Dim Pro sfz versus WAV

Author
Post
Essentials Only Full Version
DaveClark
Max Output Level: -71 dBFS
  • Total Posts : 956
  • Joined: 2006/10/21 17:02:58
  • Status: offline
2007/06/03 21:30:53 (permalink)

Dim Pro sfz versus WAV

Greetings all,

In attempting to create a new sfz instrument for Dimension Pro using my sample generator Pluckster, I found some strange behavior that I cannot explain, nor can I find any explanation in posts on this forum or in the sfz documentation.

Basically, a loaded WAV file sounds OK. An sfz that loads this same wave file at C4 sounds OK. An sfz that loads this same wave file at C5 has obvious distortion. I have exported a comparison WAV file for those interested.

Has anyone seen a similar problem? Any solution?

Thanks for your attention.

Regards,
Dave Clark

Attachment
=======================================================

1) First set of five notes:

Loaded WAV file "unison_5.wav" into E1

Played with note=60


2) Second set of five notes:

Loaded a file containing the following into E2:

<region> lokey=c4 hikey=c4 pitch_keycenter=c4 sample=unison_5.wav

Played with note=60


3) Third set of five notes:

Loaded a file containing the following into E3:

<region> lokey=c5 hikey=c5 pitch_keycenter=c5 sample=unison_5.wav

Played with note=72



WAV files:

http://mysite.verizon.net/dave_clark/unison_5.wav

http://mysite.verizon.net/dave_clark/instrument_difference.wav


sfz loaded sample with pitch_keycenter=c5 (what it should have for the root key) obviously does not sound correct. It's spectrum is also not what it should be (differs from file unison_5.wav). But loading it at c4 seems cause it to behave properly, except that it is now the wrong note (that is, one has to play the incorrect MIDI note to make it sound). Based on what I've read in the sfz documentation and in the manual, this distortion should not be occurring.


Original file is 48,000 samples/second, 24-bit; loaded into 96/24 SONAR 6 PE project; exported to 48/24 WAV file (instrument_difference.wav).

With Home Studio 4, repeated test with 48/24 all around --- audio sounds somewhat more similar, but spectra are still too different (still distorted). Looks like this *may* be a resampling problem; however, it's difficult to understand why reassigning a root key should cause it.

=======================================================
End.
#1

28 Replies Related Threads
    b rock
    Max Output Level: 0 dBFS
    • Total Posts : 8717
    • Joined: 2003/12/07 20:31:48
    • Location: Anytown (South of Miami), U.S.A.
    • Status: offline
    RE: Dim Pro sfz versus WAV 2007/06/04 08:15:35 (permalink)
    Correct me here if I'm missing some factors. I did listen to the examples and glance at the lines of .sfz, but only briefly. This caught my eye:
    pitch_keycenter=c5 (what it should have for the root key) obviously does not sound correct. It's spectrum is also not what it should be (differs from file unison_5.wav). But loading it at c4 seems cause it to behave properly, except that it is now the wrong note (that is, one has to play the incorrect MIDI note to make it sound).
    The .sfz format follows the IPN convention {see below}, and designates C4 as middle C. You'll see this on rgc:audio instruments and Roland keyboards, among others. With Yamaha, middle C will be C3 (notes from C-2 to G8), and Cakewalk hosts, for example, use C5. [You'll often see a Base Octave option in these hosts to compensate for different 'standards'.] Your controller keyboard or host may follow a different convention than the .sfz file, and leave you with an alternate set of expectations.

    In .sfz programming, I've personally abandoned octave conventions, and use MIDI note numbers directly instead. Both types are valid values. It cuts down on the confusion; both mine, and anyone analyzing the file structure. As an aside, if you want to set a lokey=, hikey=, and pitch_keycenter= to the same key, the key= input control will do it for you. In this case, key=60 (or key=c4).
    International-Pitch-Notation (IPN). A system of pitch designation agreed upon internationally under the auspices of the Acoustical Society of America. In this system the international standard A=440 Hz is designated A4 and middle C is C4. All pitches are designated with capital letters and their octave placement by an accompanying Arabic number changing with each octave C. Thus, the lowest C on the piano is C1 (the A below that is A0). The first F# below middle C is F#3, etc
    post edited by b rock - 2007/06/04 08:16:40
    #2
    DaveClark
    Max Output Level: -71 dBFS
    • Total Posts : 956
    • Joined: 2006/10/21 17:02:58
    • Status: offline
    RE: Dim Pro sfz versus WAV 2007/06/04 15:52:37 (permalink)
    Hi b rock,

    Thanks for your reply. I appreciate the advice re: note conventions. I was going to address that later.

