The Ampico Crescendo System
By Craig Brougher
The most misunderstood of all the reproducing mechanisms seems to be the Ampico, because those who rebuild them or try to study them relegate the Ampico to be simply “somewhat different” but basically very similar in most respects to the others, and here is where the big mistake is made. It is not even close!
For instance, the Duo-Art is a linear intensity build-up, based on a binary cascading system on a linear vacuum pressure chart. The intensity steps on a linear horizontal axis and the pressure changes on a linear vertical axis. The Welte expression system is based on a linear time-scale in which the expression pressure in each half of the stack rises and falls (exponentially) with crescendos which escalate the pressure up and down, but relative to the length of time the expression valve is left open. A center-point “stop” called mezzo-forte can be used to park the valve, dividing the full gamut crescendo action into “two sectors” to expedite anticipated dynamics. It is still a real-time crescendo coding system in that regard which doesn’t use discrete intensity steps at all.
The Ampico also has (what is called) “crescendos,” but never uses them to escalate the (so-called “first intensity”) pressure alone. Ampico has always been a discrete intensity system primarily, having only 6 discrete intensity steps plus its initial setting. However it does not cascade like the Duo-Art because what is wrongly called its “first intensity” (the set point from which the changes emulate) is not a step progression. Only if its initial set point were REALLY its first intensity could it do this. It is not. So because its set point is NOT an actual intensity, is why it cannot be used AS an intensity. Instead, it is just an initial setting, and never used alone to crescendo.
Now with that said, there are reasons we still find many short slow crescendos when no intensities are being used. The reasons are not, however, musical. The crescendos are being used at times to slightly bolster the pressures momentarily. For example when the sustain is being used, or the soft pedal is being used, these short slow “crescendos” might anticipate a momentary valve loss. They don’t figure into the dynamic of the music but rather the evenness of the pressure. They touch out the normal valve losses which may affect the next intensity. But as you will see shortly, the head of the Ampico roll department made it very clear that the intended dynamics were always built up by the intensities and the crescendos then were used in conjunction to “platform” them. Why? For one reason, patent infringement! Welte was making enough money off that one, already. So if Welte could prove that Ampico claimed one thing while doing another, they could get them on yet another aspect involving another patent, as well.
The Ampico crescendos were able to platform all intensity settings even prior to the #6 intensity and above. ANY TIME a crescendo exceeded its initial set point, it was directly platforming the expressions. Granted, it could also increase the set point too except it wasn’t used that way because of the Ampico roll coding standards already defined and set up. We will get into a few of those details later.
Yes—it’s called “first intensity,” but it’s really not. You’ll see what I mean momentarily. The Duo-Art’s set point for both theme and accompaniment functions is called ZERO. That means it will not cascade as a numbered intensity. Instead, it’s the set point for its fixed intensity steps, which begin at step #1.
There are 3 intensity steps in an Ampico, bass and treble, labeled 2, 4, and 6. If its “first intensity step” were really a step, then it would be used to cascade intensities. Stop and think: There would then be 1, 2, 1+2=3, 4, 1+4=5, 6, 1+6=7, 2+6=8, 1+2+6=9, 4+6=10, 1+4+6=11, 2+4+6=12, 1+2+4+6=13. That would give us 13 steps in all. But Ampico never used it this way so the so-called “first intensity” is NOT a step. They also agreed that there are only 6 combinations so the set point doesn’t cascade pressure in additive combinations. The Ampico was designed around a completely different philosophy.
The vacuum pressures in an Ampico do not add together. They MULTIPLY together logarithmically. They follow a sort-of “square law.” They do not cascade linearly (as straight lines like a Duo-Art). In combination they can make 6 discrete intensity steps together, for a total of 7 (counting the set point by itself). 8, if you count the soft pedal.
