My understanding of latency just went out the window…
So - I’ve set up Traktor DVS on my new Win 8.1 machine.
Got the latency down to 1.4ms (with some audio pops/crackles).
So I started messing around with the latency settings. As a test (and to see how it would affect the latency on the control vinyl), I set it to max. (43.5ms).
Had absolutely NO AFFECT on the control vinyl?
Feels like there is no latency between moving the record, and the track changing in Traktor.
Now - I’m not complaining! The “Load” meter is showing next to nothing (compared with hovering in the red when latency setting was 1.4ms).
Is latency ONLY linked to midi control, and NOT timecode? When I hook up my controllers (F1, Dicers, MC-1000) I’ll need to adjust the latency to get a buy off between latency and audio performance, but can someone confirm if latency setting affect timecode control at all???
There are several places where latency is introduced, and there are several KINDS of latency. The latency slider you’re playing around with is controlling the latency of the audio driver buffering - which is somewhere in the middle of the process. Here’s a simplified breakdown of the different steps in a DVS where latency is introduced…
1 - analog audio from the record is converted to digital data - .5ms 2 - data from audio interface travels via USB to computer - 5ms 3 - digital timecode is checked for record direction, speed, and velocity - 4ms 4 - digital timecode is checked for absolute position data - 6ms* 5 - audio playback is fed into the ASIO driver - 4ms** 6 - audio playback is output from the ASIO driver - 6ms** 7 - audio data is sent via USB to the audio interface - 5ms 8 - digital data is converted into analog audio output - .5ms
this can be variable depending on the quality of the timecode signal
** this can be HIGHLY variable depending on many factors
So in the best case scenario above, you can see that we’re already talking about 31ms of latency. That slider you’re moving is adjusting the audio driver buffering. This is to give the computer more time to process the audio (and thus reduces how hard the CPU has to work). If your audio buffer is too low, the CPU could end up generating underruns because it can’t keep up. If the buffer is too big, you introduce unnecessary delay into the process.
So you moving that slider all the way over is causing a little more delay, but obviously it isn’t effecting your performance. Perception of latency is going to vary somewhat between people and how they perform. But most users don’t understand what latency is, and they have ridiculous ideas about what is acceptable because of this ignorance. You’ll see some chime into this thread I’m sure. Take them for the dirty unwashed peasants they are. These people are the reason you’re posting this thread, and next they’ll tell you that a man carries the sun across the sky in a chariot.
The difference with MIDI is some of the steps in the process above are eliminated or shortened. With a MIDI controller, there isn’t an A/D conversion, or either of the steps that involve timecode interpretation. You still go through the USB buffer of course, and the application still feeds the audio driver the song data (and the audio driver spits it out to the audio interface). Generally speaking the processing time is less with a controller - although how much less depends on a lot of things.
I’ll also point out that if you’re not scratching and/or not REALLY used to actual vinyl, you’re used to the latency that nem0nic described. Changing it a little isn’t as big of a deal as introducing it for the first time.
At 120bpm, 48ms is between the duration of a 1/32nd note and a 1/64th note…or just a tiny bit of distance on the record, which your brain interprets as the sound just being a little bit over on the record.
If you’re doing really fast, accurate scratching routines, MAYBE that can screw you up. Slip cuing and beat matching…not so much. That’s also about the time it takes sound to travel about 16M. Your brain will adjust based on what you’re hearing.
I’d set it lower, but I wouldn’t worry about it unless you’re consistently screwing up.
real records will always feel and sound the best for scratching in my experience, which is probably why they make double sided skipless/timecode records.
I ran into that a while back. The “latency” number on the slider in Traktor was taken to be “gospel.”
I actually did a test. I setup a microphone where it could pick up the button press AND the sound coming from the speakers. This way I could measure the latency between the button press and the sound coming from the speaker.
My computer, on that day, with that OS and those settings, etc…was producing about 38ms of latency IN ADDITION TO whatever audio latency I choose to add with the exposed controls in Traktor.
Even with that explanation, test, and demonstration…the naysayers continued to say nay.
Exactly! I’m not saying that no one could tell that the sound was “off” a little. BUT…like any live musician, there are subtle adjustments that must be made when playing with accompaniment. The latency of the peculiar DJ instrument is not special in that regard.
The other thing to remember is that your brain is very good at dealing with latency that’s consistent. If you’re always listening to audio that’s delayed by (for instance) 40ms, you automatically compensate for that - usually without noticing. This is because sound is ALWAYS delayed by the time it reaches you. Remember that sound travels at roughly 1 foot per millisecond (yes internet detectives it’s actually a touch faster than that, and depends on humidity and temperature as well). Your brain is always dealing with this type of latency, so it’s very good at negating it.
Where you really have problems is when that latency is VARIABLE. For instance, let’s say you’re using a DVS and poorly isolated turntables to generate the timecode. You switch from house music to something very bass heavy (like hip hop) and low frequency feedback begins to degrade your timecode quality. Now it takes longer to determine the platter state, but only when the bass is really hitting. This kind of fluctuating latency can be maddening - and VERY noticeable.
I’ve found that high latency is most noticeable on very fast platter techniques like rubs or uzis. Higher latency seems to “smear” the sound a bit and make it less well defined.
I can use my DJM-850 for timecode control, and set the latency WAY HIGH, and it doesn’t effect the “latency” between my hand movements and the audio from Traktor.
But when I start connecting midi controleers (Dicers, and Zomo MC-1000) all hell breaks loose, re: hand-to-audio-latency…
I’m convinced that Traktor’s latency is just weird.
The other day, I noticed for the first time that running a track through an effects unit will induce enough additional latency to cause interference issues in Bass. I’ve never noticed it before, but I’ve also never used the particular effects I was using while layering tracks the way I was. So…it might be different b/t effects as well.
The MIDI controller thing confuses me a bit, though. I’m not quite sure why that would happen.
You got me also confused now, just got the latency of my Xone:K2 to the max (Relaxed option) and the load meter rides higher than in normal mode (Normal option)…
I think your perception of the effects of latency are more pronounced with a MIDI controller because your motions are more directly coupled with the audio than it is when using a DVS.
You could be right. You could also be experiencing confirmation bias. Ultimately, there will always be SOME latency between the movement of the control and the effect on the sound. How noticeable that delay is depends on a LOT of factors…and not all of those factors have controls in the software or hardware.
