|
|
The Welte Mignon system of recording performances for the T-100 "red" Welte roll format was a wonder of it's age. It was apparently developed in 1903 and was used extensively by Welte to record the great piano virtuosos of the age. The system allowed an artist great freedom, in that he was able to record any time he wished, and was able to hear his performance played back on the piano electrically; every note, all dynamics, and pedal nuances perfectly intact. Many technicians and engineers have theorized for thousands of collective hours as to how such a feat could have been done at the beginning of the 20th century. Professor Ludwig Peetz, PhD, has researched the possibility of such a feat at this early stage and has recently demonstrated practically, historically, mathematically, and economically how it could have been done, taking cues from old advertisements, photos, and descriptions by observers, some of whom actually saw the instrument and handled a few replacement parts. There may be those who still wish to deny that such a system existed and was not possible. There are egoists who still insist that the world is flat and will never agree that it is round regardless of the evidence. The majority of us however are willing to learn, and it should be no problem to listen fairly to the evidence with an open mind. We have duplicated the feat today digitally, but they did this analogically. Today's engineers don't know a whole lot about analog devices. The scientific attitude toward discovery should not be, "They defrauded the public and just invented this story to look good." Instead, an engineer or physicist would say, "That's interesting. I don't see yet how they did it but I'll give it some thought, because they must have done so." Getting downright practical about it, a lawyer would say, "A well-established reputable corporation like Welte, who invented the reproducing player piano, had everything to lose by involving their stable of artists in a fraudulent scheme which would certainly backfire in a class action conspiratorial lawsuit against them. They would have lost their business overnight with a claim that otherwise was not even necessary for their success." So if you appreciate good sense, ask yourself why risk everything to gain little or nothing? The stupidity is the accusation of fraud, not Welte advertisements, which not even their competition at the time doubted. Welte was truthful about it, although they refused to give away even the slightest detail that might tip off their competition, and obviously disposed of their dynamic note sheets as soon as they had perforated a master copy, so as not to clue anyone in. An original section of a note sheet drawn by this machine nevertheless has recently been examined. This is an artifact. We know not only that they did it, but pretty much how they did it, too. Corporate secrecy demands security even today when techniques are so well hidden that it would require an electron microscope to discover how something was done. The American company Wilcox and White Corporation which built the Artrio-Angelus also claimed to have a recording and playback system, as advertised in their own brochure entitled, "Genius Immortal." In the chapter "How Artrio-Angelus Records Are Made," we read about a recording piano connected by cable containing hundreds of wires, to an electric recording device. It in no way affected the piano's touch or response. The artist played, just as he would play on the concert stage, and could then hear his performance as a member of the audience, moments later. As quoted from "Genius Immortal" for the Artrio-Angelus Reproducing Piano: "And then the artist sits back and listens to the reproduction of his playing- listens as does the audience, without thought to the production of the composition- and in this regard a remarkable result has been achieved. For each time an artist plays a certain composition, he plays it with a slight difference. His mood, the temperature of the hall, any one of a thousand details may affect his playing. He finds that in certain passages he has not played it as he wished- as he actually thought he was playing it. Dissatisfied, he again seats himself at the piano- again his playing is faithfully recorded. "Ah!' he exclaims as he hears the reproduction- ‘That is what I meant'-- and when he is satisfied, and not until then, does he attest that master record with his signature." It is not necessary to know exactly with certainty how it was done in detail. Only that the technology of the day made it quite possible to do this a number of different ways. False claims were not necessary, apparently. Starting on January 28th, 2000, before I learned of Dr. Peetz, I wrote a series of articles for the MMD regarding one possible way that Welte might have built an electric recording and playback instrument, utilizing the technology of the day. As it turns out I wasn't far off. My articles, for the record, are numbers 2000.01.28.15, 2000.01.27.24, 2000.01.26.09, 2000.02.07.13, 2000.02.06.11, 2000.02.02.08, and 2000.01.29.14, in the MMD archives. One of the main differences between the simplistic drawing given of the mechanism and my keyboard sensors was that the sensing