Noob question about latency.

[Nizhny Tagil]

Tonight I was doing some tests in Reaper and I've noticed that my timing was much worse than usual.
I've improved my computer since the last time I've recorded something, so I had kind of ruled out latency as the reason; besides, I have not been practicing for some time, now.
But the more I tried be on time the more I struggled, so I decided to do something I've never done: a loopback test to see what latency I really have.

Despite the ASIO drivers reporting 5.6/5.6 ms, the test showed a gap of 780 samples, more or less, between two sounds (I've used the "click source" to test); so, if I did the math right, 17.6 ms total latency.
If really the human ear barely notices latency up to 11 ms, that doesn't seem so bad.
So, what I'm asking is: with such round-trip latency, do I have a problem or it could just be that my playing sucks more than usual?

By the way, in the guitar's FX chain I had: JS Distortion(fuzz) and JS Convolution Amp Model, for neither of which Reaper reports any added latency, so I don't think plugins are the problem.

Thanks.

[grinder]

I would recommend when recording a played instrument to a track to listen to the mix while playing but do not include your live playing sound in the mix...
Play that way and you will probably be good to go. If you are behind or late you can fix up later if you have good tempo you can just slip the recorded wave file along or back to fit perfectly or if it sounds better late or early you may want it left as it is.
I do most of my recorded playing like this and I have very little work to do afterwards.

Grinder

https://soundcloud.com/steve-maitland-1

[Nizhny Tagil]

Originally Posted by grinder

I would recommend when recording a played instrument to a track to listen to the mix while playing but do not include your live playing sound in the mix...
Play that way and you will probably be good to go. If you are behind or late you can fix up later if you have good tempo you can just slip the recorded wave file along or back to fit perfectly or if it sounds better late or early you may want it left as it is.
I do most of my recorded playing like this and I have very little work to do afterwards.

Grinder

https://soundcloud.com/steve-maitland-1

I'm not sure I understand.
How can I know what I'm playing if I'm not monitoring in some way? It's an electric guitar, and if I want to use distortion I can't even use direct monitoring. For anything else I would use that, but for electric guitar and virtual instruments I must deal with latency.

[grinder]

How do you mean you will not know what you are playing if you are playing electric guitar? I play all kinds of guitar this way and I know what chords, notes I am playing?
Work out before you attempt your recording what chords/notes you need to play for the arrangement and then once you have done that, do the recording and playing!


Grinder

[grinder]

I don't know if you have been in a band before....
In a band with no monitors, earpieces etc sometimes as a guitarist it is very difficult to hear your own sound maybe because you are jammed in the front of your floor speaker setup or what have you so you have to just rely on hearing maybe the bass or drums to get the timing of the band that is why you need to have some idea of what to play and where to play it to be in time and be in tune. Latency is forever and will always be.
Be happy....

Grinder

[thequietroom]

Originally Posted by grinder

I don't know if you have been in a band before....
In a band with no monitors, earpieces etc sometimes as a guitarist it is very difficult to hear your own sound maybe because you are jammed in the front of your floor speaker setup or what have you so you have to just rely on hearing maybe the bass or drums to get the timing of the band that is why you need to have some idea of what to play and where to play it to be in time and be in tune. Latency is forever and will always be.
Be happy....

Grinder

Tracking this way isn't practical for everyone. I've had shitty stage sound many times and I get what your saying, but trying to lay down tracks without monitoring is certainly not ideal.

I had a long go with amp sims, with 2 computers and 3 interfaces. I never got a satisfying workflow and tracking comfort level.

You may want to look around for an old line6 PodXT. I'm sure they are really cheap by now. I just gave one away to a friend. It is really easy to track a dry take while listening to the processed tone with them. You just select it in the asio driver.

This way you have the DI to use your fav software sim, but have a reasonably decent tone for monitoring.


Here is an old thread where we discussed this


http://forum.cockos.com/showthread.php?t=41161

[jerome_oneil]

Originally Posted by Nizhny Tagil

I'm not sure I understand.
How can I know what I'm playing if I'm not monitoring in some way? It's an electric guitar, and if I want to use distortion I can't even use direct monitoring. For anything else I would use that, but for electric guitar and virtual instruments I must deal with latency.