This seems to be a sticking point in the minds of some rebuilders trying to understand how 6 widely dispersed steps would not sound too artificial and mechanical and be noticed. In comparison they note that the Duo-Art (initially patented by Charles Stoddard in 1912, and the Welte Licensee patented by him in 1914, by the way) has 15 discrete steps plus its set point which makes 16, and then there’s the “crash” function to accent the loudest notes when needed. So in the Duo-Art the divisions (in their mind) are so small the ear cannot detect them, but there’s a limitation to this, too.
The limitation is in trying to play a concert grand, which has about four times the dynamic range of a home-sized grand. By halving the number of discrete steps it might seem that the Ampico intensities would be pronounced and yet they are not. The Ampico is as perfectly smooth in dynamics as the original performance. So how does Ampico A achieve this? Also, what happens when an Ampico is used to play a concert grand? (Ampico proved this as well—many times, by the way).
The Crescendo Supplemented Ampico Intensities
The term crescendo is misleading. A crescendo in music terms is a graduated increase in power, or in other words, growing louder. Ampico crescendos can do that, but it’s not their only function. The Ampico’s “crescendos” are first and foremost set point controllers with two primary functions. First, they provide the set point for our (so-called) “first intensity,” of which I prefer to use its actual provision name—“the set point” because it’s NOT an intensity. The second function is to “smooth” quick changes in expression and pump pressure equalization through an indirect connection from the stack. We could call that a form of negative feedback delivered through the amplifier to the pump. Again, that too is intended to platform the higher intensities called for by raising pump pressure through a device they called “the amplifier control block.” It does this usually (but not exclusively) through roll coding they called slow and fast crescendo, but only in conjunction with preset intensities and never alone, musically speaking, of course. [Mechanical compensations are excluded in all of these discussion particulars.]
The amplifier valve could be called a “differential valve.” Whichever side of the stack has the highest pressure at the time takes precedence over the other side, actuating the Amplifier Pneumatic on the pump to the degree the stack is calling for. That’s determined initially by…. the intensities…. of either bass or treble side as read from within the stack….stack pressure itself. The amplifier pneumatic’s own spring is set so that actuation begins to increase pump spill pressure beginning at intensity #6, and not before then.
Some Ampicos had only one crescendo for both bass and treble sides of the stack, and in those pianos the spring pneumatics were “Teed” together. This is not really much of a musical compromise because any crescendo which affects the amplifier affects both bass and treble expressions equally anyway.
Now comes the answer to the question—“why don’t we hear the individual intensity step jumps?” I will ask you to ignore the actual slow/fast crescendo functions for just a moment and think only about the intensity step divisions. Then we’ll add in the crescendo capability.
First of all understand this: The perceptible audio steps, called “audibility steps” by Dr. Hickman in his 1929 Ampico Manual (pg. 21) is what both model Ampicos were configured around from the beginning. These steps are chosen NOT BY ARBITRARY VACUUM PRESSURE linearly, but more directly—based on decibels of loudness. Doing it this way is ultimately easiest for the roll coder because a “seat-of-the-pants” approach to the final expression tweaking and artist interpretation diddling is not as crucial, although all rolls were intermediately corrected and finally touched up artistically. It was as much procedural as it was showing respect for the artist who had the final say.
A 3 db difference, which is defined on a log10-based scale, is the smallest loudness difference the human ear can differentiate. While these intensity steps are wider than that, audibly-speaking, they aren’t by much when pressures are limited to 20.” It is impossible for any piano artist to intentionally play even one note repeatedly with exactly 3db changes between, much less handfuls of notes at the same time, so let’s begin from a practical standpoint and not claim to have supernatural sensitivities or abilities to do what is not humanly possible, anyway.
Another characteristic of the human ear is that small loudness differences are more noticeable in the middle of the ear’s hearing range. So the softer the notes become, the less that intensity change is detected, and likewise the loudest notes are also less easily discerned as well. This curve looks sort-of like a half sine wave. What Ampico did was to simply “dope out” a set of acceptable loudness divisions and then by sweeping the pump pressure up and down between them could create intermediate intensities.
