The Various Versions of The Phantom of the Opera (1925/1929) with Lon Chaney

1925 Title Card

1929 Title Card

Universal Studios produced and released the first ever cinematic treatment of The Phantom of the Opera in 1925.  This silent film starred Lon Chaney, the Man of 1,000 Faces as the titular character, Erik the phantom.  Chaney has since been recognized as one of the most versatile actors in silent film, and his make up and portrayal of the Phantom has been recognzied as one of the iconic roles pre-sound cinema.  Although the role may not be his best, nor the movie which features it the most crticially acclaimed of his filmography, it is by far the one for which he is best remembered.  In 1925, he scared the wits out thousands of movie patrons when his hideous deformed face is revealed.  Before I discuss the versions of the films, a few basic facts about silent movies should be identified :

1.  Silent pictures were never silent when played during the "silent era" and often were not projected in black and white.  In larger theaters, often something close to a chamber orchestra would play music specially compiled for the film, while smaller theaters would have to make do with a house organist or pianist.  While the cameras which captured the performers used black and white film stock, prints would often be tinted during processing to reflects the mood, setting or time of day of a scene.  An outdoor daylight scene would be tinted yellow, a nighttime or cold scene blue, interiors orange or yellow, fiery scenes in red, moody scenes in green or purple.

2.  Intertitles often conveyed more than just dialog.  They provided narration, introduced characters and identified the actors who played them, described locations.  Letters, books, notes, signs and newspapers could also serve the function of the intertitle, allowing the audience to quickly soak in exposition rather than view actors "talking" to each other with many intertitles.  Intertitles made it very easy to export films to foreign markets, as they could be replaced with intertitles in the native tongue.  Intertitles were not necessarily written to match the words the actors were speaking, and before sound the intertitles could direct a story in a different way to what was originally intended to be filmed.

3.  All major motion pictures up to 1950 in the United States were shot on nitrate (nitrocellulose) film stock.  Nitrocellulose was a comparatively unstable material, very sensitive to heat and could spontaneously combust if not handled properly.  It also had a tendency to decompose or warp over time, and many films have been lost to one of these causes.  After 1950, films were shot and printed on safety film (cellulose acetate), which did not display the fragility and unstable photo-chemical properties of nitrate.

4.  By the 1950s, most studios, Universal included, deemed their silent film libraries as having no further commercial value and embarked on a campaign to destroy their silent film archives.  They would burn their films to recover the valuable silver content of the film stock and to free up storage space.  50-80% of all silent films are lost due in part to the studio's efforts to make space in their vaults.

5.  In the late 1920s, the studios were transitioning rather rapidly towards sound films.  During this time, movies would be released with talking or singing sequences.  But these films, while they would have an entire sychronized music and effects track, would not be all-talking, which was cumbersome thanks to the limitations of early sound film technology (everybody always seemed to be talking around a table).  Older films, shot for silent screens, would sometimes be converted to a "talkie".  At this time, there were two main competing technologies, sound-on-disk (such as Warner Bros.' Vitaphone) and sound-on-film (Western Electric Sound System/Movietone, RCA Phonophone, DeForest Phonofilm).  Sound on disk had the soundtrack stored on vinyl phonograph records while Sound-on-film, which won this format war by 1932, was an optical sound track printed on the film itself.

6.  In the 20's, the evolution of color cinema photography was still in its infancy, but color sequences did exist to a limited extent in several large budget features.  The most advanced of these for most of the silent era was Process 2 Technicolor, which debuted in 1922 and was used in a handful of silent films.  Process 2 used two primary colors, orangish-red and bluish-green.  This did justice to flesh tones, red and green hues, but was not truly realistic when it came to other colors.  Some films, like Douglas Fairbanks' The Black Pirate, used Technicolor throughout the picture, but other films used it only for certain sequences.  While the camera negative only used one strip of film, the resulting projection prints used two strips cemented together.  This caused no end of problems for projectionists with warped, scratched or broken color strips.  Technicolor was a slow speed system and required copious amounts of light to allow a good exposure of the images onto the camera negative.  Other processes of the period included Prizma color, which used similar principles but was washed-out compared with Technicolor, and the Handschiegl color process, which allowed color dyes to be applied to a specific portion of a film frame.

7.  More than one camera would often be rolling on the set of a silent film.  Due to the relative instability of nitrate film, having two negatives provided an important safety net.  Since all projection prints were struck from camera negatives in these days, having an extra negative was important in case the primary negative was damaged.  Often, the second negative would be taken from a second camera shooting from another angle or a second take.  If a film was going to be distributed internationally, such as in Europe, then a second negative could be shipped overseas for cheaper processing overseas.  Also, the second negative would not have the intertitles spliced in but kept separate so foreign-language intertitles could be created and spliced in for each country.  If color photography was used, there was virtually always a black-and-white version of the same footage.

8.  Silent film had no set standard for the film playback speed.  Most silent film cameras were hand cranked, so the frame rate varied with the cameraman.  Projectors often had several speeds available, including 12, 16, 18, and 20 and 26 frames per second.  16 frames per second was what all the cameramen tried to crank the film camera at, but silent films could include instructions to the projectionist to project a reel at a particular speed.  Sound films in the US are shot and projected at 24 frames per second, and when silent footage is run at this speed, it often looks comically sped up when it was not intended to be shown that way.  Slapstick comedy and chase scenes would often be under-cranked to show frenetic action, however, both in the silent and sound era.

With these principles in mind, let us turn to the film itself :

First Intertitle for 1925 Version

Chaney's Phantom is undoubtedly the closest adaptation of Gaston Leroux's novel.  Unlike later versions where the Phantom is scarred by acid (Universal's 1943 Three-Strip Technicolor "remake" & Hammer Film's 1962 treatment) transformed into a monster through a pact with dark powers (1989's Phantom with Robert Englund) or only deformed on one side of his face (Andrew Lloyd Webber's 1986 musical and 2004 movie adaptation), Chaney's Phantom was deformed at birth and presumably shunned by virtually everybody his entire life.  This is the only major film version made in the author's lifetime, although he had no personal involvement in the production.  Most of the novel was filmed at one time or another during filming, including Christine's trip to her father's grave and the inclusion of the Madame Valerius character.

So, in 1924, when Universal began production, they intended the picture to be its "Super Jewel" for 1925.  Universal had its greatest success turning out what we would call today B-Westerns during this period, but every year it would expend significant resources for one grand picture.  In 1923, that picture was The Hunchback of Notre Dame, starring Lon Chaney and a cast of thousands.  Universal thought it would replicate the success of Hunchback with another weird picture that would showcase Lon Chaney and its production resources, so it pushed Phantom into production.

The Star - Lon Chaney as Erik, The Phantom of the Opera

Of Chaney's makeup, performance and contributions to Phantom, which includes uncredited direction and designing the several masks the Phantom uses during the film, I can add nothing to the 85-years of accolades heaped upon him.  The storyboards, conceptualized by Ben Carre', have also been frequently highlighted as contributing to the atmosphere of the film.  The anonymous contractors and crew of Universal which constructed and decorated the huge sets and directed the extras are often overlooked.  Except for intimate areas such as Christine's dressing rooms, the set designs are all first-rate.  (Why would an understudy get what looks to be a huge dressing room?).  This is all especially impressive since the entire film was shot on the Universal Studios lot (called Universal City) in Hollywood.  Nobody went to France to shoot exteriors or interiors of the Paris Opera House (a.k.a. the Opera Garnier).

The film's reputation as something less-than-a-masterpiece can primarily be laid at the feet of two people, original director Rupert Julian and Universal Studios President Carl Laemmle.  Julian, who had great success with the 1918 "Hate the Hun" picture "The Kaiser : The Beast of Berlin", was brought in by Laemmle to save Universal and its over-budget film "The Merry-Go-Round" from director Erich von Stroheim.  Stroheim had a tendency to go over-budget on his pictures, financial excesses which eventually caused Laemmle to fire him from Merry-Go-Round and replace him with Julian.  Julian finished the picture and the resulting success of the film kept Universal from falling into receivership.  Julian's reputation at the studio (at least with Laemmle) was at its height and "Uncle Carl" (for whom nepotism was far from a foreign idea) deemed him to be the man to bring Phantom to the screen.

A Particularly Infamous Error - Raoul seems to be handing a lantern to Ledoux, but Ledoux is about 10 feet below him!

Unfortunately, Julian capabilities as a director were at best could be described as workmanlike.  Others have referred to him as a hack, and in Phantom his lack a firm grasp of cinematic technique or inspiration soon led to clashes with Lon Chaney.  Chaney decided that Julian was not really worth taking direction from and pretty much did his own thing.  When Julian told cameraman Charles Van Enger to tell Chaney to do something, Chaney would tell Van Enger to tell Julian "to go screw himself".  Julian was a point-and-shoot type of director, content to film a scene, any scene without moving the camera, using closeups or multiple shots.  He ludicrously demanded that the falling chandelier sequence should be shot entirely in the dark!  Fortunately cameraman Van Enger fooled him and we can actually view the impressive effect today.  To be fair, an epic like Phantom would have challenged any director, and better directors have had pictures sunk under the weight of budgets, sets, stars and hype.

