Printable View
I'm using Traktor S4 PRO as my software & all the MP3's I purchase are 320Kbps.
So what would be the best sample rate to use 44100, 48000 or 96000 when using 320kbps?
I understand if using WAV's then 96000 would be the best option and keep the highest quality, but with 320kbps mp3 do's would it be best to drop it to 48000?
At the moment i have it set to 96000, was wonder what other people have theirs set to & what bit rate the files they use are?
44100 is all you need
The reason I use it the way I do is on 96000 with a buffer of about 320 I get a latency of about 7.1ms.
Mp3 have a sample rate of 44100. The only thing you earn with increasing it in Traktor is CPU usage.
What have most you guys got your
A, Sample rate
B, Buffer size in mb
set to?
& whats the latency/processing time in ms with the setting you using?
They can also be 48000 so I guess that will be my new setting as I don't play WAV's.
But why can i achieve a lower Latency/process time with a higher sample rate?
Sample Rate: 44100
Buffer: 128
Procesing: 2,9ms
Output: 2,3ms
Overall: 5.2ms
I've just done a little experiment there and I'm not happy with NI!
If I use the S4's sound card I get a lot higher latency than what I do with the same settings on my Pre sonus audiobox usb (external usb soundcard)!!!!
For e.g.
Audio Device - Traktor Kontrol S4 (core audio)
Sample Rate - 48000
Latnecy 128 - processing 2.7 output 5.7 overall 8.4
Audio Device - AudioBox USB (core audio)
Sample Rate - 48000
Latnecy 128 - processing 2.7 output 1.5 overall 4.2
Thats a massive difference!
I have one theory that it could be to done with when using a external USB sound card all sound is handled by one USB cable and the NHL/midi by the other!
Either that of the sound card in the S4 (or drivers) are less than desirable.
The down side to using the S4 as a controller and dedicating a usb sound card is that the master volume is controller by the soundcard not the S4 and the headphones don't work =(
I want my S4 sound card to have the results the USB sound card gets. If it is to do with amount of able I wouldn't have minded the putting 2 on the S4 and saying the reason we have add the 2nd one with to handle the sound to make it shit hot!
any thoughts?
Anythoughts
did you try going lower? or is that just the default latency?
I could go lower at the risk of audio drop outs!
My main point is that my S4 sound card is 4.2ms SLOWER than running a external one on exactly the same settings!
Before slating, it would be pretty standard to lower it and find your optimum point. I've a feeling you'd probably get quite close to the other readout. Just lower it to 4.2, see what it's like. If it is dropping out, increase notch by notch until you get clear audio coming through.
Erm, don't you tell the soundcard what latency to run? I think those are just the defaults arenm't they (he says having never had an S4, and going off a dim and distant memory of how he set up his A8)Quote:
Originally Posted by DJ Shifty Sheep
This is true assuming that you don't affect the pitch, effects or internal eq.Quote:
Originally Posted by padi_04
I tend to at least affect the pitch when I mix so I find a real advantage in a higher sample rate. I have good monitors and I can hear the difference between 44.1k and 96k, but the sound will only be as good as the weakest link in the chain eg audio interface - amp - speakers -ears.
Yeh, I'm not entirely sure how the S4 latency works cos I know it's updated from Pro, but can't you just lower the latency manually to find where it actually starts dropping out?Quote:
Originally Posted by Sherlock Ohms
The latency is worked out via setting a buffer size in MB the lowest being 64mb and I have mine on 128 the second lowest setting! The processing time stays the say but its the output time that takes longer on the S4 why?
I have already test to see if it stable! The point most people are missing is that fact the none NI soundcard do's the same job a lot faster in fact 2x's faster 8.4ms was the S4 card and 4.2ms for the usb soundcard.
Do's anyone else think the latency issue is due to a bottle neck in the USB? (I already have a good USB cable I got the one DJ Techtools are selling) I.E. Running NHL & the sound over one usb. Because when I use a external sound card only the NHL is sent via the S4 USB and audio handled by a 2nd USB to the soundcard splitting the load?
I'm using mine at the same latency achieved with an audio 8dj. are you on a mac or pc?Quote:
Originally Posted by DJ Shifty Sheep
I'm on a mac book,
2.4 GHz Intel core 2 Duo
4GB DDR3
Man, you surely can decrease latency. I have pretty much the same machine but with 2g of ram and I'm getting better latency. 5,2ms leaves me a little extra headroom actually.
But its not the processing time thats slower its only the output timing that changes!
If I put it on 96000 and turn the buffer up to 320-512 I can get higher processing time but a lot lower output time, I.e. I can get the overall down to between 5.1ms (with pops and clicks) but more towards 512 it handles it with a lower overall time than if using 44100 or 48000?
What made me think was seeing how low yours was using 44100!