And then monitor from the audio interface, rather than the track. That's as low latency as you're going to get.

[jerome_oneil]

Originally Posted by thequietroom

You may want to look around for an old line6 PodXT. I'm sure they are really cheap by now. I just gave one away to a friend. It is really easy to track a dry take while listening to the processed tone with them. You just select it in the asio driver.

I have an old school Pod (the kidney bean one) with the MIDI pedal board. I'm the last guy in the world to recommend amp sims (just go buy the damn amp!) but I use it for exactly this kind of stuff, and it works great.

[grinder]

Fair call (sniffle sniffle....)

Grinder

[Judders]

Originally Posted by Nizhny Tagil

Tonight I was doing some tests in Reaper and I've noticed that my timing was much worse than usual.
I've improved my computer since the last time I've recorded something, so I had kind of ruled out latency as the reason; besides, I have not been practicing for some time, now.
But the more I tried be on time the more I struggled, so I decided to do something I've never done: a loopback test to see what latency I really have.

Despite the ASIO drivers reporting 5.6/5.6 ms, the test showed a gap of 780 samples, more or less, between two sounds (I've used the "click source" to test); so, if I did the math right, 17.6 ms total latency.
If really the human ear barely notices latency up to 11 ms, that doesn't seem so bad.
So, what I'm asking is: with such round-trip latency, do I have a problem or it could just be that my playing sucks more than usual?

By the way, in the guitar's FX chain I had: JS Distortion(fuzz) and JS Convolution Amp Model, for neither of which Reaper reports any added latency, so I don't think plugins are the problem.

Thanks.

The first question that comes to my mind is: what is your block size in preferences set to?

[Nizhny Tagil]

Originally Posted by jerome_oneil

And then monitor from the audio interface, rather than the track. That's as low latency as you're going to get.

I've thought of that too, but a decent pedal costs little less than a new interface and doesn't solve the problem with virtual instruments.

Originally Posted by Judders

The first question that comes to my mind is: what is your block size in preferences set to?

64, doesn't get any lower.

Anyway, that's a little beside my original question.
There are a million ways to improve latency, but what I'm asking is IF I'm experiencing issues because of latency (which amounts to more or less 17.6 ms according to the loopback test, as reported above) or if you can get by with such latency and it's all because of my sloppy playing.

My English is far from perfect so let me know if I haven't explained myself well.

[Judders]

Originally Posted by Nizhny Tagil

Anyway, that's a little beside my original question.
There are a million ways to improve latency, but what I'm asking is IF I'm experiencing issues because of latency (which amounts to more or less 17.6 ms according to the loopback test, as reported above) or if you can get by with such latency and it's all because of my sloppy playing.

My English is far from perfect so let me know if I haven't explained myself well.

I understand.

I think it is impossible to play with tight, accurate timing if you have noticeable latency. I also think that the better your sense of time, the more you will notice latency.

[ivansc]

Originally Posted by Judders

I also think that the better your sense of time, the more you will notice latency.

THIS!

[Nizhny Tagil]

But is it that subjective?
In your experience, can't you eventually put up with such latency as I've measured?
I mean, I read around that some can put up with latencies up to 20 ms and that until 10-11 ms it's barely noticeable.

I understand your points and I agree, but isn't it possible to draw a theoretic line between "noticeable" latency and "not noticeable" latency or it's only up to any single pair of ears?

[Judders]

Originally Posted by Nizhny Tagil

But is it that subjective?
In your experience, can't you eventually put up with such latency as I've measured?
I mean, I read around that some can put up with latencies up to 20 ms and that until 10-11 ms it's barely noticeable.

I understand your points and I agree, but isn't it possible to draw a theoretic line between "noticeable" latency and "not noticeable" latency or it's only up to any single pair of ears?

Yes it is subjective, in that I've not done a study on it or anything, but I cannot put up with latency. Even seemingly unnoticeable latency can throw my timing a little if it's a fast staccato part, especially when doubling parts in unison.

At 48kHz sample rate, 128 sample buffer is enough to interfere with my timing, and 64 samples is acceptable for me. I've never done a proper latency test on my machine though, so I couldn't give you a real world latency figure.

