[no subject]

*****************************************************************



HiFi News September 2000 Issue
----------------------------------------------




Steve Harris Editorial

WATERMARK LISTENING TESTS

Regular readers of this column (I trust there are some) will get a sense
of déjà vu as yet again I broach the ever more wearisome subject of the
Verance watermark.  But this is such an important issue that we have
devoted the first two pages of 'News' to a special report on Verance and
its London listening sessions.  Anyone who cares about the future of
high quality sound, or indeed its continued existence, should turn to
page 10 right now.





NEWS

VERANCE AUDIO WATERMARK VISITS LONDON  -  A Special Report by BARRY FOX

The SDMI, 4C and Verance gave the first demonstration of their analogue
watermarking system outside the USA in early July.  A cross-section of
recording engineers and a few journalists were offered the chance to do
comparative listening tests on four pieces of music, with and without
watermarking.

However, the point of the tests remains unclear because Verance assures
that the decision to build watermark detection into DVD-Audio players
has already been taken.

Paul Jessop (photo. left) Technical Director of the International
Federation of the Phonographic Industry (IFPI) was on hand to represent
the Secure Digital Music Initiative; David Liebowitz, Chairman of
Verance Corporation, represented the company formed last year when Aris
Technologies and Solana Technology Development merged after finding that
they were both developing, and patenting, similar technology.  Leibowitz
stressed that he had no say in the event other than to provide the
hardware.  There was no-one present from 4C Entity, the consortium of
IBM, Intel, Matsushita/Panasonic and Toshiba which ran the US tests in
the summer of 1999 on behalf of the SDMI, and on whose say-so DVD-Audio
and probably also SACD will be watermarked.

'They send their apologies and have left me to carry the flag,' says
Liebowitz.

There was no technical White Paper available on the technology.  David
Liebowitz says this is because 'We don't want another DeCSS situation;
we don't want to make it easy for people to build the tools to
circumvent the system.  The details are highly confidential.  Patents
give only a general description, and we don't use everything that is in
every patent... release of information is very tightly controlled, only
a few people know the detail'.

Stated simply, the Verance system analyses the audio waveform and makes
slight alterations which then decode as digital bits to convey a
copyright message, and trigger circuitry in a new generation recorder so
that it cannot copy protected material.

The Verance system is already used in applications where quality is not
an issue, for example, to tag music for NBC TV sports broadcasts, and in
Memory Stick solid-state recorders.  The contention is all about its use
with DVD-Audio and SACD.  From October, all DVD-A players must have
Verance circuitry built-in.

The 4C tests which led to this decision were done last year in US
studios, on 50 'golden ears' who had been chosen by the five major
record labels.  They listened in US studios to an A track with no
watermark, a B track with the same piece of music with a watermark, and
an X track either with or without watermarking.  Each listener then had
to try and identify whether the X track was watermarked or not.

The SDMI, 4C and Verance claimed that these tests proved the
watermarking system to be transparent, that is, inaudible, but the
golden ears had been guaranteed anonymity and so far only Danny Purcell
of George Town Mastering, Dave Smith (Sony Music) and Mark Wilder (Sony
Music) have been prepared to identify themselves.  Although DVD-A
providing a sampling frequency of 192kHz, 4C's test material was coded
at 96kHz/24-bit standard.  Only Aris and Solana, now merged as Verance,
passed the robustness and audibility tests.

'4C' are merely SDMI members, like other members', explains Paul Jessop.
'They performed the Phase 1 tests under contract to SDMI'.

Proposals for Phase 2 SDMI testing were due in by 3 July.  Paul Jessop
will co-ordinate tests in the USA, Japan and Europe, using methodology
based on the European Muse tests previously run by the IFPI.  Tests
should start 'within a couple of months'.

Says Jessop: 'The SDMI will now carry out Phase 2 tests at 176.4 and
192kHz.  The SDMI will not recommend anything without testing.  Nothing
will be passed for use at a data rate until tested at that rate.  We are
starting all over again, but will incorporate the UK data in the
published results.'

According to David Liebowitz, the Phase 2 SDMI tests will be on more
complicated systems, for instance to detect unauthorised recording.
'The testing for DVD-A is done and complete' he insisted.  Panasonic has
already launched DVD-A players in the USA, over the 4 July weekend.

So what was the point of the London tests, which were only arranged
after the worldwide campaign spearheaded by UK Grammy-winning recording
engineer Tony Faulkner ?

'The previous tests were covered by non-disclosure agreements; these
tests are to open awareness, create greater appreciation, and let
recording engineers judge for themselves' says Leibowitz.

