Problems of Sound for Video and Film: Concert Recording
Meeting Summary by John S. Allen
David Hadaway, BAS stalwart, preparer of meeting summaries, amp and preamp engineer/designer, and digital recordist, asks:
Is 16-bit linear PCM an aberration in that we will never see anything so good again? Will all future digital audio formats use [lossy] compressed data?
Hadaway's company, DB Systems, is celebrating its 23rd anniversary, now making inexpensive devices (as well as the original
line): an add-on phono preamp, moving coil preamp, headphone amp, attenuators, etc. He also sells accessory items, such as phono plugs (from the large
quantities he purchases once per decade); they have a ring with a single split, nicely made -- the center pin does not twist. He makes cables with these
plugs.
Hadaway wanted to talk about his recordings, Sony's "Super Bit Mapping," and audio for video and film (which
was his first topic).
The area of classical music on laserdisc is of special interest to Hadaway. Although concerts may be of limited appeal
for repeat listening, the 16-bit digital sound is of audiophile quality. Opera is somewhat problematic. In the past, it has been recorded as audio,
then the singers would mime to playback while filming the video. This causes a sense of unreality when the sound doesn't quite match the picture. Producers
can't resist "opening up" the production: in Zeffirelli's La Traviata, Teresa Stratas is singing away while running across a meadow. The voice
is always centered, loud, and close to the mike. All filmed opera, to his knowledge, is mono, except for the orchestra and perhaps the chorus.
We have had the pioneering work from John Culshaw (Wagner's Ring cycle on Decca) in creating a stereo sound picture, but
everything now is dead center. At least most opera today is filmed live, which is some improvement. (Ballet is very satisfying: no singers to upset
the balance or dynamic range, and a high-quality picture to enjoy; however, the repertory of great ballets is comparatively limited.)
Some history. At first, film was silent. Sound was recorded experimentally on cylinders and disks. Early directors were
of the opinion that silent film was the true art form. This shows an emphasis on the visual that understandably continued throughout film history but
that we see today also in the neglect of audio quality in many films and videos.
When sound came to be recorded on film, there were two systems: the single system sound, mostly for news-gathering and
home movies, which suffered from using the same emulsion for sound and picture; and the double system, with a separate strip of synchronized film to
record the sound. Costly, this was used only for serious movies.
Perhaps the pinnacle of optical sound was Fantasia, developed by RCA with Fantasound: 3-channel audio recorded at 100%
modulation with a fourth track for gain control, steered to different speakers in the theater by notches cut in the side of the film, and a separate
projector for the picture. The Philadelphia Orchestra and Leopold Stokowski were recorded using eight separate optical 35mm film recorders, 33 microphones,
and 500,000 feet of film. The first playback setup cost $30,000 and had 60 loudspeakers surrounding the audience. It became a road show transported
in 35 enormous packing cases. That was 1940. A BAS member who attended an original screening said it sounded "rather phasy" (no doubt due
to the phase shifts among the original separate optical tracks).
Unfortunately, the full quality of the original is no longer available. In 1955, with the original nitrate soundtrack film
disintegrating, Disney's engineers took the decaying film to RCA in Hollywood and played it back on the last remaining Fantasound system. They fed the
sound down three class-A telephone lines and transferred it to 3-track mag film at Disney in Burbank! This lopped off about 1kHz at the top, added hum
and distortion and more phase shift, and froze arbitrary steering decisions (this part of the story you won't hear from Disney's PR; much of this information
is taken from a short article by Barry Fox in the April 1992 Studio Sound, a copy of which I will send to anyone interested-DBH).
Fantasia was a financial flop; it finally broke even in 1952, after being released in optical mono. World War II had put
an end to further development (the movie was intended to be the first of a series). In 1982 Disney recorded it in digital with Irwin Kostal, but purists
complained, so they cleaned up the original for rerelease (for their efforts they were sued by the Philadelphia Orchestra).
In 1951, Westrex developed a 3-channel recorder using magnetically coated film, and Rangertone developed a way of recording
synch signals at right angles to the audio track on conventional magnetic tape. With either of these two machines, sound with the same high quality
as in audio music recording could be synchronized with film. But film has never achieved what we would consider audiophile quality. The fact that film
sound is commonly mixed to eight generations might be a reason. Interestingly, in 1951, optical recording using ultraviolet light on fine-grain film
was considered superior to magnetic recording, but optical had reached the end of its development and magnetic was just beginning.