Mary Philbin - Our Damsel in Distress

Reviewers of Phantom have sometimes been less than kind to Chaney's two principal co-stars, Mary Philbin as the singer Christine Daae' and Norman Kerry as her lover Vicomte Raoul de Chagny.  Mary Philbin was a quiet, shy girl in real life and it has been said that her best work was with directors who knew how to get the best out of her, not the artistically-oblivious Julian. Lon Chaney was able to get a really convincing display of terror out of her by flying into an improvised rage at her, some of which may have been caught on film.  However, her performance is constrained not only by the function of the heroine in the story but by the writer, editor and director.  Norman Kerry has been called rather stiff and while he is no Douglas Fairbanks or Rudolph Valentino, he has been castigated as a bad actor for something that was really not his fault.  In the scene where he stands outside Chrstine's door while she is talking with her "master", there is a shot of him grinning while playing with his gloves when he should be confused and jealous.  Commentators at the time and since have accused Kerry of really bad acting here, but as he never shows inappropriate behavior elsewhere in surviving footage, there is little doubt that his reaction shot was an inappropriate edit.

Norman Kerry - Our Hero being punished by the editor

When Phantom wrapped principal shooting in late 1924, the Studio decided to test out its Super Jewel with preview audiences before going nationwide.  So in January, 1925, Universal screened a preview version of Phantom in a Los Angeles theater (the Los Angeles Preview Version).  The audience's reaction was not great for primarily two reasons.  First, the picture was over three hours long, making it an ordeal for audiences to sit through.  Second, in 1925 audiences were not ready for an ending where the villain seems to avoid his just desserts.  The intertitles were described as unclear and were rough.  Universal cut the film to eliminate scenes where the stable manager complains to the managers that a horse has gone missing, where Raoul interviews Madame Valerius, where Raoul, Christine and the Phantom meet at the graveyard at Viroflay.  It also added a slightly revised ending where the Persian shoots the Phantom, but audiences were still negative.

The Original Ending - The Phantom Dies of a Broken Heart

Uncle Carl was not pleased with the reception of the film and ordered something to be done to improve the picture.  So Edward Sedgwick was brought in to direct new scenes inserted awkwardly into the existing drama where Raoul strives with the Russian Count Ruboff for Christine's affections in scenes set entirely outside the Opera.  This seems primarily to introduce comic relief in the persons of two new characters, Raoul's valet and Christine's maid.  The Phantom's sole involvement in these new scenes is when Raoul takes Christine to a shabby restaurant and the Phantom pays an Apache to kill Raoul.  The director also filmed the existing ending with the Phantom being chased by the mob led by Simon Buquet.  This version premiered in San Francisco in late April, 1925 (the San Francisco World Premiere Version) The audience's reaction was still not good, and the shlocky melodramatic additions were excoriated.

Sedgwick's Addition - Count Ruboff, a rival for Christine's hand

After the San Francisco premier bombed, the film was edited to remove the Russian Count subplot and the antics with Raoul and Christine's servants but kept the new ending sequence, and this version was more or less what the rest of the world saw, beginning with the New York premiere at the Astor Theatre in September, 1925 (the New York General Release Version).  It played well into 1926 before being returned to the vault and grossed over $2,000,000 during first run engagements against a final cost of ~$650,000.  While it was exported overseas, due to a huge public relations blunder it was not shown in the United Kingdom during this period.  By this time, the book's character of The Persian (played by Arthur Edmund Carewe) has been renamed Inspector Ledoux and Erik is no longer the Sultan's ex-torturer but an escaped convict from Devil's Island.

He knows Everything - Arthur Edmund Carewe as the Persian  Inspector Ledoux

So what is available of this 1925 version?  Unfortunately, the original 35mm negatives were lost or destroyed by Univeral by the early 1950s and no (positive) projection print has been found.  During the 1930s, Universal marketed 16mm "Show-at-Home" prints of its back library catalog and the Phantom was available. The sole visual record of the 1925 version survives in these prints, and apparently multiple sources have been used to present the film as it has been made available today.  These existing prints are dupes of the Show-at-Home print, not the actual Universal product itself.  Each generation from the negative is usually worse than the last..  As will be described later, this source is probably not 100% true to the original 1925 version as shown in theaters (but its pretty close).   The print quality is not the greatest, the image lacks clarity and detail.  Edges of the frame are indistinct, scratches abound throughout and contrast is far from ideal.  There is no tinting or color sequences on my DVD and it runs 107 minutes.

This scene between Christine and Raoul was deleted from the 1929 Version

Universal decided to resurrect the Phantom, one of its biggest moneymakers of the decade, when sound films were becoming big at the box office in the late-20s.  In 1929, the studio decided to convert the Phantom into a part-talkie, part silent version.  This would require the original actors to film new scenes with their dialogue record and synchronized to the film.  But the whole film was not to be remade, so scenes from the 1925 version would be used where possible.  Mary Philbin and Norman Kerry were available and recorded talking scenes, but Lon Chaney was not and Universal could not legally dub his voice in scenes where he appeared.  However in the scenes where his shadow is seen talking to Christine through her mirror, he was dubbed by another actor.  For the opera scenes, performances of the signers would be dubbed in.  Mary Fabian was cast as Carlotta while Virginia Pearson, who played Carlotta in the 1925 version, was redesignated as Carlotta's mother in her scenes with the managers.  John Sainpolis was replaced by Edward Martindale as Raoul's brother Comte Phillipe de Chagny.  In both cases, sound was the deciding factor as Pearson could not sing and Sainpolis was not deemed to have a good speaking voice.

The only time the Phantom talks during the 1929 Version

The 1929 version has a synchronized sound track throughout, with the sound being stored on phonograph vinyl discs.  These discs survive and give us perhaps our only idea of what these people sounded like.  (Chaney can be heard in his only talking picture, 1930's The Unholy Three and Carewe can be heard in 1932's Doctor X).  However, the disks do not sync with the available film elements from the 1929 version.  There are about eight distinct dialogue sequences spread out among the ten disks, and four of those disks (6, 8-10) have only music and effects, no dialogue.  The music score is compiled almost solely from stock music pieces and cues and plays almost throughout the picture.  Sound effects can be heard, as can unscripted chatter.  The scene with Florine Papillon, the ballet dancer and Joseph Buquet is dubbed as is Christine's singing voice.  The Faust opera scenes have been recorded with the singers.

Universal took from alternate takes of the footage shot in 1925 to keep the need to film new sequences to a minimum.  Other than the new talking scenes, the footage is almost completely from 1925.  Intertitles are still used whenever there is a non-talking scene, and these seemed to be newly created for the 1929 version.  The version released to theaters cut out some scenes in the original and the running time was reduced to about 93 minutes.  The 1929 version was successful and was released in early 1930 domestically and overseas, including, this time, the United Kingdom.  Lon Chaney would be dead by the end of the year and so was unable to star in Universal's first supernatural horror film, Dracula.  So what happened to this version?  The conventional wisdom used to be that a 35mm print of the 1929 version was struck for the George Eastman House in 1950, but the sound disks do not sync up to the print except in some isolated scenes.  It has been proposed that the print was a silent version for movie theaters which had not converted to sound.

The Curious Man with a Lantern as he appears in the print of the 1925 Version

This apparently makes sense until one encounters the Man with a Lantern sequence at the very beginning of the print.  Just after the titles a man with a lantern is shown walking in the Opera cellars and is clearly speaking, but no intertitles are shown.  Apparently he provides a spoken introduction to the film.  If this were a silent film, the Man with a Lantern would serve no purpose.  The second proposal is that this is an International Version intended for overseas and the Man with a Lantern was inserted so theaters could advertise it as a talkie (which would have been really stretching the truth).  Presumably the Man with a Lantern would be overdubbed in whatever language was necessary.  I would suggest that rather than go to the trouble of producting a domestic sound, domestic silent, international sound and international silent of the 1929 version, Universal probably used virtually the same silent print for the basis for all but the domestic sound version.  The Man with a Lantern could be deleted for any theater without sound equipment overseas.

I bought the Milestone 2003 "Ultimate Edition" DVD release (distributed by Image Entertainment), which is now out of print.  It contains both the 1929 and 1925 versions, each on a separate DVD.  This is a great DVD package with only one flaw : its1929 version suffers from horrible motion judder and ghosting/blur.  The 1929 version on the disc was sourced from a Photoplay Productions 1996 restoration for Channel Four in the U.K.  The resulting video master would have been in the PAL format, and this is what Milestone (a U.S. company) chose to work with.  At some point, they had to convert it to NTSC for U.S. TVs.  A poor conversion is one possibility for the motion issues.  The other possibility is that the technology used slow down the film when the PAL (50 fields per second) master was released caused issues with the PAL-to-NTSC (59.94 fields per second) conversion technology was applied.

Otherwise the Milestone DVDs contain a wealth of ads and promotional material, two trailers, photo reconstructions of the Los Angeles Preview and San Francisco Premier versions, and has the sound disks synchronized as best as they could to the 1929 print.  It also has a great commentary by Scott MacQueen and interviews with Carla Laemmle (Uncle Carl's niece who played the prima ballerina as Rebecca Laemmle) and cameraman Charles Van Enger.  As far as the scores go, I do not care too much for Carl Davis' on the 1929 version, but I really appreciate Jon C. Mirsalis' on the 1925 version.  The 1929 print includes the surviving color sequence for the masked ball (until the Apollo's Lyre scene), and computer colorizes the Phantom on Apollo's Lyre to recreate the original Handschiegl process and the brief sequences at the ball after the scene.  Apparently, the existing 35mm print of the 1929 version had serious nitrate decomposition issues in the scene where Christine wakes up in the swan bed in the Phantom's layer.  The DVD inserted 16mm footage from 1925 and tint it for those frames of the scene that are too damaged.  Image released an early DVD in 1997 which is reputed to have better image quality for the 1929 feature than the Milestone DVD and contains excepts from the1925 version but is otherwise rather barebones (and a port of the 1995 laserdisc release).