    I am concerned that bad or mediocre sample conversions are being done (for example, pitch up then pitch back down) and that these may be the cause of the distortion, and possibly for the reason you say about specifying notes. Still, and regardless of this possibility, one should NOT hear this distortion. To me it sounds like a kind of squeak at the beginning of the note. That is, the attack is significantly altered as an sfz, but not as a WAV. The spectra themselves are significantly different. I hope that this is a temporary problem because my reason for purchasing Dim Pro was to use for building sampler instruments via sfz.

    Thanks again.

    Regards,
    Dave Clark

    #3
    lawapa
    Max Output Level: -68 dBFS
    • Total Posts : 1144
    • Joined: 2005/01/09 19:14:51
    • Status: offline
    RE: Dim Pro SFZ versus WAV 2007/06/04 19:55:45 (permalink)
    In situations similar to what you've encountered I will guess the root. If the file plays back and sounds off/different to the original file I'll change the root designation until the original file and the SFZ load sound identical in pitch/timbre.

    Complex sounds with prominent high frequency content can be difficult to root pitch.

    But I've found if the SFZ set plays it back unchanged your good. Of course you can force the root designation.

    If your unsure what pitch C3/C4/C5 test the sound with a Dpro/Rapture patch and verify the set pitch.
    #4
    DaveClark
    Max Output Level: -71 dBFS
    • Total Posts : 956
    • Joined: 2006/10/21 17:02:58
    • Status: offline
    RE: Dim Pro SFZ versus WAV 2007/06/04 21:19:40 (permalink)
    Hi lawapa,

    Thanks for your suggestion.

    This suggestion does not apply to my situation, however. I am creating the fundamental and the harmonics, so I know exactly what the fundamental frequency is.

    I am building a complete instrument (six octaves, 72 samples), so I need to root each sample at the correct location, and I already know exactly what that location is for each sample.

    Regards,
    Dave Clark

    #5
    lawapa
    Max Output Level: -68 dBFS
    • Total Posts : 1144
    • Joined: 2005/01/09 19:14:51
    • Status: offline
    RE: Dim Pro SFZ versus WAV 2007/06/04 23:21:05 (permalink)
    I am creating the fundamental and the harmonics, so I know exactly what the fundamental frequency is.


    If the harmonics/fundamental are separate wave files you can root the Harmonics higher then use transpose down to use within the specific key range you set.

    With oscillator_on you could use the wavetables for the fundamental. You mentioned you were using wave files so I just assumed you were not using the wavetables. I bet you could use wavetables for the harmonics as well. If your constructing with wave files you could extract a 3000 sample file and make a custom wavetable.
    #6
    DaveClark
    Max Output Level: -71 dBFS
    • Total Posts : 956
    • Joined: 2006/10/21 17:02:58
    • Status: offline
    RE: Dim Pro SFZ versus WAV 2007/06/05 01:45:40 (permalink)
    Hi lawapa,

    In this case, the fundamental and harmonics are not separate WAV files. The WAV file resulted from a modelling program I wrote that solves the wave equation in 1-D, modelling a plucked string. The "harmonics" I spoke of are the eigenfunctions of a vibrating string problem which I mention in case you are familiar with that problem in physics and math. In my program, I specify the fundamental frequency (which results from the properties of the string) and the "harmonics." On edit: In an "instrument," I construct 72 separate strings and solve the wave equation. The files I provided here are only one such string to show what the problem is.