In 1965, Dolby A was developed. Tom Holman (much later) recorded one film in Dolby A but after listening to the end result,
with all the generations of copying, he was not sure it was worth the effort. Eventually we had Dolby SR, and now three digital systems, with DTS using
the highest bit rate. The latest development is Imax, with polarized shutter-glasses for 3-D and tiny speakers in front of the ears for binaural sound,
supplementing the theater system.
Home videotape recording started with mediocre linear audio tracks, then with Dolby B, and then Beta HiFi, which used a
helical scan and 2:1 compansion. Beta Hi-Fi always had a problem with noise pumping and head switching-"Beta buzz." VHS is similar, but records
the audio signal at a different depth in the tape ("depth multiplexing"). Ira Leonard: Beta buzz is worse when recording from an FM signal
with subcarrier noise, which confuses the compansion.
Video reviewers seem oblivious to the pumping problems, except for E. Brad Meyer, who compared an analog recorder with
dbx, VHS-HiFi, Beta HiFi, etc. in a 1984 article in Stereo Review.
Video on Disk: RCA Selectavision (electromechanical pickup from a grooved disk that wears out), which lost $1.5 billion
(and now RCA is owned by BMG). The laserdisc offers excellent quality, and uses FM encoding for audio, giving about 55dB S/N (limited by the wow and
flutter). Later, CX audio encoding was added, with 2:1 compansion, avoiding the "breathing" problem. This gives a respectable 80dB dynamic
range. The Reference Recordings laser test disc guide claims that CX causes "loss of depth" without explaining how.
Laserdiscs went to 16-bit PCM-encoded linear digital sound, and then added Dolby Digital, which takes away one of the FM
audio tracks -- but that has not been a major problem, since 99% of films (until recently) are recorded in mono.
Laserdiscs don't sell as well in Europe, where the PAL system (vaunted for its superior resolution in black and white)
reduces video color bandwidth to less than NTSC [unlike with NTSC, the chroma signal cannot be separated from the luminance signal without reducing
the luminance signal bandwidth-JSA]. What's worse, all movies issued in Europe on laserdisc run about 4% fast, raising the pitch of the sound, because
films are shot at 24fps and PAL video runs at 25fps (= 50 fields per second, which also causes perceptible flicker). Also, when European laserdiscs
went to the digital sound, there was no room for the FM tracks, making all previous players obsolete.
David Hadaway is happy with the laserdisc format for movies, opera, ballet, and concerts, although there is a problem with
finding which ones have good sound because reviewers so rarely talk about it.
The DVD uses data-reduced video and audio, with some artifacts. A letter to Video magazine describes momentary audible
ones: slight crackling noises or raspy dialog.
Ken Pohlmann recently wrote that DVD offers better video-audio fidelity than any other consumer medium, but didn't explain
why. This is questionable because DVD audio is bit-reduced: Dolby Digital uses perceptual coding to throw away about 75% of the data (the "inaudible"
part) based on assumptions about flat response and the level at which you are listening. An example: You have a high-level tone at 1kHz. Because of
masking, adjacent tones at lower levels are completely inaudible, so they are thrown out. But suppose you are listening in a reverberant room with a
null at the masking frequency (in this case 1kHz) or have a hearing loss at that frequency. Then you might hear artifacts. It's a matter of degree:
the more data reduction, the more likely you will hear artifacts. These systems are designed for the majority, not for a few purists.
Hadaway sees DVD as killing off laserdisc by being cheaper and providing better quality for video. He is concerned that
DVD will also kill off laserdisc for concerts, etc. He is not particularly a surround sound fan.
Problems with audio for film are often due not to technical limitations, but to the insensitivity of the producers. Michael
Fremer has described [see the
When films are viewed at home we can hear problems that are masked in noisy theaters with poor sound systems:
- Poor frequency balance - often shrill, and with no deep bass. In the optical days, bass caused intermodulation and
overload because the system was hard-limiting, so movie people got in the habit of cutting out bass. Richard Burwen wrote recently in Audio about
his equalizer; for the laserdisc of La Traviata he used an astounding 68dB of bass boost!