'Although the hardware manufacturers must put detectors in, content
owners are not compelled to use the system'.

After 4C chose Verance as the winning system, 4C created a licensing
corporation called LMI which will now administer the licences and
collect royalties on behalf of Verance.  Says Liebowitz, 'It's a
collaborative relationship'.

There is a considerable amount of money riding on watermarking.  The 200
company members of the SDMI now pay $20,000 a year, double the previous
annual fee, to participate.  The manufacturers of hardware pay a flat
fee of $25,000 a year for the source code of the algorithms used to
detect the Verance watermark, and to change the mark depending on the
number of copies made.

Those who are happy to take Verance's object code (which does not allow
'optimizing' the system to personal needs) pay $10,000 a year.  If the
company makes two types of hardware, there is an additional fee of
$10,000 a year and $5,000 on a sliding scale downwards.  The
manufacturers of mastering equipment pay 25% of the revenue on any part
of the product which is used to embed a watermark and the mastering
houses which embed the watermark pay $50 per track, with a special
'garage band' rate of $250 for the first 50 tracks.

Mid-size record companies or independents pay a flat rate of $50,000 a
year which lets them embed watermarks in up to 1500 tracks, with the
rate rising to $50 a track above that number.  The five major record
companies pay $40,000 a year or $50 per track capped at $60,000.  This
rate is based on the current situation where there are five majors and
will increase if there are mergers or take-overs which reduce their
numbers.

Malcolm Davidson of Sony Music NYC conducted the tests at the Sony/CBS
Studios in Whitfield Street.  The full results will be published within
a few weeks.  However, only those participants who agree to be named
will be named.  The Association of Professional Recording Studios (APRS)
alerted mastering and sound recording engineers to the London tests and
over 30 agreed to turn up for ABX sessions, listening to four pieces of
music, Debussy piano, Stravinsky's Petrushka, new age rock and pop
music.  The music trade press were invited and also Hi-Fi News, but no
other hi-fi magazines.  HFN/RR Editor Steve Harris was invited only on
the day, without any prior warning, and could not attend.  Technical
Editor Andrew Harrison dropped everything to go in his place.  The most
surprising discovery was that the testing process depended on a Compaq
laptop PC in the listening room: its hard drive that whirred louder than
quiet passages of the music and any subtle ambience.  The drive could
not be switched off during the tests.

Tony Faulkner flew back from Malaysia so as to be in London in time for
the tests. 'I'm forty-nine years old, I've got a cold and I had jet lag,
he said after participating the day after flying in.  'I scored 75%,
correctly identifying the watermarking in six out of eight cases.  It
was analogue source material, too, not of good quality.  They haven't
tested women, and they haven't tested members of the public.  It's not
as bad as Copycode, which anyone could hear even after a night on the
beer, but this system is not transparent, and anyone who says so just
isn't telling the truth.'

James Mallinson is a Grammy-winning producer for all the major record
labels and independents.  'Please understand I am all in favour of
effective copyright control' he says, 'but if you do anything in the
analogue domain it should be inaudible and these tests were so badly
handled that they are totally invalid.  I don't know whether it's
because they don't know about sound quality or were trying to pull the
wool over our eyes.

'The first musical piece, Debussy piano was an old analogue tape with
buckets of hiss and very constrained frequency range.  Halfway through I
refused to continue because I said there was no point in the test.  The
Petrushka recording was some of the worst recorded sound I have ever
heard; analogue hiss and you could hear the azimuth flexing, as you do
from old tapes.

'It was quite impossible to judge a system with that kind of material
and DVD-Audio and DSD are all about high quality.

'Then they played some New World music which was appallingly badly
recorded, with no low-end, no high-end above around 10kHz, and heavily
compressed.  It was of magnificently low quality.  Then they played some
dreadfully recorded pop.  It was a joke, a complete joke.  I was really
quite cross because I thought they had wasted my time.  These tests were
totally invalid'.

The two directors of Sadie, the Ely-based British company which has
cornered the market in hard-disk editing systems, participated.  If the
record companies want a watermarking system, Sadie will have to provide
the hardware.  Both Geoff Carver and Joe Bull are ex-recording
engineers.

Says Geoff Carver, obviously choosing his words very carefully: 'It goes
without saying that I am all in favour of effective copyright control.
I could detect a change consistently.  My score was around the same as
Tony Faulkner's.

'If I had heard the music in isolation I don't think I would have said
there was a technical problem - other than to comment on the nature of
the recordings used for the tests'.