    I am playing this WAV file "as is" because it is exactly what I want to play, not to create a wavetable or anything like that. The problem is that distortion is introduced when this WAV file is loaded by an sfz file operand, but not when it is loaded manually. There really should not be any difference, but I suspect that there is some undocumented or partially documented additional manipulation by the program when it loads samples via sfz operands compared to loading manually.

    I could not find anything in the operand table in the sfz documentation which would obviously affect samples loaded by sfz operands (i.e. sample=<filename>) such as that an envelope with certain parameters is impressed on the sample, yet obviously something is done to such samples --- in this case, something bad.

    Regards,
    Dave Clark

    post edited by DaveClark - 2007/06/05 01:51:36
    #7
    lawapa
    Max Output Level: -68 dBFS
    • Total Posts : 1144
    • Joined: 2005/01/09 19:14:51
    • Status: offline
    RE: Dim Pro SFZ versus WAV 2007/06/05 22:27:55 (permalink)
    If the level of the wavefile is 0 dB use the volume= -1 or greater opcode to balence the amplitude. When you render your files Lower the volume/amplitude such that it's peaks below 0 dB.
    #8
    DaveClark
    Max Output Level: -71 dBFS
    • Total Posts : 956
    • Joined: 2006/10/21 17:02:58
    • Status: offline
    RE: Dim Pro sfz versus WAV 2007/06/11 02:39:34 (permalink)
    Greetings All,

    In looking at this problem again after submitting a problem report to Cakewalk (and hearing nothing back so far, which is the typical behavior that I see....), I was very disappointed to find that programs in Dimension Pro itself also have this squeaking attack problem for higher notes. Can anyone else hear them?

    Thanks for any confirmations.

    Regards to all,
    Dave Clark


    ---------------------------------

    The following procedure produces "squeaks" with DimPro playing its own program(s) which load sfz's, running at 96/24 on my machine:

    1) Fire Up SONAR 6 PE at 96/24

    2) Create a New Project ("DimPro_Squeak" a Normal Project)

    3) Synth Rack, Soft Synths, DimPro

    4) Click Empty Program slot, 11 - Pianos, Grand Piano 1v 4th.prog[56 MB]

    5) Play C6, Note 96: Squeaks! (So do others.)

    The file at:

    http://mysite.verizon.net/dave_clark/ZWPod_DimPro_C6-Squeak.wav

    demonstrates the squeaking sounds I've been discussing. It was produced with Dimension Pro in Zynewave Podium, also running at 96/24, then downsampled to 48/24 after export from ZWPod.

    In SONAR 6 PE, Options/Audio/Sampling Rate is set at 96,000. ASIO drivers for Terratec EWX-2496 sound card. Machine is self-built 2.6 GHz P4/800 with 1 GB RAM, 1 SATA 160 GB, 2 PATA (2 x 250 GB). Windows XP Home. Pretty generic for these days and running very well for 3.5 years now....

    C6 played with TruePianos (Demo) Diamond, any style, does not exhibit this problem. (I still think TruePianos sounds great in mixes BTW.)


    #9
    René
    Max Output Level: -68 dBFS
    • Total Posts : 1103
    • Joined: 2004/01/06 13:15:57
    • Status: offline
    RE: Dim Pro sfz versus WAV 2007/06/13 10:12:53 (permalink)
    I have tried this, and can't reproduce. Actually, 90% of Dimension Pro contents is using sfz files as the one you described.
    Does it happen at any samplerate for you? Or only at 96kHz?
    #10
    DaveClark
    Max Output Level: -71 dBFS
    • Total Posts : 956
    • Joined: 2006/10/21 17:02:58
    • Status: offline
    RE: Dim Pro sfz versus WAV 2007/06/13 15:40:29 (permalink)
    Hi René,

    Thanks for your response. I appreciate your taking time to look at this problem.

    I have actually posted two separate procedures, one which loads a 48/24 WAV file that I created with a sample generator I wrote (1-D wave solver) into a 96/24 project; the other one was to merely play the C6 piano note in a 96/24 project. I'm unsure of what it is that you could not duplicate. I suspect that "tried this" refers only to the second one (piano note) because you asked if I had had problems with other sample rates, a question that I had already addressed in post #1:

    With Home Studio 4, repeated test with 48/24 all around --- audio sounds somewhat more similar, but spectra are still too different (still distorted). Looks like this *may* be a resampling problem; however, it's difficult to understand why reassigning a root key should cause it.


