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

Hi James,

Yeah, I thought it was a unit problem... Thanks for that.

And one of my units was wrong as well: I wrote 0.010416 seconds -- that should have read "milliseconds", as you corroborated.

Yes, if A/D could be sped up then it would help (perhaps by having A/D's that sampled on the 1/2 cycle of the wordclock and settled on the other 1/2 cycle) . But even still, there will always be a little bit of hardware latency thats unavoidable (as has been pointed out).

Here's some food for thought that your post cooked up in my noggin... Here's a hypothetical illustration of minimal A/D --> D/A latency:

*0 samples* = signal arrives at A/D
*1 sample* = the instantaneous level of that signal is sampled and calculated (let's call that sample "X"). Let's assume that it takes one full word clock's duty cycle to sample and then settle the A/D. No half cycles here, no oversampling.
*1 sample* = transfer time out of the A/D. _So far we've incurred 2 sample's worth of delay_. Next...

While the next instantaneous level of signal is being sampled, "X" is being transferred to the computer, incurring another 1 sample's worth of time for the computer (audio driver software) to capture that sample and route it to RAM. Maybe there's another 1 sample's worth of time to make Logic aware of it. So _now we're up to *4 sample's* worth of delay_.

Let's say that Logic is running a native (i.e., fast) EQ on that live input. There has to be some cumulative delay in order to do an FFT on the audio, so inherently there will be (compared to what we've seen so far) a pretty massive delay in order to analyze a meaningful portion of the incoming waveform. Let's say that amount is the buffer size -- 32 samples. _Total so far: *36 samples*_, and at 96K that's *.33 milliseconds*.

Let's multiply that by a nominal amount of 2 to get audio out of Logic back to the interface, D/A and thust to the monitoring system. _Now we're up to *72 samples*_, and at 96K that's *.66 milliseconds*.

Let's double that for good measure (because we know things aren't quite as simple as what I'm outlining here) and we're up to *1.32 milliseconds* of unavoidable latency.

That's at 96K in this hypothetical system. +Let's halve it for a 48K system+. Now we have an unavoidable _*2.64 milliseconds*_ of delay in monitoring live audio.

What this points out is that every single A/D + D/A on the market represents a giant step backwards (in a sense) in monitoring live audio as compared to the simplicity and immediacy of monitoring via an analog signal path. It's for this reason that I (on occasion) opine that software monitoring of live audio is a total crock. Sure, it's nice to be able to do things ITB. Convenient, etc. But that convenience has a price -- the price dependent on the capabilities of any particular system to run at higher sampling rates, the desire on the part of the end-user to use such rates, the quality of the convertors and the driver software (and thus the thruput speed).

Again, this is NOT to slam Apogee (David, that was for you, and I mean it). But the MOTU and RME guys have the right idea -- direct monitoring capability. Anything less than that:

• sparks discussions like these

• potentially generates consternation in the part of performing talent when they are asked to perform in an environment where their monitored audio is out of time (even if it's slight and annoying, like the sound of phase cancellation against their live sound that bleeds into the cans)

• gives people the errant impression that you can actually "do it all ITB", because in some cases -- especially those cases where your buffer can't be set at 32 or when you can't run at 96 -- it's tantamount to an outright lie

It seems that lots of very intelligent people are writing code to simulate the functionality of a simple copper wire. Buss wire -- the kind of wire that, when connected from input to output with a $1.00 switch, could provide instant, real-time monitoring without any delay. Sure, you won't have the benefit of pluggies to apply to live input, but that's all part of the fallacy: based on my crude analysis above, you can't help but have delay when using live pluggies on live input.

Message was edited 2x by: iSchwartz

• potentially generates consternation in the part of performing talent when they are asked to perform in an environment where their monitored audio is out of time (even if it's slight and annoying, like the sound of phase cancellation against their live sound that bleeds into the cans)

just think about that. there is potentially phase cancellation that occurs naturally from reflections, from FX incorporated into a hardware mix, from stereo mic placement, as well as the dampening and distorting effect of getting live sound through the cushions of headphones. i really don't buy this as an argument, and especially not for an instrumentalist. zero latency is a myth. i realize you are interested in not adding any more latency than already exists, but these additions are so tiny that it makes no difference. its the difference between a what a tall drummer and a short drummer will hear!