mechanism had an abstract action which allowed the una-corda soft pedal to slide the keys back and forth without taking the sensors with them. One misleading detail was to infer that the sensor box under the keybed was a "trough of mercury." The box under the keybed was tapered like a trough for clearance, not to water horses. So the term "trough" is a descriptive adjective used as a noun, referring to its shape rather than its function. You will also see in Dr. Peetz' illustration that his mechanical configuration allows the sensor system to be abstract instead of connected directly to the keys. This is the only system that would allow the key frame to slide unimpeded for una-corda action. The strongest misdirection to understanding the mechanical details of the sensor system is still due to an early oversimplified sketch which shows fine carbon rods dangling from the bottoms of piano keys into a trough full of mercury. One of the main reasons it would never work this way is that when a large grand is hammered by powerful pianists like Liszt for instance, the rhythm of the music, coupled with the power of the man's arms and wrists would shake the piano and even vibrate the stage it's on. This would set up waves in the mercury which, if the period was right, could even start sloshing, and 30 kg of mercury as shown, filling a 4 ft. long trough to a depth of 3 inches would not stop making waves anytime soon! Even baffles would be a waste of effort because of their own harmonics to the piano's tones. The piano's vibrations, transferred through the key bed would set up tiny waves which would be an appreciable percentage of the key depression displacement, ruining any dynamic sensing. The sketch was to correctly intimate the basic principle without giving away any of its mechanical details. This ambiguous scheme is common to depict unpatented, secret factory equipment required to explain a system. We already know from first-hand reports that Welte used small carbon rods equivalent to 4H pencil leads. Mr. James D. Crank has already personally examined a box of these provided by Edwin Welte through Richard Simonton, a good friend. The method of sensing is not open to speculation. We know, very basically, what they used. When someone says "carbon," lights go on (or, maybe off) and immediately some may see images of carbon piles or granules, and pressure sensors. The trouble with these is that they do not register velocity directly but only a change in pressure. The frictionless C/Hg sensor is designed to measure vector amplitudes. Its output is described mathematically as the first derivative. Also the C/Hg sensor system requires no force- just displacement- whereas the carbon pile system requires compression. This wouldn't work directly under a key during a very light, quick strike in which the key doesn't even quite make it all the way to the front rail punching and is scaled for higher pressures than the lightest key impulses. The idea of sensing by carbon rods (like pencil leads) in a row of tiny vials of mercury would prevent any kind of wave motion on the surface because of the high specific gravity (mass) versus surface area. The vials (more like sealed blown glass tubing) could hang in a cushioned rack and be very small, acting also as guides for the rods. The rod ends would be fitted or dip coated with a tiny insulative tip allowing them to float at zero resistance, automatically setting the rest position, or zero voltage position of each one. They would then be fitted at their top end with an adjustable abstract fixture to contact the key bottom. A fine wire braid with a silk center would make contact without mechanical friction. As the carbon rod assembly was bumped into the mercury by the key bottom, the level of the mercury displaced would rise around it, increasing its conductivity possibly by a factor of 4 times the normal key displacement. But even key action for quick staccato notes would propel the sensor into the mercury proportionally to the hammer blow. The key end would never have to even touch the felt to do that, and the tiny up-force provided by mercury flotation could not be sensed at the key because the key weights are many times more than that counter-force. So key impulse, strong enough to actuate a note would likewise actuate an equally fast sensor. In this configuration, no spring is required and the depth of mercury displacement might be effectively multiplied by 4, giving the system greater electrical resolution in the same way that a lever changes a mechanical advantage. It's as if the key and its end travel were made 4 times longer from the balance rail. A carbon 2mm diameter. In a vial 2.236mm ID would cause mercury to rise 4 times as fast as it plunged down. By tapering the carbon, any rate including a reverse rate gradient could be generated. Still, it would have to be tested because mercury has a mind of its own. There is always a point of diminished returns beyond which any system will not work. The tiny vials could even tend to start functioning a tiny bit like thermometers, hence the flotation system self-compensates for expansion automatically. One advantage however for a larger quantity of mercury in the box (but not filled up