    In other words, using my own sample, I see less of a problem at 48 Ksamples/sec (maybe inaudible to some people), but it's still obviously there in a 48/24 project under the conditions I mentioned in post #1.

    With the piano problem, I just tried 44.1 Ksamples/sec and hear no problem. When I close down SONAR 6 PE, restart, change the Audio to 96 Ksamples/sec, reload the project I just created, I definitely hear it again.

    Another user reported to me via PM that he/she also had this problem and changed to 44.1 Ksamples/sec to avoid it, so I'm not the only one who is having this particular problem.

    On sfz's being used: Yes, I am well aware that sfz's are used a lot. (The filenames are prominently displayed.) This seems to be a problem with higher notes, with root keys other than C4, magnified by resampling. With many of the programs and sfz's I'm not sure that it would be noticeable unless one knew what to listen for. With the unison_5.wav sample I provided, it is really obvious that there is some problem, --- somewhere that is. Many SONAR users here do not use 96 Ksamples/sec; most seem to be at 48 or 44.1, so they may never hear this problem.

    I would be very interested in whether or not you can duplicate the problem in my post #1. If not, then something really strange is going on here and at least one other machine out there.

    My workaround may be to use 96/24 samples from my solvers, if that works with root keys other than C4.... I hope to get to that later today and post the results.

    Thanks again and hope to hear back from you.

    Regards,
    Dave Clark




    #11
    René
    Max Output Level: -68 dBFS
    • Total Posts : 1103
    • Joined: 2004/01/06 13:15:57
    • Status: offline
    RE: Dim Pro sfz versus WAV 2007/06/13 18:44:11 (permalink)
    I have actually posted two separate procedures, one which loads a 48/24 WAV file that I created with a sample generator I wrote (1-D wave solver) into a 96/24 project; the other one was to merely play the C6 piano note in a 96/24 project. I'm unsure of what it is that you could not duplicate. I suspect that "tried this" refers only to the second one (piano note) because you asked if I had had problems with other sample rates, a question that I had already addressed in post #1:


    Sorry, I didn't read your report of the SR for the test, only the SR of the samples.


    I would be very interested in whether or not you can duplicate the problem in my post #1. If not, then something really strange is going on here and at least one other machine out there.


    I could not. That actually was what I reported in my previous post.
    I suggest you to contact Technical Support with this information, to see if they can reproduce the issue. Sorry I couldn't help.
    #12
    DaveClark
    Max Output Level: -71 dBFS
    • Total Posts : 956
    • Joined: 2006/10/21 17:02:58
    • Status: offline
    RE: Dim Pro sfz versus WAV 2007/06/14 16:06:20 (permalink)
    Hi René,

    Thanks for the clarification.

    As I mentioned in post #9 above, I had indeed contacted Technical Support. At the time of that posting, I had heard nothing back. On 12 June, I received this:

    Hello David,

    Thank you for contacting Cakewalk Technical Support. I apologize for the delay. The sound you're hearing is the tone going through the ADSR envelope. Try adjusting the attack curve under the AMP section of the Modulators in Dimension Pro.

    I hope that helps.


    While the C6 piano note does go through the Amp section, my unison_5.wav (loaded as WAV or by sfz file) does not, so there is obviously a disconnect about what this problem is. I tried to explain that adjusting the ADSR envelope won't solve this problem and am waiting for a reply. Meanwhile, I'll go back to using my own audio programs and troubleshooting this problem myself.

    Thanks again for taking the time to look at it and post.

    Regards,
    Dave Clark

    #13
    DaveClark
    Max Output Level: -71 dBFS
    • Total Posts : 956
    • Joined: 2006/10/21 17:02:58
    • Status: offline
    RE: Dim Pro sfz versus WAV 2007/06/17 22:16:37 (permalink)
    Greetings all,

    I finally got around to looking at this problem in a little more detail.

    1) Created DimPro track.