Sloppy players don't notice their sloppiness, so it stands to reason that they will notice latency affecting their timing less than players with a good sense of time.

[Nizhny Tagil]

Originally Posted by Judders

Yes it is subjective, in that I've not done a study on it or anything, but I cannot put up with latency. Even seemingly unnoticeable latency can throw my timing a little if it's a fast staccato part, especially when doubling parts in unison.

At 48kHz sample rate, 128 sample buffer is enough to interfere with my timing, and 64 samples is acceptable for me. I've never done a proper latency test on my machine though, so I couldn't give you a real world latency figure.

Sloppy players don't notice their sloppiness, so it stands to reason that they will notice latency affecting their timing less than players with a good sense of time.

Then I curse my apparently good sense of time.
Well, it seems like either I get sloppier or I churn out more money for new interface/hardware.

What's strange is that I've recorded with larger buffer sizes on an inferior system before and I've managed to, somehow.
Maybe that stuff was simpler to play.

Just out of curiosity, why do you use 48KHz? Video work?

[Judders]

Originally Posted by Nizhny Tagil

Then I curse my apparently good sense of time.
Well, it seems like either I get sloppier or I churn out more money for new interface/hardware.

What's strange is that I've recorded with larger buffer sizes on an inferior system before and I've managed to, somehow.
Maybe that stuff was simpler to play.

Just out of curiosity, why do you use 48KHz? Video work?

Sometimes I use 44.1kHz, sometimes 48kHz, sometimes 96kHz. It depends on a few things; virtual synths that sound better at 96kHz, if I'm mixing other people's stuff that's at 44.1kHz, what sample rate sampler instrument files are recorded at etc.. If I'm recording stuff myself I go for at least 48kHz, just so that the option to use it with video without having to upsample is there.

[ivansc]

It took me a very long time before I finally realised I am a tempo nazi!

Latency much over 5-6 ms and I struggle.
Mind you I have the same problem using wireless and wandering around the venue whilst playing.
Much more than 30 ft from the stage and I cannot deal with the delay.

[karbomusic]

Did I miss the part where we go into the sound card's control panel and lower the buffer size? I'm not understanding why 17ms is the lower limit here nor did I notice where we mention the sound card used etc.

[Judders]

Originally Posted by karbomusic

Did I miss the part where we go into the sound card's control panel and lower the buffer size? I'm not understanding why 17ms is the lower limit here nor did I notice where we mention the sound card used etc.

I already did that

Nizhny Tagil is already at 64 samples and can't get any lower.

[karbomusic]

Originally Posted by Judders

I already did that

Nizhny Tagil is already at 64 samples and can't get any lower.

Hmm... Something sounds awry. That block size in Reaper preferences doesn't work on all machines and sometimes needs to be done at the actual sound card, my RME's are that way - I see people suggest Reaper's block size and I have tested on my machine multiple times - makes no difference in monitoring latency for me. I even have a screen capture showing it here somewhere in some thread. I have always had full control of latency at the sound card properties and haven't needed to touch the block size in Reaper since I've owned Reaper. My RTL is in the ~3ms range.

And we know 64 samples != 17ms. I'd suggest opening "ASIO Configuration" from Reaper so that it invokes the sound card CP and observe/change the settings there just to be sure.

[Judders]

Originally Posted by karbomusic

Hmm... Something sounds awry. That block size in Reaper preferences doesn't work on all machines and sometimes needs to be done at the actual sound card, my RME's are that way - I see people suggest Reaper's block size and I have tested on my machine multiple times - makes no difference in monitoring latency for me. I even have a screen capture showing it here somewhere in some thread. I have always had full control of latency at the sound card properties and haven't needed to touch the block size in Reaper since I've owned Reaper. My RTL is in the ~3ms range.

And we know 64 samples != 17ms. I'd suggest opening "ASIO Configuration" from Reaper so that it invokes the sound card CP and observe/change the settings there just to be sure.

Interesting...

Hopefully this might solve the OP's problem without having to buy a new interface.

[Nizhny Tagil]

I sure appreciate your interest, guys.

