In my limited tests with Carl Haber at LBNL several years ago, IRENE performed well on pristine media, but not as well as a stylus on deteriorated media. IRENE had difficulties with electric transcription discs because the laminate was not absolutely opaque, and there could be some additional reflection from the aluminum substrate. So unfortunately, in the situation where you want most to use a non-contact solution (e.g. failing laminate), IRENE is not as useful as a stylus.
IRENE's limited resolution, and reduction of a 3D groove to a 2D representation, also tends to exaggerate noise more so than a stylus on worn grooves.
Finally, be careful not to confuse dynamic range (bits) with resolution (kHz).
~ Eric Jacobs
On Nov 19, 2015, at 8:50 AM, David Giovannoni <[log in to unmask]> wrote:
Mason's opinion is both well-informed and free from any financial ties to
any particular technology. For those reasons I consider it to be among the
most accurate and honest public assessments to date.
He mentions my study of the bit-depth issue. Last year, as a customer of
IRENE, I set out to understand why IRENE's audio is so inferior to that of
tactile (stylus) transfers. My focus was on cylinders, not discs. I looked
up the manufacturer's published specs for the probe that does the imaging.
When I asked Dr. Haber to check these, he noted that modifications can make
the probe more than twice as precise as the published specs indicate. I
accepted his performance specs without question and sat down to do the
calculations.
The math clearly shows that the issue is bit depth. Even when outperforming
manufacturer's specs, the probe can resolve the signal in a brown wax
cylinder to only eight bits of precision. Eight bits of precision in the
image domain translates to eight bits of precision in the audio domain.
Claims have been made about some sort of magic happening in IRENE's
image-to-audio conversion software. But as brilliant as the software
architects are, the laws of physics are immutable. Science says: eight bits
in, eight bits out.
Of course scanning technology will improve over time and with each iteration
of IRENE. Unfortunately, we can expect bit depth to remain a limiting factor
for decades to come. Again it's simple math. Moore's law estimates that
resolution might double every couple of years. It hasn't in the past; but if
it could achieve that velocity of improvement, IRENE would equal today's
tactile resolution around 2040 and satisfy TC-04's 24-bit standard around
2050.
The paper to which Mason refers shows the math. If you're interested the
link is below. Please understand the paper was not written for a general
audience or for publication. Rather than slog through the math you might
want to view the ARSC conference presentation that Mason referenced, also
linked below.
In July the Library of Congress held a conference on imaging. As a member of
the audience I was taken aback to hear speakers not only extol the "very,
very high resolution" of IRENE, but fully endorse the assertion that imaging
is the preferred preservation solution for entire collections!
As Mason suggests, low resolution, slow throughput, and high cost make IRENE
the solution of last resort for entire collections. Don't get him or me
wrong - imaging is a necessary and welcome preservation tool for select
items. Indeed, I was among the first to send cylinders to NEDCC for
scanning, and I continue to do so when scanning is the right way to go. But
like any technology - including tactile transfer - IRENE has its
limitations.
That those limitations have not been stated clearly, publicly, honestly and
professionally has troubled me for some time. Others confide it bothers them
too. Thank you, Mason, for raising the issue. I hope your assessment
engenders an informed discussion of the facts.
- David Giovannoni
IRENE Assessment
https://www.dropbox.com/sh/verjcwzrcdm6pm8/AADJr8_uBfmprpm2uo4VIP_ka?dl=0
Recent Developments in Cylinder Audio Preservation
https://www.youtube.com/watch?v=8OfowgD4CT0