Seen in the 1929 Version, but not from the 35mm print

Image has just released a Phantom Blu-ray with both the 1929 and 1925 versions.  The 1929 version is shown in separate 24fps and a 20fps versions.  The 1925 version is tinted, (apparently by comptuer) and has a score by Frederick Hodges.  The 1929 version offers three scores, one by Gabriel Thibaudeau (also in its 1997 DVD), one by the Alloy Orchestra and most importantly in my opinion, the old organ score by Gaylord Carter.  I first encountered Phantom when A&E (back when they actually showed movies and TV shows with artistic merit, not the trashy reality garbage they show today) showed it in 1988 on TV.  They used a print that had been circulating from Paul Killiam since the early 1970s which contained Carter's score.  The score is often moody and eerie with a dreamlike quality which fits the picture very well.  It also sounds authentic since many theaters would be equipped with an organ.   There are audio and menu issues with the disc and the discs which have been sold before the corrected discs are re-released may have to be sent back to be replaced.  Milestone has also announced a Blu-ray release for 2012.

Many issues regarding the Phantom as it currently exists :

Cast Credits for the 1925 Version

Cast Credits for the 1929 Version

1.  The opening credits of the 1925 and 1929 versions are identical for the title card, stars and director/copyright.  Then they diverge, with the 1925 showing an intertitle on a draped curtain background "Produced in its entirety at Universal City, California."  Next, they both show the cast/players, with the 1925 version on the draped curtain intertitle and the 1929 superimposed over the opening shot of the opera cellar.  After two intertitles on the 1925 version introducing Gaston Leroux and the story, both versions show the Man with a Lantern.  Nowhere in the available information for the 1925 version mentions the Man with a Lantern.  Moreover the sequence in the 1925 version is very short (11 seconds) and different angles are used.  By contrast, the 1929 version the Man with a Lantern is clearly talking to the audience (the sequence is 1 minute, 50 seconds long) and the phantom's shadow can be seen skulking in the background.

My suggestion is that the snippet of the Man with a Lantern sequence which is found in the 1925 prints was put there in error by some unidentified party.  I also believe that the title sequence on the 1925 version may have been taken from a 1929 print, probably because the original titles were in bad shape.  Otherwise, why have the cast on an intertitle card on the 1925 version and yet have them superimposed over the scene on the 1929 version?  On the 1929 version, the superimposition of the cast is seamless.

2.  In the 1929 version, what is the Man with a Lantern saying?  Phillip J. Riley, in his excellent book on the film, volume 1 of the MagicImage Filmbooks series, says that there is a transcription of a prologue in the Library of Congress that may have served as the spoken text for this scene.  He also had lip readers try to make out what the actor is saying, and one  person said he could make out Gaston Leroux and another thought the actor was speaking in German.  You can buy the book here : http://www.amazon.com/exec/obidos/ASIN/1882127331/ref=nosim/silenerafilmsond and I recommend it for any Phantom enthusiast.

3.  In a 16mm print of the 1929 version that I was able to view, the end cast credits are identical to the ones on the 1925 version.  This includes Virginia Pearson's credit as Carlotta.  I doubt that Universal made that mistake.  Milestone's DVD release does not contain the Man with a Lantern sequence in its presentation of the 1929 version.  I assume that Photoplay edited it out for showing on Channel Four, since the sequence is worthless to the uninformed.  Killiam's release with the Gaylord Carter score also has it, but when I saw it on A&E in the late 80s, the sequence was missing.  (Although I was about 11 at the time, I surely would have remembered something like that and taped it and watched it several times over the years.)  Image's releases do contain it, as does Milestone's presentation of the 1925 Version.  However, Milestone may have edited it out because there is no soundtrack disk for that sequence.  Without a soundtrack disk, the inclusion of any music from the existing disks may not be accurate.

The Phantom's handwriting improves with age

4.  In the 1929 version that is available, which sequences are from the shooting in 1925 and which are from the 1929 filming?  Most, if not all of the intertitles, including letters and cards were redone for 1929.  All of Chaney's sequences are from 1925, as are the crowd scenes.  The Technicolor scenes from the masked ball must also be original, as they were hugely expensive to film.  The Man with a Lantern sequence is obviously from 1929, as is that portion of the opening credits which lists "Carlotta's Mother".  The only obvious 1929 sequence is the scene with Mary Fabian as the new Carlotta.  Riley and others say that the scene where Christine tells Raoul about the Phantom under Apollo's Lyre is also new, but I am unconvinced.

First, the actors do not look as if they have aged 5 years.  Second, the surviving sound where they talk about the Phantom is does not come close to matching the sequence as it exists.  Third, the shots with Philbin and Kerry are virtually identical to the 1925 print.

Kerry and Philbin on the Roof of the Opera - Do they just age well?

5. In the existing 1929 version, Mary Fabian's shots seem to be the only ones that are new, the shots of Raoul, Christine, the managers and the Phantom all seem to be from 1925.  Assuming that Fabian's scenes are the only ones in the film proper that are from 1929 and shot at sound speed (24fps), what effect would this have had on the rest of the print as projected back in 1929 and 1930?  Since a relatively small portion of the print was shot at sound speed, it would be logical to project the film at silent speeds.  However, if the 1929 print was intended to be some kind of part-talkie, then it would have to be projected at sound speeds.  Did projectors designed for synchronized sound have speed settings, or were they strictly 24fps?  I would think they were adjustable, as not all films in 1927-29 would have been made available in sound.  Which brings me to my next issue :

Mary Fabian as Carlotta, in the only sequence which must have been shot at 24 frames per second

Virginia Pearson as the Original Carlotta (and restyled as Carlotta's Mother in the 1929 Version)

6.  What was the intended projection speed of the 1925 version?  The standard camera speed of silent films was 16fps, but in practice the actual speed varied widely.  Intended projection speeds gradually increased as the silent film era progressed, and 24fps would have been a reasonable speed for some later silent films.  Phantom is not one of those films.  On the other hand, a 16fps projection speed would be ridiculous.  The Image Blu-ray has a version projected at 20fps, which is probably close to the original intent.  Riley reports that the ideal film speed was 14 minutes per reel, which at 1,000 feet of film per reel gives us a frame rate of 19 frames per second.  He also reports that due to the editing, some reels were at 1,024 feet and others at 800 feet, so there was something less than an ideal standard.  20 frame per second may not be absolutely perfect, but it is an excellent choice for 1080i, which can support a field rate on Blu-ray up to 59.94 fields per second.

The Only Survivng Color Sequence, shot in Process 2 Technicolor

7.  What sequences were in color for 1925 and 1929?  First, lets start with the 1925 version.  All that truly survives today is the masked ball sequence.  We know for sure that the Handschiegl process was used for the Phantom's cape and hat as he spies on Christine and Raoul while perched on Apollo's Lyre.  The process may have also been used to color the letters in the Phantom's notes in red.  The contemporary trade journal Harrison's Reports reported that the picture had 17 minutes of color film.  It is known that the unmasking scene was filmed in color and black and white, and Chaney insisted on the black and white version in 1925 because the heat of the studio lights exposed the edge of his bald cap in the print.  The opera and ballet scenes were shot in color and black and white, although how much color footage of this variety survived to the New York General Release Version is unknown.  Establishing shots of the grand staircase also had a color version.  Finally, I read that the ending chase sequence, which was added for the San Francisco World Premiere Version, was also shot in color.  I doubt this because that sequence is rather undercranked and the color technology of the day was noted to have slow shutter speeds and required lots of lighting and relatively long exposure times.  Certain sequences were filmed with the outdated Prizma color system, but they apparently systematically trimmed from the 1925 releases because the color looked muted compared to Technicolor sequences.  The only sequence that may have been shot in Prizma and survived in the 1925 New York General Release Version are the scenes where Raoul and Phillipe appear at the opera in their military uniforms (Soldier's Night at the opera).

The Phantom's Costume Computer Colorized to Recreate the Handschiegl Process (film is tinted blue)

8.  Why is the aspect ratio so thin on the 1929 version?  The original aspect ratio for silent films was 1.33:1.  1.37:1 was the Academy Ratio for sound films after 1932.  The 1929 version in its talkie version did not use an optical soundtrack, so no space had to be taken up on one side of the print for the track.  So, in 1929 the prints being projected would have been in the ordinary 1.33:1 format.   However, when the 1929 version was being duplicated for the George Eastman House, it was duplicated with machines designed for sound.  Sound films of the time (circa 1948-50) had a monaural optical sound track on the left edge which reduced the size of the image, which had to be squeezed into the remaining area of the film stock.  Silent pictures did not have this problem, and their frame size is larger than sound pictures from 1932 onwards shot in the Academy Ratio.  If you review this discussion of Super 35mm (where the gray area in the illustration represents the optical track), you should get the idea : http://en.wikipedia.org/wiki/Super_35.  So, when they made a duplicate print of Phantom in 1948-1950, the printing machine masked the left side of the image where the optical track should be.  This is why many silent films are thinner today than they would have been back in their time.   The resulting aspect ratio is anywhere from 1.15:1 to 1.20:1.

9.  What about the Masked Ball?  The Eastman House Print has a black and white or tinted masked ball sequence.  I remember seeing the sequence tinted on the Paul Killiam release of the film (1929 version) which was done in the early 1970s.  The Image and Milestone DVDs and Blu-rays will have the sequence in Technicolor.  Film Archivist David Sheperd found most of the Masked Ball sequence in Technicolor in the 1970s and it has been added to later releases.  Unlike the material from the George Eastman House, this was not copied on color equipment, so it should show the full 1.33:1 image.