- Compression-to keep the quiet parts above the popcorn level. But then shouts are distorted. In Forbidden Planet, when
the monster Id attacks the spaceship, its roar is hardly louder than the dialog. In opera on video the soloists are louder than the full orchestra.
Vocal buffs think this is okay.
(Hadaway's opinion is that opera would be great if it weren't for the singers. When he records opera, he gets the orchestra
to peak at least 6dB above the singers, and even this is sometimes not quite realistic, as the orchestra may overpower the singers in live performance.
He uses an ORTF pair for the orchestra and cardioids for the soloists, as many as six microphones in all.)
Audience comment: Another common broadcast problem is compressing and recompressing through the signal path. ABC in New
York has its compressor, then there is one for the satellite uplink, and then one for the local station, each one typically tending to push the level
to maximize perceived and average loudness as well as the s/n ratio. With digital transmissions there is no excuse for this.
- Unnatural ambience -- often the original sound is replaced. Real ambience is complex with a dense texture; faking it
is hard work. In Tron Fremer used real sounds (recording at the zoo, e.g.) and manipulated them electronically. Sometimes producers just don't care,
so the car is accelerating on a sandy beach and the tires squeal.
- Unrealistic voice perspectives -- the dialog is often dubbed over with "looping" so the acoustics of the
recording do not match those of the image location. In Hatari, there is a scene where several people are sitting on a couch talking in natural ambience.
Next the ambience disappears, one person says a few words, and then he walks over and plays the piano. Obviously they were getting ready to overdub
the piano sound and started a little soon. Often the location sound cannot be used, owing to poor recording quality or extraneous noises. Shotgun
and radio mikes make location recording easier now.
Meyer: The movie Contact begins with at least a minute of pure digital silence and is a test for extraneous noises.
Tom Holman played the original production sound for Star Wars at an AES convention and at AES/BAS meetings, showing why
they have to replace sound. It was the opening scene when the spaceship is being boarded. You could clearly hear the clump-clump-clump of feet on the
plywood sets, cap guns were used for the laser pistols (the caps were used to synchronize the visual effects), Darth Vader with a squeaky Cockney voice....
One example of nicely done ambience occurs in Splash, where the ambience changes as people get into an elevator. If ambience and sound effects aren't
done right, you may not notice it, but it gives you a feeling of unease.
The problems can, however, be drowned out by music. Some of Hadaway's favorite film scores are by Sir Arthur Bliss (Things
To Come), Eric Korngold (Sea Hawk and Robin Hood), Max Steiner (Johnny Belinda), Brian Easdale (The Red Shoes), all Bernard Hermann (including his Hitchcock
films, the Harryhausen epics, The Day the Earth Stood Still, and Fahrenheit 451), Jerry Goldsmith (Mephisto Waltz), Vangelis (Blade Runner), and Basil
Poledouris (Conan the Barbarian).
In general, foreign films have poor sound, yet reviewers seldom mention it. Why poor? Generally they are made on small
budgets because their audience is minuscule. In Paris people flock to see Terminator 2, not the latest Eric Rohmer art film. Multiple languages in Europe
lead to shooting the film silently and dubbing everything, and since the lip movements don't match, why bother with anything else? A few exceptions
stand out: for example, Jean de Florette, obviously made for an international audience.
Hadaway played several audio selections, copied to DAT, to illustrate problems with audio for films. Such problems are
often more audible in your home than in a theater, because of the shorter reverberation time, better sound system, and better audio quality in consumer
media than in the optical soundtracks of theater films.
Eureka, a prerecorded Beta tape, had Beta buzz; Dead Poets' Society had VHS hum, not so bad. A sequence from Logan's Run
from laserdisc had synthesized footsteps that sounded like patting hands, with no reverberation, while simultaneous voices did have reverb. John Frankenheimer's
Seconds, with Rock Hudson, on laserdisc, had the sound of surf which modulated along with the dialog. This was probably because the action moved indoors
and the surf would have sounded unrealistic inside a house, so they tried to remove it with a noise gate. The director admitted on the commentary track
that 80% of the dialog was replaced because the wide-angle Arriflex cameras were very noisy and that look was paramount in his mind (sound being secondary,
as usual). Nowadays quiet cameras are readily available.
