CD/DAT with pre-emphasis: how to de-emphasize correctly?

Discussion in 'Audio Hardware' started by SolarWind, Aug 16, 2006.

Thread Status:
Not open for further replies.
  1. SolarWind

    SolarWind New Member Thread Starter

    Location:
    Amsterdam, Holland
    I know, this has been discussed a couple of times, since many rare and sought after CDs (Toshiba's Black Triangle, like "Abbey Road", early Japanese CD pressings, like "The Wall", etc.) are recorded with pre-emphasis. Still, the results, what exactly is the correct de-emphasis EQ curve to use for ripped waveforms (unless you are using a Mac where some progs are reported to de-emph “on the fly”), were either not satisfactory when listened to and/or not quite clear.

    To find out the “ultimate truth”, I measured the de-emphasis curve on the analogue output of the Sony NS999ES SACD/CD player with a self-generated reference White Noise CD burnt with pre-emphasis flag set. No matter what, here is how Sony does de-emphasis "in ES hardware" (see attached image).

    Well, it looked quite different from the other (almost linear frequency cut-off) "de-emphasis definitions " found on the Internet.

    Digging on that deeper I came across a couple of interesting threads,
    http://www.solorb.com/dat-heads/digests/V5.400/D469#Msg1
    http://www.solorb.com/dat-heads/digests/V4.100/D191#Msg3
    at last explaining CD/DAT de-emphasis (for the first time I saw) correctly.

    Reportedly, the "Sound Recording Handbook" by John M. Woram
    Publisher: Howard W. Sams & Co; 1st ed edition (August 1989)
    ISBN: 0672225832 (http://www.amazon.com/gp/product/0672225832/)
    explains in detail how the CD/DAT pre-emphasis/de-emphasis works.

    In a nutshell: to generate the correct target 15/50µs EIAJ de-emphasis (or pre-emphasis) curve for CD/DAT use the following equation:

    DE(f) = 10*log(A/B)-10.4576
    PRE(f) = 10*log(B/A)-0.9684

    DE(f) - de-emphasis output in dB at a frequency of f Hz
    PRE(f) - pre-emphasis output in dB a frequency of f Hz

    Where:
    f = frequency in Hz
    A = 1 + 1/(H x H)
    B = 1 + 1/(L x L)
    H = (2 x pi x f x t(H))
    L = (2 x pi x f x t(L))
    t(H) = high freq. time-const (15 µs = 0.000015)
    t(L) = low freq. time-const (50 µs = 0.000050)
    pi = 3.1416

    An Excel calculator is available at http://www.radonmaster.de/robernd/demp_tab.xls

    Here are some EQ reference points I generated and used for Toshiba Black Triangle "Abbey Road" or Japanse Pink Floyd "The Wall" (generated wavs - for best results: 24bit/96kHz - sound just right and just beautiful! At last!)

    Code:
       Freq.   De-Emphasis (CD/DAT)
      510 Hz  -0.1 dB
      726 Hz  -0.2 dB
      896 Hz  -0.3 dB
     1042 Hz  -0.4 dB
     1173 Hz  -0.5 dB
     1294 Hz  -0.6 dB
     1408 Hz  -0.7 dB
     1516 Hz  -0.8 dB
     1620 Hz  -0.9 dB
     1720 Hz  -1.0 dB
     2188 Hz  -1.5 dB
     2629 Hz  -2.0 dB
     3064 Hz  -2.5 dB
     3506 Hz  -3.0 dB
     3965 Hz  -3.5 dB
     4449 Hz  -4.0 dB
     4968 Hz  -4.5 dB
     5534 Hz  -5.0 dB
     6159 Hz  -5.5 dB
     6861 Hz  -6.0 dB
     7664 Hz  -6.5 dB
     8605 Hz  -7.0 dB
     9739 Hz  -7.5 dB
    11157 Hz  -8.0 dB
    13029 Hz  -8.5 dB
    15708 Hz  -9.0 dB
    20128 Hz  -9.5 dB
    21427 Hz  -9.6 dB
    30195 Hz -10.0 dB
    96000 Hz -10.4 dB
    At last it works as it should!

    And it exactly matches how Sony NS999ES SACD player de-emphasizes CDs on the analogue output, from what I measured. I am so glad I found it out now. Thought this info might be useful for some other folks out there, too.