10.  Does the Original Score to the 1925 version survive?   In his commentary, Scott MacQueen identifies two scores to the 1925 version, each by a different composer.  He notes that each had different music cues.  In 85 years no one has advertised that they were playing the original score, so preumably MacQueen was reading Universal production notes and memos describing the scores.  It is wholly unknown whether any silent version from 1929 would have had a score sent with it, as the 1925 scores were far longer.  Silent film scores seem to have a much lower survival rate than the films themselves.  The original score found on the Kino' International's Ultimate DVD Edition of Nosferatu required many an educated guess based on the composer's notebooks.  See here : http://www.gilliananderson.it/film0598.html?IDPellicola=20.  I highly recommend that release, which you can find here :  http://www.amazon.com/Nosferatu-Ultimate-Two-Disc-Max-Schreck/dp/B000VUQ4HW/ref=sr_1_1?s=movies-tv&ie=UTF8&qid=1320687866&sr=1-1.  A reconstruction of the original scores for Phantom in the same vein may or may not be possible depending on the quantity of material which has survived that describes the score.

The beginning of the 1925 Trailer

The (possibly generic) beginning of the 1929 Trailer

11.  What do the trailers tell us?  A trailer survives for the 1925 version and the 1929 version.  The 1929 trailer has a synchronized soundtrack on the Milestone DVD and is tinted yellow for the most part.  I am uncertain whether a unique disk exists for it or whether the DVD authors took it from the disks for the 1929 talkie version.  (It is not impossible that the film has an optical soundtrack.)  It begins with an intro telling the audiences that they will hear as well as see the picture.  Then it segues into the footage shot for the film itself, much of which comes from the 1925 version.  It appears to be in 35mm and runs 1 minute, 32 seconds.  The 1925 trailer seems to be missing the first few frames, including the those which give the title of the film.  The 1929 version's title seems to be the same except there is no hooded man gesticulating over the Opera House. It ends with a second showing the crowds who stood in the pouring rain to see the film at the Astor Theatre.  It seems to end extremely abruptly.  It lasts for 2 minutes, 44 seconds, is not tinted and may be in 35mm.  It may be only film footage taken of Studio President Carl Laemmle.

John Sainpolis as Philippe De Chagny - Raoul's Older, Not-Quite-Smarter Brother (but what happened to the Martindale footage?)

Virginia Pearson as Carlotta or Carlotta's Mother, depending on which version you are watching 

12.  What about the missing 1929 footage?  Those scenes where people talk to each other are lost.  This includes scenes with the managers, Madame Giry, Faust, Christine, Raoul, Phillipe and Carlotta ('s Mother).  I am not aware of any 16mm prints of the 1929 version floating around which may have the footage.  While there has to be some dubbing in the 1929 version, audiences would likely complain if the whole picture was overdubbed.  They would complain that the actor's lip movements would not match the projected image in any way and likely feel cheated.  Reviewers would certainly pan the 1929 version as cheap.  All those scenes, in which the actors appear older, are presumably lost.  Edward Martindale's contributions, as the actor who replaced John Sainpolis as Phillipe (except for scenes in long shot), are unknown outside his voice on a disk.  It is also impossible to determine how the actors aged between 1924/25 when their original scenes were shot and 1929.

The rest of the credits, common to both versions

13.  Why do the available versions have so few credits?  Both versions have few credits, and the only people credited outside the cast are Julian and Laemmle.  The cast of thousands is reduced to eight (1925 version) or nine (1929 version).  Apparently, both the Los Angeles Preview and San Francisco Premiere Versions had much more substantial credits.  Among those credited were more cast members, cameramen, writers, set designers and the like.  I would not be surprised if a frame or two with this information is missing from the 1925 version that exists today.  As the 1929 version was a very pared down version of the story, the credits may have been lost in the pruning.

In closing, I will leave you with stills from the iconic unmasking scene(s) :

The Unmasking Scene, Note the Slight Difference in Camera Angles

Thats all I have to say on this subject at the present time.

IBM PC Cards I Own or Have Owned

When considering my career in vintage computer collecting, I focused more on expansion cards than systems.  I had a hunger to acquire IBM PC ISA and to a lesser extent PCI / AGP cards.  I recently thought about what I have acquired in my career, what I have kept and what I have lost or sold/traded away.

Graphics Cards :

My early graphics card collection is among the best in terms of its comprehensiveness.  I have the following :

IBM Monochrome and Printer Display Adapter

It came with some system, and mine has a larger black bracket for IBM PC 5150 slots, so it will not fit well in other systems.  I have used it with the next card for dual monitor action.  However, only a purist would use it over a Hercules Card.  I also have an IBM 5151 Monochrome Display, which has moderate burn-in.  You can actually see images fade on the monitor, which is utterly unique.  Its parallel interface, like the standalone parallel card, can easily be modified into a PS/2-compatible bidirectional port.

IBM Color/Graphics Display Adapter

I have two of these cards, one of which I somewhat clumsily installed a pair of pin headers so I could choose the thin font.  I think it did something to worsen the video quality.  However, unless you only play text adventures, this card is an absolute must for gaming on an IBM PC/XT.  Other cards are not necessarily compatible or the speed becomes unacceptable.  I am also of the opinion that 256KB is an acceptable amount of RAM for any CGA game that does not support a superior graphics adapter.  There is no substitute for the IBM card, I also have an Epson CGA card that fails certain of Trixter's PC compatibility tests.  One huge advantage that this card has is that it can display color through its composite RCA jack.  While DOSBox, MESS and PCe Emulator can display 640x200 graphics as composite artifact color, they cannot do the same for 320x200 graphics, which usually display in color on a composite monitor or TV.  Games exist that take advantage of this functionality.

Hercules Graphics Card

This is a long card with no additional support.  Hercules later marketed a Plus card with support for user replacable text fonts and an InColor card with 320x200x16 color support.  Some games support the latter, but also support EGA with one exception (Karateka).  I had two of these cards, but sold one with a utilities disk from Hercules.  Unfortunately, I did not have a 5.25" drive at the time, so I could not image the disk.  I do not know if there is a software setting to turn the card into half-graphics mode, which only uses 32KB instead of the card's 64KB.  There is no jumper on the board for this.  The graphics tended to be a bit slow on IBM PC and XTs, however some games use the monochrome graphics well.  Look at Sierra's AGI games for example.

IBM Enhanced Graphics Adapter

Not only do I have the base adapter, but also the RAM expansion daughterboard for a full 256KB.  I have no desire to obtain a Professional Graphics Adapter, as no game I am aware supported it.  The RCA jacks go directly to the expansion connector, and the functions of the switches are not immediately obvious.  Unfortunately I have never owned a 350-line color TTL monitor, so the 640x350 mode is beyond my reach.  This adapter will work just fine with my 5151 and 5153 monitors if the switches are properly set.

IBM PS/2 Display Adapter

This is IBM's only 8-bit VGA graphics card and probably its only ISA card with its own VGA chipset (as opposed to another manufacturer).  It is a full length card with two rows of pin headers for some unknown purpose and a VGA feature connector.  I am concerned that its EPROM may eventually die, but I dumped it just in case.  It will even work in a Pentium II/III system with ISA slots, even if the system beeps.  It will also work fine in an IBM PC 5150.   It was designed to upgrade an IBM PS/2 Model 30 from MCGA to VGA, but it can work in many other systems.  (It would not fit in a PS/2 Model 25).

Unusually, it only uses 24KB of a 32KB EPROM, which is mapped from C0000-C5FFF.  It also has 8KB of sratchpad RAM, but this is mapped in a very weird way.  6KB of it is mapped from C6800-C7FFF and the remaining 2KB is at CA000-CA7FF.  There are memory holes in between, so if you are using a card with an External ROM, make sure it does not start at C8000 if you are using this card.

My later acquisitions include :

Diamond Monster 3D Voodoo Graphics

I remember owning one of these back in the day, so I bought another for those early DOS/Win 9x games which do not work properly with a a later generation Voodoo 2 card.  One day the card refused to display graphics, just some white lines on a black screen.  Eventually I threw it away.

Diamond Monster 3D II 12MB Voodoo 2

I remember replacing my Voodoo Graphics with a Voodoo 2 back in the day, but I do not believe I ever had two cards for an SLI configuration.  I am sure I replaced it by 2000 for a Geforce 256.

3dfx Voodoo 5 5500 AGP

This card was a replacement for the MAC card, but eventually it could no longer display graphical modes without severe corruption, so it too went into the trash.

3dfx Voodoo 5 5500 MAC PCI

I bought this card on eBay from a guy in China, and noticed that the card's faceplate was very rusty.  Once I flashed the card with a PC BIOS, it worked well, even its DVI connector.  Eventually the VGA output would not display the color green, and an unknown component in a set of three looked damaged, so I eventually threw it away.


IBM Cards :

IBM Printer Adapter

This card would have been used to add LPT1 to a system with a CGA card or LPT2 to a system with an MDA card.  The usual address is 0x378h, but it can be hacked and I did hack it to be selectable to 0x278h.

IBM Asynchronous Communications Adapter

This card could reliably handle null-modem transfers at 9600 baud, whereas the UART on my AST Six Pak Plus could only do 4800 baud null-modem transfers.  Has one jumper DIP block to select COM1 or COM2, another to select TTL or current loop communication.  My card has a jumper to work in the IBM XT's slot 8.  

