Differences between different SFZ versions of the same instrument

The instrument(s) available here are available in different versions

  1. Chromatically sampled wav files
  2. Minor third sampled wav files
  3. Minor thrid sampled ogg files

The table below summarizes the differences between the different SFZ versions.

Version Size Slight time stretching artifacts Marginal compression artifacts Velocity response
Chromatically sampled wav files Full Creatively utilized No 5 layers via borrowed neighbors
Synthesized effects
Minor third sampled wav files ~1/3 Yes No Synthesized effects
Minor third sampled ogg files ~1/10 Yes Yes Synthesized effects

It’s easy to guess that the bigger library is the best. But I will try to elaborate a little more. I find it easier to explain the drawbacks  of the versions with less or compressed samples. Then highlight what the extended version adds on top of avoiding such disadvantages.

Side effects of time stretching

Time stretching is the technique to raise or lower the pitch of a sample by stretching or compressing the waveform. For a simple extreme example if a 1 second waveform that has a pitch of 440Hz is stretched to 2 seconds, the resulting waveform will have a pitch of 220Hz because now it takes twice more time for the wave to make a cycle. With less stretching or (about 5%) samples can be shifted to a neighboring semitone.

Occasionally, however,  there would be a sudden jump in the high frequencies when a sample is retuned to a higher pitch that can be heard as ringing. Extreme side effects for pitch shifting such as the “chipmunk  effect” and other artifacts might be more familiar. The ringing heard with moderately stretched samples can be milder versions of these. Another related example is the notorious effects of autotune to pop singers.

Sampling an instrument every minor third means that two semitones between notes are not saved as audio files. Each octave is only represented by 4 notes. This is a good compromise that reduces the full size to 1/3rd, and this works most of the time. A sample will not be stretched more than a semitone higher or lower.

The saved space and memory occasionally comes at the price ringing and other artifacts. Perhaps because an unwanted resonance becomes highlighted. These may probably be EQ’ed out. I noticed the difference when comparing both chromatic and minor-third sampled versions of my instruments. This might also explain similar artifacts on other samples I have used in the past.

Lossy compression

Audio compression can reduce disk space requirements by between 1/3 to 1/4 while maintaining a quality that is good enough for demonstration. However, even at high quality compression settings, the lossy compressed waveform (ogg) will not be 100% identical to the original waveform (wav). The difference is hard to detect by ear, but is still there and can be isolated and “seen” by adding the wav version and a phase inverted ogg version.  There’s a possibility that the waveform values at the loop points will be marginally more different, but then again it’s hard to detect by ear. A more practical drawback of compressed audio samples can be the added processing required for decompression.

5x more expression in the extended version

All the versions have some velocity dependent changes in attack time and filter cutoff, made possible by the SFZ specification. These emulate playing dynamics (e.g. piano, forte, etc.). The extended version has more.

Besides not having some disadvantages of the smaller and compressed versions, the extended version adopts an extra feature that is only possible because of the availability of chromatic samples (individual recordings for each semitone). For the lack of a better term, I call this “neighborhood borrowed velocity layers”. This creatively uses the change in timbre when a sample is stretched to a neighboring pitch, as explained earlier. A brighter sound and slight noise-like stretching artifacts would be heard at higher dynamic levels. Meanwhile, the timbre becomes duller at lower dynamics. Effectively, each note uses 5 samples to achieve 5 velocity layers without increasing the library size by 5 times!


I hope this explains the advantages and disadvantages of each version and help you decide which one to have. Also, many of the considerations mentioned here not are not limited to my instruments or the SFZ format. I’ll be glad if this humble article can be useful in your sample based music making. 🙂


Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s