Yaking Cat Music Studios
Synclavier Hacking
formerly named "Hardcore Testing"
IMAGE GOES HERE
updated 12/12/99
Corrected information

*Author assumes no responsibility for the information contained herein*


(Now with Screen Shots!) Here is some *SERIOUS * information about the Synclavier.  You have always HEARD that it has the best sound....  How would you like to really know if it were true or not?

Would you like to see some of the audio *LIMITATIONS* I have found?

Various gain differences between input and output.
Noise.
Frequency dips in the high end....

What about *undocumented* features that exist?

Check it out!

Brandon
The Dust Kicker

go back to "Technical Information" menu
....go back to main page


 
  12/12/99 - revised (Malte Rogacki saves the day!!!)

The reason the Synclavier sounds so good - the voice cards were built "improperly".  In fact, they add their own "color" to the sound due to their "engineering flaw"  Here it is in it's entirety from XiChron's web site.  BTW, although this information has been removed from their web site, Malte just happened to have a copy of the HTML code and sent it to me.  What a guy!

----Many people say that the Synclavier was (and still is) the finest sounding sampling synthesizer. Part of this had no doubt to do with the fact that the 12 bit envelope data was applied to the reference voltage of the DACs instead of affecting the digital data before the DAC. This in effect created a kind of scaled 16 x 12 = 28 bit precision. Obviously this is in no way a 28-bit accurate DAC, and an audio purist will correctly point out that heating of the reference ladder as the envelope changes will cause nonlinearities, etc., etc., etc.. Nevertheless, in this particular situation it worked and sounded great. As Sydney Alonso used to say, "thank god I was so dumb".----

Apogee Electronics had a special card for the Direct to Disk.  It actually LOWERED the noise floor and made the sound more noticeably "rounder".  Now, my DtoD sounds INCREDIBLE.  I cannot imagine it even quieter - WHAT NOISE?!

  8/19/99 More undocumented commands
REV (name) - reverse compile a sequence
REC (1-8) - recall sequence from "Sequence button"
STO (1-8) - save sequence to "Sequence button"
  6/01/99  NED StartUp/Termulator Journaling

To log the output of your RTP screen to a file:

(command)+(option)+(forward delete) = Start Journaling
(command)+(option)+(end) = Stop Journaling

  6/01/99  DISK PATCHING (you can also WRITE in the same manner, but I ain't going to tell you how!!!)

This is an undocumented command that allows a user to see the contents of their hard drive, sector by sector.

DDU, (sector offset of current catalog), beginning Word, ending Word

Here it is broken down:
DDU - disk dump, reads the disk
(sector offset of current catalog) - the beginning sector you wish to read of the current catalog
beginning Word - start viewing at, not Hex, not Base 10, but Octal
ending Word - end viewing at (non-inclusive), not Hex, not Base 10, but Octal

Just get out that Scientific Calculator that has been gathering dust to do the conversions from Base 10 to Octal

  5/30/99  AUDIO TESTING - REV 2 (now includes screen shots at end)

I have always noticed a frequency "dip" in the upper frequencies in my 3200.  The DtoD has never exhibited this problem.  The 3200 has DDV S3.5 cards which lack the panning circuitry in the DDV S5 cards (as in my DtoD).  I have digitally lifted audio off an audio CD in the CD-ROM drive in my Mac and used S/Link to transfer the audio to my Synclavier and DtoD for accurate playback.  The DtoD *always* sounded better.  This is partially due to the lack of Multichannel Distributor and also there are only 4 tracks in my DtoD.  Less voices mean a "hotter" signal since less headroom is needed.

This is how to perform an unbiased audio test of the Synclavier's voice cards.  The actual test results are at the end of this page.  Feel free to submit your own.  Just use (command)+(shift)+(the number three) to save a screen shot.  "Attach" the 'PICT' file to an Email which includes your system configuration (PSMT vs. 9600, how many outputs, etc.).  I will post them on my page.

In SFM, create a few seconds of random noise.  We'll create more time than we need for measurement in order to allow time in the analog sample for transient responses to decay into a steady state before we start our spectral analysis.  To do this, type the following:

CRE 3   /* This creates a three second sound file. */
ADD RAN  /* This fills the file with pseudo-random numbers. */
SAV DIGITAL

Press the F2 key.  This will run the RTP and place the sample "DIGITAL" on the keyboard.  Connect the output under consideration directly to the STM input using a single cable and sample the signal.  Before saving the new sample, EXTRACT the three seconds containing signal using a crossfade of 0.  Save the sample as "ANALOG".  Press the F4 key to return to SFM and type the following:

SET FFT 9000  /* This sets the FFT length to the maximum value of 8192 samples */
SET LEN 1  /* This set the window length to the maximum value of .16384 seconds (the equivalent at 50K of the FFT length) */
SET OFF .08192  /* This sets the offset time to half of the window length.  We do this because in order to cancel out random deviations; we will average a series of running spectra.  This value causes the window of each spectrum to peak where the window of the previous spectrum reached zero. */

You are now ready to perform a spectral analysis of these files.  If you want to be able to compare the original digital sample with the analog sample, then obviously you'll need to analyze both.  If you're in a hurry and just want to see the analog frequency response, trusting in the theoretical flatness of the digital sample, then by all means skip the digital sample.  To perform the analysis, type:

OLD DIGITAL  /* or OLD ANALOG */
SPE 1 to 2  /* this should result in 11 spectra.  You might as well go get a snack unless you have the PowerPC hardware. */

When prompted with "Press RETURN for averaged spectrum" do so and type:

SAV DIG  /* or SAV ANA */

This saves your spectrum so that you will be able to do quick comparisons by typing REC ANA and REC DIG.  You can zoom in to a frequency region using SET ORI (origin in KHz) and SET RAN (range in KHz).

  5/30/99  RESULTS - REV 2

For my 3200 with DDV S3.5 cards the test in SFM revealed:

DIGITAL
@ 24.4141 Hz  the signal level was -39.10 db
ANALOG
@ 24.4141 Hz  the signal level was -49.54 db

= difference of 10.44 db.  Due to built-in padding in the Sample Inputs.

DIGITAL
@ 20007.3222 Hz  the signal level was -38.54 db
ANALOG
@ 20007.3222 Hz  the signal level was -54.21 db

= difference of 54.21 - 38.54 - 10.44 (input deviation) =
5.23 db signal drop!!! (see the screen shots below)

This happens in a linear fashion starting @ 7 KHz

This also happens on a 9600 with DDV S4 cards.
This does not happen on a PSMT with PSV voice cards.

Conclusion:

The older PSV cards, although noisier, have a much more accurate sound when played through a Multichannel Distributor.  Also, the samples tend to end in a "snap" with DDV cards.

However, the DDV voice cards in a DtoD give extremely "hot" and accurate sound playback.

One of the changes NED made was to remove the FM voices from the Multichannel Distributor and have them route to a separate stereo output.  This reduces the amount of noise that gets mixed in with the sampling voices.

Also, there are different versions of the Multichannel Distributor.  This is considered an upgrade in sound quality (as was the upgrade from PSV to DDV voice cards).  However, many users prefer the PSV voices...


 

Home


brandon@yakingcat.com