Apogee Symphony Users... 1.6 ms of latency? Hype, or the real deal?

Oops sorry I lied. This was true at one point, but now we're at 128 with no problems. Just tried 64, and it seemed to work, but I don't want to tempt fate.

Forget everything you've heard and go back to sleep : )

Ekstasis,

It doesn't surprise me that higher (or "high") buffer settings would cause a decrease in performance. Based on the performance of my system (PPC Quad, running MOTU gear):

32 = playback of pure audio tracks with various native Logic plugs (EQ, compressor) is no problem. EXS-24 playback, no problem. GPO (Garritan Personal Orchestra using the original, pre K2 player) isn't happy, requires upping the buffer size.

64 = everything sounds a little better! Can run more instances of plugs like space designer, GPO no longer complains, various other libraries like EWQLSO are a little pokey but OK. But VSL isn't happy, requires upping the buffer size.

128 = everything starts to sound even a little bit better! Everyone's happy, including VSL! But ArtsAcoustic reverb isn't happy. Timing of MIDI/virtual instrument tracks start to become a wee bit questionable. Requires...

256 = wow, everything sounds bleeping glorious! Better high end definition, rounder low frequencies. ArtsAcoustic breathes! But MIDI/virtual instrument tracks start to be come very very very noticeably out of time.

512 = system starts to go apoplectic, automation goes whacky, playback will actually stop after a few bars of intense (orchestral) playback

1024 = system will not play back anything (except for various digital farting noises)

sounds like an excellent experiment that will put an end to all of this fantastic debate. I hope !!

well no actually it won't. because it begs the question:

which one is late and which one is early? i guarantee not even iS's velvet ears will be able to tell them apart.

because whether you are analogue or whether you are digital how do you know which is which?

you will always have this problem and possibly even MORE so with analogue. the reason for this is that digital may actually offer enough delay to compensate for the sound arriving far too early if it was run through zero latency, speed of light hardware monitoring.

1.6 ms is about 50 cm. if you are a bass player or a drummer, you might be suffering from the fact that your audio in the cans is arriving before the sound in the air because the mic is closer to the sound source than your ears are.

therefore, in those scenarios the latency is beneficial.

i tell you what, i would most definitely rather hear the sound i make late by 1.6ms than early by 1.6ms.

IF latency is affecting the performance than absolutely iS and others speaking against software monitoring are completely right. BUT - and this absolutely key - when you talk about latency you are really talking about a delay from what you hear in monitored mix in your headphones to what you hear and feel when you play something.

by dint of mathematics you can clearly see that in very many cases there is NO latency AT ALL between the 2 sound sources, if the sound source is around 50 cm away from the ear and the mic is close to the source. it could actually improve sync between the 2 sources than if you were using analogue because there would be latency in the sound travelling through air. and that could be enough to put off a performer....

wow, that's really interesting. i have had my setup on 512 and haven't really ever bothered to change it. i don't really have any big problems with latency on the VI's and everything seems to work fine.

weird. i don't really get this sometimes. i know that FW interfaces cause latency issues - i am PCI as well, but i don't understand why your system would go funny going up in buffer size like that - although i have heard of this.

Rohan Stevenson1 wrote:

sounds like an excellent experiment that will put an end to all of this fantastic debate. I hope !!

well no actually it won't. because it begs the question: which one is late and which one is early? i guarantee not even iS's velvet ears will be able to tell them apart.

Rohan, with all due respect (and I think you know that in the iSchwartz household that runs pretty darn tootin' high), the purpose of this test is not to see which one is ahead or behind. It's to see if you can hear the "smearing" that occurs when two tracks -- which are supposed to be in sync -- are ever so slightly out of time.

That's the only purpose of the test. Nothing more.

If it smears, it's bad, and reflects a timing error that shouldn't be there.

IF latency is affecting the performance than absolutely iS and others speaking against software monitoring are completely right. BUT - and this absolutely key - when you talk about latency you are really talking about a delay from what you hear in monitored mix in your headphones to what you hear and feel when you play something.

First of all, I'm glad we're in agreement with this.

And now I'll tell a simple tale that I think I've told before, but perhaps it bears repeating:

3 or 4 years ago when I got my G5 and Logic 7 -- a big leap from Logic 4 on a G4 -- I went the software monitoring route (this was also back when I had my analog board). Looking to test for the optimum buffer size, I set it at 32 for starters. Playing external MIDI synths (which was all I had at the time), everything I played felt delayed, mushy, weird. Like Jim discussed above, I started to think that maybe I was beginning to suck. I thought, "maybe this better gear is revealing how bad my timing has gotten".

Then I got to thinking about the "mechanics" of it all. Audio, gets digitized, through Logic, back out of Logic, converted to analog... Can't happen instantly... Hmmm... *

So I turned off software monitoring and instead, monitored directly from my board (like I had always done previously). Bingo! My timing didn't suck anymore! My playing didn't feel mushy any longer. And the phenomenon was soooooooooooooo immediately obvious.

