5 ? and if it is 3.5 then what happened to 3.4? lol cheers

Ooh you tease.....!

talking of Cubase, i saw a good function in that software the other day which studio one would benefit from and it would have been useful for me in a recent session. The Track headers field. If you have a long title for the name of the track, if you mouse over the title in Cubase you get one of those rolling news tickertape views where by the title keeps cycling from right to left so that you can read the full title of the track. Studio One doesn't have that. It's a small feature for sure, but useful if you get other tracks sent and the producer uses long track header titles.

Steve Carter wroteOoh you tease.....!

Not teasing... Hints in a puzzle...
If one were to look at the entire content of this topic across two thread and pay close attention the Easter eggs, one might actually have a general idea of what will be revealed.

Yeah.... Been back over blogs, posts, YouTube @ 1:55 etc, my brains now hurting - never was any good with cryptic clues, guess I'm gonna have to watch the show - if I can get away from work in time.

Hope it's worth the effort 'cos I'm not in the market for a new interface currently having just ordered (and paid for - ouch!) a new PC today, specifically to run S1 (hopefully without the constant frustrations I've been getting on my old iMac due to lack of it's processing power). Gonna be mightily dissapointed if the new PC doesn't zip along smoothly so any enhancements to S1 pertaining to efficiency will be greatfully received.

Not looking forward to migrating everything from the iMac to the new PC, expecting many hiccups (or hiccoughs) and further frustrations, maybe today's 'secret revelation/s' could ease the anticipated trauma!

Regards.....

The only way I could get it to work was to send midi from the record machine over to the test host machine and then record the resultant audio with latency back.

Yes I got 410 samples at 128 buffer

I have noticed it does vary slightly depending on the virtual instrument used. Impact was a tad faster for me than Sample One. The best I could get was at 32 samples buffer size and the lowest latency I got was 200 samples with Impact. Which translates to about 4.5 mS. Which is pretty decent.

I can record at 32 samples and do so for the reasons of the best performance. Studio One is very consistent jitter wise though which is good in that respect. They could improve the speed at which a virtual instrument responds because with my thunderbolt interface the best performance is not being quite reached. The output latency is still shown as just over 2mS.

Jemusic wroteThe only way I could get it to work was to send midi from the record machine over to the test host machine and then record the resultant audio with latency back.

Yes I got 410 samples at 128 buffer

That's a legit test too because the reference and VST audio both contain the same latency when send back. Our result is pretty consistent. Over 8 trigger the average latency between the reference and the instrument on my system is 10 ms (443 samples) at 44.1kHz. Which obviously should be closer to 6 ms (40% faster) at a buffer size of 128 samples (128+128/44.1).

The same goes for 32 samples. 4.5 ms is a figure that belongs to an ASIO buffer size of 64 samples at a sample rate of 44.1kHz. While a figure like ~3 ms (30% faster) would be more appropriate at 32 samples.

Jemusic wroteThey could improve the speed at which a virtual instrument responds because with my thunderbolt interface the best performance is not being quite reached. The output latency is still shown as just over 2mS.

I doubt you can interpret the latency reported by the driver that way. After all it still takes time to calculate a real time sound, no matter how fast the communication of the ASIO driver is. But I do think Studio One could be more responsive too.

Thanks for sharing your figures.

Egg's>>>>shells>>>>>Studio One's instrument's / fx devices get a new re-skinned shell to match Presence XT / Mai Tai ?

@jpettit. You're getting creative with the animated gifts.
Lol

Keep up the good work, and thank you for correcting the location of my post.
All the best.

Lokeyfly wrote@jpettit. You're getting creative with the animated gifts.
Lol

Keep up the good work, and thank you for correcting the location of my post.
All the best.

You came the closest to guessing the core of 3 .5
You almost won the fabulous prize!

We should run our series of tests now with the new improvements. I will repeat my tests and post up results as soon as I can do it. Should be interesting.

The 16mS latency setting certainly shows up with the thunderbolt interface connected.