IBM Game Control Adapter

This card may seem useless, as it has needs a Y-splitter for two joysticks and has no speed adjustment, but its useful to have a card, the compatibility of which, is assured.  

IBM Diskette Drive Adapter

Has a card edge and can support 5.25" or 3.5" double density drives.  If you need a custom cable because you installed a 3.5" drive with only a pin connector, you should be able to squeeze on an extra connector to a cable easily.

AST Six Pak Plus

This card came with my IBM PC 5150, and it can complete a PC if everything is properly installed.  IBM even marketed it in some of their late brochures.  It can add 384KB of RAM to the PC's Motherboard's 256KB for the full 640KB.  It can also add a serial, parallel and game port.  Each of these ports can be disabled.  The parallel port requires a DB-25 female header and the game port a DA-15 female header, a N558 Quad Timer and a 74LS244 chip.  The serial port has a socketed UART.

Sound Cards :

Adlib Music Synthesizer Card

There are two versions of this card, the 1987 version and the 1990 version.  Other than an extra capacitor or two on the 1990 version, the only difference is that the 1987 version uses a 1/4" TRS output jack and the 1990 version uses a  3.5mm mini-jack.  Output is mono, and the silkscreening on the YM-3812 and Y3014 OLP2 chip and DAC is scratched out on my 1990 version card, although by that time the secret of what chip Adlib was using was out.

IBM Music Feature Card

I originally purchased this full length card for a hefty sum on ebay.  It came with a midi breakout box, which I acquired separately a year later.  Later I traded it for something, a trade I occasionally regret.  Only Sierra On-Line ever supported it in games, but they supported it for a four year period (1988-1991).  It is a combination of a Yamaha FB-01 midi music synthesizer (using 4-op FM synthesis) and unique IBM midi interface.  I put it in a PC and used the breakout box to try and replicate an FB-01, but DOSBox would not produce the correct sounds.  That was about 3 years ago, I wonder if things have changed.  I am currently looking for a Yamaha FB-01 to try drivers that it with a Roland MPU-401 interface.

Roland MPU-IPC-A + MPU-401

As I have stated previously in this blog, the MPU-IPC-A is merely a small logic card, and the MPU-401 is the external box where all the midi commands and data is processed.  It makes no music unless attached to a midi synthesizer, whether a keyboard or a module.

Roland LAPC-I

I bought this, with its MCB-1 midi breakout box, from a seller for $25 + shipping.  He did not know if it worked and this was a risk.  This was back in 2006 or so before the price of an LAPC-I skyrocketed.  I could put it in a PC and use it as an external synthesizer in DOSBox, but I needed a program that allowed sysex to pass through the midi interface to the synthesizer.  The interface by default would block sysex, which would eliminate the synthesizer's ability to receive custom sounds from a game.  I traded it for something good after I had acquired a CM-32L, which has the exact same synthesizer capabilities.  My card had ROM v1.02, not EPROM, which was v1.00.

Creative Labs Game Blaster

My most recent acquisition came as a part of a trade for a Tandy 1000 TL.  It came in its retail box with driver disks on 5.25" and 3.5" disks and the Sierra game Silpheed, also on both disk formats.   I wanted this because there are games which I have confirmed will not work with a Sound Blaster with C/MS chips, they obviously are looking for something inside that big CT-1302 chip.  The next widely-available CL card using RCA jacks would be the AWE64 Gold.  It is a stereo card, but that was probably the only thing for which it was known.

Creative Labs Sound Blaster 1.5 w/CMS Upgrade

The earliest Sound Blaster cards came with a v1.xx DSP, but mine came with a v2.00 DSP.  I added the C/MS chips, which are Phillps SAA-1099s.  That is the only part of the card in stereo.  With the v2.00 DSP (which adds auto-DMA support among other things), I have cajoled Trixter's 8088 Corruption demo to work with the card.  I read that the v2.00 DSP was necessary for MPC-1 compliance with Windows 3.x multimedia features.  I may keep it around only to check whether C/MS games will work with the card, but DOSBox now supports the Game Blaster.

Creative Labs Sound Blaster 2.0

I sold this card because I could not upgrade it with C/MS chips because it uses an extra PAL chip, a 16L8N, the programing for which nobody has ever been able to replicate.  It would have worked as well as any Pro in any game except for stereo and mixer support.  Its abilities are firmly encompassed by other cards.

Creative Labs Sound Blaster Pro 1.0

I kept this card because there are games that support its dual OPL2 chipset.  It is also necessary if you wish to use a Sound Blaster with any Tandy system with a PSSJ sound chip.  The PSSJ only works with DMA1 and if a Sound Blaster is also set to DMA1, games will freeze when playing sounds.

Creative Labs Sound Blaster Pro 2.0

Although I sold this, it has one big advantage over the Pro 1.0.  The OPL3 chip on this board is not not particularly sensitive to system speed, whereas the OPL2 chip is (like the game port and a rev 0 MT-32).  If you run an older game on say a Pentium system, the game may send the data to the OPL2 chip so fast that it cannot process all the data, and the music will be incorrect.  However, the Sound Blaster 16s I have also have true Yamaha OPL3 chips.  Although Windows 9x does support the Sound Blaster and Pros, their 8-bit limitations will show themselves in garbled 16-bit audio playback.

Creative Labs Sound Blaster 16 MCD ASP & Creative Labs Sound Blaster 16 SCSI-2 ASP

This card is the bare minimum which modern programs can use, since it can do stereo 16-bit playback at 44,100kHz.  I originally got the SCSI-2 CT-1770, but I did not like the fact that the SCSI interface used an extra high IRQ and eventually traded it.  I acquired a MCD CT-1760, and the proprietary CD-interfaces on that card can have their IRQ usage disabled.  Both my cards have a soldered ASP chip (TFX uses it) and use DSP 4.05 for error free midi playback through the waveblaster or the external midi out.  The waveblaster port does no favors for the output of a midi daughterboard.

Creative Labs Sound Blaster AWE32

When I bought this card, a CT-2760 rev. 3, I soon discovered that I had no case in which to put it.  My Pentium II/III system uses a modern case with a hard drive cage going all the way down the case.  This blocked the card, which is a full-length 13" ISA card.  Also, the plastic tabs that held the SIMMs in place were broken.  Gravis used much more durable metal clips in their PnP.  I only used it once or twice by removing the motherboard from the case, which was unwieldy to say the least.  I eventually donated it to a friend of mine who was interested in the rev. 3.

Creative Labs Sound Blaster AWE64 Gold

This card was superior to the 16s and the AWE32s in almost every way.  I did not care about the lack of SIMM slots or waveblaster connector.  However, ISA PnP is no fun and its FM Synthesis is lacking in quality.  Unlike the early AWE cards, which used genuine integrated Yamaha OPL3 core, Creative was using its CQM substitute by this time.  Although the sound is close, it is not the same and usually sounds harsh.

Gravis Ultrasound ACE

This card I picked because it had perfect GUS playback capabilities.  My board came with driver disks and thick manual.  It also came with the maximum 1MB of RAM.  It has one soldered SOJ chip and one socket for an additional SOJ chip.  Unlike an original GUS or GUS MAX, it does not support a daughterboard for 16-bit recording, but I have no interest in recording anyway with an ISA card.  Its Adlib emulation, which uses I/O 388/389h can be disabled, which eliminates conflicts between the real thing.  It does not have a joystick/midi to further reduce configuration headaches.  Unfortunately, this is the 1.0 version, which means the output jacks have reversed stereo.


Gravis Ultrasound PnP

This card embodies the true 2nd Generation of Ultrasound sampling technology.  Unfortunately, games did not really support the extra features of this card.  Mine (v1.0) does not have any RAM soldered on the board but has both SIMM slots populated.  2MB of RAM is on my card, which ensures full Ultrasound compatibility and not much else.  I prefer the ACE over this card because this is a dreaded PnP card and its RAM usage is far from unique in Windows 9x.

Roland SCD-15

This is the official name of the marketed combination of a Roland MPU-401/AT ISA MIDI Interface Card + Roland SCB-55 daughtercard.  It has these mini-DIN connectors which were always hard to find.  Creative Labs had mini-DIN connectors for their Live! and Audigy breakout boxes which worked, but their support site did not always have them in stock.  The daughterboard was almost as big as the card itself and the card hung off it.  In fact, although the daughterboard had four standoffs, only two actually connected to the board.  This did not seem like a stable long-term solution, so I traded it away.

Yamaha DB50XG

I also got one of these, although the DB60XG is more common on eBay.  The former is a retail product, while the latter is an OEM product.  The former has the advantage of at least partial support for Yamaha QG300 synthesis custom voices.  The Fat Man really expoused the virtues of this card, and some games either supported it or were attuned to sound good with it.  I avoided the SW60XG because it had no external midi port.  When I finally acquired a MU10XG, I traded this away.  