Ampico first of all needed a smoothly graduated curve between these discrete steps which they could subtend to, which will then allow an approximately linearly graduated controller to oppose it. Since this curve is mathematically generated and negative feedback from the stack through the spill is generated, no guesswork for expression scaling is needed to initially code the rolls. The controller (crescendo) pressure is, of its own, linear so it doesn’t impose its own curve on the natural percentage ratios created by the design leverages created by the three counter-forces attempting to close the expression valve.
Those become roughly logarithmic ratios with each other because of their positioning along the wooden expression lever they’re connected to. That approximates the ideal division, and by combining this pressure ratio with the calculated delay (as in a graph’s horizontal axis) they’re able to develop an editing standard (a scale stick, if you will) that will put each in-between intensity smack on the ideal loudness curve, which is already drawn (See The Ampico Service Manual 1929, pg 20).
The crescendos therefore could also be said to operate individually as intermediate intensity dividers for the first 3 intensities which would be between the set point (1), step #2, and step #4. But when the music requires step #6 or above, something else happens when the player is set to “Brilliant.”
The amplifier in the pump begins to tighten the spill spring proportionally and the vacuum pressure increases. That means the last 4 intensity steps, multiplying upon themselves will also increase pump pressure a little (on brilliant), yet without any crescendo codes added! That’s a bit of positive feedback, if you will. When that happens, a coder can, if desired, take advantage of that time lag too and without any crescendo at all momentarily select an intermediate intensity as well. That can happen anywhere between #6 step and 20” of water (normal pump pressure). So for what it’s worth, the coder also has some options of intermediate intensities as well, even without crescendo.
The “Crescendo Feature” of the Set Point Controller
These set point controllers are named crescendos but it’s only half of what they are designed to do. This set point controller actually holds its set point as we’ve mentioned, but is built with additional crescendo valves to platform the intensities. Like they say on TV, “But wait folks, that isn’t all you get.” Yes, it’s BOGO time again.
You might wonder where I ever got the term, “set-point controllers?” One of my jobs at Honeywell years ago was designing test equipment for their dual set point controllers so I kinda know ‘em by heart. But that was then and this is now. It’s just a simple little thing that someone can set to maintain several separate levels or an acceptable range of latitude between the set points. They are useful in many industrial control processes.
In this case, it is the central device controlling the Ampico dynamics. Back in Stoddard’s day, they did things pneumatically (same principle) and used a leaking bellows tensioned by a spring with a sensitive bumped pallet valve to keep exhausting it. (Don’t worry though—it never really gets ‘exhausted’ doing that). That “up pressure” was then used to counter the 2,4,6 square pneumatics’ “down pressure” and that became the necessary set-point value at which the piano would faintly play all of its notes reliably.
The 2,4,6 square expression pneumatics are normally vacuumed down when off, tightly closing the expression valve supplying the stack. The spring pneumatic opposes them, preventing them from completely closing off the stack, since they are vacuumed down when “off.” The spring pneumatic has the same pressure as the crescendo, being supplied directly from the crescendo pneumatic, whose pressure is adjustable by the spring tension of its “first intensity adjustment screw.” That’s the spring tension I call “the set point,” as long as the pallet regulates the setting it centers around for producing that pressure. The set point is the level at which the normal scale curve begins. It’s the minimum pressure supplied to the spring pneumatic and is never calculated as an intensity step. (And yes, I keep repeating myself on purpose).
The expression valves and their respective crescendos are supplied directly by pump vacuum, and the “negative feedback” from the stack comes mainly from each side of the stack to the amplifier control box. There is however one other element that provides a bit more stack equalization and that’s the supply force differential exerted on the 1” dia. expression valve disk. If stack pressure (vacuum) drops below its intended percentage there is a small “up” force (to open the valve) exerted on the disk. Not much in comparison but it can be measured. This is why Stoddard included his lightly sprung “equalizer pneumatics” bass and treble intended to even out unwanted small fluctuations at low intensities which might allow the expression valve to momentarily close on reverse overshoots.