    2) Created a 96/24 version of unison_5.wav and loaded that into DimPro as WAV and as SFZ.

    3) Drew in C5 (#60) and C6 (#72) notes into the MIDI track at velocity = 100.

    4) Mixdown WAV with C5 note; mixdown SFZ (loaded WAV) with C6 note.

    5) Compared and found differences in both time and frequency domain (listening and FFT).

    6) Saved cwb (*Right-click* and SAVE LINK AS if you want this):

    http://mysite.verizon.net/dave_clark/DimPro2.cwb

    I hope that the new unison_5.wav loaded into DimPro was saved here --- appears to be about the right size.



    Please note that this was created purely with software, eliminating numerous possible causes.

    If you are confused by my use of C4, C5, and C6, please see post by b rock above. These notes should sound identical as I wrote them in but the C6 note still causes an incorrect attack and distortion. Although it these may sound the same to some people, the spectra verify visibly that there is a difference.

    Regards to all,
    Dave Clark

    #14
    DaveClark
    Max Output Level: -71 dBFS
    • Total Posts : 956
    • Joined: 2006/10/21 17:02:58
    • Status: offline
    RE: Dim Pro sfz versus WAV 2007/06/20 16:51:40 (permalink)
    Greetings to all once again,

    Small update: DXi and VSTi Pro and DXi LE all exhibit this "squeak" problem.

    Regards,
    Dave Clark

    #15
    DaveClark
    Max Output Level: -71 dBFS
    • Total Posts : 956
    • Joined: 2006/10/21 17:02:58
    • Status: offline
    RE: Dim Pro sfz versus WAV 2007/06/21 17:10:22 (permalink)
    Greetings all,

    On another machine, DimPro 1.1 (not updated yet) does NOT exhibit this problem in Zynewave Podium using unison_5.wav and the sfz. The speakers are different, so it's hard to tell for sure, but it seems to be fine. Next is the 1.2 update.

    Regards to all,
    Dave Clark

    #16
    DaveClark
    Max Output Level: -71 dBFS
    • Total Posts : 956
    • Joined: 2006/10/21 17:02:58
    • Status: offline
    RE: Dim Pro sfz versus WAV 2007/06/22 20:38:08 (permalink)
    Greetings once again,

    On the second machine, now with DimPro upgraded to version 1.2, the problem IS audible. Therefore, I have managed to duplicate this problem on a very different machine. Either the problem was introduced with version 1.2, or it got worse with version 1.2.

    The only thing in common with these two machines is Windows XP (although one has Home and the other Media or whatever it's called) and DimPro 1.2. The sound system is different, the processor and motherboard are different, the speakers are different, and so on. As a reminder, I have also duplicated this problem with two different VSTi hosts: SONAR 6 PE and Zynewave Podium.

    Also, I received an email from Tech Support suggesting that I try the Cubase fix version. Unfortunately this version also has the same problem.

    I must say that I suspect that people who cannot duplicate this problem may simply not be able to hear it as well as I can. It is a little hard to hear for unison_5.wav loaded as an SFZ, but when many notes are played when more WAV's are loaded by an SFZ, it becomes really obvious.

    Regards to all,
    Dave Clark

    #17
    DaveClark
    Max Output Level: -71 dBFS
    • Total Posts : 956
    • Joined: 2006/10/21 17:02:58
    • Status: offline
    RE: Dim Pro sfz versus WAV 2007/07/05 19:40:23 (permalink)
    Greetings all,

    I've now verified that Session Drummer 2 has the same behavior as DimPro 1.2 on my main machine, so now it's SONAR and Zynewave Podium, two machines, two programs that I believe are very closely related (SD2 and DimPro). This isn't looking good.... It's not the sound card, it's not the speakers, apparently neither something easy nor something straightforward.

    To review, the problem is that a file unison_5.wav plays OK loaded as a WAV, but has a distorted attack portion when loaded via an SFZ file and re-assigned a different key pitch_keycenter than middle C. It's not necessarily easy to hear, but it is most definitely there.

