Ableton Forum

LoopStationZebra wrote:What do you mean by full experience? I can think of some features I'd like to see that would truly mean never looking at my laptop screen again, but what are you after specifically?

Midi clip step sequencer in TouchAble, or mixer+macros etc in Liine.

ze2be wrote:

LoopStationZebra wrote:What do you mean by full experience? I can think of some features I'd like to see that would truly mean never looking at my laptop screen again, but what are you after specifically?

Midi clip step sequencer in TouchAble, or mixer+macros etc in Liine.

The mixer got added to the latest version of Griid. Macros is probably something they're investigating now as Live 8.2.2 removes all device navigation limitations.
If you have MFL, I got something even sweeter that'll work in rack presets, sets and Live clips.

As for syncing external apps with a computer, its never going to happen using MIDI.
MIDI has a latency of over 100ms (1/64th at 120bpm). OSC has like 3ms in a round-trip.
Trying to sync with MIDI Clock isn't going to work unless coremidi gets faster.

hoffman2k wrote:The mixer got added to the latest version of Griid. Macros is probably something they're investigating now as Live 8.2.2 removes all device navigation limitations.
If you have MFL, I got something even sweeter that'll work in rack presets, sets and Live clips.

Sweet, ill have a look at Griid.
Cant afford MAX before I cut down on working with music. Ironic, isn't it! (This thread aint the place for the full story.)

Yeah, about midi syncing apps. I don't really care that much about it. What really matters is getting a great, full fledged DAW control instrument for Live. One that knocks Mashine out.

hoffman2k wrote:
As for syncing external apps with a computer, its never going to happen using MIDI.
MIDI has a latency of over 100ms (1/64th at 120bpm). OSC has like 3ms in a round-trip.
Trying to sync with MIDI Clock isn't going to work unless coremidi gets faster.

touchAble and Griid are essentially syncing now, aren't they? And taking pings or something from Ableton at regular intervals to stay in sync? I guess I'm not following you here because clearly something is connected that allows these apps to respond perfectly with Live's tempo.

LoopStationZebra wrote:

hoffman2k wrote:
As for syncing external apps with a computer, its never going to happen using MIDI.
MIDI has a latency of over 100ms (1/64th at 120bpm). OSC has like 3ms in a round-trip.
Trying to sync with MIDI Clock isn't going to work unless coremidi gets faster.

touchAble and Griid are essentially syncing now, aren't they? And taking pings or something from Ableton at regular intervals to stay in sync? I guess I'm not following you here because clearly something is connected that allows these apps to respond perfectly with Live's tempo.

They're not using MIDI and they're not really synced. 90% of what these things do is done in the python scripts that are directly connected to the Live API. Anything timing related is done by polling Live. You get the visual info like position indicators, but there isn't actually a clock running on the iPad.
These apps are completely different from a Step Sequencer that does its timing on the iPad.

hoffman2k wrote:

LoopStationZebra wrote:

hoffman2k wrote:
As for syncing external apps with a computer, its never going to happen using MIDI.
MIDI has a latency of over 100ms (1/64th at 120bpm). OSC has like 3ms in a round-trip.
Trying to sync with MIDI Clock isn't going to work unless coremidi gets faster.

touchAble and Griid are essentially syncing now, aren't they? And taking pings or something from Ableton at regular intervals to stay in sync? I guess I'm not following you here because clearly something is connected that allows these apps to respond perfectly with Live's tempo.

They're not using MIDI and they're not really synced. 90% of what these things do is done in the python scripts that are directly connected to the Live API. Anything timing related is done by polling Live. You get the visual info like position indicators, but there isn't actually a clock running on the iPad.
These apps are completely different from a Step Sequencer that does its timing on the iPad.

Yeah, okay. That's how I understood it to work with my 'ping' reference - there's no clock involved. Why couldn't that work with any kind of app? Sequencer? Synth with tempo based FX? The app would be connected to a script - just like touchAble and Griid - and not rely at all on the clock.

I'm still lost on why you think sync would not be reliable? CoreMIDI is already fast, lol.

Where do you get your 100ms number for MIDI latency? That's a huge number.

The iConnectMIDI folks are reporting less than 1ms latency between their device and the iPad. They didn't say if anything else was connected on the other end, however. But, let's say it's iPad-->iConnectMIDI-->MacBook-->AbletonLive. I'd expect no more than 2 to 3ms total, if that.