Again, this was at a buffer setting of 32!!!


"*" = denotes iSchwartz deep in thought

Rohan Stevenson1 wrote:
weird.

Yup. Equally interesting for me to see that you don't have this problem. Warrants further investimigation, for sure. Could it be a MOTU PCI/Audio Units thing (he asked rhetorically?)

BTW, that's running L7.1.1. I haven't yet tried it with L8.

Cheers!

iSchwartz wrote:
Again, this was at a buffer setting of 32!!!

A buffer setting of 32 doesn't tell you what the latency is, it could have been very low or it could have been much higher, especially if it was a firewire box instead of a card.

Tests are all good and well, but have you tried tracking while monitoring through a system with 1.6 ms of latency?

The question isn't whether you can hear phasing when you delay a track against itself, it's whether you can hear a delay between the acoustic sound and the headphones. They are two completely different things.

ok, well
we can keep on arguing so that one of us claims the win and say " I told you so.. I was right" :))
anyway, the point is I don't think that I seen any player who can play in perfect sync to 0.8 ms before, but that could be my own problem.
why 0.8 ms ? well because 1.6 reflects input and output, so a middle ground could be :
when over dubbing an audio track, monitor current instrument being played through hardware, while monitoring previously recorded material through software (i.e playback) with whatever plugins, this way you are closer.
of course you can argue that what ever latency the system introduces is added to the imperfections produced by the player him/herself. but I think that at such low latencies we can be close enough to be able to perform efficiently and perfectly.
Now we regard to playing back stuff with plugins, as had become repeatedly evident, different plugins require different buffer sizes and will introduce varying levels of latencies if the buffer sizes are not large enough. so you have to use common sense before trying to play back such plugins and perhaps you should freeze these tracks before hand.
for example the linear phase EQ will introduce more latency if you use higher resolution for the FFT, makes sense ?
so typically, i would freeze such a track after I adjusted the EQ
then lower the buffer size to 32 again
then record new tracks
then increase buffer size back to 256 or so
then unfreeze EQ track
then carry on mixing and adjusting.

This process is working for me on a macbook pro C2D with MOTU traveler, and although the traveler doesn't have pre amps that tough Apogee, their AD/DA are reasonable quality (lower jitter than similar interfaces in same price range) and they also offer one of the best drivers short of Apogee and RME (as I heard from people)
I'm generally happy with these process, but of course there are scenario's when the latency thing is a pain in the arse and I agree with IS's point that in many scenarios you can't ignore it.
for me I'm not willing to give up all that ITB have to offer for a complete OTB solution that will coast a lot more than I can afford now, but some combination of both seems to work best at this stage in my studio.
I look forward to the technology that can make ITB work exactly like OTB.
I also wish someone from Apogee could come in on all of that and tell us their thoughts and not simply try to defend a product.
either way it is always great debating with you guys, so much to learn from all of you
cheers

Ha.

I'm not a hopeless guitarist but I'm not going to win any polls either. I actually prefer playing thru software monitoring with 64 or 128 sample delay. I enjoy the feel it replicates of when I was playing at a venue with the amp a good 20 feet away. That...delay is liberating somehow.

This is all a tempest in a teapot, imo. We deal with delay and phasing all day everyday and we have been dealing with delay induced by analog devices in the signal chain for as long as there have been such things. The fact that we can now quantify it and make it so -visible- within the digital domain makes it easier for us to gnash our teeth and shake our fists at the lords of digital audio for creating this evil source of new delay & phasing and whatnot.

I don't dispute that anecdotal issues exist with latency in certain circumstances. But I am extremely dubious that any inherent difference exists between 1.5ms of digital latency vs. simliar latency induced by distance from the sound-source or analog devices in the signal chain.

The whole mix-and-monitor ITB vs. external monitor-mixing debate is academic and the ultimate technical configuration is forgone. Does anyone seriously doubt that we are heading, quickly, towards a 100% In-The-Box music recording/mixing/production world?

Sorry man, but the buffer size directly affects Latency and although Latency is a combination of that plus other aspect such as AD/DA converters and Drivers, at the beginning the formula

Latency = [ (IO_Buffer Size/Sample rate) * 2 ] + All other latencies introduced by drivers, AD/DA, etc..

this means that as the unit is better designed we can make the later part of the equation approach 0, but we still have the first part of the equation as our limiting factor..

by the way, this formula is not from my head, it is from Apple Notes with regard to Latency, and makes perfect sense as I have used it during my programming endeavors in Max/MSP and Java.

@ Jim: thanks for telling me about your system/buffer sizes.

@ Mike:

A buffer setting of 32 doesn't tell you what the latency is, it could have been very low or it could have been much higher, especially if it was a firewire box instead of a card.

Of course it does! Figure 32 in, 32 out. That's 64 samples. At a sampling rate of 48K that's 1.3 milliseconds. At 44.1K it's slightly more delayed. No, that doesn't include additional A/D and D/A latency. But that small # of milliseconds is a quantifiable amount of time and it made a difference. That was the moral of my story from my previous post. In short, YES, I did the test.

