48k

[Tony Ostinato]

hi, i havent tried mobius in awhile and i was wondering if 48k support is working?


i remember from tryin it at 44 that it was awesome but live i run my laptop at 48k.

[Sjaak]

I run my Motu audio interface on 44 kHz but I didn't know about problems when using a 48 kHz interface. Just curious: why would you like to run on 48?

[Jeff]

The problem with 48K (or any other sample rate other than 44.1) is that the Mobius synchronization code was first written for the standalone version, and some assumptions about 44.1 are in the code, for example to take the
the length of time between clocks and calculate the ideal loop position
in samples.

If you aren't using any form of MIDI synchronization (In, Out, Out User Start,
MIDI Beat, MIDI Bar) then you should be able to run Mobius at other
sample rates, you just can't sync. You should be able to use Host Beat
and Host Bar if that helps.

There may be a few functions that don't work properly, one that comes
to mind is Auto Record which uses the Auto Record Tempo parameter to determine how long the loop should be.

This will be fixed someday, but for most people it hasn't been a big problem. If you play live the audience can't tell the difference between 44.1 and 48 anyway. For studio recording it's more of an issue.

[Tony Ostinato]

ahh ill have to try it, i dont even run my sequencer its just a realtime vsti/fx rack. mobius would be used loop and layer audio jams.

heres all of what im up to:

http://www.youtube.com/user/TonyOstinato


ive tried 44, 48, 88, and 96 and for trumpet techniques that ive migrated to windsynth like double and triple tonguing 48 is the sweet spot, 88 and 96 had too little latency and 44 too much.

[Jeff]

It sounds like you're confusing latency and sample rate.
They're related but should be approached differently.

Latency is normally controlled by the size of the buffers being used
to get packets of audio from your audio device, through the OS and
back out to the device. The smaller the buffers the lower the
latency.

The sample rate effects this because the higher the sample rate
the faster these buffers are get filled and in theory the faster
they move through the system. So in a perfect world, 96Khz with
a typical buffer size of 128 will give you lower latency
than a buffer size of 128 at 44.1khz.

The problem is that buffers are there to give the computer enough time
to get the audio through the system, if the buffers are too small then
you'll hear "clicks and pops" in the output. Raising the sample rate
to 96khz just means the system has to work harder since the buffers
are filling faster, and if it can't keep up you'll get clicks.

So in other words, in a system tuned for the minimal buffer
size at a sample rate of 44.1, you may have to use a buffer up to
twice as large to run at 96.

Instead of adjusting the sample rate look around in your sequencer
for an options window that lets you set the ASIO buffer size. Typical
values are 128 or 256 depending on the host and how many other
plugins you have running. There should also be a place nearby where
you can set the sample rate to 44.1. Try 128 first and see if you
get clicks.

If you tell us the host you're using I can provide more detailed instructions.

[Tony Ostinato]

setup is listed in the video.

7T29uWWjmCI

nuendo 3 with an asio buffer of 64 samples on a m-audio firewire 410.

i dont route in or process any input audio, i just trigger off vstis via midi input.

right now i use angstrolooper when i do looping and thats worked well.


trumpet itself has a little latency because of air resistance and thats where i learned tonguing so 48k@64 samples ended up being the perfect feel. 88 and 96 just tripped me up by being way too quick. 44 again was too slow and 48 i can play all the classical stuff i can on trumpet, better even.

my vsti loadout is setup so i could run at any of those settings but 48 gives me a little headroom which is nice playing live.

[Jeff]

Wow, you're the first person I've ever met that complained
about not having *enough* latency :)

I think I understand though. Accoustic wind players train themselves
to play slightly "in front" of the note to compensate for the inherent
latency of the instrument. When you switch to an electronic instrument
that latency is gone so the compensation you're accustomed to
overcorrects.

I noticed this when swapping between a sax and a WX7 way back when I
still played sax, but I guess I found the difference so slight it was
easy for me just to adjust my mental compensation.

By my calculations, at a 64 sample block size we're talking about
the difference in in-to-out latency of 2.9 milliseconds for 44.1
and 2.65 milliseconds for 48K. Or an effective difference of .25
milliseconds. That is extremely small and for most people
beyond the limits of perception.