like a beer mug) would be to actually normalize heat gradient caused by the electric current of repeated notes. Mercury expansion therefore would be prevented in each vial, sinked away by a mercury heat sink in which the vials were immersed. This then would be a contact reservoir to which the vials were emptied and refilled. What is so very interesting to me about professor Peetz' mathematical analysis is that since the compression of the front rail felt is about 3 mm depending on the power of the strike, that 3 mm is really all the rod needs to dip in order to resolve all dynamic levels by magnetic flux in the registering magnets except ppp, anyway, so he has made the system sensitive to 5 mm, total. He has shown that electrically, he can resolve everything from the lightest touch to the most powerful strike in 5 mm with sharpened soft rubber inkers, and since his system is also abstract from the piano's action, quick key action that will not touch the front rail felt will actuate the sensor as long as it actuates the hammer, just the same. So the vial idea isn't needed electrically except as guidance and higher resolution. I invite you to read Dr. Peetz' MMD article now, understanding that refinements will suggest themselves. He is building an actual working model of his system as envisioned, from the sensor to the inker, in order to demonstrate the efficacy of the Welte system, using only their early technology. Below the MMD "teaser" is an earlier article sent to me personally by Dr. Peetz which goes into much greater detail for those interested. It might answer the technical questions which a general magazine overview could not. Craig Brougher
For the record, the only other dynamic record that we actually have proof of is from Dr. Clarence Hickman of Ampico Corporation in the middle '20s. He decided that the best place to pick off the dynamics of a piano was at the hammer shank. That is not because it was the only place to do it. It was primarily because he intended to measure the dynamic in relation to the speed of the hammer shank by measuring the time between two contacts as the shank passed. This was more convenient than measuring it at the key and was simply marked on paper passing a spiral high voltage wire on a rotating drum. A fine spark passed through the paper and was "developed" in dye water, later. The Welte company it is believed, translated the key impulse force as a resistance. A lockout connection in series with it, as soon as the knuckle left the jack in the piano action, probably caused that resistance to be registered as a force by varying the width of an ink line on the paper note sheet. The dynamics are the same, the math is the same, and the relationship is one to one with hammer shank velocity, but key velocity and after-touch depression is from 5 to 8 times too fast and almost 50 times too close to measure with the Hickman spark chronograph. For Dr. Hickman, the speed of the shank was easier to register with two marks through paper against a spiral wire spinning on a drum. The spacing between marks in a serial raster determined dynamics. For Welte, their version was an amplitude mark in parallel format and strictly analog. The Hickman recording horizontal line therefore was angled up from left to right across the note sheet, because the traveling spiral line (cursor) required a finite time to make one complete horizontal pass. This did not hurt its accuracy, since the editor's square lay at the same angle and the drum spun at a very high rate of speed. This system was used exclusively for all new Ampico B rolls coded by Ampico from the day it went into operation until some very late popular music was again hand-coded by Frank Milne reputably when Aeolian-American could no longer justify their roll editing department. So for all purposes, the spark chronograph was a highly accurate guide and indisputably successful, as all first hand accounts would indicate, but it would not allow a replay of the performance. Below is a copy of Dr. Peetz' article concerning the physics of a recording system using carbon rods and a mercury trough. It is only sketched out and you may have questions about the details. The following is a physicist's view of the system. I have in my collection of e-mails the mathematics which document the likelihood that this system works as described, and works well. Craig Brougher
The Mechanical Music Digest
Subject: Welte-Mignon Recording Technology
-- non-subscriber, please reply to sender and MMD --
There is no need of any speculation:
Most probably the Welte-Mignon T-100 recording system worked exactly
as Richard Simonton and Jim Crank described it (see, e.g., several
MMD contributions in January and February 2000):
Under each key a carbon rod was fixed by a spring. The carbon rod
was dipping in mercury, the deeper the harder (and faster) the key
was depressed. This resulted in a varying resistance and varying
electrical current: the deeper the key was depressed the higher
the current was, which acted on an electromagnet. Forced by the
electromagnet, a sharp-edged rubber disk printed a line on a running
paper roll during the time when the key was depressed. The harder
the key was depressed, the broader the line was.
