Latency - will it getting better?

  • I don't know how the new quoting works but this is in response to Froschn.


    I disagree entirely our ears and brain can detect 3ms latency as our brain and ears are adaptive to the environment we are in. For example, a symphony orchestra can play in unison despite a wide distance between players. There is no conscious adjustment for latency. When you're on a stage and singing, the sound within your head (the only case of 0 latency which exists) and what's coming from the monitor is not in synch but you're mind corrects. I mean technically, any guitarist which plays in a group is able to reconcile the multiple latencies due to differing distance from the drum kit and variable human response time and distance from amp to play "in time". If latency is so destructive to playing, how can a guitarist possibly change position on stage and function? We do not think about this, we just do it. Our brain adapts. Splitting hairs about the minuscule amount of latency introduced with a modeler is being a bit obsessive I think. If it weren't easy to measure I often wonder how many would really notice it at all.

  • Our senses are also easily fooled and the influence of placebo is impossible to ignore when doing accurate measurements. Therefore it's always best to do A/B blind tests to see if we are correct when we can't see or know what we are playing through. Many are surprised with the results when doing such sound tests.

  • Latency isn't destructive or alike. Somewhere we are used to it and somewhere not, that's all. Close your eyes and let a friend drop something onto the desk. You can say in what direction and how far away, right? You are used to that enough, your brain learned that. Little children couldn't do the same, they still need to learn. They sometimes don't notice a car coming nearer. Bad. The longer you played over a certain amp the less tolerance you have for the littlest changes. Nothing destructive, it needs time to get used to new stuff in that way, that's all.

  • With your eyes closed, attempting to determine distance based on sound is based solely on amplitude unless one has a reference. What you are are alluding to, the comparison of an amp and FRFR and modeler, there are a great many differences other than the potential of a small additional latency. In fact, depending on where you're monitoring from (off axis), some frequencies with an FRFR based system are going to be hitting your ears with less latency than a traditional amp (where they are typically broadcast past the user and bounced of a wall, and masked) due to the angle of dispersion when standing equal distances from both sources. You'll get no argument from me that an FRFR is a different experience than a traditional guitar cabinet, but I don't think latency is the biggest culprit.

  • Everybody can check that himself. Put a delay in the rig, sound uncoloured, mix at 100%, then change the time between 1ms and 20ms. The point where you start to dislike it may be different personally and there might be some placebo believe going on...but the try is for free.

  • You can calculate distances in milliseconds, you can counter-check artificial latencies with a delay, you can do many things but in the end there will be always something metaphysical about this latency thing. I come from classical guitar, then strat-cable-champ purism and now I am fine with the Kemp.

    www.audiosemantics.de
    I have been away for quite a while. A few years ago I sold my KPA and since then played my own small tube amp with a Bad Cat Unleash. Now I am back because the DI-profile that I made from my amp sounds very much convincing to me.

  • Froschn That test is superb...as long as someone else is controlling the delay amount and you as the guitarist are blind to how much delay is actually happening. I'd bet no one (or very, very few) would detect anything up until around the 7ms+ mark.

  • ... and just to remember you guys, that the OP was talking about a 1.5ms difference between 2 devices.


    @OP: I wouldn't bet a single cent on any latency improvements in the current KPA. It's simply not possible without a hardware upgrade. Limited processing power, high internal sample rate, pretty complex DSP ... no room for the improvements you're asking for. ;)

  • Quote by hand: Froschn That test is superb...as long as someone else is controlling the delay amount and you as the guitarist are blind to how much delay is actually happening. I'd bet no one (or very, very few) would detect anything up until around the 7ms+ mark.


    Quote end.




    Never tried that test with a 3rd person that controlled the time parameter. I fully trust in the placebo effect :)


    Glad you had fun with the delay!

  • @OP: I wouldn't bet a single cent on any latency improvements in the current KPA. It's simply not possible without a hardware upgrade. Limited processing power, high internal sample rate, pretty complex DSP ... no room for the improvements you're asking for. ;)

    Why? I think its posible, when we can have all the effects on at the same time and 6 super fast harmoniser at the same time. Would be cool to have a latency paramether like the soundcards have, going to 1.5.ms. :)

  • About latencies in DAWs and audio interfaces:
    When you set the buffer size to a certain number of samples, the latency will be at least double as long, AD and DA not included.


