By Craig Brougher
|[ Editor's Note:|
[ The following is an article written by Craig Brougher which
[ addresses some misconceptions about the Ampico A and B
[ expression systems. Robbie did the editing, with input
[ from Craig and Terry Smythe. There are 4 drawings which
[ are available as a suppliment to this article (but are
[ not required to understand it) . I have FAX resolution TIFF
[ images of these drawings and will try to get them onto the
[ FTP server in the next couple of days. I'll announce
[ the pathnames at that time.
- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
The purpose of this article is to put to rest many of the misconceptions regarding the supposed differences existing between the Ampico model A and model B. There seems to be so much confusion when it comes to understanding the differences between their crescendos, that an in-depth analysis is sooner or later obligatory. However, this paper will deal only with an overview of the common misconceptions as they relate to the worthiness of the Ampico model B. It should stand to reason that if the prime requisite of the design was compatibility with the model A, insofar as roll performance was concerned, then mechanical differences between the two machines are academic. Our subject then, will be to examine some misconceptions and show why some rebuilders mistakenly believe that the model B doesn't come up to 'A' standards.
A valuable history reference is the book, "The Ampico Reproducing Piano", edited by Richard J. Howe, published by Music Box Society International (MBSI). For information contact:
Angelo Rulli, 887 East Orange Ave., St. Paul, MN 55106
For materials and information about the Ampico, contact The Player Piano Co., Wichita, KS at 316-263-3241. Reprints of these service manuals are available:
"Ampico Inspector's Instruction Book, 1919 (with 1920 Supplement)"
"The Ampico Service Manual 1929"
-- A Self-Compensating Mechanism
In contrast to the Duo-Art, the Ampico is a self-compensating mechanism. For example, if a number 1 intensity in a Duo-Art is to be set for three particular notes of a chord, the roll will set it up from the 1 intensity hole on the tracker bar. But if 5 or 6 notes are called for at the 1 intensity, the 2 intensity hole is used in order to play the 1 intensity. The reason is due to "stack transients." Losses and undesirable changes in pressure due to the playing of notes. In the Duo-Art, these changes are anticipated on the note sheet. In the Ampico, they are automatically compensated for by the physical way in which the reproducing mechanism is designed.
For example, in the model A, the three intensity pneumatics for each side of the stack are supplied, not by pump vacuum, but by regulated stack vacuum, and are opposed by a so-called spring regulator pneumatic, which is slaved to the crescendo and driven by pump vacuum. This means that as the crescendo increases its pressure the regulator valve opens more, the stack also increases vacuum, and therefore the intensity pneumatics get stronger, too. As the notes make their demands on the pump through the expression system, the intensity pneumatics are changing accordingly, even at the same unchanging intensity, allowing the spring pneumatic, which is not supplied by the stack, to become a greater or lesser percentage of the opening force on the regulator valve. This alone is how the model A is able to compensate for played notes. The explanation is made very clear also in the Ampico Inspector's Instruction Book, 1919 (with 1920 Supplement). See pg. 8.
-- Stack Transients
It has been said that: "the model B crescendo system does not initiate corrective action for stack transients as do most earlier stack crescendo systems."
In actuality neither the model A or B allows the crescendo to take corrective action for stack transients. They are totally and completely isolated. How can they, since in both model A and B, the crescendo system is connected directly to pump vacuum? And always have been. Nor can they be alluding to the action of the A amplifier on pump vacuum, because then they would have to agree that when the Modify switch is set to "Normal" or "Medium" (both the same), or "Soft," the amplifier is switched out of the reproducing circuit. It has already been said in a previous article by one rebuilder, erroneously, that when the model A modifier switch is in the center position (Normal), the player provides "unfettered operation." So by this criteria, the model A without its amplifier becomes a direct-to-the-pump-spill crescendo without an ability to raise pump vacuum over the nominal mezzo-forte 18 in. water. So you can see that either way we try to understand this from their point of view, it is incorrect, and contradictory.
Were the crescendo required to compensate for stack transients, as we have read, the machine would not be an "Ampico," but a totally different concept altogether. This will be fully explained.