And what of hardware? The interface that one is using factors in heavily with latency amounts; which is why people are so excited over iConnectMIDI.

The main issues I've seen on apps that support clock are:

1. Poor implementation in the app itself. I know of no app that utilizes OSC for the clock, but even with apps that use wifi OSX MIDI the latency is horrid - way beyond what should be normal (2 to 5ms). It's faltering at the 2nd bar, ffs. lol. That translates into seconds of latency instead of miliseconds. Clearly the implementation is incorrect.

2. Hardware. With the CCK/USB, you are totally at the mercy of the unreliable USB-to-MIDI cable - which have all shown to be wildly inconsistent in performance and results. Many of those (including my own M-Audio UNO) don't support clock at all.

So, to summarize, we should be getting 2 to 5ms for OSX wifi MIDI, 1ms from the Camera Connection Kit / USB, and hopefully less than 1ms with the iConnectMIDI box. What we don't know is the latency of the iODock from Alesis.

I can't see any reason why reliable clock sync wouldn't be possible.

Just try this simple test: Route a MIDI clip from one track to another using the IAC driver. Add that result to the 3ms latency of the iPad.
Virtual MIDI drivers are causing these big delays. That is why it has always been a pain in the ass to sync 2 laptops. The same bottleneck remains.
Seems like CoreMIDI is more efficient on the iPad than the Mac, but the only way to verify that is to be able to set it up as a MIDI thru and check the delay of the roundtrip.
Maybe Charlie's app can help here?

How are each of you determining the response times? What applications are being used to monitor these times?

I guess my question comes from my networking background. If you "ping" something from a command prompt or whatever on a local network, the "ping" command may say "<1ms" which means under 1ms. If you actually run a sniffer trace on the packets, those times are exponentially lower.

Like .001 ms between hosts. I see it. Right now. On my screen. In wireshark.

IMHO, the network "medium" is not the bottleneck. What, then, is the issue?

I'm pleading ignorance on the Midi specific protocols on the machines in question, this is more for my own education, so educate me

I agree, its not networking that is the issue. Its MIDI conversion I think.
I'm still looking for the right tools to perform some tests. Looks like only the Control app would allow you to use it as a diagnostic tool. Most of the other apps I've tired so far don't have a throughput.

I was way off with my earlier guess of 100ms with the IAC driver. 1/64th at 120 bpm is 25 milliseconds. Forgot the basic math, but here's a verification.
This is simply a note recorded from one clip to another using the IAC driver. The note should fall on 1.3.



Any javascript wizards reading this? With a little script we can probably test OSC and MIDI. It would be a useful tool to have, if only to compare the timing of an Ad hoc with a home network.
This free app can be used for this: http://charlie-roberts.com/Control/
The timing can be tested just like the test above.

Yes I wonder about test results as well. When the iConnectMidi guys say they are getting 1ms or less latency, they don't go into any detail as to how that was measured. When asked about it in threads they don't respond.



Yeah, I wonder if Charlie would be willing to write that script?

I'm on vacation right now... steadfastly avoiding all javascript. But hopefully in a few days... what would the tests look like you guys have in mind?

bigbadotis wrote:I'm on vacation right now... steadfastly avoiding all javascript. But hopefully in a few days... what would the tests look like you guys have in mind?

Hey Charlie,

We could use 2 scripts. One for testing the OSC, a message comes in and is sent out immediately.
The same for MIDI, but using the network MIDI.
If the iPad has a CPUclock, any data from it would be nice too.

Cheers

- Bjorn Vayner

Yeah, thanks Charlie!

OK, sure... should be pretty simple. I've used Control for doing percussive stuff with Ableton (via MIDI) and I would be surprised if the latency was 25 ms; for me jitter seems to be a bigger problem than latency, even using an adhoc network.

I just repeated Bjorn's test only using IAC. I get slightly less than a 32nd note offset at 999.00 BPM, consistently, no jitter. That's about 1.87 ms of latency. Not sure what could be causing the discrepancy.

bigbadotis wrote:I just repeated Bjorn's test only using IAC. I get slightly less than a 32nd note offset at 999.00 BPM, consistently, no jitter. That's about 1.87 ms of latency. Not sure what could be causing the discrepancy.

Weird... How did you measure 1.87ms? At 999bpm that works out to a 1/16th delay for me. Which is still almost 25ms.
I'm using OSX 10.6.7.