Since I’ve tested that hypothesis decades ago and have never experienced even one instance while purposely creating worst-case conditions I never experienced an overshoot, so I have always since rebuilt them but disconnected them. (The parts are all there to reconnect.) Otherwise, all they manage to do is add latitude to the all-important small stack changes at low intensities when they start to fluctuate. They supposedly compensate for low intensities but “No likeums!” It was an engineering presumption I call “The law of the Medes and Persians.” An absolute with absolutely no reason for it. That said however, all it takes for the expression valve rod, movement-wise, to make very large changes in air pressure is only a thousandth or three, when the stack and all the valves are restored and tight again. I’ve also measured these variations on a test jig in my shop years ago. So, “close, but no cigar.”
For a long expression lever to accurately control a few thousandths of an inch at its business end requires a PERFECT HINGE at its rear end without play by even a thousandth of an inch over the decades. That hinge should be replaced—always, if the Ampico is going to give nice, clean, discrete expression steps at the lowest intensities. That means new canvas with rock hard hide glue, and zero vertical play between the lever and its hinge block. Some might ask the musical question, what about the compressible variations of all the little felt washers in-between on the valve stem, and the answer is that it’s all relative because of the constant tension they are under all the time. No harm, no foul, but they give us flexibility and that’s most important.
Regarding the main way in which Ampico has an infinite number of intermediate intensities to select from is by means of the crescendo coding. The discrete 3 steps merely play the role of “stations along a train track,” you could say. They define the overall ideal loudness curve as determined by decibels (named for Alexander Graham Bell). Each decibel increase of 3 db doubles the power required to generate it, which sounds like (linearly speaking) twice the vacuum for each audibly discernible power step, except that (real) “power” in a pneumatic system is NOT MEASURED as vacuum pressure! (Yes, I know—surprise, surprise, but I can’t help it. That’s what I do.)
This principle is not understood by many, and certainly was not known at the time by Mr. Stoddard when he designed the Duo-Art. But just as apparent power in electricity is EI (volts times amps), and actual power is E2R, actual power in pneumatics is similar—P2 (Q). (Q is my quick way of saying something more complex, involving cu. ft/min, delay lag, and impedance). All we need to know is, it’s an exponential function just like db so it follows the square law, and that’s just old log10 stuff, OK? (a log is a power of 10 whose product becomes the answer.) Anyway, it works out on the db chart and also the Ampico book (pg. 20) perfectly as long as we are working in a log system to the base 10 (instead of ln).
When you get right down to it, Hickman related db loudness to hammer velocity (V). Vacuum pressure (P) is just an abstract linear quantity that we all relate to, but loudness is directly proportional only to the square of hammer velocity. We don’t need that here (pg.136 in the Howe book), but you should know about it. However, were you to substitute velocity for pressure in Hickman’s chart pg. 20, loudness would be linear.
There is another misconception about Ampico as well and it is that “Ampicos play popular music better than Duo-Art and Duo-Art plays classical music better than Ampico.” That is truly baloney. As a matter of fact, it was Ampicos (not Duo-Arts) that filled the concert halls of the cultural centers of the world, giving “comparison performances” (as they called them) and filled the halls. The newspaper music critics loved them. Ampico had 4 concert grand Knabes on tour all the time, and then I was told by a fellow who actually assisted in these concerts that he seemed to recall that they added 2 Concert Chickerings (of that I wasn’t aware). That’s a lot of money to invest, plus paying world famous concert artists travelling with the tour, when all you’re getting are poor reviews for your pianos when they played classical music! Somehow, I don’t think that any of us would have been asked to lend our amazing ears to critique an Ampico comparison concert performance anyway, but I could be wrong.
