It’s not only petty but is not how life fits... while it is mathematically correct that at 44,100 segment resolution you can figure out the number of samples that equal 4 measures. Believing that your music actually fits into such nice neat time slices is the weakness in your view.

I’ve been here, done this and have earned many t-shirts on this topic. Both digital and analog versions of the t-shirt!

Here’s what I mean... the idea of dividing music into a specific number of time slices is very similar to the old days when you used to edit tape with a razor blade. We are dealing with what is small enough to be “negligible” - negligible is a very important concept to grasp for any discussion of this type. Negligible simply means small enough not to matter.

If you cannot except this concept, I highly recommend you get another hobby. There is no perfect, there is only “human” (“human” is often used in conjunction with the concept of negligible when it comes to certain measurements).

Long story short, the value you get with your calculator will be good to get you in the ballpark, use it for that only. Then use your ears to micro manage the actual Start and End Points. At a resolution of 44,100 time slices per second... you’d be amazed at just how much wiggle room there is in what we accept as four measures... and once you recognize this fact, you’ll begin to relax if the number is a few segments longer or shorter. As a “human” it is negligible when it’s too small to hear.

A click or unwanted spike can easily be found in your very first time segment. Ultimately, in the Hitchhilker’s Guide to the Universe, so what, what is it? You are assuming it is musically useful stuff, it may not be, it could be a noise unrelated to the music. At that resolution if you assume anything that causes level is a musically useful sound, again, it doesn’t have to be that at all.

Thing is a click or pop occurs when enough signal fills the segment to cause a digit... you are assuming it’s some musically useful data, fact is it may not be at all. Cueing your data to the exact time slice that is supposed to be the next downbeat or Measure line, is not really mathematically predictable using resolutions as tiny, tiny, tiny as dividing a second into 44,100 equal segments.

And it matters not if a mechanical device is counting this stuff. To think that every downbeat lands perfectly, a set number of time slices apart... not going to happen. If it does... it’s the exception not the rule.

You refer to a large jump in the segments, which makes me wonder if you are aware of the 100ms “wiggle room” added to the front end of your Start Point. If you set Sampling to begin at a specific Measure and Beat, the sampler actually begins Sampling 100ms prior to you starting the recording.

What? Yes it’s true. Samplers are always Sampling. The Record and Stop functions only define what gets kept.
A sampler is always Sampling Data... much like a microphone is always processing sound. The Record and Stop buttons just define what gets kept. So when you set Record to Start (whether manually, by measure marker, by level) the sampler keeps 100ms prior to the START POINT. This is why at a Sample rate of 44,100 your Start Point is always listed as 4,410 which precisely 100ms

This way you have room to Trim the Start Point in either direction... if you’re late, you have data just prior to your Record Point. And if your early punching in, you can shift the actual Start Point to where it belongs.

You are removing this, right? You want to extract just the data you want to keep. This 100ms must be removed before you set your number of time segments... otherwise you’re adding a 100ms hiccup in front of your first segment.

You'll get no disagreement with me on the "not how life works" part. I'm totally aware of how unorthodox this method is, and I've pretty much dropped it at this point. I'm responding late because I've been busy so sorry about that. Though I did manage to read what you said on the 21st.

I'm continuing the thread because I still have a couple of related questions regarding how samplers work.

You refer to a large jump in the segments, which makes me wonder if you are aware of the 100ms “wiggle room” added to the front end of your Start Point. If you set Sampling to begin at a specific Measure and Beat, the sampler actually begins Sampling 100ms prior to you starting the recording.

I'm aware that the Start Point is by default set to 4410 samples (or 100ms) after the beginning of the entire audio clip. I thought that the Integrated Sampler would record exactly 16 beats (352800 samples at 120BPM) and then prepend an additional 4410 samples before them. Those additional 4410 samples could be empty (in the case of the very first 16 beats in the Song) or filled with data from the previous 16 beats. However, if this were true, the full untrimmed audio clip would equal 352800 + 4410 samples, which it (most of the time) wouldn't. Typically, instead of the clip being exactly 357210 samples which would back up this theory, I'd get anything from 356200 to 357400. It seemed random. On top of that, I discovered that there are additional samples being appended to the 16 beats, not just prepended. If it weren't for that last fact, I'd be able to calculate exactly where the real audio begins. But I can't figure out how many of the additional 3400 (with 356200) to 4600 samples are prepended, and how many are appended. Although it always seemed like more were prepended than appended, I still couldn't get a mathematically accurate figure. If there's a set ratio of prepended to appended somewhere deep in the hardware, let me know. I dare say it would solve the problem completely.

I will say though, every solution would require some level of certainty as to how many samples actually get recorded to the resultant audio clip. At this point, I'm skeptical that certainty can be obtained. Although it was clear that the entire 16 beats were encapsulated within the resultant audio clip regardless of it's length in samples, I'm guessing that the length in samples is somewhat random and depends on the universe around us. What I mean is, perhaps the additional 100ms offset of the Start Point is compensating for recording inaccuracies as best as it can. I guess the manufacturers could guarantee a record latency of 100ms maximum with Integrated Sampling recording.

These tiny inaccuracies don't matter much to me when I'm sampling individual notes or sounds, but your analogy summed it up perfectly. I really was trying to cut tape with a razor blade. Requiring the sequential audio clips to sound as clean as sequenced MIDI on a single track is virtually impossible without knowing exactly how many samples are prepended and appended to the 16 beats.

Unless I'm wrong above, and there is some way to calculate the samples, I've decided against using the Integrated Sampler to bounce MIDI and will probably just bounce the aforementioned 16 beats in Cubase. Obviously, I can't bounce the MIDI separately like I was trying to do above (the razor blade method) for the same reasons. Honestly, I haven't really tried the razor blade method in Cubase because dealing with all the different forms of latency seems convoluted and I just can't be bothered, but I simply presume it will be the same story,

If I decide to bounce the entire MIDI track in a single operation, I won't have to deal with sample-counting. But there will still be a small gap at the beginning of the audio clip due to regular record latency between the Motif and the PC. I'm wondering how to most accurately detect when the sample value exceeds 0 so I can cut the empty space before the song begins. The downside of this, is that signal that exceeds a sample value of 0 isn't necessarily intended to be recorded. Unlike recording directly on the Motif, it seems like noise now has to be taken into account.

I guess that you're the best person to ask on this topic. So I wonder two things:

1. Is the razor blade method even achievable? Not just on the Motif, but anywhere. Is it physically impossible at this point or could some really intricate tweaking do it?
2. Was I right about the 100ms Start Point inaccuracy compensation? Is it to allow for record latency within the Motif? If so, are the prepended and appended samples fixed to a ratio?

   I'm using this "problem" as an excuse to learn more about sampling and synchronization, etc. I won't bother you more about it. Just want to see what you have to say for now before I embark.