    Here is a sequence of notes from Zynewave Podium played through SD2 using WAV and SFZ in a pattern:

    http://mysite.verizon.net/dave_clark/SD2_Bounce_4416.wav

    The pattern is not strictly alternating; if you cannot hear the difference in the attack portions of the notes in this sequence, then you may have the problem I do with DimPro/SD2 but not know it.


    Regards to all,
    Dave Clark

    On edit: This particular file started life as a 48/24, then became a 96/24, then I resampled it with my own resampler program to 44/16 to fit better on my web pages. It sounds the same regardless of format (unfortunately). The purpose of posting it is to ensure that those who say they cannot duplicate it actually understand what I'm claiming about the attack portion and can actually hear it.

    post edited by DaveClark - 2007/07/05 20:06:55
    #18
    DaveClark
    Max Output Level: -71 dBFS
    • Total Posts : 956
    • Joined: 2006/10/21 17:02:58
    • Status: offline
    RE: Dim Pro sfz versus WAV 2007/07/12 21:45:59 (permalink)
    Greetings all,

    ===========================================================
    On edit: This was actually version 1.10 versus 1.2 not 1.0 versus 1.2 Sorry for
    any confusion!! I took the liberty of editing the URL as well as editing below.
    I won't change the window caption in the JPEG unless it really bothers someone.
    ===========================================================

    I've prepared some spectra for those interested:


    Dimension Pro version 1.10 SFZ loading versus WAV loading, first 6000 samples of Unison_5.wav:

    http://mysite.verizon.net/dave_clark/DimPro_1.10_Bounce_Unison_Spectrum.jpg

    Same.


    Dimension Pro version 1.2 SFZ loading versus WAV loading, first 6000 samples of Unison5.wav:

    http://mysite.verizon.net/dave_clark/DimPro_1.2_Bounce_Unison_Spectrum.jpg

    Definitely different.


    These spectra are from what I have called the second machine, a garden-variety Dell Dimension E510 (that is, not my audio workstation where I first discovered this problem).

    The waveforms analyzed were rendered with Zynewave Podium with the Bounce facility using one drawn MIDI note, so it was all software. This is the same type of result I obtained on my audio machine using SONAR 6 PE, but this was a very careful apples-to-apples comparison of the SFZ versus WAV loading of Unison_5.wav, with two different versions of Dimension Pro, illustrating that something went wrong with 1.2 on two completely different machines, different VSTi hosts, different audio systems, and so on.

    Because this problem is not particularly easy to hear, I felt somewhat obligated to show that it definitely exists and that it exists on more than one machine. If you suspect that the attack of some instruments doesn't sound quite correct for some notes, this SFZ loading problem may be the reason.

    Regards to all,
    Dave Clark

    post edited by DaveClark - 2007/07/13 01:02:06
    #19
    rabeach
    Max Output Level: -48 dBFS
    • Total Posts : 2703
    • Joined: 2004/01/26 14:56:13
    • Status: offline
    RE: Dim Pro sfz versus WAV 2007/07/12 23:57:32 (permalink)
    assuming i understood your procedure correctly i have one suggestion; profile the frequency spectrum again without upsampling and then downsampling the data. if you get the same results when i get time i will look into it also.
    post edited by rabeach - 2007/07/13 00:07:56
    #20
    DaveClark
    Max Output Level: -71 dBFS
    • Total Posts : 956
    • Joined: 2006/10/21 17:02:58
    • Status: offline
    RE: Dim Pro sfz versus WAV 2007/07/13 00:46:57 (permalink)
    Hi rabeach,

    Thanks for looking at this, posting, and offering to compare.

    For the case of comparing frequency spectra, no resampling was done. It was all 48000 Samples/sec from start to finish. It started as 24-bit, was converted to single-precision floating point in Zynewave Podium and remained at 32-bit floating point through bouncing, export, and analysis.