    Explaination:
    When you have set the buffer size to 64 samples, they will all processed in one block.
    Thr process will start, when the last (!) sample has arrived from the input.
    The process will take up to 64 samples duration.
    In the meantime the next 64 samples from the input are received and buffered (sic!)
    The first sample of the block in question will not be sent to the output before the whole block is calculated.


    I don't have the full picture, but I was told that by several reasons even more buffers are in the chain.


    The buffer size does not necessarily guide to the true latency of a DAW or audio interface.
    Does anybody have links to real latency measurements?

  • What it seems to me has to be said, is that the latency setting in a soundcard doesn't serve the purpose of lowering it, but rather raising it when the system can't get far enough to process all the data. Otherwise, all the cards would just be gifted with the smallest possible buffer, for the lowest possible latency.
    IOW, raising the buffer gives the system time enough to process the data, somehow smoothing the need for a "real time" processing (i.e., each frame by one).


    So the point is not how small the least latency value is in the software, but how low you can afford to set it given a certain system.

  • If you like to have true latency measurement, then you'll need some reference signal like a PING in networking or timecode in broadcast video: Problem here is, that these run through a given system unaltered, so they are easy to detect and measure.


    One possible way to measure would be to load a rig of your choice, mute input and output of the KPA, reset the whole signal chain to 0, generate a max gain spike right behind the AD section and then measure the time difference right before the DA section.


    Since you don't know how much these spikes would be altered by the signal chain, you'd just measure the first occurance of signal <> 0. Since you don't know how much reverb and delay (time based effects) are introduced to the signal, you'll have to reset the whole signal chain to zero before repeated measurements.


    Not sure how to do precise measurement in an unmuted system though.

  • I think the latency measurement can be fairly simple. Use a microphone to record an analog of the acoustic sound of the guitar string to catch the attack while simultaneously recording the KPA. Run both signals through the same interface so they both have the same latency to the DAW. Compare waveforms and measure the difference of the leading edges in time.


    To me, it matters very little. I'm not concerned with 3ms and don't think it matters in my performance. I just don't feel it at all. With all the sensitive people, makes me wonder how guys on a large stage ever manage with their back line and monitors 20 or 30 feet away. Somehow they do just fine.


    bd

  • You guys are crazy. The mic example also introduces the natural latency of the soundwaves moving though the air. In fact, monitoring the KPA via headphones will have less latency than standing 5 feet from a tube amp. There is no such thing as zero latency as it is naturally occurring. In a DAW, unless you are trying to monitor real time effects on a performance, low latency is not required. In fact, the ability to max out the latency in a DAW is a huge benefit. The best practice would be actually to crank up the latency to darn near max and then set up direct monitoring off the inputs. This way your computer can take all the time in the world to process all the effects and samples in your project with out sweating and you don't have to wait for a round trip to monitor your playing.

  • Being concerned with the inherent latency of the KPA with normal use of effects is misplaced.


    If you have an awesome sound card and fast computer I doubt you'll ever even notice it.


    It also depends on the DAW and settings. For example, I literally NEVER hear latency with Reaper using anything like Superior Drummer and multiple Omnisphere instance while overdubbing on my Macbook from 2010. But I do with Ableton Live 9, and have to do a lot of buffering to avoid it, and even then, have to have mixdowns or else any running plugins just are additive in that program, cause it's programmed like a bog unlike Reaper. Ableton Live 9 on my recently bought Lenovo and Presonus F10? No latency I can detect audibly. Programmed like a water skier!


    So the KPA is the LEAST of my concerns. Perhaps if you have a few Pitch effects running? Haven't tried that. But a profile and a few regular effects?


    **IMPOSSIBLE** to detect naturally. Impossible. And the additive few ms is not the cause of the issue. That will be elsewhere in the chain.