Yamaha YMF-724/744/754 Cards

I do not talk about PCI cards much here, but these were a very good solution for backwards compatibility with the Sound Blaster.  Most late ISA and PCI card emulated the Adlib OPL2/3 FM synthesis poorly or inaccurately, but these cards came from Yamaha and incorporated a true OPL2/3 core into the chip.  They also supported the PC/PCI connector found on some TX/LX/BX motherboards or D-DMA for TSR-less digital Sound Blaster compatibility.  Finally, some cards also supported S/PDIF output for crystal clear sound.  Pure FM recording with these cards is quite possible.  I have not tested it, but 4-speaker sound output is available in 744 and 754 cards.  But there are some drawbacks :

While the digital sound blaster emulation is good, it is not perfect and only goes up to an SB Pro.  Fortunately, very few games require a Sound Blaster 16 or better. Fallout for DOS had broken SB/SB Pro drivers, but I used drivers from another other Interplay game to get the SB/SB Pro sound working again.  It will not emulate Sound Blaster ADPCM 8bit-3bit and 8-bit-2bit modes, which Duke Nukem II among others use for some sound effects.  Since games use direct I/O access for the Sound Blaster and Adlib, the card may not work in Windows XP or other NT machine.  The card supports DirectSound and DirectSound 3D and emulates EAX 1.0 through Sensura, but the surround sound causes system performance issues.  Finally, most motherboards for the Pentium/II/III have at least one ISA slot, so why not use a true ISA Sound Blaster?

Aureal SQ2500

The ultimate card for A3D support, I picked it because it represented the last card to support a widely used but eclipsed technology.  No other card except the Aureal AU8830 supported A3D 2.0, other cards only went up to 1.0.  It supported 4-speaker output, but the rear speakers were not as widely used as the front speakers.  Games that support A3D 2.0 include Half-Life, Descent 3, Unreal and early versions of Quake III.  It also has a waveblaster header.

MIDI Modules :

Roland MT-32

My MT-32 was something of a late purchase as I did not fully understand the need for one or the unique character of the LCD display.  It is very convenient to be able to reset the module by pressing Master Volume and R at the same time.   Other modules require a shutdown or sending a reset command via midi.  Viewing messages on the LCD which games display is always neat.  Mine is a rev 0 ROM v1.07, which is the last ROM version before the 2.x versions, exclusive to rev. 1 boards.  The MT-32 works great in DOSBox, which can easily adjust transmission speed to be slow enough for an MT-32.  I would say 3,500 cycles is the limit if the game is transmitting custom patches.  Since I am a big fan of Sierra games, I want to know how these games sounded, and some of them exploited bugs of the rev. 0 boards.

Roland CM-32L

Being unhappy with my CM-64 and CM-500, I turned to this, simpler model.  No slot, no mode switch, only an on button.  Necessary for games that causes errors on the MT-32 regardless of speed or use the extra sound effects of the rhythm/percussion part.  

Roland CM-64

After learning of the CM-500's vibrato issue, I turned to this module, a true combination of CM-32L and CM-32P.  Unfortunately another issue reared its head.  Sierra's games music synthesis engine broadcasted  MT-32 data on midi channels 2-10 and Adlib data on channels 11-16.  The MT-32/MT-100/LAPC-1 and CM-32L did not care, as they did not use channels 11-16.  The CM-32P does, and wrong sounds would constantly be heard.  Later Sierra drivers eliminate the issue, but they will not work with all games.  Also, this may occur in other games, although this is unlikely.  Although the CM-32P was supported in some Japanese NEC-98xx and X68000 games, my primary interest is DOS, so a better solution was found in a simple CM-32L, and this got traded.



Roland CM-500

When I was first collecting vintage hardware, this module was seen as the Holy Grail of Roland LA Synthesis and expensive and rare even then (2005).  It does support the Roland SC-55 GS (and later General MIDI) synthesis engine and the Roland CM-64 (emulating the CM-32P).  Unfortunately, not only does it share the same issue as the CM-64, it was pointed out that it had annoying Vibrato.  So it got traded.

Yamaha MU10XG

This external synthesizer used the same synthesis engine as the DB50XG and SW60XG, so I knew it was a quality card.  It was also hard to find, I guess it was not very popular.  Unusual for an external module, it has a battery compartment.  It also has two 1/4" audio input jacks to which the module can apply reverb and other effects.  It requires a +12v adapter, so I used an adapter from something completely different that fit.  It has not gotten much use because DOS games generally composed for Roland LA or GM.

Roland SC-55

This was a relatively recent purchase.  It came with a remote control, which I have somewhere.  Like the MT-32, it has a display and there are games that take advantage of it (Lands of Lore).  Mine is a GM/GS module, which came later than the original, GS only modules.  It makes no difference in functionality whether the module supports GM and GS or just GS.  Also, there is the stargame.mid, which uses the equalizer to display graphics.  Its only downside is the 24 voice polyphony, but the quality of the sounds with effects more than makes up for the deficiency.

Roland SC-55ST

I thought this was better than the original SC-55 in every way, but it turns out not to be the case.  The original SC-55 had a Capital Tone Fallback feature that if a game tried to play a variation tone which the module did not have, the module would play the capital tone instead.  Yamaha also used this technology and forced Roland to remove the feature from the 2nd and later generations of Sound Canvases.  Unfortunately, there are games that use this functionality (Might and Magic IV & V).  It gets little use as a consequence, as the SC-55 has much more character.

Tandy 1000SX - The Best Overall Choice for IBM PC Gaming of the 1980s

Back in the 1980s, if you wanted to play games on a computer, there were many incompatible options available.  If you were on a budget, you may have had a Commodore 64.  If you were a real computer geek, you may have appreciated the open system of the Apple II.  Of course, your parents may have made the unfortunate choice of buying a Coleco Adam, Mattel Aquarius, TI-994A, a Commodore PET or TSR-80 or Color Computer.  None of these machines were particularly good for gaming.   You may have had the awesome parents that would have bought an Atari 8-bit computer or the VIC-20, which were great for games and very little else.  But later in the 1980s there were fewer options, in no small part due to the acceptance of the IBM PC in the business world.

In the second half of the 1980s, the 6502-based systems like the C64 and Apple II were really starting to show their age and the Atari machine was virtually dead.  In response, Commodore, Atari and Apple introduced new computers, the Amiga, ST, Macintosh and IIgs respectively.  The first two never caught on in the North American market, the Macintosh suffered from a lack of a color screen and gaming hardware early in its life and the IIgs withered on the vine due to a perceived slow processor and apathy from Apple.

From 1981-1986, the IBM PC line had become a huge success in the business world, but the high prices of the offerings from IBM did not make for much enthusiasm from the home market.  Even in 1987, just before IBM discontinued the line, the most basic configuration of the IBM PC cost $1,165.00 This configuration only came with 64KB of RAM, no floppy or hard disk drive or adapter, no monitor or graphics card, no software, no game controller adapter.  To get a working configuration you would practically have to double the price.  For that year, that was a slow machine.

IBM's one effort to market a machine to the home user during this decade was the PCjr from 1983-1985.  The Jr. failed for reasons that are widely known, namely that people were expecting a fully PC compatible machine and got something significantly less than compatible, had a terrible chicklet keyboard, limited and proprietary expansion options, and was not especially competitive with other home computers at the high price point of $1,269.00.  If the PCjr. could not escape the shadow of its big brother and its own shortcomings, it did have some influence.  First, it was designed to be easier to get up and running.  If you got the expanded model, there was little need to open the machine as everything attached to the back or was bolted onto the side.  It had lots of built in hardware that were extras on a PC.  It had a built-in custom graphics controller that had better capabilities than the Color/Graphics Monitor Adapter found on the PC and could connect to a digital RGB or analog composite color monitor.  Comparable but incompatible graphics were available on the Enhanced Graphics Adapter, which was hugely expensive at the time.  It had a three-voice sound chip in addition to the PC Speaker and could mix external audio from a sidecar and had an audio RCA output.  All the above did not exist on the PC.  It had a serial port and a game port with plugs for two joysticks, which were extra add-in cards on the PC.  Perhaps in part to offset speed issues with RAM shared between the CPU and the Graphics Controller, the machine had two cartridge slots for games and programs.

Tandy Corporation, which operated the Radio Shack chain of stores, had a very active home computer division which had already produced the aging TRS-80 and not-exactly a sales leader Color Computer lines. Tandy had designed an MS-DOS machine called the Tandy 2000, which was at best semi-compatible with the PC and was hardly a best-seller.  The home market was crying out for affordable PC-compatible machines, and Tandy liked the PCjr.'s features and thought they could do better at the price point.  They developed a machine that supported the PCjr.s features and offered mostly better PC-compatibility.  Unfortunately, in 1985 as the machine was about to be released, IBM discontinued the Jr.  So Tandy began to emphasize MS-DOS and PC compatiblility for the machine, the Tandy 1000.

The Tandy 1000 came with a 8088 running at 4.77MHz, just like a PC, 128KB of RAM (shared with graphics),  the 90-key Keyboard used in the Tandy 2000, two joystick ports using joysticks from the CoCo, the same graphics and sound capabilities as the PCjr., composite video and digital RGB support, a printer port.  Its keyboard and joystick ports were at the front of the unit whereas IBM's were always at the back.  It had a reset button whereas IBM users had to press Ctrl-Alt-Del.  It ditched the cassette and cartridge ports of the Jr., the wireless keyboard and avoided proprietary ports to a greater extent than the Jr.  It came with a 5.25" floppy disk as standard and had the floppy controller circuitry built into the motherboard.  MS-DOS 2.11 and GW-BASIC came with the machine (no separate purchase required) as did DeskMate (Office App).  Unlike the PCjr. with its one drive bay, the 1000 came with two.  Its weaknesses included :

Only 3 XT/8-bit ISA slots

RAM upgrades only by special Tandy upgrade cards took up 1-2 ISA slots to get to 640KB.

No Serial port

No ability to scan for bootable ROMs on expansion cards (no booting hard drives from a disk controller)

No DMA built-in, came on one of the RAM upgrades.  This gave a stock 1000 the same kind of annoyances as the Jr. (CPU servicing disk drive, leading to ignored keystrokes, keystrokes interrupting serial data).