    Regards,
    Dave Clark

    P.S. I'll go edit my previous post. For some reason, I got it my head that this was DP 1.0; not true, it was 1.10.



    post edited by DaveClark - 2007/07/13 02:26:29
    #21
    DaveClark
    Max Output Level: -71 dBFS
    • Total Posts : 956
    • Joined: 2006/10/21 17:02:58
    • Status: offline
    RE: Dim Pro sfz versus WAV 2007/07/14 19:06:09 (permalink)
    Greetings all,

    Because rabeach commented on upsampling and downsampling, I thought I might extend the discussion a bit and show what happens to the "attack portion" (first <8000 samples at 96/24) of Unison_5.wav in SONAR and DPRO when Unison_5.wav is resampled.

    Because I use it below: First to establish that my own resampler does not modify Unison_5.wav beyond the limitations of precision, I upsampled then downsampled using my own resampler. The following plot is at 48/24 (Nyquist 24,000 Hz) comparing the original 48/24 to the up-down 48/24:

    http://mysite.verizon.net/dave_clark/Unison_5_Sampster_up_down.jpg

    The original is cyan; the up-down one is magenta. Note that at -40 db, there is no cyan. At -60 db, there are some cyan dots; at -80 db, more dots and some short dashes; at -100 db, some longer dashes. This is due to the limitations of machine computation rather than the resampler.


    Second to establish that SONAR introduces slight distortion when upsampling, I compare importing of already resampled Unison_5.wav and SONAR-resampled Unison_5.wav. The rest of the plots are at 96/24:

    http://mysite.verizon.net/dave_clark/unison_5_Sampster_vs_SONAR.jpg

    There is a cutoff filter and some additional harmonics.


    Now to show that Dimension Pro and SONAR together introduce significant distortion, even with the better case of WAV loading, I compare already-resampled to DP/SONAR-resampled:

    http://mysite.verizon.net/dave_clark/Unison_5_DPro_WAV.jpg
    http://mysite.verizon.net/dave_clark/Unison_5_DPro_WAV_logF.jpg


    And that the SFZ loading introduces even more distortion:

    http://mysite.verizon.net/dave_clark/Unison_5_DPro_SFZ.jpg
    http://mysite.verizon.net/dave_clark/Unison_5_DPro_SFZ_logF.jpg


    Regards,
    Dave Clark

    post edited by DaveClark - 2007/07/14 19:17:10
    #22
    rabeach
    Max Output Level: -48 dBFS
    • Total Posts : 2703
    • Joined: 2004/01/26 14:56:13
    • Status: offline
    RE: Dim Pro sfz versus WAV 2007/07/15 13:46:16 (permalink)
    by the way what program are you using to resample.

    maintaining consistency in frequency ranges with respect to the snapshots will allow for a more accurate comparative analysis.

    if you are performing noninteger resampling generally the interpolator and the decimator will be band limited with a lowpass filter. if you are performing integer resampling generally the interpolator or the decimator will be band limited with a lowpass filter.

    the spectrum in cyan appears to show frequencies beyond the nyquist in some comparisons.

    when interpolation and/or decimation are involved in noninteger/integer resampling approximations are being made no matter what device or algorithm is being used.
    post edited by rabeach - 2007/07/15 16:12:43
    #23
    DaveClark
    Max Output Level: -71 dBFS
    • Total Posts : 956
    • Joined: 2006/10/21 17:02:58
    • Status: offline
    RE: Dim Pro sfz versus WAV 2007/07/15 16:21:40 (permalink)
    Hi rabeach,

    Thanks for taking an interest in this and for your reply.

    ORIGINAL: rabeach

    by the way what program are you using to resample.


    Sorry if I was unclear:

    I used my own FFT-overlap resampler to produce a 96/24 version of the 48/24 original to stand in for the original which I loaded into SONAR "as is" at 96/24, then wrote back out at 96/32FP. To eliminate any question about the capability of my resampler, I produced the first plot to compare the original 48/24 to the up-down 48/24 created by my own resampler.

    All other resampling was done by SONAR and by SONAR + DPro, so you would have to ask Cakewalk what they used. Any additional problems beyond what you see in the first plot were introduced by SONAR or SONAR + DPro.


    maintaining consistency in frequency ranges with respect to the snapshots will allow for a more accurate comparative analysis.