Remember that after the Second World War Richard Simonton was well
befriended with Karl Bockisch and Edwin Welte, both the inventors
of the Welte-Mignon T-100 technology (see, e.g., MMD 2000.04.03 -
2000.04.06). After the 1930s the T-100 recording system was of no
economical value any more, so it can be understood that the inventors
informed Dick Simonton about the once very secret T-100 technology.
Dick Simonton's friend, Jim Crank, confirmed that he was a proud owner
of some carbon rods, of some feet of an original recording roll with
lines of varying width, and a very old and poor photo of the mechanism
of the mastering machine and of the factory recording piano's trough
under the keyboard. In addition, Jim Crank explained that the widening
of the lines was taken into account for the evaluation of the dynamics:
a slow key depression gave a gradual widening of the line, a hard key
stroke gave a rapid expansion of the line.
There is a proof that Welte worked with inked discs (or wheels) in
a writing system. The American Welte Philharmonic Organ recording
system still exists in near-complete state in the Schweizerisches
Landesmuseum (Weiss-Stauffacher collection) at Seewen, Switzerland.
The corresponding Freiburg recording system, which worked somewhat
differently, is described by Kurt Binninger (a former Welte employee)
and published in "Organologica Acta", vol. 19, Berlin, 1987, pp.
179-207. From the nice drawing on page 200 it can be seen that it
worked also with discs.
For one and a half years I've been working on the reconstruction of the
T-100 recording system strictly based on this historical description.
Recently (20th of September 2003) I presented my very encouraging
results during the annual meeting of the German mechanical music
society GSM in Triberg (Black Forest). These results will be published
next year in the GSM Journal, "Das mechanische Musikinstrument".
I discussed a lot on this topic with Craig Brougher, Mark Reinhart,
Hans-Wilhelm Schmitz, and many other Welte-Mignon experts who provided
me important information, for which I thank them very much.
There is only one parameter which defines exactly the loudness of
a piano tone: the momentary key velocity at the HLP (HLP: hammer
let-off point), where the hammer leaves the lever in order to move
freely towards the string. Nothing else!
I found the following points (by own experiments, theoretical
considerations and historical studies):
1. Based on the historical piano dynamics recording system by Binet
and Courtier ("Recherches Graphiques sur la Musique", in "L'Annee
Psychologique", Paris 1896, available at the Freiburg University
library), an equivalent system to the later "seismographic"
Welte-Mignon-Licensee system was state of the art already in 1896.
Note that the Licensee system developed in USA in the 1920 and the
T-100 system developed 1901-1904 in Freiburg are completely different!
2. If high quality and authenticity is requested, dynamic dependent
time delays must be taken into account for each tone during the
translation of the original recording roll of the production master
roll. These time delays cannot be discriminated by systems which
record the dynamic information only summary for bass and treble.
Therefore -- on reason of accuracy -- the dynamics of each key must
be recorded separately, as Edwin Welte and Karl Bockisch did.
3. The harder a piano key is depressed the deeper it goes into the
felt of the front rail punching. There is a key distance travel
difference of nearly 3 mm between ff and pp tones. The progressive
non-linear force-distance relationship is based on the elasticity of
the felt punching below the key. The "constant key stroke hypothesis"
(see MMD 2000.02.04.07) clearly is wrong.
4. The MHRM (minimum height of the repetition mechanism) must be taken
into account. The best measuring condition is to write lines only in
the zone where the key is at or below this point. In this case it can
easily be decided whether a tone is repeated (interruption of the line)
or held (continuous line). The MHRM point is about 2 mm above the pp
and 5 mm above the ff key position, therefore the measurement range of
the mercury-carbon (Hg-C) sensor must be around 5 mm.
