This forum is now locked and moved (back) to Plogue's forum
Analogue Drums Tape Series 2 - SFZ mapping
  • pljonespljones
    Posts: 73
    OK... it's been a long time since I did an SFZ drum kit and this one has taken a long time (started, uh... October last year...).  This is a commercial kit, so you will need to go buy the samples!  (Go buy them! - might as well get the Tape Series Two bundle... ;))  The (completed) mapping is for the Gorilla kit.  [Mappings live here].
    [MP3 demo] (4.8Mb; there's a FLAC link on the page above; no complaining about my drumming... ;))

    Comments on the mapping are invited!
  • pljonespljones
    Posts: 73
    Right, this is probably a good point to provide an update.

    I've redone Gorilla a couple of times and it's schedule for a third re-write.  I was about a third the way through my initial analysis of the Kingpin samples earlier this week when I finally cracked.  Google found me a hint to how to get the information out of that Swiss Army Knife of sound analysis, SoX.  (Of course, now I can't find the forum post to give proper credit -- although it was an answer to a different question, it was enough to crack the problem.)

    Since then, I've got drafts of Kingpin and Monotown done using the new method (first cut -- going to try them out over the weekend to work out what needs adjusting and whether the method actually worked...).

    For background information: the samples in the Analogue Drums Tape Series 2 set (Gorilla, Kingpin, Monotown and Plastique) are supplied normalised with Kontakt mappings.  Normalised means the peak levels of the samples are all at 0dB (more or less) and the difference in levels when triggered is controlled by using velocity to reduce the amplitude.  Simple...

    Well, simple if you know how loud each loudest sample is in comparison to the others and equally how quiet each quietest sample is in comparison to the loudest, across the kits.  Which, of course, you can't tell from the samples, as they've been normalised.

    The saving grace was that they were recorded to tape and the noise floor is significantly high that normalising the quiet samples brings it up a lot.  You can see how much louder samples have been made by looking at the differences between the noise floors!  And that's exactly what I'd been doing.  Going through and looking.  Very, very time consuming.

    sox has a "stats" command that returns, amongst other things, the "trough" level of the input waveform.  After trying this out across a number of samples and finding it was getting similar numbers to what I was getting visually, I decided I'd happily rely on it for that part of the work -- that is, the part that had been taking months!!

    You can get SoX from SourceForge.  You can get the Windows script I wrote below.  Drop a directory onto the batch file and it will generate a file containing the samplename, peak level and trough level in comma-separate values format, ready for importing into a spreadsheet.

    I called the script "soxstats.bat":
    @echo off
    SetLocal EnableDelayedExpansion
    set sox="%ProgramFiles(x86)%\sox-14-4-1\sox"
    set stats=%CD%\%~n1-stats.csv

    echo Output statistics to "%stats%".

    echo sample,RMS Peak,RMS Trough > "%stats%"
    for %%X in (%1\*.wav) do (
        set dco=0
        set minl=0
        set maxl=0
        set pkl=0
        set rlev=0
        set rpk=0
        set rtr=0
        for /f "tokens=1-4,*" %%a in ('"!sox! "%%X" -n stats 2>&1"') do (
            if "%%a %%b"=="RMS Pk" set rpk=%%d
            if "%%a %%b"=="RMS Tr" set rtr=%%d
        echo %%~nxX,!rpk!,!rtr! >> "%stats%"

    goto end

    rem This passes all samples in a directory to sox to extract the RMS Peak and Trough values.
    rem The idea is then to use these in place of manual inspection to determine dynamic range
    rem and relative levels.


Howdy, Stranger!

It looks like you're new here. If you want to get involved, click one of these buttons!