The model B regulator compensates for stack transients as does the model A. That is obvious, otherwise the roll would be used to anticipate its transients like a Duo-Art, and the B could no longer be used to play early Ampico rolls. In fact, the model B intensities are quicker and more precise in compensating for stack transients than any of the earlier models of Ampico. It does this automatically by a proportional curtain valve called the regulating curtain.
This, by the way, is technically not the so-called original "C pouch" referred to by Dr. Hickman in his diary, as some believe. See page 239 in the book, "The Ampico Reproducing Piano". Also see pp. 64, 67, 76-78, 283, 289, 303. The "C pouch" had a different configuration, and was eventually changed, beginning about July, 1926, and finally the regulating curtain is born in Sept. 1927. All this is according to the Hickman diary kept during the development of the model B.
There is nothing more basic to the operation of the Ampico than the fact that it is a self-compensating mechanism. Unless a reproducer is able to counteract stack transients either by roll coding (Duo-Art) or in the physical mechanism (Ampico and others), it cannot be called a reproducer. It would be "at the mercy of its own extremes" -- an impossible device.
If a Duo-Art, for example, was given a self-compensating mechanism and then a Duo-Art roll played on it, the performance would be over-compensated and unrealistic. Likewise, if the Ampico note sheet were re-coded to compensate for what it does automatically, it too would play too loud when it wasn't supposed to, and if readjusted to play very soft, would then be unrealistic.
-- The Model B Crescendo
The crescendo mechanism in the model A acts as a reference against which the intensities are expressed. Therefore, it must be unaffected by the stack expression intensities, or it would become dependent on the intensity value, and the Ampico could no longer reproduce.
The model B uses a proportional "curtain regulator." This is a thin, flexible sheet of pneumatic cloth which seals off the supply to the stack, proportional to the percentage of the vacuum pressure under it to that which is above it, which opens it proportionally. At the first intensity, which is the "set point" of the Ampico, the vacuum under the regulator pouch is fixed so that the pressure above the pouch in the stack will be balanced with exactly the same amount of vacuum regardless of the number of notes being played (all the way to the limits of the pump, of course).
The set point vacuum, or control vacuum which opens the pouch or curtain makes the demand that the same pressure in the stack corresponds to the pressure on the other side of the curtain. It does this by "peeling" the curtain off the supply grid holes to the exact percentage required to exactly balance the vacuum pressure on the other side. This is done "automatically" without adding more intensities under the curtain. It's just the nature of the device.
There are two reasons that this is an improvement over the model A:
1) The stack vacuum acts directly on the regulator valve
instead of external pneumatics controlling the regulator
2) the regulator curtain of the model B is much lighter and
faster than the A mechanism.
The reason that (1) is important is because the model A regulator is also affected by the stack pressure upon the regulator valve disk itself which creates a small self-defeating counter-response. In the B system the regulator curtain is much lighter and faster than the combined mass of a spring pneumatic connected to the large intensity pallet and its three pneumatics used by the model A. That means (2) that the regulator curtain would never overshoot or overcompensate due to mass as the model A might conceivably do (although this is actually anticipated and compensated for on the model A by another device called the "equalizer" pneumatics, mentioned below).
In regard to the comparison of differences in speed between the model A and B Ampico crescendos, some believe that the has a faster crescendo than the A, implying that the two could not possibly be compatible. Just because the model B completes it travel in half the time does not make it incompatible, since it must travel faster in order to maintain about the same rate of change as a direct crescendo operating on the intensities themselves. They are effectively equal.
I have documented by quotes in my article, "Unraveling the Ampico Mys-tique", in the AMICA BULLETIN, that the model B was designed to be compatible with the late 1921-22 roll editing procedures which were then standardized and used ever since (with the exception of the Hupfeld roll series which was a frustrated project creating an "experimental" set of rolls, most of which were never reissued). It was not designed to be equivalent to the model A, nor did it have to be. The model B crescendo rate is about equivalent to the A. It goes at twice the speed in order to have the same affect, dynamically. That is because of the direct nature of the crescendo pneumatic on the A, making it effectively faster exponentially (starts higher on the same exponential curve if you wished to graph it).