Another misconception has been that “Ampico A’s don’t play B rolls well.” That is also very wrong. I’ve always used B rolls primarily to test the model A’s that I’ve restored! It was the initial criteria of Ampico’s front office that the model A had to be able to play B rolls without any compromise in expression. If you recall, the model B had only ONE CRESCENDO and that was located in its own pump! So the rebuilding community has been laboring under many a misconception for many years hence. See my article entitled “The Great Ampico Myth.” When I mentioned this fact to one rebuilder he wrote this: “Your rhetorical appeals to authority to try to validate your indefensible assertions by quoting Stoddard and Delcamp don't carry weight with me, since it is clear that you don't understand the basics of what they were actually saying.”
(sniff-sniff), Aw Gee whiz—and here I thought I did! Well as you can see, nothing the inventors themselves said about it carries any weight at all with some people. If they had a contrary opinion, then even the inventor’s own explanation is all just hearsay. When you quote it to them, then “it’s a fallacy of argument called appeal to authority.”
Actually I’m guilty of that all the time, I’m pretty sure. After all, I’ve taken most of these things from the book, “The Ampico Reproducing Piano,” edited by Richard J. Howe, so shame on me. I think I have now been put in my proper place by a REAL rebuilder/expert on the Ampico who knows better, even when I’ve measured everything I’ve claimed here and verified everything I read verbatim from the builders.
Apparently, modern opinion today was firmly established even before Nelsen Barden’s comment on pg.110 in Howe’s book. This was during an interview with Dr. Hickman in regard to the way the Ampico A played B rolls (and vice versa). I guess the common wisdom Barden expresses will never be changed by the facts. I quote:
Barden: “Peter Brown said that you fought with Stoddard a little bit over having the “B” rolls work on the “A” piano and vice versa.”
Hickman: “Yes, I didn’t get anywhere with him. Well, he couldn’t get into it because that was policy set up by the president and vice president. But it broke my heart to think that we actually had to make the quality on the new piano rolls less so they would 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.” (Barden here was going by hearsay. It wasn’t dynamics that were the problem. It was because the model A sustaining pedal was too slow to track a B roll. However, this is easily correctable by proper damper regulation and a small change in valve bleed size. I’ve been doing it that way for many decades, now.)
Hickman: No, I think we did a pretty good job on that. It looked like it was a pretty tough problem when it was first patented.”
So Clarence Hickman flat-out disagreed with Barden. Now I guess you could say, “Well, Hickman was just defending his opinion and that’s normal,” but wait—he wasn’t defending anything. What’s normal (i.e. “human nature) is for Hickman to have said:
“Yes, I know (sigh). I tried to convince them I was right but they wouldn’t hear of it, so I’m glad to see that people today can finally tell the difference. Yaknow, none of the front office seemed to hear it, Delcamp missed it too, and the music critics in all the papers and elsewhere who wrote about the differences between the old and new Ampicos couldn’t hear any difference, either. Well, it just goes to show you, I was right, all along. Thank you for agreeing with me.” Now, THAT’s what’s “normal.” That’s the same thing as saying, “I’ve never been wrong in my life.” Dr. Hickman’s clarification to Barden was just the opposite: “I was wrong about it.”
The question, “were the A roll dynamics changed once the dynamic recorder went into operation?” was dogmatically answered by Angelico Valerio, the head of the roll editing department when Delcamp left. The answer in pg. 186 is that NOTHING AT ALL was changed in the roll editing department, except that it made coding a lot easier. Now all very sensitively played classical rolls were ALWAYS touched up by ear, but that was more of a note timing adjustment for effect, and part of it was necessary to fix small errors, not to mention the “clinkers” and woopsies” played by the artists themselves. Let’s not imagine that they were all unquestionably perfect performances, by definition. (Anywhere people and fingers are concerned, nothing is perfect.)
Barden: “With the old rolls, it looks as though the editors thought in terms of the crescendo mechanism first and then the intensity steps second.”
Valerio: ”No, it’s the other way around.” (Wow, now that’s perfectly clear! Did you get it?)
Barden: “Did you always think first in terms of intensities and then the crescendo to kind of polish it off?”
Valerio: “Yes, that’s right. To smooth it up.” (To platform, not to “crescendo” anything.)