No ability to disable graphics circuitry (no upgrades to EGA or VGA)

No socket for a Math Coprocessor.

The Tandy 1000A added a math coprocessor socket, fixed some bugs and allowed the use of bootable hard disk controller cards (Tandy offered a version called the 1000HD with a 10MB hard drive and a 640KB RAM & DMA upgrade board that came on 1 board instead of 2).  The IRQ for the hard drive was 2 for these Tandys, with IRQ 5 used for the video.  IBM used the opposite designations for the PC line.

The next models in the series, released towards the end of 1986, are the Tandy 1000 EX and 1000 SX.  They are functionally the same machine, but the EX is a compact model with only one drive bay and the need to use Plus card for expansion with room for only three and an external drive for a second disk drive.  The RAM and DMA upgrade comes on a Plus card.

The Tandy 1000 SX, the machine in question fixes virtually all the above issues.  It has 5 XT/8-bit ISA slots, can be upgraded to 640KB of RAM on the motherboard, can support an EGA or VGA card and a hard disk controller and a math coprocessor.  It came with 384KB standard RAM, a DMA chip and used standard 256Kx1 chips to upgrade to 640KB.  It came with MS-DOS 3.2, which added support for 720KB drives.  It could boot from either drive, which is useful if you have bootable 5.25" and 3.5" disks.  It ran is 8088 at 7.16MHz or 4.77MHz for a decent speed boost.  This speed is selectable on bootup for software expecting the slower speed.

Later models in the 1000 series added support for a 286 processor (TX), DOS-in-ROM (HX), 768KB of RAM so graphics memory would not conflict with conventional memory (TX), high density floppy controller (RLX & TL/3), 8-bit IDE interface (TL/2, RL), 3.5" drives as standard (TX & HX), improved graphics with 640x200x16 mode (TL & SL), enhanced sound with Digital-to-Analog and Analog-to-Digital conversion (Tl & SL).

Here is how I have my Tandy 1000 SX configured

Monitor :
Tandy CM-11 High Resolution Color Monitor

Keyboard :
Tandy 1000 Keyboard

Drive Bays :
5.25" 360KB Drive
3.5" 720KB Drive

ISA Slots :

1 - Serial Card
2 - 286xpres Accelerator
3 - Roland MIF-IPC-A w/ Roland MPU-401 & Roland MT-32
4 - Adlib Music Synthesizer Card of Creative Game Blaster Card
5 - ADP-50L 16-bit IDE Controller

I feel no particular need to go 100% Tandy, but the floppy drives come from, if not manufactured by, Tandy.  I used to have a 20MB Tandy Hardcard, but the drive was very noisy and soon died.  So, in my only concession to modern hardware, I have a 1.0GB Compact Flash card mounted instead.   Unfortunately, the ADP card, which supports 16-bit IDE drives on an 8-bit bus, can only utilize 504MB of that due to the infamous limitations of the straight INT 13h/ATA addressing it uses.  This limitation is independent of any DOS limitations.  Compact Flash cards are virtually IDE devices, requiring only a passive pin converter to work with an IDE port.  The ADP card is a bit finicky about which CF drives it will boot, having not been designed for that.  The Board does not use IRQs or DMAs.  Two drives can be used for a total of 1008MB. The ADP board relies on memory mapped I/O, not port mapped I/O, so it should be faster than, say an XT-IDE card.  (The XT-IDE card does allow for use of hard drives at least up to 8GB, however).  I may also be able to use a Trantor T-130B 8-bit SCSI interface card but that requires a separate SCSI-IDE bridge.  On the plus side, it does support and can boot up to 1GB storage devices.

Tandy MS-DOS 3.2, like all DOS versions 3.0-3.3, allows a primary DOS partition of only 32MB.   I am uncertain whether DOS 3.2 supported extended partitions, but 3.3 does.  Tandy MS-DOS 3.2 does support up to three other DOS partitions on the drive (up to 32MB each), but it requires loading a driver in config.sys on startup and seems non-standard.  Ultimately I chose MS-DOS 5.00 for its features (edit mainly) and it uses less conventional memory and hard disk space than DOS 6.22.  Also, while DOS 3.3's fdisk supports an extended partition up to 736MB in size, it has to be divided into 32MB logical drives from D:-Z:.  DOS 4-6 supports a primary partition of up to 2GB, all of which is the C: drive.  So MS-DOS 5.00 with a few additions from Tandy DOS (the Mode command, a device driver or two) works just fine.  If games start complaining about lack of free RAM (only 624KB is available due to sharing with the video), then I may have to go back to 3.3.

What kind of games would be played on this machine?  Well, all the important PC originals from 1981-1990. I will include any game whose maximum resolution was 320x200x16 and supports Tandy graphics or supports the Tandy sound chip for music and/or effects that does not support an Adlib, Sound Blaster, Roland MT-32 or Game Blaster.  Some examples include :

Thexder - Although the game supports an EGA 640x200x16 mode, which is appropriate considering its Japanese origins on the PC-88, it supports Tandy Sound for its music (all of two pieces).

Secret of the Silver Blades, Champions of Krynn & Death Knights of Krynn - All support Adlib music, but if you select Adlib you get PC Speaker sound effects.  Tandy gives better sound effects than the speaker.

Games that require an EGA card and do not support Tandy, like Commander Keen, Duke Nukem, Sorcerian) are not going to be on this machine.  Similarly, games with a VGA version and an EGA/Tandy version aren't going to be either, unless the EGA version came out first (Indy 3, Monkey Island).  Nor would games that support a high resolution color EGA mode (Thexder II - Fire Hawk, SimCity, Zeliard, Silpheed).  In short, if there is no real benefit from running the game on a Tandy instead of an IBM PC or clone, then it has no real use in this machine.

The other hardware should be fairly obvious.  The Sound Blaster will see little use for its digitized sound and none for its midi or gameport capabilities.  Its there because I wanted an Adlib and a Game Blaster in one card.  The Roland is the ideal setup, nothing more need be said about it.  Chiefly only Sierra's games would have put it to any real use (custom sounds).

The 286xpress Accelerator requires a little explanation.  This device is an ISA card that can be used in a Tandy 1000, 1000A or 1000SX.  It has an 80286 CPU running at 7.16MHz and 8KB of cache.  It has a ribbon cable running from the card's daughterboard to the 8088 CPU socket on the motherboard.  It has a spare socket for a 80286 running at 4.77MHz.  The 8088 is plugged into the daughterboard.  This board does nothing unless activated when DOS boots either by a program loaded in autoexec.bat or a device driver in config.sys.  So it will not interfere with PC booter programs expecting 8088 speeds.  You can turn the cache off for better compatibility with programs.  With cache off, the system is just a little faster than the system running at 7.16MHz on the 8088.  With the cache on, it is almost as fast as a Tandy TX or IBM PC AT @ 8MHz.  While an TX provides a serial port (freeing up a precious slot) and CPU upgrades fit into the slot (freeing up another slot), even at its slow speed it should still run faster than the SX at its fast speed.  So for older games that run like they are on amphetamines unless the CPU is an 8088 running at 4.77MHz,  it just isn't as good a choice.

Things to watch out for :

The RCA composite video output tint is off compared with the IBM PC in CGA modes, and the colors displayed, compared to the IBM colors, are off.  For example, in Winnie the Pooh in the Hundred Acre Wood, Pooh appears in brown on an IBM and blue on the Tandy.  You can try to use the tint control on your monitor or TV to adjust the colors, but most TV monitors only allow a shift between red to green, which is not enough to make the colors accurate.  A composite monitor designed for computer use may allow a greater range of adjustment.  I believe this is present throughout the Tandy 1000s that have composite color output.  One game, Indianapolis 500 - The Simulation, by Electronic Arts actually allows the player to select an IBM or a Tandy palette in composite color mode, allowing the proper colors to be displayed with either type of machine.

The RCA audio output is not turned on by default.  (The Tandy 1000/1000A suffers from a similar problem, the EX, HX, TX and later models do not) The default is for all sound to go through the PC Speaker.  Some programs like Maniac Mansion fail to recognize that the SX is a little different and do not turn the audio output on.  A program called tdyspkr.com can be used in your autoexec.bat to turn the output on to the RCA out and/or off to the internal PC Speaker.

The default 80-column text mode on the Tandy annoyingly uses 225 lines.  IBM uses a 200 line text mode.  All graphics modes in these machines use 200 lines.  In order to get a proper aspect ratio with the graphics modes, the vertical size control has to be set almost to the point where the graphics are almost hidden by the monitor bezel.  When you return to the text mode, this winds up cutting off either the top, bottom or both lines of the display.  Use Tandy MS-DOS 3.2 mode.com command : mode 200, to fix this.  This does not work in the TL machines.  You can use the Tandy mode command in a later, non-Tandy DOS.

Tandy floppy disk drives require a ribbon cable without a twist!  They also require setting jumpers or switches on the drives to tell the machine which is drive A and which is drive B.  On an IBM PC, both drives are set to DS1, and the twist turns the drive at the end of the cable to DS0.  If you are using a modern 1.44MB drive in these systems, they are generally soldered to DS1, so they will be the B drive.  5.25" drives should always allow you to set DS0 or DS1.  But one of the great things about a Tandy is that they allow you to boot from the B drive by pressing the F3 key at startup.  If you have a standard "universal" floppy cable (with 3x pin connectors and 2x card connectors), you have to untwist the cable by prying off the end connectors, flipping the wires around, then snapping the connectors back on again.  Tandy did not support high density drives until the RLX and TL/3, but can use most 1.44MB floppy drives with double density/720KB media.