I didn't bother listing my hardware because I know it's prehistoric, and my original question wasn't how to lower latency but if I had to worry about it in the first place.
Anyway, since we're on the subject, the interface is a Lexicon Alpha and the processor is a Core 2 Quad Q9550 @ 2.83GHz. I'm about to expand the RAM to 8GB, I'm using Win10 on an SSD with multiple drives for audio files and I have applied all possible tweaks and optimization.

About the block size: I do exactly what karbomusic said, which is Preferences>Audio>Device>ASIO Configuration>interface's driver control panel and set the buffer size to the lowest=64.
I've tried to use Reaper's "Request block size" option, but it didn't help.

Until now I've tried not to worry too much about latency because I thought it was a marginal problem, but it's interesting to know that for someone even a few ms make a difference, let alone 17 ms.

[karbomusic]

Originally Posted by Nizhny Tagil

Until now I've tried not to worry too much about latency because I thought it was a marginal problem, but it's interesting to know that for someone even a few ms make a difference, let alone 17 ms.

Yea, that's pretty nasty latency in my opinion. I'd offer more but I'm not sure where the latency is coming from if you have your actual sound card down to 64 samples. Let me ask this fearing I didn't read closely above..

Is the latency is what you hear while your are playing or is that OK but when you listen back it is out of whack timing wise?

If the latter, you can adjust for this in preferences > recording then make sure "User audio reported latency" is checked, then to the right of that there are input and output manual offset settings. These are supposed to be used with a loopback test to account for any disparity between what the driver reports and what you actually see in the recorded result. Keep in mind this has nothing to do with what you hear when you are playing but where the recorded notes actually land.

[Nizhny Tagil]

Originally Posted by karbomusic

Yea, that's pretty nasty latency in my opinion. I'd offer more but I'm not sure where the latency is coming from if you have your actual sound card down to 64 samples. Let me ask this fearing I didn't read closely above..

Is the latency is what you hear while your are playing or is that OK but when you listen back it is out of whack timing wise?

If the latter, you can adjust for this in preferences > recording then make sure "User audio reported latency" is checked, then to the right of that there are input and output manual offset settings. These are supposed to be used with a loopback test to account for any disparity between what the driver reports and what you actually see in the recorded result. Keep in mind this has nothing to do with what you hear when you are playing but where the recorded notes actually land.

No, the problem's only about what I hear when I play.
You seem pretty knowledgeable about this, so may I ask you: where else could latency come from, if not from bad/old equipment and/or buffer size too large, or plugins adding latency etc.?
Because I think I've checked off every possible option, but who knows, maybe I've missed something?
Thanks, by the way.

[Judders]

Hopefully Karbo will be able to steer you in the direction of how to affect latency on Windows machines (I'm a Mac user for music), but just for the record your processor should in no way be a barrier to low latency recording (people were doing it before the days of multi-core anything, including myself).

[karbomusic]

I don't know actually but a quick search on that unit resulted in a number of threads showing that ASIO4ALL was required to get the latency down. You seem on top of things so I'm assuming your test was with no VSTs etc. just to simplify things. Though I've never needed ASIO4ALL and don't generally recommend it until someone has a problem only it can solve - it might be a good troubleshooting step.

So you could try it and uninstall if it doesn't help. I don't think that is an ideal solution but might by you some time (pun) until something better comes along. Another idea might be a completely different USB buss on the machine but that is a longshot.

[stratman]

I used to use and old M audio PCI card which had about 7ms latency on the 128 setting, which was the minimum I could use without getting pops and clicks. That was based on a loopback test. Even then I could feel the latency when listening through my monitors which were about 3 feet from my ears, which would add another ~3ms. But things were acceptable when I used headphones, so I guess that I start to be put off at somewhere between 7 and 10ms latency.

With my new computer I use a Zoom UAC 2 USB 3.0 interface and the latency is imperceptible.

If your computer isn't powerful, you could use a line 6 interface and monitor with the built in sounds, as has been mentioned earlier.

[Nizhny Tagil]

Originally Posted by Judders

Hopefully Karbo will be able to steer you in the direction of how to affect latency on Windows machines (I'm a Mac user for music), but just for the record your processor should in no way be a barrier to low latency recording (people were doing it before the days of multi-core anything, including myself).