Whats slowing mine down as I said I'm using a good cable?
Are you turning off wifi and bluetooth before a gig? I'ts the only tinkering I do before each gig. Btw, I'm using the bundled S4 USB cable.Quote:
Originally Posted by DJ Shifty Sheep
I'll do some experimenting in a little bit see if turning off the wifi helps (bluetooth already off), I have turned it off before when recording a mix because I'd seen it suggested but never checked the latency.
But I have a feeling this will make no difference as its not the processing time where its slowing down, but i'll check back in a bit and let you know.
While messing around on the S4 last night I found that turning tha WiFi off did nothing to help. Also 8.4ms is the lowest I can have while on 48100, Don't get me wrong even on 8.4ms I have a very responsive controller but...... But it's still pissing me off that I know a 3rd party USB soundcard will out put in 4.2ms :confused:
Also the fact we have (padi) both have mac books and the S4 yet your time is so low :confused:
Peace out guys!
I really think you're searching for something that may not even be necessary. 8.4ms is still incredibly small. You should be fine when playing music especially considering that human error (for pushing buttons) is more than 100 times that amount. (This is for your thumb to a stopwatch :P )
keep in mind.
the s4's usb cable is also carrying high rez midi data along with the audio signal.
your audiobox has the luxury of only carrying audio.
I have suggested that this be the case in a earlier post.Quote:
Originally Posted by xtianw
Also anything over 10ms is noticeable to me when using the jogwheel, but what is bothering me is that I know there are people out there running pretty much the exact same set up but getting a lot lower latency.
Also if I lower the sample rate the output time go's up :confused:
I can get the lowest latency on my system running 96000 with a larger buffer than with 44100 with the smallest buffer. To me it makes no sense, I was hoping someone could explain whats happening.
"Also if I lower the sample rate the output time go's up"
http://imagemacros.files.wordpress.c...6/lol_wut1.jpg
[ame="http://en.wikipedia.org/wiki/Digital_audio"]Digital audio - Wikipedia, the free encyclopedia@@AMEPARAM@@/wiki/File:Text_document_with_red_question_mark.svg" class="image"><img alt="Text document with red question mark.svg" src="http://upload.wikimedia.org/wikipedia/commons/thumb/a/a4/Text_document_with_red_question_mark.svg/40px-Text_document_with_red_question_mark.svg.png"@@AME PARAM@@commons/thumb/a/a4/Text_document_with_red_question_mark.svg/40px-Text_document_with_red_question_mark.svg.png[/ame]
The reason increasing the sample rate reduces the latency is simple math.
You've got 48000 noises a second or 98000 noises a second.
Now, your buffer is 128 - traktor stores 128 noises and plays them all.
SO traktor empties it's buffer at a rate of 48000 or 98000 giving...
37.5 (48000) or 76.5~(98000) new buffers each second.
Meaning...
When I press play traktor stores 128 noises in memory BEFORE any sound is sent.
When I say scratch, traktor adds this new sound to the end of the buffer and it gets played but not until after the other 128 noises have played.
The time it takes to get to this sound (the latency) is decided by the sample rate as I showed above.
Did that make any sense kids? Or do I get another talking fruit?
I understand that but mine do's the opposite of normal!Quote:
Originally Posted by Remote
If I LOWER the sample rate the output time go's UP!
absolutely. let me try to formulate it in different words.Quote:
Originally Posted by Remote
The buffer in Traktor determines the number of samples that are stored. so a buffer size of 128 means that 128 samples are stored.
now, if you work at 44.1kHz, there are 44,100 samples per second. Let's say we start with a full buffer. If the computer cannot process audio (suppose it's because it's busy doing other stuff), the audio can continue to play without interruption until the buffer is emptied.
how long is the buffer in units of time? well, we have 128 samples / (44,100 samples / s) ~ 0.0029s = 2.9ms. Thus, at a sample rate of 44.1kHz, the music can continue to play for just under three milliseconds until we get a dropout.
now suppose we double the sample rate to 88.2kHz. if we keep the buffer size at 128, you can go through similar math to find that the music can continue to play for just under 1.5ms until we get a dropout.
thus, when we increase the sample rate without increasing the buffer, the buffer becomes less generous and the chance of dropouts or glitches increases. to illustrate this: in the first case (with 44.1kHz), if the computer is "busy" for 2ms and cannot replenish the buffer, we don't get a dropout. in the second case (with 88.2kHz), if the computer is busy for 2ms and cannot replenish the buffer, we DO get a dropout.