• gives people the errant impression that you can actually "do it all ITB", because in some cases -- especially those cases where your buffer can't be set at 32 or when you can't run at 96 -- it's tantamount to an outright lie

well, i don't see that follows. PT systems have been doing it in the box for years - that essentially is how they operate, and that has been a fairly standard way to work in the industry for a long time. the way they have gotten around it is to distribute audio processing onto dedicated hardware, just the same as if you have any digital effects on your monitored mix in hardware monitoring situation - any digital FX such as reverb will have latency. all that happens in a software situation is that world is modeled by the computer. instead of PT farming the reverb to a dedicated chip, it is handled by the host computer, because they are fast enough now to do the same thing in the same time.

It seems that lots of very intelligent people are writing code to simulate the functionality of a simple copper wire. Buss wire -- the kind of wire that, when connected from input to output with a $1.00 switch, could provide instant, real-time monitoring without any delay. Sure, you won't have the benefit of pluggies to apply to live input, but that's all part of the fallacy: based on my crude analysis above, you can't help but have delay when using live pluggies on live input.

the reason this has been sought after is simply because with fast computers, people have wanted an alternative to a protools rig, and because it is much faster and more convenient to be able to do everything ITB, requiring less cables that are prone to failure, a simpler outboard setup, reliability, monitoring what is actually being recorded (off the playhead principle which i am assured is still very important to engineers) rather than what is being monitored.

the symphony system was being looked at really hard by a major studio here in london last time i was in recording there for tracking in logic. yes (to all those 'you can't track in logic' nay-sayers), logic is quite commonly used to track here, and DAE was something of a pain. the latency would be not really any different to what they experienced with PT. i don't know if they went for it or not though, but they were talking a serious look.

i just think the latency issue once it gets down around the amount you get with symphony is non-issue. it just is not a reason NOT to go down that road.

here are my reasons to not go down that road:
• cost, i don't need anything as sophisticated as symphony. i have software monitored through my less sophisticated set up with satisfactory results. i can also hardware monitor if i so choose.
• i don't track more than one player often enough to justify it, and i have never had any complaints from singers or instrumentalists software monitoring when i have done that.
• i don't think it matters that much to listen listen to what is being recorded (software monitoring) or to have exactly the same effects as what you will use mixing, an argument i have heard from engineers. not to me anyway. i think hardware monitoring is fine.

its about practicality - for me a combination of soft or hard monitoring works fine depending on my needs, but it may not be for more serious recording engineers. i think ringing hands over latency, once it gets down to those tiny levels is not the reason to discount software monitoring.

Message was edited by: Rohan Stevenson1

zero latency is a myth. i realize you are interested in not adding any more latency than already exists, but these additions are so tiny that it makes no difference.

That's my point. In particular, the point of my last post was to illustrate that there's an inherent latency in the A/D - D/A process that cannot be overcome. And especially when you can't (or don't want to) run at 96K, the "near zero latency" becomes "mondo big latency". And direct monitoring has to be the ultimate way to implement the desire to "\[not have] any more latency than already exists".

Not to belabor the point, but rather to clarify, zero latency may be a myth even in real-world situations (i.e., your tall v. short drummer was a great analogy), but in digital audio product-land it is a huge selling point AND a myth (or, at best, a deceptive sales pitche) that causes many people big problems. Consider this ad:

"Can't sleep at night? Thoughts of low-latency monitoring keeping you awake? Try new Insomni-Whack, by Pfizzier. Yes, one teensy Insomni-Whack caplet before bed will ensure you get a good night's sleep. But remember, consult your doctor before taking Insonmi-Whack. Side effects can include death, dyspepsia, colic, nasal hemmorhaging, bleeding from the ears, and a penchant for writing long posts."

