Print

Print


On Wednesday, March 10, 2010 5:52 AM, Parker Dinkins wrote:

> on 3/9/10 8:42 PM US/Central, Eric Jacobs wrote:
>
>> Transfering FLAT will cost you about 6-7 bits of dynamic 
>> range - an audible loss.  EQ in the digital domain will 
>> not recover that lost dynamic range.
>
> For a different conclusion, you might want to read
>
> Filter Reconstruction and Program Material Characteristics 
> Mitigating Word Length Loss in Digital Signal Processing-
> Based Compensation Curves used for Playback of Analog 
> Recordings
>
> By R.S. Robinson, Channel D Corp., Audio Engineering Society Paper 
> 7185, Oct. 2007.

I'm not sure that the conclusion is different, but rather
candy-coated to some degree...

Keep in mind that the entire paper and its arguments are built
around modern vinyl LP media.

On Page 3, under "4. Program Characteristics", it reads:

    "In the high frequency region of the spectrum, the
     degree of extra headroom needed for recording
     uncorrected program material will depend on the
     frequency balance of the program material. The
     amplitude of program content above 1 kHz is critical,
     because that frequency range is pre-emphasized and
     will impact the effective dynamic range of lower
     frequencies. Therefore, it's useful to know about the
     frequency balance of typical, actual music LP
     recordings that have been emphasized with the RIAA
     compensation curve."

Keep in mind that shellacs and older recordings when 
transferred FLAT have a frequency balance that is significantly
biased above 1 kHz when compared with a modern LP due to their 
high frequency noise content.  Therefore, the dynamic range of 
lower frequencies on older recordings is going to be more 
impacted than on a modern LP.

On Page 7, under "7. Upper Bound of Treble Headroom", we find:

    "For the exceptional example (Bob Marley and
     the Wailers) among the recordings surveyed, and
     shown here, this was a maximum of 17.5 dB (2.9
     bits). There also is the possibility that other, more
     extreme such examples exist and could be located, but
     they will be uncommon."

More extreme examples among LPs probably are uncommon, but
among shellacs and transcription discs, more extreme examples
are the norm with their clicks, crackle, and pops.

    "Given the results from the variety of recordings
     surveyed (most of which aren't presented here), and
     considering that the above figure represents an upper
     bound, a more likely typical loss of digital dynamic
     range due to treble emphasis, from the peak
     responding / peak hold analysis, is less than one bit
     (6 dB). It appears to be unlikely, in practice, to exceed
     this figure with LP music recordings on a consistent
     basis."

Again, all references in this paper by Robinson are vinyl
LP.  Shellacs and transcription discs, with their noise 
characteristics, will likely result in more than one bit
of loss.  And 6 dB is audible.

On Page 8, under "6. Conclusion", it reads:

   "The reconstructive properties of digital de-emphasis filtering, 
    as shown here, in conjunction with the characteristics of most 
    program material, will cause a typical overall bass resolution
    truncation of only one bit or less..."

And again, this one-bit loss is what you'll find with LP program 
material.  I would expect more with pre-LP grooved media.

Finally, the article states:

   "Exceptional, uncommonly encountered program material may 
    cause a worst case bass truncation of approximately three bits, 
    which is negligible considering the 24 bit resolution capability
    of modern analog to digital converters."

This statement just glosses over the fact that losing three bits
in the bass (18 dB) is actually audible - not sonically negligible.

   "In cases where bass word length truncation does occur, the
    disadvantage is balanced by the complementary enhancement of 
    digital resolution, due to treble preemphasis, in the frequency 
    range where human hearing is at its most sensitive."

Does increased treble really make up for bass loss?  This really
seems like a non-sequiter.  But the article is called "Word Length
Loss *Mitigation*" - and mitigation is different from recovery.

This paper argues that for modern vinyl LPs, that digital EQ can
come very close to the very best analog EQ for less cost.  It 
does not say, or even imply, that digital EQ can match or exceed 
the best analog EQ.  It only says that the shortcomings of digital
EQ are relatively minor (for vinyl LPs).  The gap between analog
and digital EQ is greater for pre-LP grooved media due its 
greater high frequency noise component.

Food for thought anyway. <smile>

Eric Jacobs

The Audio Archive, Inc.
tel: 408.221.2128
fax: 408.549.9867
mailto:[log in to unmask]
http://www.TheAudioArchive.com 
Disc and Tape Audio Transfer Services and Preservation Consulting