The Ampico B vacuum pump contains an all-important coiled extension
spring in the vacuum regulation section of the 4-lobe pump. One end of
this spring is connected to the long edge of a rectangular curtain of
pneumatic cloth, a portion of which is guided to cover the perforated
spill valve grid of the pump. The other end of the spring is connected
to a control rod which is attached to the movable board end of an
accordion-like collapsible crescendo pneumatic.

One end of this spring is connected to the long edge of a rectangular
curtain of pneumatic cloth, a portion of which is guided to cover the
perforated spill valve grid of the pump. The other end of the spring
is connected to a control rod which is attached to the movable board
end of an accordion-like collapsible crescendo pneumatic. [See the
illustration "Fig. 9" at the link below.  -- Robbie]

The box-shaped crescendo pneumatic is comprised of 4 separate wooden
boards, the middle two of which serve as alignment and guide spacers.
The mechanical characteristics of the attached spring exert a strong
influence on the playback performance of the reproducing piano into
which the pump is installed.

Rebuilders of original unrestored Ampico B pumps (1928 1941) have
often observed that these coil springs vary in appearance, size (coil
diameter), number of turns, diameter of wire, etc. They have also noted
that the coil spring appears to become 'weaker' with age. Based on
these persistent observations, one has to wonder why? For example,
why do the springs come in the same form (tubular extension), but
with differing physical parameters? And what accounts for the changing
mechanical characteristics of the springs over extended time periods
(many decades)?

In order to address the question(s) meaningfully, we need to consider
at least two governing factors. Though it may seem inconsequential at
first, it is critically important to understand how an extension spring
is fabricated. The second factor is the microstructural properties of
the material from which the spring is constructed. Although it is not
obvious, it turns out that the microstructure properties of the spring
material significantly affect the 'long- term' behavior of the spring
(under tension).

Let us look at the first factor. An extension spring is commonly made
of metal wire in a tubular (solenoid) form. The wire is guided (by hand
or machine) to form a helical coil around an axially spinning metallic
rod (called a mandrel). This operation is usually done on a lathe or
similar motor-controlled system.

To illustrate some details, the wire is attached (via a hole or slot)
at one end of the rod while maintaining some opposing tension on the
wire supply from a spool as the mandrel spins. From this geometric
configuration, and the exercise of reasonable care, it should be clear
that a tubular helical coil can form on the spinning rod (on its long
axis). During the winding process, the feed position of the wire supply
has to be incremented slowly (advanced axially) along the mandrel. This
takes into account the lateral accumulation of wire on the mandrel in
such a way that only one layer of coil turns is formed.

The operator of this spring fabrication equipment has an important
decision to make. He needs to specify the angle (feed angle) at which
the wire is fed toward the spinning mandrel. If the wire is directed
in a perpendicular manner, an extension spring without any preload is
produced. For an extension spring, preload refers to the axially applied
force required to begin to separate all the turns of a uniformly wrapped
coil spring where each turn is initially in direct contact with its
adjacent neighbor.

If the 'feed angle' is slightly less than 90 degrees, then an extension
spring with some preload is produced. If the feed angle is slightly
more than 90 degrees, then a spring with zero preload is produced.
Such a spring could be used in compression or extension. What is the
significance of these engineering distinctions?

For a feed angle slightly less than 90 degrees, all adjacent turns
will be touching their immediate neighbors. However, some 'internal
compressive force' will also be produced by the action of forcefully
'squeezing' the turns together in the winding process. This additional
force is caused by internal material stresses built up within the metal
wire during the wire wrapping process. For a spring made with a feed
angle slightly greater than 90 degrees, convince yourself that all the
turns will form (advance in position) on the mandrel without touching
each other.

Visual inspection of a typical Ampico B vacuum pump spring at rest
(removed from the pump) tells us that the corresponding 'feed angle'
was not greater than 90 degrees. As mentioned before, if the feed
angle were greater than 90 degrees, none of the coil turns would be
in contact with their adjacent neighbors. But the coil turns of the
Ampico B spill valve spring (at rest) do touch each other.

Gently stretching the spring by hand suggests that the feed angle
was probably only slightly less than 90 degrees, if not at precisely
a right angle. This is equivalent to saying that the preload of the
spring is relatively small, possibly near zero.

The logical way to properly assess this situation is to measure
the force versus deflection characteristics for each spring. Such
a characteristic curve can be generated by plotting the force required
to axially stretch the spring (as a function of deflection), without
exceeding the elastic limit of the wire material.

The elastic limit corresponds to the internal mechanical stress
that would cause the wire material to suffer permanent deformation
of the helical structure of the spring at rest. As a practical
guideline, do not stretch the spring any further than it would go
when the accordion-like crescendo pneumatic is fully collapsed. If
this condition is satisfied, one should not expect the material's
elastic limit to be exceeded.

With the above information as background, it may not be apparent
that different extension springs which all exhibit the same force
versus deflection curve could be made. This follows because so many
variables, e.g., coil diameter, wire diameter, heat treatment, and
specific metal determine the overall force vs. deflection curve of
a spring.

From the point of view of the dynamic performance of the vacuum
regulator, all that matters is the force vs. deflection curve of the
Ampico B vacuum pump spring. On the other hand, finding such a variety
of pump springs today could simply reflect requests made by customers
(or salesmen) back in the 1930s to modify the overall sound intensity
of their Ampico B pianos. This might have been done to suit individual
tastes and/or to satisfy the 'needs' of the intended sites of piano
installations.