Add to this that it is about twice as computationally intense to do audio processing at 88.2kHz than at 44.1kHz. So even if you doubled the buffer to 256 samples, the chance of dropouts is still higher in the 88.2kHz case than in the 44.1kHz case.
of course, all of this is abstracting from a bunch of things but i think this illustrates the core issue. OP, you're lying to yourself, nothing's for free ;)
It is simple,
Buffer size divided by sample rate = latency
eg:
256 samples / 96000 Hz = 0.0026 seconds or 2.6 ms
256 samples / 44100 Hz = 0.0058 seconds or 5.8 ms
The bottom line is that 96KHz has significant advantage in audio quality when using Traktor and it provides much lower latency. The only disadvantage is that it has a higher CPU cost, but if you have a modern CPU and your laptop is optimised well this shouldn't make any difference in use.
As a demonstration of how using 44,100Hz loses information:
If your source track is 44,100Hz and you play it at +4, in one second you will have:
44,100 x 1.04 = 45,864 samples
If you then play that out at 44,100Hz you lose 1,764 samples every second.
If you play it a high sample rate you lose nothing.
When you apply EQ and effects and key lock there is a huge amount of extra information generated.
Most people are completely missing the point, I would expect the higher sample rate i.e. 96000Hz(more info) to take longer to process then 48000Hz (less info). But i'm getting it the other way around.
48000Hz with a buffer of 128 = 8.4ms
96000Hz with a buffer of 128 = 4.2ms
Please people stop trying to explain the basics and look at my numbers! It doesn't make sense to me, for the 48000Hz to have a larger latency.
^^^ you still don't get it. probably the reason is traktor is reporting things in a confusing way. traktor does NOT show your system's actual latency but it shows you what are essentially buffer sizes. please re-read my post.
what you observe is a reduction in the amount of music (in units of time) that is stored in the buffer.
to give a drastic example: I install traktor on some 10 year old, super-slow computer. after 30min *sweat*, the installation finally goes through. i now set the buffer in the audio settings panel to 64. And WAAAAAAAAAAAAAAAAAM!!!! Traktor now shows super-low latency on my lemon of a computer.
if you think this is the computer's actual latency, you are wrong. if you think the actual latency is much higher and at current settings we'll get nothing but glitches and drop-outs, you are right.
Your numbers are correct, that is the way it is. What is wrong is your assumption. The answers have been given to you, it is up to you if you want to understand them. :)Quote:
Originally Posted by DJ Shifty Sheep
The setting are different in the S4 version of Traktor.Quote:
Originally Posted by itskindahot
You can set the sample rate & buffer, traktor then tells you the:-
Processing time & the output then adds them to give you the over all. (and its fairly accurate) also I have tested the settings on my machine to make sure its stable (so I know the latency is correct).
The 96000Hz setting with the same buffer size OUTPUTS faster then the smaller sample rate of 48000Hz with the same buffer size.
"Latency is a measure of time delay experienced in a system" from (wikipedia) I.E. how long it takes from me moving the jog-wheel for it to be processed and outputted.
Why would it make sense for the higher sample rate 96000Hz (more infor and higher cpu usage) be more responsive!
Also please explain how Padi use's a macbook with less RAM & a S4 but has a much lower output time.
If I could get my output time down my numbers would read:-Quote:
Originally Posted by padi_04
Sample rate 48000Hz
Buffer 64
Processing: 1.3ms
Output: 2.7ms
Overall: 4.0ms
But this is what I have:
Sample rate 48000Hz
Buffer 64
Processing: 1.3ms
Output: 5.7ms
Overall: 7.0ms
OUTPUT 5.7ms when I was using Serato scratch live I had a stable latency of 5ms (my output time is more than that now without adding the processing time)
To reiterate, Why is my out put time so high?
Okay I'll give this another bash but let's try an analogy this time...
Say traktor is a car, the sample rate is the speed traktor is traveling at and the buffer size is the tracks length.
Traktor if traktor goes at 48000 or 98000 on the 128 track, which one is going faster?
On the padi_04 thing, I too get a different number to him. I get:
48000
64
1.3 5.7 7.0
Meaning I probably have a different Macbook to him (I have a 09 unibody 13)
The way I see it is, 96000Hz is more info to process than 48000Hz so it should take longer & I'm right it do's take longer to process but...... even though it processes 48000Hz faster it still takes longer to output (I know sample rate just sends more info, but its signal isn't any faster traveling down the wire), So why the massive slow down in output time on what I see as the smaller task(less info).
96000Hz = More information to process
48000Hz = Less information to process
I just want someone to explain whats happening at the output stage & not the processing stage, as the processing stage acts as I would expect (the larger sample rate of 96000Hz takes longer to process) but it outputs faster than the 48000Hz?
Surly a longer processing time should equate to a longer output time!
Also the buffer should make no difference once its been processed as the buffer is the safety net for pre processing.
Why, tempo adjustment, stretching, at 44.1 your sound will tear up. 96k, you can get away with larger tempo adjustments. Studio techs will use 88.2 because it dithers down to 44.1 well.