Here is a screenshot of the tempo map in Logic - here you can see the overall tempi changes.

This is a comparison of the recorded overall sound resonance (audio spectrum) of two stradivarius violins in the range of G3-O6. The vertical axis is the energy in decibel and the horizontal axis is frequency in midi notes (middle C is C4=60). These recordings are Itzhak Perlman and Sandor Vegh performances of Bach's solo violin sonata in G, the Adagio.

©1999, Nemerical Sound. All rights reserved. No part of this document may be reproduced or transmitted in any form by any means, electronic, mechanical, photocopying, recording or otherwise, with the prior written permission of Numerical Sound. For Information on getting permission for reprints, excerpts or further details, contact Ernest Cholakis at Numerical Sound, ernest@numericalsound.com, 414-444-6644, (FAX 416-444-7045).

An Analysis of Swing Characteristics of 16 well known Jazz Drummers

Although we may not agree on what is swing, we know it when we hear it. The following analysis deals with swing as the rhythmic placement of the offbeat. The common assumption is that at a medium tempo, up to 200 beats per minute, most drummers will play the offbeat as the third note of a triplet. At a slower tempo, it will move closer to the next beat, more like a 16th. As the tempo increases, the offbeat will gradually approach a straight eight.

It has been difficult to the explain why a traditional jazz ride felt so different in the hands of the following drummers:

• Art Blakey
  
• Art Taylor
  
• Bernard Purdie
  
• Buddy Rich
  
• Donald Bailey
  
• Ed Thigpen
  
• Max Roach
  
• Mickey Roker
  
• Philly Jo-Jones
  
• Jack Dejohnette
  
• Elvin Jones
  
• Steve Gadd
  
• Roy Haynes
  
• William Kennedy
  
• Tony Williams
  
• Dave Weckel
  

I decided to look at how these drummers actually played their ride cymbals at tempi 200-251 bpm. One representative two bar pattern was extracted from a recording featuring each drummer. The total duration of the two measures was divided into eight equal beats, represented as 1000,2000,3000, etc. All the timing are compared to idealized beat to establish when the drummer was playing ahead or behind the beat, and how the he is placing his offbeats. Amplitude (dynamics) was also measured (in decibels) with the loudest attack equal to 100.

In comparing the performances, the first factor to consider is the number of swing points, or offbeats followed by downbeats. Here the variance ranged from 3 (very open) to 8 (very busy - intense). The placement of the swing point is an indication of how hard a given drummer swings. The range here was from 23-505 (250 being a 16th, 333 a triplet and 500 an 8th note).

Although there is an overall tendency toward the eight note at faster tempi, it turned out to be much more an individual matter that predicted. Particularly, notice that Art Taylor's offbeats were nearly 16th notes at a tempo when most drummers were on the 8th note side of the triplet. Interesting, Taylor was (with Rich & Purdie) one of only three drummers who played behind the beat. Among the busier examples (evident by the number of swing points) Steve Gadd and Roy Haynes averaged predictable near the eight while Tony Williams was on the 16th side of the triplets and Dave Weckl averaged perfect triplets. Gadd and Weckl were boiling at the high end of variability.

The above mention observations are just a few musical observations -- download the pdf file and see if you have a few others -- I would be interested to hear your observations. The full list of the swing data with graphs in acrobat pdf format is available simply click on the next line.

Introduction:
When attempting to create a realistic MIDI performance using drum samples, I discovered huge discrepancies between the MIDI performance and the actual original acoustic performance the MIDI file was based on. To get to the bottom of this, I decided to graph the dynamic range of real-world instruments versus the dynamic range of samplers. This allowed me to visualise the nature of the transfer function (the relationship between the output and input signal). I was quite surprised by how varied the results actually were.

Place the Guitar Graph and the Snare graph here

To illustrate the results, I have provided graphs and audio examples to document the comparative differences between the acoustic performances of electric guitar, drums and piano, and the same 'digital performances' triggered from the sampler. From this research, I've explored the use of new technical features and the creative use of existing features that might be incorporated into current samplers to correct the existing limitations that are discussed in this article.

The following software samplers were analyzed in this article :

• GigaSampler
  
• Kontakt
  
• Halion
  
• EXS24
  
• Reason Redrum & NNXT
  
• MachFive
  
• SampleTank
  
• Logic's EVP88
  

Here are velocity versus output graphs of popular software samplers. The closer the velocity output is to the red line the more accurate the sampler can represent real world acoustic sounds - but the minimum level is also critical. You will have to read the article to find out which sampler they represent !

© SOS Publications Ltd 2005. All rights reserved. Permission obtained to post excerpts from article "Sample Velocity".

Since this article was written Digidesign release a sampler in 2007 called Structure that can accurately reproduce the dynamics of acoustic instruments because it has a linear decibel transfer function (1 to 96db) the the user can program dynamics on a per note basis.

Accurate velocity reproduction in a sampler is essential for an accurate re-creation of feel. Here are the actual output of the hi hat/Snare parts in three well known grooves. Bob Marley's Waiting in Vain (Carlton Barrett), Toto's Rosanna (Jeff Pocaro) and Steely Dan Song "Chain Lightning". From the graphs you can see that the dyanmics require at least a 30 decibel dynamic range. If output of the velocity versus midi value for the HiHat cannot cover the 30 db range and is not a straight line (linear db) then even scaling the velocity values will ruin the groove.

Output of the Hi Hat/Snare parts in three well known grooves