This author is not aware of relevant documentation that displays
measurements of the force vs. deflection curves for any extant Ampico B
pump springs. One would think that such a specification would have been
generated by The American Piano Company, so that the designated spring
manufacturer would know what design requirements to follow. Perhaps
the springs were made in-house, and the specifications were treated as
a trade secret. C. N. Hickman's original laboratory notebooks and
technical papers may be a valuable resource to help clarify this issue.

In the current absence of original technical documentation, it would
be helpful, if not illuminating, to compile sample sets (say 25 or
so) of force vs. deflection curves for original pump springs. These
springs could be temporarily removed from original Ampico B pumps
for dimensional inspection and nondestructive extensional testing.
Correlating these individual curves with the corresponding set of
associated spring geometry parameters could also help address why
the springs appear to vary dimensionally from piano to piano.

We now move on to consider the second spring factor, the microstructure
properties of the metal wire that comprises the spring itself. Metal in
the solid form consists of large numbers of microscopically tiny grains
of metal bonded together at intergranular surfaces. This can be verified
by microscopic examination of polished cross-sections of metal samples
that are chemically etched to enhance the viewability of the grains and
their interfacial boundaries.

The chemical and mechanical characteristics of the intergranular
boundaries between the minuscule metal grains have a profound influence
on the mechanical strength and long-term behavior of the metal under
load. Most metals from which springs are made are not pure. That is,
impurities (sometimes called contaminants) commonly accompany the raw
metal of the spring. The metallurgy of this composite material is such
that the impurities typically settle within and along the intergranular
joints between the microscopic metal grains.

When the metal is subject to a mechanically impressed load (like
stretching the spring), some of the stress is applied to the joints
between the metal grains. As a result, these joints can become distorted
and the metal grains can 'give' somewhat. Because the metal grains are
not all the same size, the microscopic structure of the aggregate
material is complicated and highly variable.

Each intergranular space between the grains can be characterized as
having a mechanical response time associated with it. One response time
could be about 0.1 second and another could be on the order of years or
even decades, with all the others having intermediate values of response
times.

Because there are so many individual metal grains (of different sizes),
the metal itself is macroscopically described as possessing a broad
distribution of time constants. That is, distinct microscopic portions
of the metal react differently to the same applied load in terms of
their locally individual response times. However, the overall (global)
response of the spring to an applied load is what we observe in
practice.

This state of affairs makes it mathematically difficult to predict how
a spring will behave under tension over the long term. But the analysis
has been done (1930-1950). A summary of the review of the work shows
that metals will gradually stretch (and become weaker) when loaded
continuously over long time periods. This phenomenon is known as
anelastic relaxation and occurs in many metals, especially when they
contain impurities.

For the mathematically inclined, one component (creep) of the total
stretching behavior of a spring under load will follow a long-term decay
characteristic curve approximately described by the difference between
two exponential integral functions. The degree to which such creep
occurs depends on the particular chemistry and amount of impurities
(and the local temperature) in the metal of the spring.

Homeowners with electronically-controlled garage doors that incorporate
extension springs will probably recognize this annoying mechanical
characteristic. After about 10 or 15 years ordinary garage-door springs
become noticeably weaker with time or fracture altogether, usually with
a disturbingly loud (and sometimes dangerous) bang. One common aspect
of garage door springs and Ampico B pump springs is that they are both
under continuous (and nearly constant) tension almost all the time.
When the garage door is closed, the extension spring is under tension.
When the Ampico B piano is not being played, the pump spring is under
tension.

Higher quality garage-door springs that last longer (more than 15
years or so) without breaking and that do not weaken as much with time
are available, but they are more costly. Why? Such springs are made
of a metal (commonly steel) whose chemical composition contains fewer
contaminating materials. Also, the steel is usually heat treated
according to proprietary time-temperature profiles. The fact that Ampico
B pump springs have already lasted about 90 years is a testament to
the high quality with which they were initially made back in the 1930s.
Nonetheless, they do weaken with time, but not as much as they would if
their quality were significantly less.

The reader might recognize that the stretching behavior of piano
strings under tension (about 160 lbs. per string) is essentially the
same as that described above. The metallurgy of piano wire has evolved
substantially over the years to yield a very high quality product.
Unintentional impurities in piano wire have been reduced to a low level,
indeed.

The elevated quality is reflected in the relatively short period of
time (say, less than a year) within which a piano string will 'settle
down' to a stable vibration frequency without the need for an excessive
number of tunings. This offers circumstantial evidence that Ampico B
pump springs may have been made from readily available piano wire.
A chemical/metallurgical analysis of B pump spring samples could help
resolve this conjecture. 

We do not know what the design specifications of the original Ampico B
springs were. Consequently, one is now faced with the problem of how to
incorporate new ones (when needed to replace damaged original or missing
springs) into rebuilt Ampico B pianos. A brute force approach to this
difficulty is by trial-and-error. It may be that The American Piano
Company used this method to select an appropriate spring for each piano.
If one had a large selection of springs available from which to choose,
the task would not be unreasonable.

Alternatively, the B pumps could have been mass produced with 'standard'
high quality springs, and then subsequently swapped out for others that
better satisfied the musical and environmental requirements of the piano
customer. Today, the solution to the problem of choosing/designing and
installing a 'new' B spring in a rebuilt original Ampico B pump would be
another story waiting to be written.

Bill Koenigsberg
Concord, Massachusetts

 [ Illustration from "The Ampico Service Manual, 1929",  page 16:
 [ "Spill Valve, Crescendo and Amplifier -- Fig. 9"
 [ http://www.mmdigest.com/Attachments/19/03/13/190313_180210_Fig_9.jpg