So here at least the company is disclosing the pitfalls up front. But not so in digital audio land. The term zero latency -- which some manufacturers can rightly claim (if they employ direct monitoring) -- has been coyly and deliberately slanted in its other form, "near zero latency". Here, "near zero" can mean absolutely anything. "What's near zero?" asks the unsuspecting digital audio consumer? "Hmmm... One is near zero, right? Maybe even two. But heck, it's near zero so it must be close enough, right?"

With this assumption... I mean "knowledge", and cash in hand, our intrepid DAW customer buys a shiny new audio interface and computer. He hooks it all up, if but only to find that he has to set his buffer to like 125,000 millionbillion because his system won't run at a lower buffer setting. Then he complains that he can't hear himself in time when he uses software monitoring, at which point he, like thousands of others, flock to the forums asking, "why do I hear myself late"?

ok, well that is a slightly different argument but one i agree with broadly. i am deeply skeptical of audiophilia, just one step on the ladder down from another type of of -philia.

stuff like 'sample rates up to 192k supported', etc etc. so much stuff that makes not an iota of difference to how good your tune is.

so no, i wouldn't buy in to hype, but i would want to inform myself about the practicalities of my setup. my setup is so old and creaky it is virtually steam powered. but i running out of room in my studio to keep all the coal...

my next move is likely to be an ITB setup, and i would be ditching a mixing desk for anything other than hooking up lo-fi gear. in which case i would track directly into an interface and yeah, i would use software monitoring, because my experience here tells me even relatively high latencies compared to what we are talking about here are fine.

therefore it is not hype but hard cold practicalities.

now, i am going to say something that might shoot down everything i have hitherto said:

one thing i have noticed is that patching my external reverb directly into the interface does not yield as satisfactory (to my ears) results as mixing on the desk. it doesn't 'glue' as well to the sound. it is a subtle thing but i felt happier mixing the sound on the desk. latency? quite possibly. i think the difference here is a much subteler effect than whether or not you notice any latecny in a feed back to yourself over headphones.

Rohan Stevenson1 wrote:

• potentially generates consternation in the part of performing talent when they are asked to perform in an environment where their monitored audio is out of time (even if it's slight and annoying, like the sound of phase cancellation against their live sound that bleeds into the cans)

just think about that. there is potentially phase cancellation that occurs naturally from reflections, from FX incorporated into a hardware mix, from stereo mic placement, as well as the dampening and distorting effect of getting live sound through the cushions of headphones. i really don't buy this as an argument, and especially not for an instrumentalist. zero latency is a myth. i realize you are interested in not adding any more latency than already exists, but these additions are so tiny that it makes no difference. its the difference between a what a tall drummer and a short drummer will hear!

I definitely agree with this. There seems to be an assumption that with analog monitoring, there's no time delay between the two sources and no phase issues, and that simply isn't true. There will virtually always be phasing and timing issues when playing an instrument and listening to the mic through headphones.

A 2ms delay is equivalent to the guitarist moving two feet further away from his amp. Is that really something that the guitarist is really going to notice, much less cause him to perform worse?

iSchwartz wrote:

And especially when you can't (or don't want to) run at 96K, the "near zero latency" becomes "mondo big latency".

What is the total latency of the apogee at 44.1? Isn't it about 4ms? Just curious what number you'd consider acceptable, and where "mondo" starts for you.

I do agree that "near zero latency" is not a good term to use in their literature, but that's more a marketing issue. But they do have the real numbers, and they make it clear that 1.6 is at 96k and that they're using a mac pro. It would be nice if they put up the numbers for 44.1/48.

Mike,

I definitely agree with this. There seems to be an assumption that with analog monitoring, there's no time delay between the two sources and no phase issues, and that simply isn't true. There will virtually always be phasing and timing issues when playing an instrument and listening to the mic through headphones.

There is no assumption. It's important to distinguish between naturally occuring delay and artificially induced additonal delay. (Rohan and I seem to be in agreement on this point).

Is that really something that the guitarist is really going to notice, much less cause him to perform worse?