The EX and SX are the last Tandy 1000s that does not use the floppy connector port to power the drives.  The HX, TX and all later Tandys, (inlcuding the SX's replacement, the SL), do run power through the drive cable.  If you are using floppy drives on a Tandy 1000 system that do not correspond to Tandy's power-in-drive cable pinouts, you must cut holes in pins 3, 5, 7, 9, 11 and 29, 31, 33 for any portion of the cable going to the regular drive.  I used an Xacto knife and a pair of clippers to do the trick, and only small holes are needed.  Tandy did this themselves with their cables so an individual could connect a standard 5.25" drive into an empty bay.

In the TL/SL/RL series, there is a way to disable the Tandy joysticks and use an IBM joystick adapter.  On the early Tandys, there is no way, you you have to use Tandy joysticks.  Tandy used the same joysticks for the TSR-80 Color Computer and the 1000 series.   They come in three varieties : the one-button (cheap) black non-centering Joystick, the two-button Kraft-style Deluxe Joystick, and the flight simulator pistol grip joystick (four buttons, only two useable).  You can also use a Color Mouse (one-button) or Deluxe Color Mouse (two-buttons) in a joystick port and a special driver.  The Deluxe Joystick is virtually identical to the Kraft IBM PC joystick or the IBM PCjr. stick, and can be either set to self-centering or or free floating in one or both axes.  There are some subtle differences in the control interfaces between the PC and Tandy joysticks which may manifest themselves in compatibility issues with older games.

Now, while you could use a Tandy Color Computer mouse, this is not an ideal method because of the attention the CPU must spend on the reading the joystick ports.  This can slow the PC down.  A much better alternative is to use a serial mouse, especially as serial mice are easy to find.  So are 8-bit serial cards.  Serial cards use IRQs to get the CPU's attention so excess CPU time is not wasted in polling the serial controller's I/O ports.  You will need to use an 8250B based serial card, which is typically what 8-bit cards use.  I always use ctmouse as a driver as does virtually every other vintage computing enthusiast.  Its compatibility is excellent and it only uses 3K of precious memory.  In the SX there is no serial port, Tandy fixed this in the TX by replacing the useless Light Pen port with a serial port.  (Some games, like the Gold Box games referred to above, do not support using a mouse with Tandy graphics).

Using a printer requires a special adapter cable.  For some reason, probably TSR-80 compatibility, Tandy used card edge printer port on all 1000s until the RLX and TL/3.  The ribbon cable would have a card edge connector on one end and a Centronics port on the other end.  Fortunately, building a compatible cable should not be a problem, especially if you are connecting it to an IBM compatible printer.  One thing to note is that all card edge printer ports are unidrectional and devices that use the port for data transfer in (like a Backpack CD-ROM drive or a parallel port ZIP drive) will not work, even in nybble mode.

Games and programs requiring BASIC may be an issue if they require ROM BASIC.  ROM BASIC is virtually unique to IBM PCs and PS/2s of the 1980s, and even highly regarded compatibles like the Compaq Portable failed to run programs that required it.  Tandy DOS comes with GW-BASIC, which incorporates the BASIC that would otherwise be in ROM and supports the Tandy graphics and sound capabilities.  GW-BASIC may work for some games.

The Tandy 1000SX has two speeds, the default 7.16MHz and the 4.77MHz speed.  The latter speed is for compatibility for PC software that was only intended to be run on an IBM PC or XT.  Pressing F4 at boot will put the computer into slow mode, which is great for PC booters.  The commands mode fast and mode slow can be used in DOS to switch back and forth.  Also useful is F1 for mono mode and F2 for TV mode to kill the color text or to make the text more legible on a TV.

With my Compact Flash card, the only noise from the system is from the Power Supply fan.  The PC Speaker is a large 3" cone, even bigger than the 2.25" cone of the IBM PCs.  PC speaker sound can be output to the RCA audio out.

The Sound Blaster 1.0-2.0 all use DMA1 and do not play nicely with the PSSJ sound chip in the TL/SL/RL, which also uses DMA1.  Lockups and freezes are commonplace.  However, I have a Central Point Software Option Board, which requires DMA1 in a Tandy system (it can share DMA2 in an IBM system).  So if that is installed, the Sound Blaster may not play nicely with it.

The Tandy keyboard is a funny thing.  It has 90 keys, and it is much more compatible with the IBM keyboard layout than it first appears.  Since an IBM 83 and 84 key keyboard used the numberpad for directional keys, use can use the Tandy's if the dedicated cursor keys are not working in the game.  The small shift keys are not much fun, there is no separate * key, the Alt key is in a funny place, and there is no scroll lock equilavent (use Alt Break).  Beware the Hold key, it will seem like the machine has frozen until it is pressed again.  Its nice that the Num and Caps Lock keys light up, which did not on an IBM PC or XT.  The \ key is also less than ideally placed.  The fat enter key is far superior to the IBM 83-key keyboard.  However, the IBM 83-key keyboard is probably the most solid and clicky keyboard I have ever typed on.  By comparison, the Tandy keyboard is mushy and prone to registering a keypress with only the slightest pressure.  It does not use rubber dome keys, it has springs in the keys, but still the action leaves much to be desired.  PC-compatible keyboards do not work on the Tandy 1000s prior to the TL/SL, but there was an adapter from Tandy that would allow a PC keyboard to work.  The Northgate Omnikey keyboards are also compatible with a special cable (which you may have to make yourself).

The SX and TX are the first Tandys that allow their graphics to be disabled if an EGA or VGA board is installed.  (The EX and HX can as well, but require upgrade cards with the PLUS adapter).  There is a special program that improves the detection of these cards in Tandy 1000s.  An MDA or Hercules compatible card should be installable in any Tandy 1000 except the 1000 and 1000A.  You can even install a CGA card in the machine, but some games may detect a Tandy 1000 machine and set their graphics to use the Tandy Graphics Adapter, making the use of other cards ineffective.

On the SX, Adding a math coprocessor is as easy as installing it into the empty socket and removing a jumper.  The machine came with 384K and was upgradable to 640K by installing eight 256x1 RAM chips into the empty sockets by the drive cage and removing another jumper.  There are four switches.  Switch 1 is useful if you have a MDA/Hercules card, switch 2 for hard drive controllers that are hard wired to IRQ2, switch 3 and 4 disable IRQ6 and IRQ7, respectively.

The last issues which must be addressed relate to the expansion slots.  Original Tandy 1000s did not come provide -5v to the expansion bus.  Sound cards like the Sound Blaster 2.0 require this voltage, as do many VGA adapters.  The SX has the voltage on the bus, so this should no longer be an issue on the later machines (except for the EX and HX, which also do not have this voltage on the PLUS expansion port).  More concerning is that fact that only 10" boards or shorter will fit inside any Tandy 1000 case.  The Roland LAPC-I, for example, will not fit, so I used a bare MPU-401 interface box with a small card.  Nor will most EMS memory boards.  Hard cards may be difficult to come by that will fit.

Tandy 1000s are supremely easy to open, all you need to do is to unscrew two screws from the front or the side and pull the cover forward.  Working inside the first generation models is not quite so easy, due to the disk drive cage and the screws used to hold the expansion slots in place.  Inside, the screws are all of the hex nut, standard screwdriver types, no Phillips head screws.  I use a hex nut on a flexible screw driver handle to get at the screws, and it works pretty well.  The expansion slot screws are a nightmare because they are so small and covered a bit by the back plastic piece.

In late 1986, the SX was released in its 2x5.25" drive version for $1,199.00.  A similarly configured PC or XT, which did not come with a printer adapter or a graphics adapter, would have cost much more and not have been as much fun.  Like the PCjr, it supports 160x200x16, 320x200x16, and 640x200x4.  (I think only Deskmate ever supported the last resolution).  Unlike the PCjr., it has register level compatibility with CGA, whereas the PCjr. has only BIOS level compatibility.  Also, while the PCjr. and the Tandy 1000 support the PC Speaker sound generation, the PCjr. only has a piezo tweeter, while the 1000s have the biggest PC Speaker cone I have ever seen in a compatible.  This means that games that tweak the speaker, like Access Software's Realsound games, will sound far superior on the Tandy than the PCjr.  The Tandy 1000 does not require device drivers to support more than 128KB, unlike the PCjr.  256KB and greater booter games will work on the former but not the latter.  Many games use these graphics and sound capabilities to give great improvements over competing systems.  There are several games that support Tandy but not EGA graphics and games that support Tandy sound but not sound cards or midi devices.

PCjr. specific software (almost all of which was released by IBM) may not run on the Tandy or support Tandy graphics and sound.  King's Quest and Touchdown Football (PCjr. versions) will not work in anything other than a Tandy 1000 with 128KB of RAM.  Fortunately there are Tandy versions available of these games.  The near mythical M.U.L.E. port for the PCJr. is an unknown quantity.  Cartridge software, even when dumped, will not work because it expects to find itself in the D000 and E000 segments.

Unfortunately, the SX requires the 286xpress accelerator board to play those games of the late 80s that run slowly on 8088 machines.  The 286xpress is not an easy find as it was very expensive back in the day.  Also, finding an ADP-50L is not easy either, as it was a niche product.  A fast hard drive is an absolute necessity with these machines.  If you cannot find a 286xpress accelerator, then I would definitely recommend a Tandy 1000 TX instead, which has an 8MHz 80286 processor.  It is the last system to be truly Tandy 1000 compatible, as the later systems use a PC compatible keyboard and lose composite output.