Question about samplers and sample library combination

[BezO]

Regarding the MIDI timing issues, this is a problem that plagues both software and hardware alike.

I meant to ask this yesterday since a few folks in here seem knowledgeable about timing and jitter. Would clocking a sampler or synth to a good master clock help MIDI timing at all? When jitter was brought to my attention and I started to research it, I don't remember seeing anything about MIDI. But samplers have word clock inputs...

I'm already in the market for a Rosetta Big Ben to clock my 002 Rack and converters to.

but for a similar price two S5000s would give that, and would outperform the Z-series (128 total poly, 4 MIDI ins, etc).
The S5000 can play back monophonic samples direct from disk too; as far as I know the Z-series can't do that. It seems to me that the S5000/S6000 are more professional machines than the Z-series.

S5/6000 seem to be going for almost as much as Zs on Ebay. But more polyphony and streaming directly from disk is always a good thing.

Regarding size of the samples, memory alone is not the answer to creating expressive, good-sounding instruments. A fully expanded S5000 (256MB) can store almost an hour of mono CD quality audio -- that's rather a lot of samples!

Since you don't have to load the samples to RAM in the 5/6000s, how would the performance of 256MB of RAM under those conditions compare to loading 512MB of RAM in a Z? I'm sure it would depend on the arrangement, but I would guess streaming with only 256MB of RAM would be comparable in most situations.

Good info. Thanks!

[Spaceman Sounds]

http://www.soundonsound.com/sos/mar01/a ... sician.asp

http://www.soundonsound.com/sos/mar00/a ... ditime.htm

These explain midi jitter

However, once recorded into your sequencer, MIDI events are passed to VST Instruments with sample-accurate time-stamps, so there can be no loss of timing resolution between the host and instrument. This means that anyone who subsequently quantises their MIDI tracks or enters them in step time should have sample-accurate timing. This is far better than hardware MIDI synths, that have a gap of about one millisecond between 'simultaneous' notes, because of the restricted transmission rate of the MIDI standard.

And this is what I've been trying to albeit long windedly explain. Source:from sound on sound question and answer page

That said, with MIDI's timing being about as accurate as a sundial, trying to pull off sample triggers smaller than a 32nd note will sound like hell

from remix magazine.

[Rodent]

AUTO-ADMIN: Non-MP3, WAV, OGG, SoundCloud, YouTube, Vimeo, Twitter and Facebook links in this post have been protected automatically. Once the member reaches 5 posts the links will function as normal.

I meant to ask this yesterday since a few folks in here seem
knowledgeable about timing and jitter. Would clocking a sampler or synth
to a good master clock help MIDI timing at all? When jitter was brought to
my attention and I started to research it, I don't remember seeing anything
about MIDI. But samplers have word clock inputs...

Shouldn't make much difference. Problems come from the MIDI link (including the computer), and faulty sampler firmware. Old Akai firmware had some trouble, but it's user upgradable.

For the MIDI link, spread the load, remove unnecessary device drivers, and avoid audio on the same computer that's doing MIDI.
If your system can't achieve low audio latency (~3ms) without crackling, that may suggest an interrupt latency problem, which may also cause MIDI timing problems.

Since you don't have to load the samples to RAM in the 5/6000s, how
would the performance of 256MB of RAM under those conditions
compare to loading 512MB of RAM in a Z?

Depends on the samples you're using. I've never tested streaming, but the manual suggests that for each sample, there can be only one instance playing at a time.

This would be fine for sax, clarinet, drums, where you'd only play one sample at a time.
It'd also work well for polyphonic sounds if you had a different sample for every key: a given sample would only be playing once, regardless of how many other notes are pressed.
Limitation: It can only stream about 10-20 samples at once due to disk speed; that'd be a problem for piano parts (damper pedal), but fine for organ, etc.

For very short samples, it's of little benefit. Like disk-streaming softsamplers, the Akai loads the "nose" of every sample into RAM, so it's got something to play the instant you hit a key. If the loop points and filter envelopes have been used carefully, the sample might be smaller than the nose anyway.

If you're printing your arrangement to tape track-by-track, then it's moot -- you could use all 256 megs for one instrument, record it, then load the next instrument, etc.

{quote from SOS explaining the VST synth timestamps}
And this is what I've been trying to albeit long windedly explain.
Source:from sound on sound question and answer page

Interesting articles you linked there. It's fairly difficult to design a software synthesizer that places the notes at the correct points in the buffers based on timestamps, and quite CPU intensive too. It's much easier to have it "sample" the input data before producing every buffer, triggering all new voices at the start of the buffer. Some are better than others in this regard.
It's easy to design hardware firmware to give tight timing: a "note-on" merely involves setting a few hardware registers. Given that firmware programmers sometimes mess this up, just think how often softsynth authors mess it up given that their task is even harder. ;)

The big problem with softsynths is playing them "live". While a good hardware sampler aims to trigger the note the instant it arrives (asynchronously), a software sampler played from an external device can cause timing jitter (even if the keyboard is USB). e.g., all note-on messages are quantized to an artificial "grid" that has no bearing on musical time.