Even monitoring through software? Well, that's good to know because I have a collection of old games that require old hardware.

karbomusic, I did the loopback test without anything going, no anti-virus, no firewall, no internet, no plugins. By the way, I've run LatencyMon and it didn't find problems either.
I remember trying Asio4All but they weren't an improvement. I'll try that again.
The interface right now is plugged into the rear USB bus. Aside from one port that gave me BSODs, the others seem fine.

Slightly unrelated, but is it possible to do what jerome_oneil suggested, i.e. using a distortion pedal and monitoring through the interface, and then apply ampsims and cab impulses on the recorded track in the box later?
That would help with virtual instruments, but I can play only very simple stuff with the keyboard anyway.
If that's a viable option I could borrow a pedal or find a used one for cheap.

[karbomusic]

Originally Posted by Nizhny Tagil

Slightly unrelated, but is it possible to do what jerome_oneil suggested, i.e. using a distortion pedal and monitoring through the interface, and then apply ampsims and cab impulses on the recorded track in the box later?

It would work, I've done it before. Monitoring won't have the benefit of the cab emulation while you are laying the track so YMMV on whether that's a problem. FWIW I did some tests last year comparing an OD pedal of mine with an amp SIM, minus the cab simulator and setting both to similar settings they were pretty much identical on an oscilloscope. That tells me you can probably do a lot with pedals because most of the magic is the cab SIM etc.

[Nizhny Tagil]

Originally Posted by karbomusic

It would work, I've done it before. Monitoring won't have the benefit of the cab emulation while you are laying the track so YMMV on whether that's a problem. FWIW I did some tests last year comparing an OD pedal of mine with an amp SIM, minus the cab simulator and setting both to similar settings they were pretty much identical on an oscilloscope. That tells me you can probably do a lot with pedals because most of the magic is the cab SIM etc.

That's interesting.
I'm thinking: what about effects that usually go before the distortion? I use a noisegate and wah pretty often, would they work after it?
Come to think of it, a real wah pedal could save me the hassle of automating it note for note...

[karbomusic]

Originally Posted by Nizhny Tagil

That's interesting.
I'm thinking: what about effects that usually go before the distortion? I use a noisegate and wah pretty often, would they work after it?
Come to think of it, a real wah pedal could save me the hassle of automating it note for note...

I'd try it just be conscious of setting the gain levels on the Lexicon since the signal level will potentially be a bit hotter coming out of the pedal.

[Nizhny Tagil]

Originally Posted by karbomusic

I'd try it just be conscious of setting the gain levels on the Lexicon since the signal level will potentially be a bit hotter coming out of the pedal.

Understood.
I've never seriously considered it, but this might be a way around if I can't make the latency work software-wise.

[Judders]

Originally Posted by Nizhny Tagil

Even monitoring through software? Well, that's good to know because I have a collection of old games that require old hardware.

Faster/more CPU's will allow you to run more plugins at any given audio buffer size, but they don't reduce latency per se, if you get the distinction?

EDIT: wait, I'm not really explaining myself well. A faster CPU will process auto faster, but my only point is that recording one channel of audio while a few more are playing takes so little CPU that your processor is definitely not the bottleneck in your latency problem.

If you use a huge fx chain of virtual amps when you record guitar, or if you're recording CPU heavy virtual instruments, then CPU is very relevant in allowing you to use a smaller buffer.

[Nizhny Tagil]

Originally Posted by Judders

Faster/more CPU's will allow you to run more plugins at any given audio buffer size, but they don't reduce latency per se, if you get the distinction?

EDIT: wait, I'm not really explaining myself well. A faster CPU will process auto faster, but my only point is that recording one channel of audio while a few more are playing takes so little CPU that your processor is definitely not the bottleneck in your latency problem.

If you use a huge fx chain of virtual amps when you record guitar, or if you're recording CPU heavy virtual instruments, then CPU is very relevant in allowing you to use a smaller buffer.

Yes, I never thought of it but that makes sense.

