KVR Audio

grease · Post by **grease** » Mon Sep 19, 2005 12:45 am

I have a Presonus Firebox, and while I have no complaints with its handling of audio, I believe it might have slower MIDI handling than other interfaces. I tested it with the utility called Miditest available here:

http://earthvegaconnection.com/evc/prod ... sults.html

Unfortunately I am getting reported MIDI message latencies around 17-18 ms:

Message latency: 18.21 ms
Message jitter: 1.29 ms
Message max deviation: 4.70 ms

For the most part, even latencies this high aren't too bad for real time playing, but I am considering getting a simple MIDI interface to use apart from the Firebox if I can get much lower latencies. So I am interested in your test results!! There are several test results posted at the Miditest site, but I don't see any data for some of the newer devices, such as a few small MIDI-to-USB devices made by Edirol and M-Audio and perhaps others.

Besides, maybe you would be interested in measuring your device with Miditest anyway...

Please share your results if you do!!

And send your results to the author of Miditest - he posts them on his site.

grease · Post by **grease** » Mon Sep 19, 2005 3:33 am

One thing to note, I've realized, is that when controlling VST instruments, one might experience only half of this reported MIDI latency (but there will be additional contributions to latency from audio drivers, etc.)

As far as I can tell, when testing something like the Firebox, Miditest measures the time it takes for:

1. Firebox receives data from firewire port and sends to its MIDI out port,
2. Firewire receives data from its MIDI in port and sends to its firewire port.

I would guess that these two tasks take about the same amount of time. Only task (2) is relevant for VST playing. Maybe this would only be about 9 ms if Miditest reports a message latency of 18 ms. Still, that's 9 extra milliseconds I'd rather not have!

Anyway...test those interfaces you have...the results would be interesting...

grease · Post by **grease** » Tue Sep 20, 2005 3:41 am

OK< I seem to be having a conversation with myself, but oh well...

I tested the Firebox again but with MME mode chosen instead of DirectMusic, and the results were a bit better:

Message latency: 10.74 ms
Message jitter: 1.14 ms
Message max deviation: 17.22 ms

Strange that the max deviation is so high but the jitter is decent. In any case, the Firebox is not the lowest MIDI latency interface available, but I suppose it is decent.

OK, I'll probably post another reply to myself later sometime just for fun...

Lunch Money · Post by **Lunch Money** » Sun Sep 25, 2005 9:26 pm

I always find it mind-boggling that there's any MIDI latency at all, in ANY system. I once calculated the potential throughput for USB 1.1 (as part of an old thread) and it was something like... you could send the data of an entire 3-minutes song arrangement 17 times in one second... That's a fully-arranged GM file with multiple parts.

I'm not saying MIDI latency DOESN'T exist... but I am saying that it SHOULDN'T exist. The data packets are simply too small and insignificant to be producing anything more than a half-millisecond latency.

Sorry... I'm no help (especially since I currently have no MIDI gear connected) but I find that high of latency to be completely unimaginable. What the heck did Presonus DO?

n-04:05:03:05:7 · Post by **n-04:05:03:05:7** » Sun Sep 25, 2005 9:57 pm

Hi grease,

Just did the test to compare your results with my Edirol DA-2496 PCI + 1U rack (8 IN/OUT audio + 1 MIDI I/O).
Default settings used in Miditest 3.1, tested with a good quality Korg MIDI cable.

DirectMusic

Message latency: 0.33 ms
Message jitter: 0.00 ms
Message max deviation: 0.04 ms

MME

Message latency: 0.84 ms
Message jitter: 0.22 ms
Message max deviation: 0.49 ms

As you can see, DirectMusic offers the best timings here.
Always had great real world MIDI timings/Audio latencies with this interface btw.

grease · Post by **grease** » Sun Sep 25, 2005 11:12 pm

Thanks for the replies in my thread! I was beginning to feel neglected.

n-04...damn, sorry I can't get your name right...thanks for posting your results - wow, what a difference. I'm wondering what contribution the firewire interface brings to this. I know firewire is supposed to be slower than PCI, but should it be that much slower? I don't know. It does seem strange as Lunch Money pointed out that there should be any kind of latency here...MIDI is such a tiny bit of data.

Well, I e-mailed Presonus about it a week or two ago but have heard no response from them...I'm not holding my breath either. If they just got back to me and said, yeah the choice of hardware components just doesn't allow for faster MIDI processing, sorry about that, then I'd be pleased with the honesty and then decide if it's really a problem anyway. I'm just wondering if it's a hardware or software issue, or whether the measured latency from Miditest is the same as what is obtained when I'm using my host program, etc...just looking for some more information from them but I haven't heard anything.

In the meantime, it does seem that I'm getting a fairly natural playing feel when playing something like the piano in GPO with the Firebox audio latency set to 1.5 ms. (Glad to have a fast system!) Whatever latencies I have don't seem to be adding up to something I can feel.

But still...it would be nice to remove all unneeded latencies!

flugel45 · Post by **flugel45** » Sun Sep 25, 2005 11:38 pm

Grease,

I'd heard that firewire adds to the latency time. I wonder if that's the case here, or at the very least, the firewire is throwing the numbers out of whack.

I took a look at the results chart (I'm assuming the two last entries are yours?), and it's hard to believe that this fairly new unit gives the worst midi performance of every other interface listed, going back to even an ISA card!

Ironically, I just ordered this unit this morning, and so as you can imagine, this is of great interest to me!

You might consider emailing the developer of the midi test, and get his thoughts on this.