Notice how Barden is still riding the same horsie he was on when challenging Hickman? Now to add a comment here about crescendo vs intensities, it is certainly true that a few Ampico rolls, notably those later transcriptions coded by Suskind from Welte master rolls they were licensed to copy, were probably the very ones that Valerio was mostly referring to. Milton Suskind should have used the Delcamp standard editing procedures to transcribe their dynamics. Apparently it was too time-consuming so I’m guessing he just compromised and used large crescendos, some 8-10” long, I’m told (I’ve never seen one). I was also told he got fired over that.
Now some people would say that firing him would have been an extreme overreaction, but Ampico, as a result of imitating Welte’s own expression coding on the Ampico, (which can easily be done) might have also been headed for a big lawsuit with Welte whose expression system worked exactly that way—timed crescendos without step intensities! By firing Milton then, and not selling many copies, they showed good conscience to a court of law and would have avoided the inevitable lawsuit!
There were always some old model A rolls around which didn’t adhere to the Delcamp editing standard, and there always will be. These do not even challenge the bulk of the Ampico collection, the same way that the Welte transcriptions are considered, which had only one cutting before they were trashed, so let’s not make a big deal over a few unstandard exceptions. They are not representative. I have a fairly large roll collection myself and have yet to find a Welte transcription but that doesn’t mean they don’t exist. As to modern recuts of obscure, very early titles prior to their reediting by Ampico, those may still be available.
In Angelico Valerio’s own words then, the crescendos were used merely for creating intermediate intensities (smoothing) which would lay on the same curve with the 3 intensity steps and their combinations. It wouldn’t be called “smoothing” if they were not perfectly in line with the log curve, or if they were musical crescendos. This int. step PLATFORMING causes the log curves to change their tangent (slope) as the pump pressure escalates, but in a perfectly relative, smooth and unbroken fashion.
This is why the Ampico model A can play a large concert grand to greater upper dynamics than the model B could produce, unless it was especially calibrated! The B is limited by the scaling of its pump crescendo. In most cases, the top pressure achieved by most baby grand model B pumps is 40 inches of H2O. Now you’ve also read “45 inches,” too. And that is also true for some. As a matter of fact, I have measured the pull of many Ampico B springs over the years and have found 4 basic spring rates. Many differed in their number of coils, and I’ve personally recorded 3 different wire diameters. These were all from the original factory and seem scaled to the size piano.
But 45 inches is only half of what you would need to play a concert grand to its full power. Usually 90-100 inches was required (no load) on full power for accents. Some might think that sounds just like a Duo-Art pump, but frankly, the Duo-Art, having only 16 discrete steps between their zero intensity and their “crash” intensity works well only for small grands. There’s no way to create intermediate steps between them in the Duo-Art, whereas in the Ampico, that’s the entire philosophy of the mechanism. Besides, it’s a little more relational to cu. ft./min anyway.
So then, how do we get to 90” of vacuum in an Ampico A concert grand? No problemos, we speed up the pump and from there, everything is still in perfect proportion, from the softest note you can play to the loudest, db ratio-wise, because it all lies along the same exponential curves, at an even distance from each other. The only thing that changes are the heights (the scaling) of the pressure divisions along the vertical axis! It’s all done by percentages or ratios to each other, not by fixed steps.
So just as the Duo-Art was always designed to play home grands (even though they did debut their piano in a Steinway concert grand at Carnegie Hall, once only), the Ampico was designed to play any size piano, with or without the artist.
Then came along the model B—which limited the size of the piano it could realistically play, exactly like the factory Duo-Art. I would have said now that’s kinda ironic, but actually, it was mostly just wood, leather, and cloth—and together with that, fully restorable as well! So although computers and electronics have perhaps an 18-20 year MTBF (mean time between failures), pneumatic reproducing pianos are measured in lifetimes, depending of course on the thoroughness and materials of its rebuilder.
Design by Player-Care
407 19th Ave, Brick, NJ, 08724