Daah! It was latency all along

[ThinCrappyTone]

This is a tale of ignorance, lack of faith in one's abilities, and self-deception.

Back in the analog recording days, I always felt that my timing was solid. A few years back, I started recording with a digital interface, and realized my timing wasn’t as good as I thought. The recordings were so clean that I was hearing things I never heard before. I would roll through a song, feeling like I had nailed it, only to play it back and it just didn’t feel locked on a micro-timing level. Other musicians said it sounded fine. I wondered if my playing was always this way, and I just couldn’t hear it, if I just didn’t possess the rhythmic control that I used to, or if I was just tripping out.

Now, I knew about latency, and could detect it when it was really bad. But most of the time, what I was experiencing was subtle enough to be undetectable to me, so I wrote that off as the probable culprit.

A few weeks back, I picked up a new thunderbolt interface with *very* low latency. Fudging cripes, man. I want to punch myself. It has become instantly apparent that the imperceptible (at least to me) latency of my old system is what had been affecting my recordings. I can now see the issue for what it was, because when I record stuff with the new box, if it feels locked on a micro-timing level during the performance, then it actually is locked on a micro-timing level during playback.

This led me to think back to the digital pedals I have owned, and especially, guitarists in my band have owned, and i wonder how much the stacking of almost imperceptible timing delays caused by the d/a a/d conversions subtly affected things without us knowing it.

Anyway, I tell this slightly embarrassing story because I would guess there are others out there being affected by latency without entirely being aware of it. How does the saying go? If you your life can't be an example to others, let it at least be a warning.

 

[bholder]

I've never been entirely happy with my timing, and starting to record myself in the last year hasn't made me any less suspicious - but I have to wonder how much of that is latency further mucking up my already questionable timing.

 

[12bass]

Almost all audio interfaces have a near-zero latency monitoring mode which should work well for tracking a direct bass signal. Latency only becomes a significant issue if tracking through the DAW using plugins (e.g. amp simulators). FWIW, old PCI interfaces from the early 2000s had better low latency performance than most of the USB interfaces which replaced them. Thunderbolt can provide excellent performance, but no better than a good PCI/PCIe card.

 

[ThinCrappyTone]

Almost all audio interfaces have a near-zero latency monitoring mode which should work well for tracking a direct bass signal. Latency only becomes a significant issue if tracking through the DAW using plugins (e.g. amp simulators). FWIW, old PCI interfaces from the early 2000s had better low latency performance than most of the USB interfaces which replaced them. Thunderbolt can provide excellent performance, but no better than a good PCI/PCIe card.

True. My early experience, when I was having problems, was not with PCI, but usb and then firewire. But I guess the point I'm trying to make is regarding what we consider acceptable amount of latency, and how much that amount affects performance, without us really knowing it. On the one end of the spectrum, there is significant, obvious latency, like when plugins are used. On the other there is true zero (or as close as possible) latency of analog. But what I am musing about is that grey area in between. The amount of latency we accept because we don't hear it, but is actually causing very subtle timing issues. So subtle, that the tracks sound fine, even to the person who recorded them, but really are just a hair off.

 

[JRA]

yeah: latency.

glad you got the problem resolved!

it happened for me in my 40's.

 

[ThinCrappyTone]

I've never been entirely happy with my timing, and starting to record myself in the last year hasn't made me any less suspicious - but I have to wonder how much of that is latency further mucking up my already questionable timing.

It really might be. Its a subtle sneaky thing. And I think it's worse for those of us who are not 100% confident in our timing to begin with. We might tend to accept the slightly jacked up playback, assuming we just played it a little worse than we thought, rather than explore the possibility that the gear is doing something to make it worse.

 

[heavyfunkmachin]

Just to be clear: Distance from the amp gets you latency as well... 1,125ft/s
So, when recording and not direct monitoring, adding SOME latency is not a big deal...
Let's say you're 15 feet away from you cab... a nice stage, that's giving you about 13 ms of delay!

So, As loon as the latency is not TOO big, it's not a big deal, all you have to do is noodle a bit before recording. Your brain will soon adapt to the latency (human brains are great at this, you won't notice it), and you'll play slightly ahead to make up for it... Just remember to add the balance the latency you heard to your DAW's track, so it's recorded as you meant to record it (as you listened to it) /Ableton is super easy with this... Invalid Link Removed

 

[12bass]

True. My early experience, when I was having problems, was not with PCI, but usb and then firewire. But I guess the point I'm trying to make is regarding what we consider acceptable amount of latency, and how much that amount affects performance, without us really knowing it. On the one end of the spectrum, there is significant, obvious latency, like when plugins are used. On the other there is true zero (or as close as possible) latency of analog. But what I am musing about is that grey area in between. The amount of latency we accept because we don't hear it, but is actually causing very subtle timing issues. So subtle, that the tracks sound fine, even to the person who recorded them, but really are just a hair off.

