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

Question

Level 5

5,333 points

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

Was wondering if any Apogee Symphony users could give me their impressions on using it, and Apogee converters with Logic.

The claim is there is 1.6 milliseconds of latency monitoring through Logic.

I'm about to pick up an Intel Mac (finally), and am seriously considering adding a symphony card and Rosetta 800 (to use in tandem with my Lavry converters) in order to get this kind of performance from my system.

Does this really work as advertised? 1.6 ms throughput latency would change everything for me personally. No more using latency-free software outside of Logic... that's REALLY tempting to me.

Thoughts?

Dual 2 gig G5, Mac OS X (10.4.11), 4.5 Gigs of RAM-Logic 8.01 -Lynx AES-16/Lavry 4496 converters/RME Fireface 800

Posted on Jun 25, 2008 5:36 AM

Reply

Answer 1

Jun 25, 2008 9:14 AM in response to Jim Frazier

i know of one person who has tried and tested this and confirmed the results. you have to run logic at 32 samples buffer and 96k in order to get this latency but it all worked flawlessly. at least he wasn't complaining about anything.

you use this in conjunction with non-latency inducing effects (that much is self-evident) and or with LLM switched on. if you have anything latency inducing it would simply bypass those effects.

i don't remember what flavour of mac it was but this was about a year ago or so. likely to be new but not brand brand new. someone else was telling me that the next batch of macs are going to be significantly faster - something to do with memory management and hyper threading (god i almost sound like i know what i am talking about), so if you wait a little bit that really could be the answer for you.

1.6 ms is around the time it takes sound to travel about 3m through air. that's pretty close to negligible latency.

Reply

Answer 2

Jim Frazier Author

Level 5

5,333 points

Jun 25, 2008 9:35 AM in response to Rohan Stevenson1

Thanks Rohan.

I guess a higher sample rate reduces latency? That seems backwards to me for some reason. Can you (or anyone) explain why that is?

I still can't fathom working at that sample rate when I get into track counts 50 or higher, which does happen fairly often. Can the 8 core handle that, with a reasonable amount of native plug-ins? And a buffer of 32? That seems unreal.. but I'm not used to what the 8 cores can do, CPU wise. Naturally, the low buffer would only need to be used when recording...

For various financial reasons, I'm gonna need to move on the new computer within the next week or two... so waiting for the "latest and greatest" is probably not gonna happen this time around.

Reply

Answer 3

Jun 25, 2008 12:44 PM in response to Jim Frazier

Hi Jim,

Yes at 96k you will get 1.6 ms of latency. At 44.1 you get about 4 ms. So the higher the sample rate, the better the latency.
I record everything at 44.1, monitor through Logic, and the 4 ms of latency is never a problem for me or any of the musicians I work with.

We have of course used Symphony on the new 8 Core machines but not yet done a "push to the limit test"
But here are the results of the last test we did using a 2.66 dual core. (Sorry this test is a bit outdated, but we are working on an 8 core test and will post on line when complete.)

Computer configuration:
Mac Pro, 2.66GHz Dual-Core with three internal 500 GB, 7200 RPM drives configured as a RAID and 4 GB of ram.

Software:
Logic 7.2.2 session running at 96k.

Track count during latency test:
32 tracks playback, 32 tracks recording

Plug ins engaged during latency test:
10 Adaptive Limiters, 10 Linear Phase EQs, 6 Space Designers (default preset)

Buffer setting:
32

Also keep in mind that we did these tests using an Apogee AD-16X and DA-16X. Since all AD and DA converters have different latency times, the latency may be slightly different using the Lavry.

The other thing to keep in mind is if you are using the Rosetta simply as a way to get digital in and out of the Symphony card, then It will not work. You can go digital in or digital out of the Rosetta from the Symphony system, but you can't do both at the same time. If you are going to use Lavry just as your A to D and use the Rosetta as your D to A (or the other way around) then you would be fine.
The only way to get digital in and out of the Symphony system at the same time, is to use a AD-16X and DA-16X.

Hope this helps. Please let me know if you have any other questions.