Is that for you (or should I say, "anyone") to judge and alter based on the artificial (and all-too-convenient yet false) notion that small increments of time don't matter? And is the timing relationship between their sound and their physical performance something for you or anyone to condone altering for the sake of ITB "convenience"? Are all guitar players cut of the same cloth to where those 2ms wouldn't affect any of them?

When you and others proffer arguments such as "2ms = guitarist 2' further from the amp", you are trying to reinforce the notion that small increments of time don't matter to musicians. Think about that one without going "duh?!?!?!" You are also saying that it's OK to artifically change the timing relationship that player has with his amp.

_Neither argument is valid._

To the first point (2 ms doesn't matter), if that were the case, why would Apogee have developed low latency gear? Think about it... their 1.5ms benchmark is LOWER than your 2ms/guitar player example. So I don't think you need to take it from me where "mondo" starts. Look at what Apogee is doing and work backwards towards the ideal -- the copper wire that takes the sound from 'here' and puts it 'there' with no artificially induced delay.

You can't rightly justify the performance of technology as superior when it doesn't perform up to the standard of the technology it's attempting to replace. I don't know of any copper wire that has a built-in 1.5 ms delay. Do you? And when you sweep meaningful discussion of monitoring latency times under the rug as "inconsequential" because they're inconvenient to your argument doesn't say much for your argument ab initio...

Message was edited 2x by: iSchwartz

iSchwartz wrote:
There is no assumption. It's important to distinguish between naturally occuring delay and artificially induced additonal delay. (Rohan and I seem to be in agreement on this point).

Sure, they are different, and sure, the digital delay can be greater. My point is that even with analog monitoring, you're going to have phase and timing differences.

Is that for you to judge? And is the timing relationship between their sound and their physical performance something for you or anyone to condone altering for the sake of ITB "convenience"? Are all guitar players cut of the same cloth to where those 2ms wouldn't affect any of them?

Why don't you just try it out? Have your guitarist play, then have him stand two feet further away from his amp. Does that mess him up or not?

Do your guitarists insist on being a precise distance from their amp? I have never known one who wasn't completely fine with it in quite a range of distances. Same with singers and other musicians who regularly play on stage with monitors varying distances away.

When you and others proffer arguments such as "2ms = guitarist 2' further from the amp", you are trying to reinforce the notion that small increments of time don't matter to musicians. Think about that one without going "duh?!?!?!" You are also saying that it's OK to artifically change the timing relationship that player has with his amp.

Not at all. Small amounts of time can matter. The point is that musicians are already dealing with time delays, often ones bigger than the ones induced by digital monitoring.

If you have a guitarist six feet tall standing next to his amp, he's going to hear the amp with a 5 ms delay. If you have a mic right up to the amp and passing through a digital system with 4 ms of latency, he's still going to hear it in the headphones BEFORE the sound coming from the amp and passing through the air gets to his ears.

So when the sound from the cans gets there EARLY compared to the acoustic sound, the problem is still that it's too LATE?

To the first point (2 ms doesn't matter), if that were the case, why would Apogee have developed low latency gear?

Because other gear has higher latency than 2ms, and at some point latency becomes unacceptable. For you I guess that point is either something less than 1.6, or anything that's not zero.

You can't rightly justify the performance of technology as superior when it doesn't perform up to the standard of the technology it's attempting to replace.

I never said it's superior in every way. There's no question that the latency of digital recording is inferior to analog monitoring. But for me the latency is low enough that it's not a big deal, and for me the other advantages of digital outweigh that disadvantage.

And when you sweep meaningful discussion of monitoring latency times under the rug as "inconsequential" because they're inconvenient to your argument doesn't say much for your argument ab initio...

I'm not sweeping anything under the rug, nor did I describe them as "inconsequential".

I'm just saying that:
1) Analog monitoring can have timing and phase differences between the acoustic sound of the instrument - using analog monitoring generally isn't going to automatically put things perfectly in phase or time.
2) Musicians already play with time delays whenever the instrument isn't right next to their ear, so I think playing with a small delay on headphones can be acceptable, and if it's small enough, not even noticeable.