By comparison "speed of sound" latency is around 1 millisecond
for 34 centimeters (at sea level). .25 milliseconds corresponds
to 8.16 centimeters or 3.2 inches.

So to bring this back around, the difference between a 44.1 and 48K
sampling rate is the same as standing 3 inches farther away from your
monitor speaker. If this throws off your technique, congratulations
on your super-human accuracy :)

I have to wonder if Nuendo isn't doing something funny under the hood
when you change sample rates that makes it have a much more pronounced
effect on latency. What happens when you bump the buffer size
up to 128 and leave the sample rate at 44.1?

[Jeff]

Just to be clear here, are you talking about latency just playing a wind controller through VSTi's without Mobius involved, or are you talking
about the alignment of overdubs with the backing loops when you
use Mobius? If the later, since you are apparently monitoring "through the computer" you will want to disable Mobius latency compensation.

Bring up the Audio Devices dialog and find the two latency override
boxes, enter 1 in both boxes and save. Mobius latency compensation
is used when you monitor outside the computer, either with an
accoustic instrument, with an analog mixer side chain, or with the
"zero latency" monitoring option of the audio device. In that case
Mobius has to compensate to keep overdubs aligned with what
you hear in the montor mix. But when you monitor from the output
of your audio device you don't need this.

[Tony Ostinato]

im not using mobius yet so ill try those settings when i do. i last tried it like 2 years ago and i seem to remember some problem with loops lining up or something so i thought id ask before trying again. i have been using angstrolooper in the meantime but i remember mobius having a ton of extra features.

i cant speak much to the physics or anything superhuman. i just havent been able to do this kind of tonguing:


http://www.wallanderinstruments.com/DownloadDepositoryFile.php?id=412&mode=preview

on any other settings. definately not 128 samples at 44. i have to admit i didnt try 128 samples @88 or 96 because i reckoned it would be the same latency roughly as 64 samples @48


ive seen comments about latency with grand piano players and that they like 8ms because thats close to the mechanical action times.

what it was for me was a about 2 months of setting things up and then running classical pieces like concert etude and the haydn and hummel concertos and what would happen at other settings is id end up stumbling all over it, i dont even think i could get out 8 notes in a row, something about the muscle feel just wasnt lining up with the sound the way i expected and that would throw off the next note and blam.

double tonguing and triple tonguing were always tough for me anyways, its a lot easier on windsynth now.

[Jeff]

This sounds like sixteenth-note triplets at 140 BPM (nice playing BTW!,
and I really liked the video).

That's about 71.4 milliseconds per triplet and what we're talking about
is sliding those notes forward or backward by .25 milliseconds that's
almost 1/300th of a triplet.

Look, I'm sure you're hearing/feeling *something* but I'm skeptical
that it is being caused just by changing the sample rate from 48 to 44.
It's hard to say what, there are several parts to your system each contributing their own amount of latency (it takes at least 1ms if not 2 just to get a MIDI event from the WX5 into Nuendo). The attack envelope
on your VSTi's usually have far more effect on "feel" than .25 ms.

Whatever the reason, you should be able to use Mobius at 48K, just
click Ok on the little warning window. The only thing that won't
work correctly is MIDI clock sync, which it doesn't look like you
need anyway.

Let me know if you find any problems at 48K, other than sync they
shouldn't be hard to fix. You will want to disable Mobius latency
compensation however (set both input/output latency overrides to 1).

Good luck!

[Jeff]

BTW, here's a good article from Apple on calculating and adjusting latency,
it applies to any host.

http://support.apple.com/kb/HT1314

[paradiddle]

Just wanted to post my experience:

I'm using a p4 with sblive (cuz my main pc with the delta and audiophile crashed) and it's running at 48khz. I tried running mobius at that sample rate and when I load it up (version 1.40) in audiomulch, I get some problems. Like if I load it and delete right away, it will crash Audiomulch. I tried the same thing in Energy XT2 but it wouldn't load it complaining about the sample rate. When I set Audiomulch to resample at 44 khz then Mobius would work ok no problem.

Nothing really special but if somebody has crashes with Mobius, it might just be related to the sample rate.

That's all folks!!!

[Newbie Brad]

Thanks so much for the look at what you are doing. Forwarding your YouTube to my partner Pantha the Akai 4000S player.