5. Upon dipping the carbon rods into mercury, surface waves are
created which disturb the contact accuracy of the Hg-C sensor. The
wave amplitude must be minimised by dipping the rod with its sharp edge
exactly perpendicular to the surface into the mercury. Any rotation
or horizontal movement component gives rise to a much bigger amplitude.
Because the piano key moves in a rotational manner about the balance
rail fulcrum, a direct fixation of the rods at the keys is forbidden.
The only useful method is to transfer the vertical component of the
key movement with a well-guided prolongation device, e.g., a metal rod
moving a metal tube. To ensure that the carbon rod follows exactly
the key movement, it must be pressed against the key by a spring.
Advantageously, the spring allows to transfer the current from the
moving carbon rod to a fixed point from which it can be transferred by
a connecting cable to the writing apparatus without any problem. Such
a system fits exactly the description of Richard Simonton.
6. In addition, the sharp edge of the carbon rods minimise spark
effects due to the inductance of the electromagnets: the conductivity
(inverse electrical resistance) and thus the current varies continuously
from zero to the maximum value. To avoid oxidation effects (due to
dissolved metal impurities), a very pure grade of mercury has to be
used; following Simonton, the Hg surface was protected with an oil,
a very common technique in that time.
7. A theoretical calculation of the electrical resistance of a
pyramidal edge of round section shows that it varies very sensitively
with the dipping height. Special conditions regarding the resistance
of the connecting cables and the electromagnet can be chosen that the
conductance (and the current through the electromagnet) varies linearly
with the dipping height.
8. This theoretical resistance calculation could experimentally be
proven by a small device using a normal pencil lead of 2 mm diameter.
If a 12-volt DC source is used a very high sensitivity of around 0.8
amperes per millimeter, within a range of 5 mm, can be achieved.
This Hg-C sensor, in the exact form as described by Richard Simonton
and confirmed by Jim Crank, can be used as a very sensitive analogous
key-travel measurement system.
9. By experimental tests of elastic discs of different types (one
wheel with a round section of the elastic material and one with a
sharp-edged elastic part), I confirmed that the line width varies with
the force with which the wheel is pressed to the paper. In both cases
there is a non-linear relationship between line width and force.
10. By small modifications of the Welte Philharmonic Organ writing
system as described by Kurt Binninger, a very useful Welte-Mignon
T-100 recording system can be reconstructed.
11. The momentary value of the key velocity is given by the first
derivative of the key-travel distance versus time. With an analogous
key-travel measurement system, the momentary velocity value at the
hammer let-off point can be determined exactly (by graphical derivation
methods). Other systems, like the Nystroem system or the Ampico-B
recording system, are only able to determine medium values by the
two-point [travel-time] method. Medium values (which can easily be
determined with the above described system as well) take no account of
acceleration or slowing down effects which a pianist may perform within
the two-point measuring zone.
12. Therefore the Welte-Mignon T-100 system must be regarded as the
principally best of all known dynamic recording systems.
Best regards to all Welte-Mignon friends!
Ludwig Peetz
Pirmasens, Germany
P.S: Within the next months I plan to build a one-key model of this type.
[ Dr. Peetz teaches polymer and textile technology at the Pirmasens
[ campus of Kaiserslautern College, and I believe he is also a master
[ at chess! ;-) Danke sehr, Prof. Peetz, for contributing your
[ technical article to MMD. -- Robbie
|
Some of the articles mentioned above, or portions of those articles, first appeared in the Mechanical Music Digest, and are used herein with the express permission of the authors and editors listed. All references are copyrighted and cannot be used for any purpose without the express written permission of the authors and the editors. Other references used in Dr. Peetz's treatise are similarly copyrighted and cannot be used without permission.
|
|
|
|
|
|
|
|
|
|
  |
© 1996-2010 Player-Care.com |
  |