If you play a 10 note chord into a single MIDI port of a hardware sampler, you'll hear sound start to appear the instant the first "note on" message is processed, which fools the brain into thinking the 6ms delay between the first and last note is much shorter than it really is.
With a software sampler being played live, there's an overall latency that may be quite high (possibly higher than the worst-case 6ms above!), and on top of that often a jitter of perhaps a few milliseconds for every note, even the first.

If you're recording each part live and leaving them unquantized, the delay and jitter may be bad enough to throw your timing off when performing.
Conversely, if you're doing quantized techno, softsynths may have an advantage providing they're designed to respond to the timestamps.


I was curious about how large the variation between good and bad MIDI gear was, so today I performed some tests on a few devices:

http://asdfasdf1102.tripod.com/ (http://asdfasdf1102.tripod.com/)

The first mp3 shows 60 note on messages on a single-voice piano patch being sent one after another, to the following devices in sequence: Korg X5DR, Roland XP-50, Yamaha P-90, Korg X5DR (again). All devices are 64-note polyphonic, interfaced with standard MIDI cables.
The test was conducted using a purpose-written program under a real time operating system and a homebrew MIDI interface to eliminate as many variables as possible.
Note how the Korg X5DR appears to sound the notes at almost the same instant. By contrast the Roland XP-50 (internally a JV-1080 with a keyboard) "strums" the 60 notes quite badly. The Yamaha P-90 is incredibly slow responding; perhaps as much as 5-10 times slower -- I could play a glissando with my finger that'd easily outpace it, if I weren't worried about breaking keys or fingers. The Korg was manufactured mid-1995, and the P-90 was manufactured late-2004. Progress, eh? =P

The second mp3 shows a short rhythm and chords loop played back on both the Korg and Roland (the P-90 has no percussion samples).
Again, the rhythm loop was generated from a purpose-written program; the ANSI C sourcecode is supplied in the third link and should compile under any Unix-like OS with some tweaks (gives a whole new meaning to "programmed drums" ;)
The program outputs the rhythm notes first, followed by a four note chord on piano (left channel), then 40 notes on a pad (two voices per note), followed by an 8 note chord on piano (right channel).
All patches were entered using the front panels of the synths; the program sends no patch change messages (which would cause the Roland to choke horribly).
Again, the Korg is recorded first (which is essentially a budget O1/W with a few extras), followed by the Roland (which was their flagship workstation released in the same year as the Korg).

Despite severely overloading the polyphony limit, the timing is still reasonably tight on the Korg -- the delay between both piano parts is small enough for it to sound more like a chorus effect than an echo, despite 40 note-on messages separating them! The rhythm is not affected at all, and remains tight.

The Roland is diabolical here; the note triggering takes so long that it actually messes up the following rhythm notes, and choruses the pad. One last dig at Roland (sorry!) -- most of the hiss in the recording comes from the Roland, and the rest from my lashed up mixer; the Korg is completely, utterly silent.

Disclaimer: I don't work for Korg! And probably never for Roland after posting this ;)

I've always felt the Roland was a little "sloppy" responding to MIDI; the JV1080 too, conversely the Korg has always seemed pretty tight in actual use. However, I wasn't expecting the results to turn out anything like that!
This is where MIDI gets its bad name from: really, really sloppy firmware design in popular synths.
As you can hear in the Korg demo, a good piece of MIDI gear is actually pretty good timing-wise, despite being taken beyond the limit of polyphony and being deliberately abused. I don't imagine the X5DR is the best out there by any means, either.

It'd be interesting to know how tight the Akai samplers are compared to this, though it'd be difficult to arrange a test where dissimilar setups can be compared fairly.
Anyone know if the newer Roland XV series is any better for timing than the JV?

[BezO]

Ended up with an Akai Z4, maxed memory & 60GB drive.

I didn't realize the digital & word clock inputs were optional on the Z4 until it was too late. Would it be helpful if the sampler was clocked to my Digimax LT (master clock at the moment)? Would that help to control jitter, or is it not necessary since I'm recording from the sampler to the Digimaxt LT?

And I desparately need the 8-output expansion. If anyone knows where I can find one for less than $350, I'd appreciate the heads up.

And thanks all for the info!