-- How The Crescendos Really Work
Another mistaken belief is: "During those periods when the crescendo driver <that means, "Crescendo" in Ampico terminology> is not actively engaged by an external signal (programming from the roll), the action of the pallet applies inverted feedback from the stack to driver control, thereby continuously tailoring the strength of the spring pneumatic to the action of the main regulator valve."
The regulator pallet portion of the Crescendo is called a "set point controller," in fluid mechanics or industrial engineering. It doesn't work by negative or "inverted" feedback from the stack at all. That is a term reserved for a return of the output of a machine to the input, from a later stage (output) to an earlier stage. This is just a mechanical compensating mechanism.
The more objectionable point here is not in bad terminology, but the authoritative teaching that the action of the pallet valve on a crescendo applies inverted feedback from the stack. We have already seen that the crescendos MUST BE ISOLATED from stack pressure in order to work. This is not a technical opinion. This is a fact. There is only 4 tubes going to each crescendo. You have a pump supply hose, two tracker bar tubes, and the connecting output tube (which is crescendo/ first intensity vacuum) slaving the spring pneumatic. There is no physically possible way of tubing the crescendo so it can be influenced by stack vacuum. Again, this is not a matter of opinion, but a physical fact. (We will talk about the model A amplifier in a moment).
Were the crescendo somehow even slightly influenced by the stack pressure, then to the degree to which it is influenced would be the degree the reproducer could not respond to the roll. It is the job of any set point controller to maintain a constant pressure against a varying one. Were the set point to vary as well, then it could not be the reference. It's like having a crescendo-decrescendo analog ramp device attached to an imaginary spring. Without this basic understanding of all self-compensating reproducing pianos, we could not have the first inkling of how they could do their job! To begin to understand the Ampico, you must first understand this basic concept of regulation with a fixed reference point.
Visualize a mechanical spring replacing the spring pneumatic (which is exactly what they did in the Marquee Ampico). It would be a compression spring normally set for the first intensity. Adding crescendo action to the spring is the same thing as moving its mount up and down accordingly, increasing and decreasing the tension. Now you have mechanically the exact analogy to the model A crescendo system. You might suppose why Stoddard called it a "spring pneumatic." It takes the place of an imaginary, continuously adjustable compression spring.
The next mistaken belief is this: "Ampico's highly stabilized expression system combined with its spring loaded stack equalizer pneumatics make the task of keeping up with normal stack operating demands (especially at soft playing levels) much less dependent upon the expression coding."
-- The Equalizer
The equalizers' task in the model A is definitely not to keep up with stack demands in any way, shape, or form. As a matter of fact, they actually add to the stack demand just a little. You can't hang a "winker pneumatic" onto each end of a stack and not create more demand. (No engineer would try to keep up with "demands" by hanging another load on the thing that needed the help.) So why are they really there?
The equalizers are there for one purpose -- to compensate for regulator valve overshoot. This is a logical assumption, but frankly the only one that makes sense at all. There was a fear that because of the speeds at which the expression valve was forced to shoot up and down, that wide excursions would cause it to momentarily close the regulated stack vacuum for a split second and a note could miss. So these lightly sprung equalizers were added to become tiny little vacuum supplies all their own, storing a little energy in each spring to oppose the closing off.
The problem is, they are a "compromise" between cleaner low intensity reproduction of dynamics and a mechanical flaw of the regulator valve. The irony of the story about the winker pneumatics is, they were not needed anyway. It was an unfounded fear and then possibly an inability to admit to the error.
(They say the proof of the pudding is in the eating. So why not take a very sensitive, difficult roll to play like "Gardens In The Rain," and try it with and without the winker pneumatic springs. Just unhook the springs. That's a five minute job at the very most. Record it both ways. Then listen after the performance has gotten cold in your own ears, and decide for yourself which is the best one. Some say they don't hear any difference at all (which is not surprising either). I personally like it better without the "even-ing" effects, but historically, restore the piano completely without changes. Then if you want to make that change, do it. It's your piano!)
It was also said that the crescendos are controlled by a "binary" code: "By means of a binary code -- the crescendos are controlled." And, "(the model B) employs the same binary language using two tracker bar ports."
Binary coding is done in a combination of off and on codes. Analog coding is done relative to real time by determining how long an input remains open, closed, or in combination with the other input. As long as one or the other tracker bar holes stay open, is as long as that crescendo function is active.
