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

well good luck jim.

as for the 'controversy' you can generally trust iS and myself to opine with fervour and delight if the opportunity presents - and we don't often disagree, so its nice. hardly anyone is wrong on the internet these days...must be the recession.

Rohan Stevenson1 wrote:
2) whether or not latency, or the differential is perceivable at distances of under 2 meters. i am deeply skeptical about that - that it makes any sort of difference at those levels. and this is probably why i sound defensive of software monitoring.

No need to have a hypothetical discussion about latency, though. Just hook up an analog board and compare. Set up a vocal mic, give yourself a healthy headphone level, then try singing with analog monitoring, then through a low-latency digital system (figure latency in the 1-3ms range).

I'd be very surprised if anyone didn't hear a difference. Most people will probably describe that the digital signal sounds "muffled" or lower in level, even when the level and eq are the same. The latency causes phasing with the direct sound in your head, and it's not exactly a subtle effect, even when the latency is just a few milliseconds.

Not to say that digital monitoring is always bad, or anything like that. But digital monitoring latency is real, and it's not something that requires super-human hearing to notice. If it doesn't bother you, then great! And, personally, I understand the difference and still use digital monitoring in some circumstances. But this reminds me of the "no-one-can-hear-the-difference-between-44k-and-96k argument" -- generally, the skeptics haven't actually listened for themselves. When was the last time you heard someone say, "I've tried both digital and analog monitoring side-by-side, and I couldn't hear a difference ?"

James

'd be very surprised if anyone didn't hear a difference. Most people will probably describe that the digital signal sounds "muffled" or lower in level, even when the level and eq are the same. The latency causes phasing with the direct sound in your head, and it's not exactly a subtle effect, even when the latency is just a few milliseconds.

think what you are saying. you are asking people to differentiate between 2 sounds 50 cms apart. if you have done this test and heard a difference then i suspect you are encountering a completely different issue. even phase is not a particularly big issue because you will encounter all sorts of phase issues in the real world (ie through air) as well, such as reflections from the room. i bet you wouldn't be able to tell them apart from a blind A/D.

Rohan Stevenson1 wrote:
well good luck jim.

as for the 'controversy' you can generally trust iS and myself to opine with fervour and delight if the opportunity presents - and we don't often disagree, so its nice.

'Zactly.

hardly anyone is wrong on the internet these days...must be the recession.

LOL!!

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...

FWIW, I have an 8-core machine, and have used a buffer of 32, but only at 48khz. The benchmark I used involved about 25 EXS instruments, and I just added instances of Space Designer until I hit the ceiling. I got up to 48. I know that doesn't exactly meet the track setup you need, but I imagine that at a less absurd number of SDs, you could boost the track count significantly and probably get away with 96khz.

Oh, also, I'm running LP 7.2.3, meaning that only 4 of my cores are being recognized. These 8-core machines are crazy...

Rohan Stevenson1 wrote:
think what you are saying. you are asking people to differentiate between 2 sounds 50 cms apart. if you have done this test and heard a difference then i suspect you are encountering a completely different issue. even phase is not a particularly big issue because you will encounter all sorts of phase issues in the real world (ie through air) as well, such as reflections from the room. i bet you wouldn't be able to tell them apart from a blind A/D.

Has nothing to do with 50cm. In the test I described, the phasing is between the sound of your voice that's heard immediately (mostly through direct bone conduction), and the monitored sound coming back through your headphones with digital latency.

In the case of instruments other than voice, the phasing is between the latent monitored signal and the acoustic bleed that comes through the headphones.

With quieter instruments and better isolated headphones, it's less of an issue. But for vocals, I've always found the phasing very distracting. Especially when recording and engineering myself at the same time--it's much easier to judge your own vocal tone in your headphones when there's no digital latency.

Again, this is something I've heard, many, many times. If you don't believe me, try it. Speculation is kinda pointless.

James

Jim Frazier wrote:
This is my main concern, to be honest. Before adapting a set-up that allowed me to take advantage of the RME's Zero-latency" software, I simply monitored through Logic, with an audio buffer setting of 128.
...
Based on it's sound, and what I know if chorus's and flangers, I'm guessing it was around 30-40 ms.

I don't know for sure, since I've never measured with RME hardware, but I don't think you were hearing that much latency. For comparison, with my MOTU devices, a buffer setting of 128 translates into about 400 samples of latency... at 44.1k, that's around 9ms. I try to record at a buffer setting of 32 whenever possible, and that gives me about 4ms of latency with my MOTU hardware. That's about the maximum I find acceptable, both for tonal and timing reasons, but I'm admittedly picky about this stuff.

In an A/B test, 9ms latency is enough to cause serious dulling of the monitored sound on vocals. Based on the way I hear things, that's about the maximum latency I'd expect anyone to tolerate without saying, "Hey, my headphones sound funny." But hey, maybe your clients were tolerating 30-40ms of monitoring latency... I've been wrong before 🙂

In fact, does anyone know what the latency is in Pro Tools? Because that's the number I need to be chasing, if I'm going to monitor through Logic.

Depends on what you mean by "ProTools." PTLE, for instance, will give almost identical latency performance to Logic, given the same hardware--in both cases, you're looking at software buffering of at least 32 samples in/out, plus the latency of the hardware conversion.

I assume you mean ProTools HD, though. I tested a PTHD system for latency once, and I think I remember it was around 2ms A/D/A, which is typical for a well-designed hardware-only digital system.