Please forgive me, I'm not exactly a newbie, I've recorded a lot of stuff as clueless as I am, but I still can't get, with such low buffer size, what else adds up to the latency I've reported.
Couldn't the bottleneck be USB bandwidth/speed? The Lexicon Alpha is USB 1.1.
I'm sorry if that was stupid, but I thought that if you can have a low buffer size without glitches, then basically you're good to go.

[Judders]

Originally Posted by Nizhny Tagil

Yes, I never thought of it but that makes sense.

Please forgive me, I'm not exactly a newbie, I've recorded a lot of stuff as clueless as I am, but I still can't get, with such low buffer size, what else adds up to the latency I've reported.
Couldn't the bottleneck be USB bandwidth/speed? The Lexicon Alpha is USB 1.1.
I'm sorry if that was stupid, but I thought that if you can have a low buffer size without glitches, then basically you're good to go.

Not stupid at all

But USB 1.1 is not the bottleneck either. While you're recording it only has to pass two channels of audio out (the DAW main output stereo channel) and one in (the channel you're recording).

I think the problem must either be the interface itself or the drivers for it.

I'm as lost as you when it comes to setting up audio drivers on Windows though. I expect to plug things in and just have them work

[Nizhny Tagil]

So I've tried Asio4All. I've left the native ASIO drivers installed, otherwise the interface won't be recognized by Windows, but I've used Asio4All for the loopback test.
Plus, I've disabled everything in the BIOS. Network device, onboard soundcard, anything. I've also disabled every C-state related option.

With Asio4All and the buffer size at 64, I get around 10 ms.
It's not 100% stable, though. A couple of times, while playing guitar, when hitting the record button or shortly after, I got a dropout (i.e. no sound for a very short time, and then back to normal).
However I did manage to play pretty smoothly for a while and it feels like I can play again.
I've noticed some crackling when the strings were muted, but I think it's noise produced by the guitar because if I moved the cable plug in its socket it would go away; it must be the guitar's electronics acting a little funny, rather then glitches due to the CPU struggling.

Besides, despite all that disabling, the CPU was far from maxed out.
So I think the problem lies all in the interface and its native drivers.

What do you think?
Anyway, even with the occasional dropout, that's surely an improvement, so thank you.

[Fabian]

Originally Posted by Nizhny Tagil

So I've tried Asio4All. I've left the native ASIO drivers installed, otherwise the interface won't be recognized by Windows, but I've used Asio4All for the loopback test.
Plus, I've disabled everything in the BIOS. Network device, onboard soundcard, anything. I've also disabled every C-state related option.

With Asio4All and the buffer size at 64, I get around 10 ms.
It's not 100% stable, though. A couple of times, while playing guitar, when hitting the record button or shortly after, I got a dropout (i.e. no sound for a very short time, and then back to normal).
However I did manage to play pretty smoothly for a while and it feels like I can play again.
I've noticed some crackling when the strings were muted, but I think it's noise produced by the guitar because if I moved the cable plug in its socket it would go away; it must be the guitar's electronics acting a little funny, rather then glitches due to the CPU struggling.

Besides, despite all that disabling, the CPU was far from maxed out.
So I think the problem lies all in the interface and its native drivers.

What do you think?
Anyway, even with the occasional dropout, that's surely an improvement, so thank you.

I have the Lexicon Lambda, big brother of the Alpha, and I agree that the native drivers are crap. I get much lower latency and jitter with ASIO4ALL, so I only use that now, never the native drivers.

[Nizhny Tagil]

Originally Posted by Fabian

I have the Lexicon Lambda, big brother of the Alpha, and I agree that the native drivers are crap. I get much lower latency and jitter with ASIO4ALL, so I only use that now, never the native drivers.

Fabian, I remember you were the one that first suggested me to use Asio4All with that interface.
Frankly, to me the native ones are dead stable, never caused a glitch, but now I suppose I also know why: they never really get to 64 samples even if you tell them to.

[zeekat]

Did you experiment with WASAPI mode? Wasn't very stable in my case, but apparenty some people get okay results.

17m at lowest buffer is pretty lousy indeed. You could add it to the latency thread for posterity, hope the DB is still updated. IMO it should be stickied and updated forever, there's never enough info about it, manufacturers hate sharing this information for some reason.