pz wrote:This latency number for computer/audio latency was very useful in the days when it was 200 ms, but now appears to, in my opinion, makes people worrying about the wrong things. The main point I had is that some people do not seem to realize how short 1 ms really is; perhaps we should go back to writing it as 0.001 second.
Hello All,
Regarding latency issues, I completely concur with pz's observation that people might be worrying about the wrong things. Are you aware that 20ms is one-fiftieth of a second and that a "very large" 70ms is approximately one-fourteenth of a second? From some of the rather dire concerns I am reading in this thread, it sounds as though you believe the latency is on the order of half a second or more.
Here is my take on latency of a real grand piano:
Although the literature states the response time of a real piano is on the order of, say 30 to 40ms from the onset of depressing a key, I believe the latency mentioned in this thread is defined as the timing delay one experiences after the midi keyboard has bottomed out and the sound is heard.
When one plays a real grand with the intention of creating a pianissimo sound, the player is pressing the key down so "slowly" that the hammer strikes the string BEFORE the key bottoms out in the keybed. In other words, the hammer shank is in free flight after the escapement and has already contacted the strings while the front of the key is still moving downward by the player's finger. Typically the hammer contacts the string approximately 10ms (1/100th of a second) before the key front bottoms out in the keybed.
When one attempts to play fortissimo, the hammer of a real grand piano typically contacts the string AFTER the key front has bottomed out in the keybed, because the high force of the action upon the hammer shank has caused the shank to flex (e.g., the hammer itself remains stationary by its own inertia for a minute fraction of a second AFTER the hammer shank has started to move). In addition, bass note hammers have more inertial mass than upper octave hammers. Typically, this difference in timing -- delay, if you will -- is on the order of about 10ms difference (one-one hundredth of a second).
Somewhere between pp (where a real piano string is contacted before the key bottoms out) and ff (where a real piano string is contacted after the key bottoms out), there must be a loudness level (say, mezzo forte or forte) where the string is contacted at precisely the time the key bottoms out in the keybed.
* * * * * * * * *
The above discussion was how a real grand piano works. Note that it is different when one plays a midi keyboard, where no signal will be initiated until the key is at or very close to contacting the keybed.
Now, since a midi keyboard generally sends a note-on midi signal when the key bottoms out (or very nearly so), the "clock" for delay really doesn't start until about 20ms after the key is first pressed.
Let's now go back and review the difference between a real grand piano and a midi keyboard: You will recall for a note played pianissimo on a real grand, the "latency" between bottoming out the key and hearing the sound might be "negative 10ms", because the hammer contacts the string before the key hits the keybed. Well, clearly the sound cannot precede the midi keyboard bottoming out, because the nature of an electric keyboard mandates that a switch must be closed in order for a note-on midi message can even be sent.
While I am aware of the so-called latency of X number of buffered samples, say 64, 128, 256, 512, 1024 samples or more in a midi system, we must remember that Pianoteq also requires some time to "compute" the sound in real time. This is dependent upon the cpu speed as well as the speed of your computer's busses. There has been some confusion in terms regarding sample rate and samples to be buffered: Here, we are talking about breathing room for your computer to calculate the sound before sending it on. By increasing the buffer size, the cpu is given a slight amount of breathing room, but it comes at the expense of taking more time to make the calculations.
Larger sample buffer size = less pops and clicks, but longer latency.
Smaller buffer size = less latency, but at the expense of working your cpu harder.
Once the sound gets calculated, it must also be panned, sent through reverb, and ultimately to the audio amplifier of the sound system. There are amplifier rise times associated with components of hardware associated with an audio system.
Hopefully you have read this far -- as you know, I tend to become long-winded, especially while typing at nearly 4AM local time on New Year's Day.
Here is what I believe this latency issue boils down to:
An electronically midi enabled keyboard physically cannot match the latency of a real grand piano --- if one wants to measure the difference in units on the order of milliseconds -- thousandths of a second. Hopefully the amount of latency / delay you experience is tolerable in the music you perform. Everything in the electronic chain of depressing a note on a midi keyboard, the midi protocol itself, whether you use a USB connection versus midi in and out, the speed of your computer's processor, the number of processors in your computer, your computer's bus speed, hard drive rotational speed, hard drive data transfer speed, etc, etc, etc, WILL contribute to an amount of delay from the time your finger bottoms out on your midi keyboard and the time you first hear a given sound. Please be aware that many (most?) laptop computers are equipped with 5400RPM hard drives, when drives ranging from 7200, 10000 and even 15000 rpm's are commercially available, but probably not for laptops.
This response is rambling, especially at nearly 4AM local time on New Years Day 2010. Your questions and comments are welcomed.
Sincerely,
Joe
P.S. Happy New Year 2010 to all
Last edited by jcfelice88keys (01-01-2010 19:05)