Chris Lawson
Apogee Electronics

Reply

Answer 4

Jim Frazier Author

Level 5

5,333 points

Jun 25, 2008 1:25 PM in response to clawson_apogee

Chris,

Thank you so much for jumping in.

Here is my (current) plan. Let me know what the pitfalls are.

I plan on getting a new 8 core computer (probably the 3.0 model). I then planned on getting a Symphony card, and a Rosetta 800.

I was told I could digitally connect my Lavry 4496's to the Rosetta, via the AES I/O.

My plan was to have the Lavry's as I/O 1-4 (because I have 4 I/O's of Lavry 4496's), and the Rosetta as I/O 5-8.

Outputs 3-8 are primarily to allow the talent to have some control over their cue mix, with those outputs being sent to a Hearback headphone cue system.

But you're saying I can't use my Lavry's A/D, while monitoring through their D/A's, if connected digitally to the Rosetta's AES i/o? That's certainly not a deal breaker, as I'd be more than happy to monitor from the Rosetta's D/A while tracking.

Can you see anything wrong with this approach? Am I missing something obvious? (I can't afford 2 16X's right now, and besides, I am rather smitten with the lavry's, as you can tell 🙂 )

Reply

Answer 5

Jun 25, 2008 1:31 PM in response to Jim Frazier

the reason higher sample rates reduce latency is because the buffer, which is 32 samples is over a shorter time. there are more samples per second at 96k ergo the 32 sample buffer is proportionately a shorter length of time.

Reply

Answer 6

Jun 25, 2008 3:30 PM in response to Jim Frazier

Jim,

No I don't see any problem with the set up you described. As long as you use the D to A of the Rosetta (as you mentioned) while you're tracking.
So the first 4 channels of A to D could be the Lavry going in AES and 5-8 would be the Rosetta. All 8 channels of D to A would be from the Rosetta, and if you haven't already heard the Rosetta, I think you will be more than happy when you hear it. You may even be "smitten" about the Rosetta.

Please don't hesitate if you have any further questions.
And as always with all Apogee products, our tech support will be more than happy to help you with the set up.

Chris Lawson
Apogee Electronics

Reply

Answer 7

Jim Frazier Author

Level 5

5,333 points

Jun 25, 2008 4:18 PM in response to clawson_apogee

Thanks again Chris. I do appreciate you taking the time to post here.

And Rohan, that explanation makes perfect sense now.

So let me follow this up with another question...

How many of you out there are recording, editing, and mixing, in the box, at 96k? What's your track count like? How many plug-ins are you using (on average)?

I personally hadn't planned on moving up to 96k yet. I'm on the side of the fence that just doesn't see the advantages of it in the big picture (when staying ITB)... but this latency card might be enough of a reason to consider it...

Reply

Answer 8

iSchwartz

Level 5

4,331 points

Jun 26, 2008 5:59 AM in response to Jim Frazier

Hi Jim,

Three things...

1) FWIW, the Apogee's latency specs are good, but there are audio interfaces that (per their published specs) exhibit less latency at 44.1/48 than the Apogee at 96. These include MOTU, RME, and Lynx, to name three.

2) As I believe has already been pointed out (or at least inferred), what's true for many (if not most) digital audio processors is that when you run them at higher sample rates, the overall latency thru those devices is reduced. For example, the Yamaha 01V96 has a published latency spec of 1.6 ms through that device at 48K, but only .8 milliseconds (exactly half) at 96K. Not that you can necessarily compare the sound quality of the A/D between the Apogee and the Yamaha, but right there is a better latency spec. So in short, that increasing the sampling rate lowers overall latency is not at all unique to the Apogee.

3) Jim, personally I don't think 96K recording is worth the extra overall "cost" in hard drive space, backup time (and expense), reduction in track count, time spent sample rate converting your 96K track to a deliverable medium...

So, as to "hype or the real deal"? The fact that Apogee touts this low latency is hype to the extent that the 96K sampling rate 'requirement' isn't made known up front in their advertising. Otherwise it's a real spec. And as I said, there are other systems that exhibit better latency specs at lower sampling rates, so IMO, therein lies the hype.