    You never stop learning!
     

    Attached Files:

    Robert C, paulisdead and Chooke like this.
  2. MartinGr

    MartinGr Well-Known Member

    Location:
    Germany/Berlin
    Thanks for these interesting post! As I'm converting a lot of my older CDs for my portable, I'm still looking for an easy way. I'm not sure if the iTunes story is just a rumour, as it doesn't do anything to my CDs with deemphasis.
    But almost none of my CDs with preemphasis have set their flag correctly in the TOC - only in the subchannel, as reported in an earlier thread. So maybe iTunes looks only in the TOC?
    I made an impulse response with samplitude, that I use in the convolver of foobar2000 while converting. I tried to made an analog impulse resonse with my CD player, but the digital impulse response (made with Samplitude's FFT-Filter and its preset CD-Deemphasis) sounds much better to my ears.

    My preemphasis finds so far:

    Genesis - Genesis *
    Paul McCartney - Give my Regards To Broadstreet *
    Paul McCartney - Pipes Of Peace *
    Lionel Richie - Can't Slow Down
    Ultravox - Vienna *
    Mahler Symphonies (Inbal, Denon) *

    (*Preemphasis not detected by any program except Exact Audio Copy 0.95 until prebeta 5)

    Martin
     
  3. SolarWind

    SolarWind New Member Thread Starter

    Location:
    Amsterdam, Holland
    The author of Exact Audio Copy, Andre Wiethoff writes:

    Myself, I never encountered pre-emphasized CDs with wrong TOC (not flagging pre-emphasis).

    I am not an expert on digital filters, but I’ve always tended to think, that in case of de-emphasis, the convolution (pulse/response estimation) - foobar2000 plug-in and WaveEmph (VB tool that crushes:)) quoted in earlier threads - would produce a slightly inferior quality results, with some (subtle) artifacts (linear phase distortion, pre-echo), which are associated with this type of digital filters - I am not saying you would immediately hear that, but it might be unnecessary injected into the archive. It can be avoided completely when filtering is done in the frequency domain only by using FFT instead of convolution.

    I'd always opt for FFT de-emphasis, and ditch convolvers when archiving my records, to be on the safe side. There must be a reason why professionals use FFT for serious eq studio work and convolvers just to mock up / hypothesize / approximate sound of something (fun/play stuff, audio effects mainly). I may be wrong, though.
     
  4. MartinGr

    MartinGr Well-Known Member

    Location:
    Germany/Berlin
    I couldn't believe this TOC problem, too. But I've checked it with all Plextors and Pioneers available - at home and at work. Plextools Professional as well as the current EAC detect the preemphasis only at my self-burnt CDs and Lionel Richie - Can't Slow Down.
    But "Detect TOC manually" in EAC prebeta3 detects the preemphasis from the subchannels in all the CDs mentioned above - and one can easily hear that this is correct.
    I did an out-of-phase-compare of the results of my foobar convolver and the FFT-Filter in Samplitude (which I used for the impulse resonse). The result was "nearly identical". I don't hear a difference.
    But still, I don't think that FFT-Filtering is the best way for this purpose. Preemphasis and deemphasis-filters are originally analog - and not phase-linear. So, in a perfect world the deemphasis-filter reverts like a mirror all the phase-shifting done by the preeemphasis-filter. Which means it must be designed like an traditional analog filter.

    Martin
     
  5. SolarWind

    SolarWind New Member Thread Starter

    Location:
    Amsterdam, Holland
    That's too bad... I've been using EAC v0.95 beta 3 (also with different Plextor CD drives SCSI&ATA) without any such problems. But like I said I don't have that many pre-emphasized CDs.

    In retrospect, the idea to adopt pre-emphasis (designed to battle the tape hiss) for digital recordings, was probably one of those "most awesomely useless inventions", giving minimal added value to the sound (if any; yes, I know about "quantization noise") but causing a lot of troubles, confusions, awkward CDs with invalid(?) TOCs and different sounding on different CD players :D even more than 25 years after.

    IMHO, they would've battled quantization noise much more efficiently just by recording those early CDs at proper levels (and not -6 dB max peak) - I understand the need for headroom - but 25 dB below maximum in average was just a waste. :(
     
Thread Status:
Not open for further replies.

Share This Page