Hmmm... which specific interfaces are you talking about? There are some FireWire interfaces with excellent low latency performance, and a few USB units as well. RME is pretty much the king of low latency performance regardless of type of connection, though their PCI/PCIe interfaces are the fastest.

As to the more general point, I hear what you're saying. I don't feel quite in the pocket if the latency gets much over 7ms or so; many USB interfaces are not even capable of getting down to 7ms with any stability. Direct hardware monitoring or "near-zero" latency is generally so short that it is never a problem, usually 2ms or less (whatever the converter latency happens to be), and is essentially as good as analog monitoring IME.

 

[seamonkey]

The A/D D/A takes microseconds. Effectively zero latency.
What trips up latency is the stuff gets buffered in the DAW
The DAW chews on chunks at a time, software effects chew on it more.
Many PC's these days can handle very small buffers so it shouldn't be a problem.
But you have to know where to change it in the ASIO panel.

 

[DirkP]

If you use direct monitoring while recording - and nearly all interfaces offer this since decades, no matter if it is PCI, USB, Thunderbolt or Firewire - you have nearly zero latency while recording.
It is smaller than the latency you get, if you move a few feet away from your speaker!

It is different if you are monitoring your processed signal- going through effects or EQs in the DAW.

 

[12bass]

The A/D D/A takes microseconds. Effectively zero latency.
What trips up latency is the stuff gets buffered in the DAW
The DAW chews on chunks at a time, software effects chew on it more.
Many PC's these days can handle very small buffers so it shouldn't be a problem.
But you have to know where to change it in the ASIO panel.

While it is generally true that modern multicore PCs have improved low latency performance, many current audio interfaces have very poor performance at low latency, due to both hardware design and driver inefficiency. Ironically, quite a few old PCI audio cards from over a decade ago outperform modern USB interfaces while under stress at low latency. This is largely because USB was not originally designed for bulletproof realtime audio streaming; FireWire is better because it supports isochronous data streaming and has lower CPU overhead, while Thunderbolt is better yet. However, PCI/PCIe potentially offers the best low latency performance at high CPU load. This is all dependent on hardware/driver design and there are a few outliers (e.g. some USB interfaces like RME outperform some FW and PCIe devices, though RME designed their own USB interface hardware).

See here: Invalid Link Removed

 

[Gideon352]

What I sometimes do while recording (which might sound a little whacky, but you can try it for fun) is to turn myself off completely and just listen to my bass acoustically with the other tracks at a low volume. If you're really sensitive to latency (I am - even with my UAD Thunderbolt interface) it might help.
I find that I play by feel more this way too. I know if I've made a mistake by how the track felt, and if I have, I just do another take immediately.

 

[seamonkey]

This is largely because USB was not originally designed for bulletproof realtime audio streaming

USB, is not the problem. Even for USB 2. interfaces with 2 to 18 tracks no problem. Just don't use a "Hub"

 

[seamonkey]

What I sometimes do while recording (which might sound a little whacky, but you can try it for fun) is to turn myself off completely and just listen to my bass acoustically with the other tracks at a low volume. If you're really sensitive to latency (I am - even with my UAD Thunderbolt interface) it might help.
I find that I play by feel more this way too. I know if I've made a mistake by how the track felt, and if I have, I just do another take immediately.

It always good to listen in different ways.

Again though, thunderbolt, USB, ... is not the source of latency. It's the buffers. If you have the buffer size set to 2048 Samples, then it's the sampling rate that dictates the latency. At the current sample rate how long does it take to collect or assemble 2048 samples - no need for math, if you don't like the latency then Make it smaller - PC's can handle it.

 

[12bass]

USB, is not the problem. Even for USB 2. interfaces with 2 to 18 tracks no problem. Just don't use a "Hub"

Take a look at the link provided above and read through the thread. RME's PCI/PCIe interfaces clearly outperform all tested USB interfaces, a fact well known among recording professionals. Those who require the best low latency performance in modern machines choose PCIe. USB does work; however, glitch-free performance at ultra-low latencies is uncommon with USB interfaces. Again, look at the data provided. Even RME is not able to get performance comparable to PCI/PCIe from their custom USB chips. Some recent Thunderbolt interfaces are close.

Further, often hidden buffers are used, which are not adjustable by users. Thus, setting a buffer at 64 samples on one interface will not give an overall RTL (round trip latency) that is the same as another. In some cases, hidden buffers can double RTL.

 

[seamonkey]

a fact well known among recording professionals.

PCIe, Thunderbolt, USB - it's the buffer size that trips up most people.

 

[Gideon352]

It always good to listen in different ways.

Again though, thunderbolt, USB, ... is not the source of latency. It's the buffers. If you have the buffer size set to 2048 Samples, then it's the sampling rate that dictates the latency. At the current sample rate how long does it take to collect or assemble 2048 samples - no need for math, if you don't like the latency then Make it smaller - PC's can handle it.

Check out the UAD console application and how it works.