    ??? I did exactly that for the plots that matter. I can go back and forth between the linear freq plots, and the only things that move are the data.


    if you are performing noninteger resampling generally the interpolator and the decimator will be band limited with a lowpass filter. if you are performing integer resampling generally the interpolator or the decimator will be band limited with a lowpass filter.


    Again, the resampling was done by SONAR and SONAR + DPro. Yes, certain types of resampling demand filtering, windowing, etc. because they create (essentially) noise. Again, you would have to ask Cakewalk what they are doing. It doesn't look good to me, but perhaps they have an explanation.


    the spectrum in cyan appears to show frequencies beyond the nyquist in some comparisons.


    It is definitely not in the 96/24 file that I created to stand in for the original because I write zeros for anything beyond the original Nyquist frequency when upsampling. It is either a product of SONAR loading "as is" and writing out again or an artifact of analysis and plotting. I hope it's that latter because I don't want SONAR to do anything (especially filtering) unless I explicitly tell it to.


    when interpolation and/or decimation are involved in noninteger/integer resampling many approximations are being made no matter what device or algorithm is being used.


    This doesn't explain why the SFZ loading method suddenly caused DPro to introduce even more distortion in version 1.2 than it did in 1.1. That is the problem under discussion. Resampling makes it worse.


    Regards,
    Dave Clark


    #24
    stratcat33511
    Max Output Level: -39 dBFS
    • Total Posts : 3636
    • Joined: 2004/12/27 09:48:37
    • Location: Tampa FL
    • Status: offline
    RE: Dim Pro sfz versus WAV 2007/12/06 13:24:32 (permalink)
    anything derived from this ?

    http://www.myspace.com/edmclaughlin
    http://cdbaby.com/cd/exilecollection3
    http://www.exilecollection.com
    Exile Collection available on iTunes
    #25
    DaveClark
    Max Output Level: -71 dBFS
    • Total Posts : 956
    • Joined: 2006/10/21 17:02:58
    • Status: offline
    RE: Dim Pro sfz versus WAV 2007/12/06 14:09:12 (permalink)
    Hi Ed,

    Nada. Zip. Zero.

    Hope all is going well --- uh --- other than this problem!

    Best regards,
    Dave Clark

    #26
    :10:
    Max Output Level: -38.5 dBFS
    • Total Posts : 3678
    • Joined: 2004/10/24 21:31:38
    • Status: offline
    RE: Dim Pro sfz versus WAV 2007/12/10 10:36:21 (permalink)
    I think I just figured the problem out....

    I used my own FFT-overlap resampler to produce a 96/24 version of the 48/24 original


    .....lol

    long day at work.... Im bored.

      
    #27
    stratcat33511
    Max Output Level: -39 dBFS
    • Total Posts : 3636
    • Joined: 2004/12/27 09:48:37
    • Location: Tampa FL
    • Status: offline
    RE: Dim Pro sfz versus WAV 2007/12/14 16:55:20 (permalink)

    ORIGINAL: DaveClark

    Hi Ed,

    Nada. Zip. Zero.

    Hope all is going well --- uh --- other than this problem!

    Best regards,
    Dave Clark




    Hi Dave

    Not too bad
    My DAW is on the fritz
    building a new one over teh holidays
    busy at work
    busy on a few projects as well
    Have a happy and a merry if I don't talk to you before then
    E

    http://www.myspace.com/edmclaughlin
    http://cdbaby.com/cd/exilecollection3
    http://www.exilecollection.com
    Exile Collection available on iTunes
    #28
    kk6pr
    Max Output Level: -90 dBFS
    • Total Posts : 18
    • Joined: 2004/09/09 05:10:06
    • Location: Bay Area, CA
    • Status: offline
    RE: Dim Pro sfz versus WAV 2007/12/14 19:40:39 (permalink)
    As per another thread, it sounds like he's checked himself into rehab...

    http://forum.cakewalk.com/tm.asp?m=1111166&mpage=3


    P.S. I'm probably going to have to ignore the Cakewalk forums for awhile, so please do accept my apologies for any lack of response in the near future. I'll try to at least remember to check my PM box.
    #29
    Jump to:
    © 2024 APG vNext Commercial Version 5.1