Reply

Answer 9

Jim Frazier Author

Level 5

5,333 points

Jun 26, 2008 6:43 AM in response to iSchwartz

iSchwartz wrote:
1) FWIW, the Apogee's latency specs are good, but there are audio interfaces that (per their published specs) exhibit less latency at 44.1/48 than the Apogee at 96. These include MOTU, RME, and Lynx, to name three.

Good timing on your post IS. I was wondering this myself since starting this thread.

Here's the deal... I currently use a Lynx AES-16 PCI card to connect to 4 I/O's of Lavry 4496 converters. I also use an RME Fireface 800 (as an aggregate device in Logic) to have access to it's 8 D/A's, so I can create headphone mixes.

The lowest I can set my buffer in Logic with this set up is 128. Latency is tolerable, but noticeable. For most musicians, it isn't an issue, but a few feel the "disconnect", and over time this has started to really bother me. It's far from "ideal".

So a few months back, I decided I would use RME's total mix cue software, and monitor through the RME when tracking. I use my Lavry A/D's (with a SPDIF converter box) into the SPDIF input of the RME Fireface. This allows me to get the benefits of my Lavry "sound" on input, while monitoring entirely from the RME, thus allowing me to use it's latency free software.

I am now completely hooked on NO latency for tracking. I can't go back, you know? But this set-up is far from ideal, as there is some tricky routing going on to pull this all off, and I would REALLY love to be able to monitor through Logic completely, eliminating the need for latency free software.

So... I started looking at options a few months back, and the whole Apogee thing came to my attention. At the time, I was simply going to get a Lynx AES-16 PCIe card, for the new Intel Mac but Lynx hadn't finished developing the PCIe card yet.

But since I started gathering info, the Lynx AES-16 PCIe card has become available, and I started wondering... if I bought that, and then, maybe got 4 more channels of Lavry D/A's, to have enough outputs for my cue system... if I could get record with my audio buffer in Logic at 64, or even 32, surely the latency would be negligible... to the point where it sounded non-existent, even if technically it was still there.

So is anyone running buffers that low in Logic, getting non-noticeable latency, AND not getting clicks and pops in the audio as a result of such a low buffer setting?

Reply

Answer 10

iSchwartz

Level 5

4,331 points

Jun 26, 2008 7:05 AM in response to Jim Frazier

Hi Jim,

So is anyone running buffers that low in Logic, getting non-noticeable latency, AND not getting clicks and pops in the audio as a result of such a low buffer setting?

For sake of comparison, though perhaps not a perfect A/B... I see you're on a G5. I'm also running a G5 (a Quad) and with my MOTU PCIe-based system I could run at a buffer size of 32. The only reason I'm currently running at 128 is because of VSL. In other words, if I was recording straight audio, I'd be able to use 32. But then again, I don't monitor through Logic. Maybe you've read my views on this before, but IMO the whole notion of software monitoring is a complete crock, and the only way around it is to use some kind of device which provides direct monitoring, as you're doing with Total Mix (RME). And clearly you see the benefit in this.

In my system I monitoring everything "live" through my Mackie D8 (which, BTW, not to slam Apogee, but it has better thruput latency specs at 44.1 than the Symphony system at 96, i.e., on the order of ~.6 ms at 44.1, and ~.5ms at 48K). So those are my "global" latency specs for monitoring everything I play. For some reason I'm not bothered by that, though I'm pretty sensitive to timing errors.

Anyway, back to the Buffer I/O settings --- I wasn't getting any pops and clicks in recorded audio at 32. But audio playback sounds much better at 256. The only downside to this is that changing the buffer size for mixdown throws off the timing of any live-running MIDI or virtual Instrument tracks. So it's necessary for me to print all of my VI's/outboard MIDI gear before going into mix mode.

Finally, and although a little OT, if you haven't done so already, I'd like to recommend that you do a loopback test to calibrate your recording delay setting in Logic. Doing this will let you confirm that your current recording delay setting is correct or not; and if it's not, setting that parameter to its optimum value will only further enhance the feel of the playback of your recorded audio tracks. (Post back if you want details).