JD

flugel45 · Post by **flugel45** » Sun Sep 25, 2005 11:58 pm

Okay, all of this got me thinking that if firewire did indeed add a large amount of midi latency, it would have been discovered by now, so a bit of Googling... and I found this SOS article:

http://www.soundonsound.com/sos/apr04/a ... cnotes.htm

Using the same utility mentioned at the start of this thread, Martin Walker measured a couple of firewire devices (not the Firebox), and determined that the added latencies to midi were negligible (certainly nowhere near the numbers Grease is showing).

So, either the Firebox is particularly bad in this regard - at least in measurable numbers - or the miditest utility has been changed since last year, and is now giving questionable results.

I'll run the test also when I get my unit later this week, and post back.

In the meanwhile, are there any other Firebox users who could take a few minutes and try it?

JD

grease · Post by **grease** » Mon Sep 26, 2005 1:18 am

Hi flugel,

I did e-mail the Miditest developer with questions, along with my test results - yes those appear to be my results. He didn't seem to think the readings were incorrect. But who knows! One thing he offered some explanation for was the DirectMusic/MME difference - here's what he said:

You might want to try running MidiTest in MME mode. You will probably get lower latency that way. Because the Presonus apparently has no real DirectMusic drivers (that's why they have the word 'emulated' behind the name) the latency is increased when accessed through DirectMusic. That is normal behaviour.
However, the application you are using will probably use MME anyway (unless it's Cubase SX or Nuendo).

Until now I have not seen many results of firewire devices so it's hard to tell what they are capable of. My guess would be that they perform about the same as USB devices.

I really have no idea how this stuff works...

But I must emphasize that I haven't really noticed the latency. Again note that if you are controlling VSTs then the timing of a full MIDI out to in loop is not relevant - it's only the MIDI in latency that matters. I'm guessing that this would be half of the total measured loop time, but that may not be correct.

Again, I haven't noticed the latency, and I don't know if the Miditest utility is accurate, but it is still a bit disconcerting to see such large numbers appearing from the test.

Otherwise the Firebox is quite nice - perhaps you'll never have any problem with it!

grease · Post by **grease** » Mon Sep 26, 2005 1:21 am

Oh - one other piece of information I just found...Presonus is located in Baton Rouge, LA, and was apparently hit hard by Katrina. They say on their site that they are back in operation, but this might explain their slow response!

flugel45 · Post by **flugel45** » Mon Sep 26, 2005 4:33 am

grease wrote: ....Again, I haven't noticed the latency, and I don't know if the Miditest utility is accurate, but it is still a bit disconcerting to see such large numbers appearing from the test.

Otherwise the Firebox is quite nice - perhaps you'll never have any problem with it!

I'm not particularly worried, as I do have a Midi interface I can use just in case, but you're right, those numbers are enough to make one wonder what's going on here...

And, like you say, what's important is that you don't "feel" the latency (I haven't read of any reports from anyone who has), so we're ultimately talking about a bunch of numbers that may mean nothing.

I've been in the market for an audio interface for a couple months now, and I'm glad I waited. The Firebox was on my short list, but the recent price drop made this decision a no-brainer.

Looking forward to making some music again !

Regards,
JD

C00kie · Post by **C00kie** » Mon Sep 26, 2005 9:04 pm

flugel45 wrote:those numbers are enough to make one wonder what's going on here...

In general it's no wonder that there is some delay, because the midi messages have to transfer over the firewire (wait for a data package to be constructed & sent, transmittal & decoding at the other end) and be translated by the drivers.

The measurements Grease has shown are clearly noticable delays when playing it. Maybe there are other things hogging the FireWire bus? Does it make a difference if the Asio buffer size is made smaller?

grease · Post by **grease** » Mon Sep 26, 2005 9:22 pm

C00kie wrote:In general it's no wonder that there is some delay, because the midi messages have to transfer over the firewire (wait for a data package to be constructed & sent, transmittal & decoding at the other end) and be translated by the drivers.

Yes, I suppose this is a good description of the origins of latency.

C00kie wrote: The measurements Grease has shown are clearly noticable delays when playing it. Maybe there are other things hogging the FireWire bus? Does it make a difference if the Asio buffer size is made smaller?

From my understanding, the ASIO buffers have no relation, as ASIO only controls audio. The MIDI part of these interfaces is completely separate in hardware and drivers. The data is transmitted down the same wire, of course...but I have no idea how this works!

Again, I must emphasize that the latency measured by Miditest measures the total time taken for a MIDI message to be sent out of the computer, received by the interface, sent out its MIDI out, received by its MIDI in, and sent on to the computer. Within this loop the interface must receive and send data twice (from firewire port to MIDI out port, then from MIDI in port to firewire port). When controlling a VST with a MIDI keyboard, the interface must send and receive data only once. So the 11 ms reading may not be experienced in full - perhaps only half of that is experienced during the playing of VSTs (in addition to plugin latency and audio latency).

C00kie · Post by **C00kie** » Mon Sep 26, 2005 9:41 pm

Since it goes through the same cable: if audio is playing, the midi messages must squeeze inbetween that stream. If it's not the same device but IS on the same cable, then you got your answer there. Replug in another FireWire slot that's not shared.

grease · Post by **grease** » Mon Sep 26, 2005 11:33 pm

C00kie wrote:Since it goes through the same cable: if audio is playing, the midi messages must squeeze inbetween that stream. If it's not the same device but IS on the same cable, then you got your answer there. Replug in another FireWire slot that's not shared.

Although my understanding is that only one of these firewire devices can be used at a time (but I'm not sure). Instead, maybe a USB MIDI interface can supplement a firewire audio interface. It is interesting, though - ultimately the computer can only do one thing at a time, regardless of how the data is received...it's unclear to me where the bottlenecks actually reside! Confusing stuff for a novice like me...

MIDI message latency - Test your interface please!!