Martin Colloms was not invited so he phoned, asked to be invited and was
slotted in.  Says Colloms: 'I withdrew from the tests.  The quality of
the material was appalling.  It was sub-Walkman standard.  They were
using very old analogue material; the Petrushka dates back to 1962.  The
piano had a dynamic range of only around 40dB.  It was awful music,
lifeless and anaemic.  There was a mess of hiss.  The material had been
copied from CD-R on to computer hard-disk, and then processed through a
96/24 converter, so it didn't start as 96/24.  The room acoustic was
bad, with metalwork ringing.  The PC hard-drive was whirring in my ear
and a transformer was mechanically humming.  I told Sony it was an
embarrassment to their company and I couldn't avoid wondering whether
they had chosen material that was bad to conceal what the system was
doing.'

In an open letter to the profession Faulkner says: 'I believe that the
strategy of watermarking high-quality material on high-quality carriers
is fundamentally flawed if the watermarking is audible on high-quality
systems.  The testing so far has been inadequate in terms of sample size
and quality of test material and methods.  If the system is audible now
with a 2-bit copy management payload, how will it sound with a 72-bit
full-identifier payload?

'The myth about the watermarking being optional is becoming very
tiresome.  How will it be optional for listeners to major label output?
How is it optional for performers?  How will it be optional for
producers and engineers generating regular releases for major
international participating labels?  How will it be optional for DVD-A
player and recorder manufacturers to choose not to build in, and to pay
for the technology?'






WATERMARK WASH-OUT  by Andrew Harrison

It would be nice to report that the analogue watermark that is liable to
appear in almost all future music releases is entirely inaudible, and
that music lovers with an ear for quality have nothing to worry about.
Unfortunately I cannot confidently say this.

I attended the first of two one-day listening tests held in July.  After
an hour of small-talk with Sony, Verance and SDMI representatives
Malcolm Davidson, David Leibowitz and Paul Jessop, I was told that there
would be no time available to actually listen to watermarked material.
At a second visit I was disappointed to discover that the choice of
listening environment and music material was not conducive to showing
subtle differences pre and post watermarking.

Playback of material was from computer hard disk controlled by a Sonic
Solutions workstation.  Quality of playback was deemed at best average,
but well short of the quality expected for critical listening for
watermark distortion.  For example, I found the listening position too
close to the ATC SCM70 active speakers, these placed too close to the
rear wall of the small monitoring room.

Ambient noise level in the room was not low.  A ticking clock was easily
removed, but the same could not be said for the whirring hard-drive of a
PC laptop, used to log and collate the ABX decisions; loud enough even
to mask tape hiss in the older recordings.  This disturbance was
compounded by the sound of loud classical monitoring in an adjacent
room, the sound leakage being high enough to break concentration.

The music selection, four pieces chosen presumably by Verance or
Verance-licensee Sony to illustrate diverse styles, was not felt to be
of particularly high-fidelity quality considering the stakes.  I was
expecting at least native 24-bit/96kHz recordings, in lieu of DVD-Audio
24/192 recordings.  Despite this, and the poor ambient conditions, I was
determined to see the test through and resisted the idea of walking out.

I was told I had spent too long listening to the first extract, and I
began to appreciate why previous sessions had over-run pre-allotted
time.  Interestingly, my score from the first piece of 62.5% correct
identifications was virtually inverted on the three remaining extracts,
where I scored 25%, 37.5% and 25%.  A score of 50% would suggest
statistically that watermarked and raw music were indistinguishable.
But at the risk of accusations of crying sour grapes given these
inconclusive ratings, I remain unconvinced that these or others' results
could demonstrate the inaudibility of Verance analogue watermarking in
the London listening sessions.  If the US tests of last year that
allowed the system to be sanctioned were anything like these, I fear for
the bright future of high-resolution music that is promised to us.
Furthermore, I hope I'm not alone in doubting the conceptual validity of
allowing a company that stands to make its fortunes from selling this
system to the world to be allowed such control of these listening tests
which strive to prove the company's own claims of inaudibility.

As a final thought, for listeners and engineers that feel that digital
audio cannot match a good analogue recording, despite digital's near-
ubiquity in this wired world: Verance watermarking can only be applied
in the digital domain.  In other words, a recording that strives to be
deliberately digital-free from microphone to mastertape must now pass
through A/D and D/A converters if it is to carry a Verance watermark.
Without digitizing the music, all the supposed benefits of radio air-
play accountancy, duplication prohibition and copyright management will
be unavailable.