The Ampico Crescendo is an analog ramp device, not a binary
controller. Its control is more like a faucet with two knobs, than two telegraph keys. See again pp. 4-6 of the Inspector's Instruction Book, 1919. In contrast, the Duo-Art's expressions do compound in true binary code, even though its tracker bar signal is also part analog by duration. True, one can utilize a logic state table to show the relationship of any two-port combination during operation, but logic tables apply to analog devices as well as digital. You can do the same for lock-and-cancel valves which are also not binary coded.
Wrong terms create wrong premises. I would just encourage the use of an original instrument term, or if unavailable or inapplicable, to check a technical dictionary to be sure we have it right. For example, the company used the different terms for the same thing, like "Normal" or "Medium." Therefore, both are correct when referring to the Modify Switch center position.
-- To Crescendo, Or Not Crescendo
"All independent dual crescendos described above directly affect suction levels in their assigned portions of the stack, i.e. they are true crescendos." Another misunderstood concept.
The term "Crescendo" clung to the Ampico terminology ever since its patent description-- sometime around 1909. That was long before any standard coding techniques had really been worked out to everyone's satisfaction. It became immediately evident that using these "set point controllers" to actually swell and diminish the dynamics irrespective of intensities was the wrong thing to do because it made the note sheet highly tempo dependent!
Consider that in only 1.5-2 seconds, that piano could go from first intensity (ppp) to a triple forte (fff), via the crescendo if the roll were so coded. If the roll speed were 65, only 1.3 inches of travel and the expression would be off the chart! If you liked the tempo, say 15% different than printed on the leader, the roll expression would vary exponentially for each linear inch of travel of the fast crescendo, so 0.02 inches difference, plus or minus, could change the roll expression considerably, when the difference is taken at the top of the expression curve. In general you can say that a 15% difference in tempo will average out to be about 4.5 in. water error, during fast crescendos used exclusively. That's a whale of a lot of difference, and is probably why Ampico decided not to use the crescendos for crescendo-ing. Even slow crescendos sort-of "swept up and down" like waves, very melodramatic-like.
You don't have to take my word for this. You can know this for a fact from an interview between Angelico Valerio, Ampico roll editor from 1923-1930, and Nelson Barden, the now well-known interviewer who also interviewed Dr. Clarence Hickman and others connected with the Ampico in its heyday. Here now they are discussing this very problem regarding roll coding techniques with crescendo.
"The Ampico Reproducing Piano", p 128:
Barden: "Weren't most of the rolls before 1926 very
Valerio: "Yes they were. And some of them sounded
Barden: "Terrible, just awful."
Clearly, they are talking about some of the strange rolls, like from the Hupfeld series which were desperately re-coded by chief editor Milton Suskind (just before he was fired), because except for these, there are very few others that did not follow the 1921-22 standardization. That's easy to prove. The crescendos on rolls between 1921 and the present can actually be measured (again, not a technical difference of opinion), and it will be seen that the vast majority of them are NOT crescendo oriented, as Barden prompted Valerio to respond.
But the point is this: Crescendo-oriented rolls were considered by Ampico to be "Terrible. Just awful. Atrocious." That means, no "Crescendo" on any Ampico can be a "true crescendo," unless they were used for that purpose.
I really don't see two sides to this discussion at all. This is not a point of view. It is an historical as well as a proven physical fact, documented in Ampico manuals, measurable on a gauge, and proven in rolls that don't play well.
It is certainly true that the crescendo in an Ampico B doesn't change the intensities directly, but requires intensities to be on so it is able to modulate them, whereas in the A, the Crescendo overcomes the intensities directly.
-- The "True" Crescendo
The point still remains this: There is only one thing which makes either system a "true crescendo" -- the roll coding formula! It makes "true crescendo" from either system but never allows either system to crescendo itself as a function of roll speed! Do not fall into the belief that only one of these systems has a "true" crescendo. That is incorrect. Just keep in mind the reason why they could not be used that way, for the most part. It made the rolls too greatly tempo dependent, but it did something else, too. They sounded humorously melodramatic, as though someone were spoofing an artist instead of reproducing his performance.