Regards,

-=iS=-

Reply

Answer 11

Jun 26, 2008 7:39 AM in response to iSchwartz

iS,

i have to pull you up there - the apogee latency is talking about *software monitoring*. all converters add a small amount of latency, but what is being talked about here is *round trip latency* from mic to mac to logic through logic effects and back out to the headphones. you will not get this low latency with anything less than 96k when software monitoring and i doubt you will get it with anything other than a symphony system as it has been specially designed to do this.

the discussion as to whether to do this or not is a completely different issue. but this has long been a sort of 'holy grail' to those who would like an alternative to TDM and PT rigs. now, you can achieve similar latencies to PT HD rigs natively using native plug-ins and FX. and there are people who really want this.

while i agree that 96k is overkill in terms of quality - the optimal sample rate is something like 68k, which is a bit better than 44.1 - but the sample rate is how to unlock the low latencies natively. a lot of people favour 96k (i personally think 44.1 and 48 is fine) and generally there is plenty of bandwidth on modern systems to cope with pretty high track counts and big projects.

Reply

Answer 12

iSchwartz

Level 5

4,331 points

Jun 26, 2008 8:09 AM in response to Rohan Stevenson1

Rohan,

I think we can all agree that direct monitoring (or hardware monitoring) is the ideal way to go. Musicians for umpteen dozens of years have had the benefit of hearing what they played exactly in time with reference tracks by virtue of analog electronics and the virtually instantaneous transmission of sound source--->monitoring system. It's with digital systems that latencies are introduced. Even a 1.6 millisecond latency will cause the recorded audio of the world's best, tightest-to-the-click drummer to sound like he's flamming with the click in his headphones. And if you choose not to work at 96K, you're going to hear a much more distinct flamming against the click. Direct monitoring means no futzing, no BS. You hear what you hear without guesswork after the fact.

So unless I'm mistaken (and please correct the err of my ways if I'm missing something), when I'm comparing latency specs I'm talking about analog-->analog monitoring. Again, just for sake of example, with my D8B, the total monitoring latency for any live audio is a better spec than the Apogee at a lower sampling rate. I'm not dogging the Apogee. But their superlative software monitoring spec isn't any kind of free lunch (i.e., "hype") because it requires you to work at 96K to achieve that performance; and it's still not better than direct monitoring.

Reply

Answer 13

Jun 26, 2008 8:53 AM in response to iSchwartz

well, and i really am not defending one method over the other - i too prefer hardware monitoring (although actually in reality i am able to record musicians quite happily with software monitoring and a lot of latency compared to what we are talking about here - it depends on the material), but i think what you are missing is just how little 1.6 ms is, and that it is comparable to PT HD system whereby you can monitor with effects which i believe is a pretty standard way of working.

1.6 ms is how long sound takes to travel 55 cm at room temperature. about half a meter.

thats roughly the distance between a snare being hit and the ear receiving the sound. as far as i can see that equates to negligible latency. and this with software monitoring - ie passing through your computers cpu and adding effects to the signal, so you can have a little reverb or compression or eq or whatever.

Reply

Answer 14

iSchwartz

Level 5

4,331 points

Jun 26, 2008 9:14 AM in response to Rohan Stevenson1

To me, 1.6 ms is a timing hole big enough to drive a truck through. And that 1.6 ms would be a wholly artificial, additional latency on top of that 1.6 ms of sound-through-air that you wrote about. So now we're talking 3.2 ms of total latency. But to your point about using ITB effects... sure, there's an unquestionable advantage there, one to be taken advantage of most successfully at 96K which may or may not be desirable. YMMV, thanks in advance, my 2-cents, IMO, just sayin', blah...

😉

Reply

Answer 15

Jun 26, 2008 9:38 AM in response to iSchwartz

seriously mate, 55 cm is the distance sound has to travel from the instrument to the ear! now if you had your mic say 10 cm away from the drums then it would be no different than the difference between a drummer with short arms or a drummer with long arms!

there is latency every where. even your analogue desk introduces latency - and your converters if you are hardware monitoring. orchestral players deal with latency as routine.

staying with our drummer, if you measure the distance from the kick drum to the ear, it is llikely to be well over a meter. that means a mic can pick up the sound and send it to the computer, put a little eq on it, and have it back into the musicians headphones fractionally BEFORE the sound got there through air!!!!

it would be a pretty interesting truck that would fit through that little hole....

Reply