I think the real question is just playing and listening and asking yourself whether the latency is acceptable.

Have you actually played tracking to a mic and passing through a system with 1.6 ms of latency? And you could hear that delay and it bothered you? Have you tracked on a system and had other musicians complain about latency at 1.6 ms?

I agree 100% with iS.

To add a few thoughts...

iSchwartz wrote:
When you and others proffer arguments such as "2ms = guitarist 2' further from the amp", you are trying to reinforce the notion that small increments of time don't matter to musicians. Think about that one without going "duh?!?!?!" You are also saying that it's OK to artifically change the timing relationship that player has with his amp.

I will say that I've experienced exactly this, and onstage with guitar amps, I've always noticed a compromise between locating an amp in the backline, versus right up behind my head. When the amp is behind you (say 10-15 feet), the sound spreads better and you get a better idea of the full projecting tone. But I've always thought I played better on smaller stages where the amp was up on crates right behind my head, even though the tone was artificially bright, and I typically had to wear earplugs to keep from blasting myself. The difference, I think, was latency.

Later on, I moved to in-ear monitors, and one of the things that immediately grabbed me about using the IEMs was the immediacy of the sound. From the first moment I tried it, I preferred hearing my amp through a mic right up against the speaker grill. There was a palpable sense of being more connected with the strings of the guitar. In some sense, this is "artificially-reduced" latency, but what can I say... it feels better to me when I'm playing.

And when you sweep meaningful discussion of monitoring latency times under the rug as "inconsequential" because they're inconvenient to your argument doesn't say much for your argument ab initio...

Absolutely. And not that I don't fall into this trap at times, myself. The way I see it, there are two very distinct ways to choose gear:

1) Listen blindly and pick what sounds the best.

2) Read the specs, look at the photos, plan how the gear fits into rig and workflow, then try to convince yourself that the gear sounds "about as good" as the other choices.

I don't say this to be insulting to anyone... as elementary as it might seem at first glance, I think it's shocking how many times we all choose gear via #2 rather than #1.

Analog monitoring takes more effort and is less elegant (from a gear/setup standpoint, at least). But it simply sounds better, to me at least. If other people prefer the sound of digital latency, or can't hear the difference, then great... they should be thrilled monitoring ITB. And maybe some of the people seeking lower latency are doing so in response to marketing. But I am certainly not the only one who has tested monitoring and virtual instruments with varying degrees of latency and buffer sizes, A/D/A converters, etc., and who has found that--all else equal--lower latency always "feels" better while I am recording/performing.

And now that I've experienced that for myself, I think it would be kinda disingenuous for me to suggest that clients I record "probably won't hear a difference" if I make some latency compromises to make my job easier.

James

James, what's the lowest you have tracked at? Have you been able to try it with 1.6 ms of latency?

Mike Connelly wrote:
Have you actually played tracking to a mic and passing through a system with 1.6 ms of latency?

Yes. That's about the latency introduced by an outboard Lexicon reverb. Patch one in and out of an analog monitoring chain and you'll hear exactly this latency difference.

And you could hear that delay and it bothered you?

Yes. I hear it as an obvious eq difference when singing, and as a subtle feel difference when playing stringed instruments. "Bothered' is a loaded word. Could I perform adequately in spite of this latency ? Yes. Did I get used to it and forget about it after working a bit ? Yes. Did I prefer the direct analog sound in an A/B test ? Yes.

Have you tracked on a system and had other musicians complain about latency at 1.6 ms?

No. Fair point. People record everyday on ProTools HD systems monitoring ITB... it's basically the industry standard way of working. Virtually everyone is used to it, and no one complains. But, we're also all used to the sound of a full band being played back through a 4" television speaker, and we don't complain. But given the choice... 🙂

James

Mike Connelly wrote:
James, what's the lowest you have tracked at? Have you been able to try it with 1.6 ms of latency?