This is why Valerio said what he did about crescendo-dependent rolls. Nobody who is truly musical could stand them for very long. A musician would really get a chuckle out of some of them. They are reminiscent of something you might hear in silent movies.
"All pre-model B Ampico expression systems, except the 1A, can superimpose full crescendos over any preprogrammed intensity level including zero. Their crescendo systems are fully independent of their step systems and can achieve full effect at the stack at all times (the model B expression system does not have this capability)."
In the first place, you don't have to worry about the so-called "Ampico model 1A." Only _one_ of them was ever built! It was considered unsatisfactory by both Dr. Hickman and Mr. Stoddard.
The fact that the Ampico roll coding system was never intended to utilize crescendos separately from the intensities is actually borne out by the fact that neither the model A or the model B used the so-called "independent" Crescendos. This becomes an ipso-facto statement which requires no proof. It is self-evident in Ampico rolls.
As an aside to this fact, the longest slow crescendo I have found on any of my rolls, in a modest 600 Ampico roll collection, is a little over 2" for a slow crescendo. It is in a 1915 march roll. The model B has absolutely no problems with this at all and plays the roll perfectly. Now what does a crescendo 2.3 inches long amount to? Well, with the pump spill set to 20" and the roll traveling about 65-70, the change in pressure is 5% or about 1" water difference. And why did they do that? Because the model A had to have a little boost in the compensation department, anyway. While it was almost a perfect compensator, it wasn't a compensator to the degree the model B was. With the model B, we could just leave those excursions in. They wouldn't hurt anything, and they didn't help either.
As a matter of fact, you could tape over most of the slow crescendos in 99% of all Ampico rolls and hear very little difference indeed. After all, most slow crescendo holes are between 1/4" and 3/4" long. At a speed of 80, not even the best ear will hear it. It was for mechanical, not musical reasons. That is why they were always added last of all. This is also elaborated upon in the interview with Mr. Valerio who explained that it was just used to give it a "little backbone." Ampico apparently used them to raise the "margins" on a few possibly weaker pianos.
Also concerning the terminology used by Ampico Corporation, there was never such a term as "zero intensity." The set-point reference on an Ampico is always called the _first_ intensity. Some tend to confuse the Ampico with the Duo-Art.
As a final observation on this point: There may be some confusion regarding the amplifier in the model A and the crescendo function. There is no connection here. The amplifier is totally independent of the crescendo, and raises pump pressure above the nominal 18-20" when a full solo performance is called for. It is true "positive" (not negative) feedback from the stack to the pump, driven indirectly by the expression system's overall requirements and is just the opposite of negative feedback, which, were the Ampico to have instead, would render it useless.
Another comment we have seen is this: "In the second amplification position, (referring to the model B) the crescendo capability is nullified."
Why some rebuilders try to question the wisdom of the second amplification, claiming that it is a stupid idea designed to "nullify" the crescendo, is conceptually wrong and a moot point, but demonstrates the depths to which a wrong premise can be extended to create many vastly wrong conclusions.
The amplifier and crescendo are combined in the model B. In the model A, when the piano is playing on Normal, the crescendo can raise the level in the stack to only the same level as the top power available from the pump (18-20"). Setting the Modify switch to Brilliant, the crescendo still is only able to raise the pressure to the top level of the pump (30-35"). Whether model A or B, the crescendo (or the pump amplifier section) of either can raise the stack to the full pressure of the pump for whatever setting you wish to play at. Once that happens in the model B they say, "the crescendo capability is nullified!"
Nullified, if that's the word to use, means to invalidate. But it makes it sound like -- while the model B can only rise to the top pressure of the pump, the model A can rise above the top pressure of its pump. Well, perhaps in the Mad Hatter's piano(?)
-- Compatible Intensities
"Why the three curves (these are the expression curves in the 1929 Service Manual) are depicted in the manner that they were is open to question I suspect that the author(s) did not consider the effect of the two switched stack spills on stack suction at lower intensity levels. Or perhaps, the lack of linearity at the low end was minimized in an attempt to show the discrete amplification stages."