The ballet Romeo and Juliet on laserdisc had noise-gate effects. Whenever the music stopped, the sound disappeared into
a black hole. Evidently the producers felt that since it was digital it had to be noiseless.
The Loves of Emma Bardac, on laserdisc, had voice-overs that were tubby with a grotesque bass emphasis and hum, and a surprisingly
high level of hiss for a digital production. It got excellent reviews with no mention of any audio problems. The picture is gorgeous.
The ballet Raymonda, in high-definition video and digital sound, showed the wide dynamic range possible on laserdisc. The
climaxes showed the additional headroom of the digital track over the analog track: analog zero may be -8dB, while digital zero is -20dB.
Concert Recording
David Hadaway started his recording career in 1963 with a Lafayette Radio tape deck [gray plastic-coated cloth on a rounded
wooden case, 3-3/4 and 7-1/2 ips, 7-inch reels, 4-track stereo; nice quality for the price, I had one too -- JSA]. It was a $100 tape deck without speakers
or amps. It worked well for recording FM, though the heads wore quickly.
In the '70s Hadaway started recording local musical groups, first with 1/4-track Dolby B, then 1/2-track Dolby A. In 1982
he bought a Sony PCM F-1 for recording 16-bit PCM onto Beta tape. This worked well except for occasional dropouts. He wrote an article for Audio Amateur
(now Audio Electronics): "How To Build a Dropout Detector." Much later he found that a few dropouts escaped the detector, so he added a loss-of-tracking
detector. This caught all but the most minor, inconsequential ones. Compatibility among machines was a continuing problem (Peter Mitchell recommended
the Toshiba M411 as making particularly compatible tapes; Hadaway bought one, but it never performed that way for him). The F-1 and its successors had
a rise in distortion at -50dB, probably not audible.
In 1982 David Hadaway got a DAT machine that had better A/D and D/A conversion, no 1% distortion at -50dB anymore, but
it did have dropout problems, including a head clog that went on for 30 minutes. He is still archiving on Beta. He has a lot of Beta tape to use up,
and Beta tape uses oxide particles. Hadaway doesn't trust the archival properties of DAT, which uses metal particles that lose their magnetism if they
oxidize. He did find that he could get perfect digital edits (not the same as a perfect audible edit) on the Panasonic 3700 DAT. He found an error-concealment
flag on a chip on the Sony DTC700 DAT and wired it into his previous dropout checker, so he could check a concert tape without listening for an audible
errors. The intermission break was seamlessly joined by the Panasonic.
Micha Schattner: A mastering house will return any tape with an error even if it is inaudible, or any tape with more than
three samples in a row of clipping.
Hadaway now has the ability to produce perfect digital copies without listening to them. By adding a sensor for it to the
dropout detector, he overcame the problem that the muting relay of the receiving DAT sometimes would click even with no problem on the source tape.
When he had dealt with all these problems, he could copy through 25 generations with no deterioration.
In 1995 his PCM-F1 died and Sony wouldn't repair it. He advertised and was able to purchase three PCM-601s with digital
in/out. His only problem with them is that they record and play back at 44.05 instead of 44.1 (because they use the color video standard, with a slightly
slower frame rate). He can now create high-quality audio and archive it on video tape on $99 VCRs on $1 tape.
One reason Hadaway is now using VHS is that the machines and tape are new, but he is still archiving on Beta. One potential
problem with VHS is that there is no way to disable a VHS deck's dropout compensator. This copies one video line to replace a missing line. That confuses
PCM audio, which has its own dropout compensation as part of its digital decoding algorithm. However, modern VHS tape has very few dropouts, so it doesn't
seem to matter.
PCM audio uses every part of the picture including the part you can't see, and is fussier to record than video. The VHS
decks have auto-tracking, one less thing to fiddle with. There is still one Beta machine being made, the Sony SL 2000. Hadaway will buy one sometime.
By going digital-in, he now bypasses the antiquated 1982-vintage A-D converters. He uses a $500 Sony Super Bit Mapping
(SBM) converter, which they want you to plug into a $200 adapter that has all kinds of functions he doesn't need. David Griesinger told Hadaway how
to cut off the special plug on the SBM and replace it with an RCA plug, foregoing these functions but saving hundreds of dollars. Hadaway found no measurable
distortion from the SBM converter, compared with a test CD direct digital input to the 3700 DAT recorder.
|