On a daily basis, when I'm working quick-and-dirty, I have 4ms of latency. When I take the time to patch in analog monitoring, it feels like a breath of fresh air.

4ms is as low as my system will go. But most of us have experienced 1.5-2ms of latency, since that's what ProTools HD gives you. Between 4ms and 1.5ms, lower wins... no question about it. But I still think analog sounds better than PTHD in a direct comparison.

Sure, you can rationalize that adding 1.5ms of latency is like moving the mic a few feet farther away from the source. But, really, once I've got the mic in a place where it sounds good, the last thing I'd want to do is move it back a few feet... who would ? It's a false choice. Compare the analogy of a 0.02" thick piece of latex... I don't think I have to be explicit to make this point...

Is 1.6ms "acceptable" ? Sure. Is 4ms "acceptable" ? Sure. Not to be harsh about it, but if your goal is to be "acceptable," then just buy any decent digital system, and you're there. If your goal is to create the best recording environment you can afford, maybe you want to compare analog and digital monitoring and choose based on which sounds and feels better. Latex or no latex in a direct A/B comparison. I bet I know which one you'll pick...

James

For the sake of comparison, what's the latency on protools tdm?

Mike, thanks for your reply.

Mike Connelly wrote:
Do your guitarists insist on being a precise distance from their amp? I have never known one who wasn't completely fine with it in quite a range of distances. Same with singers and other musicians who regularly play on stage with monitors varying distances away.

What I'm getting at, I guess, is that making musicians concerned/aware about their distance from other players, mics, etc. because of the abstruse phenomenon of monitoring latency in an acoustic recording situation (amp-->mic, guitar-->mic, whatever) shouldn't be part of the music-making process. It's not unlike talking on a telephone. On a landline there is, for the most part, immediate connectivity between the parties. On a cel phone, however, the delay is enough to throw off even the rhythm of a conversation! Smaller timing discrepancies can be just as unnerving to musicians.

In analog/direct monitoring situations, WYH (hear) IWYG for the most part. But for some people, the indirect monitoring fosters room for doubt (which I'll get to in a second...). A musician's timing shouldn't have to be discussed in terms of anything other than how they're playing against existing tracks. To make that job more difficult by virtue of indirect monitoring has, IMO, no place in a recording studio. Which brings me to a very common phenomenon -- what Jim said about questioning whether or not he was beginning to suck...

One of the other problems exacerbated by DAWs is the fact that audio is inevitably going to arrive late to the DAW (by virtue of the latency inherent in the A/D--->computer process). So even in a zero-monitoring situation, if your system isn't calibrated to compensate for that delay, perfectly played parts will be played back perplexingly late. Or early! Viz:

Some RME systems seem to try to compensate for this hardware delay and overdo it in the process, which is why many people find that their tracks seem early when played back in Logic. The more common scenario, though, is that tracks end up being played back late in Logic. (And not just Logic -- many other DAWs as well). So it's extremely important to calibrate one's system to compensate for hardware latency vis a vis Logic's recording delay setting.

iSchwartz wrote:
In analog/direct monitoring situations, WYH (hear) IWYG for the most part.

That's the part I disagree with.

In ANY recording situation, the player will hear the sound twice, and in most cases the two will be out of sync (the main exception being vocals). This is inherent to recording and monitoring on headphones and happens with analog recording as well, it's not unique to digital recording with latency. The only way to avoid it would be for every instrument to be immediately next to the player's ear, which is generally not the case.

In a perfect world, the player would hear both the acoustic and headphone sounds exactly as they play them. But that's not even the case with analog monitoring. Anyone playing an acoustic instrument is always hearing a delay between making the sound and hearing the sound. It would be cool to avoid that, but it's not possible, so the player gets used to it.

I guess what I don't get is the objection to a latency delay that's shorter than the delay caused by the acoustic sound travelling through the air - why someone who hears a three MS delay of the sound traveling through the air and is OK with it has a harder time playing when the sound travels through 1.6 ms of latency - when they're actually hearing the sound earlier.