The intensity chart takes everything into consideration. It has much of the information one would need to show the normal, first and second amplification curves at whatever intensity setting one wished to set up. How else could it be done in simple fashion? It is from these three curves that the roll coding format may still be extrapolated! I believe that a more clear and concise attempt to be open and honest about the piano could not be had. Everything is fully and honestly labeled. The person responsible for providing the curves for publication to begin with was Dr. Hickman, himself. In other words, the designer of the Ampico thought it important enough to provide these for publication and authorized them. Perhaps there is information in them that an average rebuilder may not appreciate.
The interview and the speech by Dr. Hickman in Philadelphia in 1979 or '80 showed he disliked trying to fool people or cover up things by making them more complicated than they were. It sounds very much like him, in my opinion.
There is no reason to accuse the model B of having "incompatibilities" and the inability for the piano to play other Ampico rolls as well as the model A does. The first design criteria, according to the Hickman diary, was that the new piano had to be fully compatible with the earlier roll coding system for those people who might want to trade up, but who had a library of favorite rolls to play. It had to play those earlier rolls, which complied with the standardized coding at least as well, and preferably better than its predecessor. And a good model B does just that.
In the Barden/Hickman interview (The Ampico Reproducing Piano, pg. 110):
Hickman: "..it broke my heart to think that we had to make the
quality of the new piano rolls less so that they could work on
the old one. But we finally doped it out so that we didn't
lose too much."
Barden: "But it's noticeable, and there are some sections where
there is a problem."
Hickman: "No, I think we did a pretty good job on that."
The model B does not skip notes at any intensity level, on any roll that I have ever played on it. It does not use elongated staccato notes because the valves are too insensitive, but because of the occasional use of the sub intensity, which is far below (30%) the vacuum level that any other player ever built can manage, so the valve actually requires an extra few perforator steps to give it time to rise. The difference cannot be the basic mechanism then, but perhaps small points overlooked during a rebuild. Nor can you disregard a jaded expectation of the rebuilder to his piano's performance.
In summary, there are two main points to consider. It is essential to understand that the crescendo must be completely isolated from the stack and powered directly by the pump. It is influenced by the amplifier through the spill valve indirectly, but not influenced directly by the stack or its transients as claimed by some. Unless a regulator system has somewhere a solid unvarying reference to balance against, which is *not* manipulated and bounced around by stack intensities, the player cannot regulate and maintain a known pressure to develop intensities upon.
It's like a balance-beam scale. You can place a standard weight in one tray, and then as you change the amount of grain or weight in the other tray, the beam pointer swings back and forth proportionally to the difference in weight. This changes the position of the reference weights, but not the weight of the reference itself. Unless the reference remained the same, the pointer position would mean nothing, however. Hence, no reproducer!
The second point seems to be an attempt to discredit the model B by comparing it unfavorably to other Ampicos. I have never heard of or read an official comparison of reproducers which presumes to evaluate the quality of one over another. That really isn't in the spirit of collecting or restoring vintage instruments, in most people's opinion. Part of the fun is in accepting and displaying an instrument as an original restoration of the designer's intent. We don't seriously compare Hupfeld Phonolists and Mills Violanos, for example, and run one of them down at the expense of the other. I believe that each mutually excels the other. And so it is within the Ampico series of pianos as well.
There are a number of special considerations required to properly restore a model B to its former glory. These details are not necessary and/or different with other reproducers, and possibly the failure to observe some or all of these aspects of this decidedly different instrument mechanically, is the reason that some wish to create doubts about the genuineness of the model B. They are perhaps the most difficult of all reproducers to rebuild well, but the effort is worth the time and expense.
The Ampico model B is a glorious, wonderful reproducer player piano that should make the goose bumps stand up on anyone who loves heartfelt music. It plays Ampico rolls gorgeously, without fault, without missed notes and weak passages, without any problems whatever, and sounds as eerily human as it is possible to get. Its dynamics are absolutely explosive, and yet it can play softer than most human artists are able to play, reliably, themselves. It is a profound musical experience, frankly, and the more that is understood about this piano, the more profoundly incredible the new arrangements coming out for it will sound, because we have yet scarcely begun to take advantage of such power of expression. Like they say, "You haven't heard anything, yet."
(Message sent Sun 10 Nov 1996, 00:58:35 GMT, from time zone GMT-0800.)