Tests are all good and well, but have you tried tracking while monitoring through a system with 1.6 ms of latency?

Why are you asking this question? I already posted that I've had this experience, and the results.

Now, to my test... If you would please perform my simple test (which I'd encourage others to do as well) and report back what you hear, then we could all speak from a place of experience instead of (potentially) coming across as a bunch of arm-chair scholars.

The question isn't whether you can hear phasing when you delay a track against itself, it's whether you can hear a delay between the acoustic sound and the headphones. They are two completely different things.

No, they're not different things. Delay is delay is delay.

Rohan Stevenson1 wrote:
IF latency is affecting the performance than absolutely iS and others speaking against software monitoring are completely right.

We happily agree on this, then... the proof's in the pudding, so to speak. If you hear latency and it has a negative effect on performance (I answer "yes" to both of those from experience), then it's a problem, regardless of how one might try to explain it away. Maybe if we can get the latency low enough, I won't be able to hear it. But that day has not come yet. Personally, I am "deeply skeptical" that my ears will ever not hear a difference... when I can be fooled in an A/B test, I'll concede the point... but I've heard 2ms, and I'm still sticking to my story 🙂

BUT - and this absolutely key - when you talk about latency you are really talking about a delay from what you hear in monitored mix in your headphones to what you hear and feel when you play something.

This is true, but it's not the whole story. We haven't even touched on the subject of monitoring versus what's actually being recorded. That's why I object to the idea that analog monitoring is ever "early." True, it's "early" in the sense that you hear something through a close mic sooner than you might hear it acoustically in the room, and that might cause phase problems with headphones. But when the monitoring is analog, what you're hearing in the headphones (and in the control room) is exactly what's being recorded. If you monitor with digital latency, playback will actually sound slightly different than tracking--the part just monitored will play back earlier than it sounded in the monitors while tracking. That, in my opinion, is a real problem, both for performers, and for producers/engineers trying to judge timing and feel from the desk. And I think it's why players with precise timing tend to have more problems with rushing when monitoring through digital systems--the slight latency in the phones causes you to compensate to make things sound good in the moment, and then upon playback, your part gets pushed forward in time by a few milliseconds.

I'd argue that maybe the "delay compensation" feature should not be set the same way if you're going to use digital monitoring. In all facets of musical performance, timing is about adjusting your technique to make things sound good... relative to the downbeat, you blow a tuba earlier than you press a piano key, for instance, since there's more natural "latency" in the tuba. Good technique means knowing how to make the sound come out at the right time.

So, if I'm going to hear my instrument sound late in the headphones, I'd rather it at least was going "to tape" that way, because it's affecting the way I play the instrument.

Things are much simpler with analog monitoring--you can argue all day long about mic and instrument position and "natural" latency, but the fact is, with analog monitoring, what you hear is exactly what's getting recorded. That's not true for digital monitoring.

(And this is all without getting into ProTools, which doesn't have recording delay compensation to adjust for A/D and D/A time... on ProTools, if you monitor in-the-box, your recorded track gets placed late relative to the actual event, but it actually gets placed early relative to the latent monitored sound... who knows, maybe this is actually a good compromise, although it means you have to slide tracks around when you patch through external gear).

James

BC Productions wrote:
I'm not a hopeless guitarist but I'm not going to win any polls either. I actually prefer playing thru software monitoring with 64 or 128 sample delay. I enjoy the feel it replicates of when I was playing at a venue with the amp a good 20 feet away. That...delay is liberating somehow.

I can totally relate to this. The problem, though, is that the sound is going "to tape" earlier than you're feeling it. In a way, it's similar to standing onstage 20 feet from your amp, while your sound is going into the PA via a mic touching the speaker grill. I think that common scenario probably causes us to sound slightly rushed through the PA, as it similarly causes the recording to sound rushed even though it didn't in the digital cue mix. And in an age of in-ear monitoring, the 20-foot-amp-delay is becoming less and less common even onstage.

The whole mix-and-monitor ITB vs. external monitor-mixing debate is academic and the ultimate technical configuration is forgone. Does anyone seriously doubt that we are heading, quickly, towards a 100% In-The-Box music recording/mixing/production world?

Fair enough. But does anyone doubt that we're also heading toward a world where the iPod mp3 is the gold standard for audio quality ? That doesn't mean that those of us who believe in the beauty of sound are going down without a fight 🙂

James

I don't dispute that anecdotal issues exist with latency in certain circumstances. But I am extremely dubious that any inherent difference exists between 1.5ms of digital latency vs. simliar latency induced by distance from the sound-source or analog devices in the signal chain.

I wouldn't for a second dispute your experience -- that you prefer monitoring with delay. Actually, who could argue with that? You like what you like. There can be NO argument.

However, when you say that you are dubious of the 1.5 ms latency experience you utterly contradict yourself, and your point becomes invalid. Why? Well, y'know that buffer size you spoke of -- 64 samples? That IS about 1.5 milliseconds!