Speaking of ProTools, though, when I was testing the latency, I also tested the recording latency (that's the amount of delay you see on a recorded track--i.e. it's the amount a recorded track is "off" from where it should be on the timeline). In Logic, there's a setting to adjust for this--if you adjust Logic properly (search the forum for "delay compensation" if you're unfamiliar with this), you should have ZERO recording latency. Last time I tested ProTools HD, I saw over 1ms of recording latency, even when recording with no plug-ins active and ADC disabled. I was kinda shocked, and as far as I know, ProTools still doesn't have any kind of recording latency compensation... every time you overdub, a track gets placed slightly late. Bummer 🙂

James

I have a Quad 2.66 Mac Pro using the MOTU PCI card with 2 2408s and a 308 running into 2 Yamaha 02R96s. I don't use the converters in the MOTUs (I use a Crane Song Spider and Aurora converters). I use to use MOTUs cue mix direct monitoring software before I got the Mac Pro and the monitoring was very good. The problem is that the routing can be a nightmare in larger sessions.

When I got the Mac Pro Quad, I tried a buffer setting of 32. Using software monitoring, I found it very usable, still not as good as direct monitoring, but usable and with less routing frustration. But, and this is a big but, I started get stuttering in the audio. This stuttering would happen on playback, recording and when the system was idle just monitoring through a mic. It didn't happen often, maybe once every hour or two, but it became a deal breaker. If in record, the stuttering would be recorded and ruin that take. My Mac Pro Quad is not hooked up to the internet and does not have an airport card.

I am now going back to cue mix and direct monitoring. Very frustrated.

In an A/B test, 9ms latency is enough to cause serious dulling of the monitored sound on vocals

well, i take your point about phasing causing an issue with vocalists and buy it - just. but i was talking about apogees latency of 1.6 ms for round trip software monitoring at 96k and i had made the point earlier in the thread that 5ms (maybe revise that down to 4ms to be safe) is as near to zero as makes no difference - and i count phase with that as well. so it doesn't actually sound like we are that far apart in our opinions after all.

your figure of 2ms for PT was what i seem to be recall being the case, hearing that from my friend who has the symphony system and also a PT HD system.

just to be clear - most of my opining here comes from second hand but reliable information regarding the symphony and PT systems, but my views on latency and their detectability are my own.

...I agree.. I checked PT Mix plus once and it gave me 1.6ms recording latency. I suppose it´s the AD conversion, but still find it unacceptable for a professional recording system. Logic is better, but not always... If you rerecord a digital loop then it´s early...not Logic´s fault IMHO, but it happens.
BUT, if you have fxs in the busses or masters and activate the lousy PDC we suffer then ALL your recordings will be early for a mystery amount and also your MIDI and Video (and automation). This doesn´t happens in PT AFAIK. And please I knoe I shouldn´t record with PDC activated... But that´s because it´s wrong and that´s just an excuse from Apple.


Regarding hearing latency: I can work @4ms, but I always prefer to hear what I´m playing thru the console. ...and 40ms... give me a break... 😀 that´s unusable. I remember trying to play sofsynths in the 90s with that latencies and was impossible... not to mention to sing or play ANY instrument.

gpiccolini wrote:

BUT, if you have fxs in the busses or masters and activate the lousy PDC we suffer then ALL your recordings will be early for a mystery amount and also your MIDI and Video (and automation). This doesn´t happens in PT AFAIK. And please I knoe I shouldn´t record with PDC activated... But that´s because it´s wrong and that´s just an excuse from Apple.

The PDC system will have to be reworked from the ground up, not just tacked on as it has been. Unfortunately, this is one of Logic's features that doesn't "Just Work". I've used software manitoring a few times as this RME PCI card does pretty well at 64 or even 32 buffers... never found it to be an enjoyable experience and always end up going back to desk monitoring.

pancenter-

Jim Frazier wrote:
So.... if you were me (which you're not, but play along...) Would you:

The third option is an analog mixer... I see the tradeoffs as:

1) Monitor through Logic: most convenient, easiest routing options, most latency. Figure ~4ms with typical hardware and a fast Mac.

2) Monitor through your audio interface (the so-called "zero-latency" feature). Convenience level and routing options vary by device. Eliminates software buffering latency, but I've never seen an audio interface that mixed in the analog realm, so you're still going through A/D/A stages... figure 1-2ms latency.

3) Monitor through an external analog mixer. Requires more patching and knob-tweaking, so less convenient, but this gives true zero-latency.

Personally, I have an Ashly LX308B mixer (~$500) that I use when I care enough to want zero-latency. That gives me the most live and realistic headphone mix, but it requires an extra minute or two to patch the analog splits and set up a mix. When I'm writing or recording non-critical demos (probably 80% of the time I'm using my system), I just set my buffer to 32 samples, turn off any hoggy plug-ins, and accept the ~4ms of monitoring latency. When I'm doing real album tracks, I'll run splits off my mic pre outputs into the Ashly, which does sound better, especially when I'm engineering myself (without the latency, it's much easier to position mics in real time listening via phones).

I've never used a "zero-latency" mix feature that I thought was easy enough to use to be worth the effort. If I'm going to take a little extra time to get my latency down from 4ms, I'd rather use an analog mixer and get zero latency, rather than use a digital hardware solution to get 1-2ms of latency. You mileage may vary, of course!

James