Initial Trials –
Breaking Soundboard Samples
Rick
Kemper,
A sample of redwood in
deflection, 55.8 lb load
In
this experiment, I broke 18 sample sections of Sitka Spruce, Western Cedar and
Redwood. Each sample was build into
a fixture that approximated a cross section of the soundboard near the bass end
of the harp. With a one inch
pine liner on each side, each 17” sample had an open span (liner to
liner) of 15 inches. Each sample
was made from clear, tight grain stock with minimal run-out.*
The
fixture was suspended from an overhead beam, a dial indicator was attached to
measure the deflection at the soundboard at the center of the sample, and progressive
loads added to the eye bolt to simulate string tension.
This experiment had
four goals:
1. It attempted to
determine how soundboard deflection corresponds to its load
2. It attempted to
measure whether screws through the soundboard and into the liners measurably
improve the sound board’s load bearing capacity
3. It attempted to
measure the relative breaking strengths and failure modes of Spruce, Western
Cedar and Redwood.
4. To Determine whether
soundboard samples broken in this type of fixture fail the same way as real
sound boards do
A few
practical tips for anyone wanting to try this in their own shop:
·
When sound board
sample fails, it can be a sudden explosive bang. Work carefully.
·
Suspend the weights so
they only fall in inch or two to the floor. Wear steel toed shoes and keep your feet
toes out of the way.
·
Make sure the tie- ins
to the overhead beam are secure and rated to carry the loads you are testing
·
You can get a 110 lb
fish scale and a 2” dial indicator on E-bay for ~$35 each
·
You can use plastic
buckets of Sand calibrated to the nearest tenth of a pound for a load
·
Set up the dial
indicator so the tip is pre-loaded.
The sample should break away from the calipers delicate tip
·
Use a secondary
lanyard to tie the dial indicator and holder to the overhead beam so it does
not fall four feet to the floor
DATA:
For
each material, samples #1-3 were simply glued to pine liners. Samples #4-6 were glued and
secured at the edge with a .210" cherry wood batten and number six by
¾ inch steel square drive screw.
Testing stopped when the sample broke. For example, the #1, #2 and #3 Redwood
samples were able to sustain a 25.8 lb load, but failed when they were loaded
with 35.8 lbs of weight. There are
no deflection data points for loads greater than 25.8 lbs for glued redwood
samples.
The
samples all broke at some common failure points.
It
is interesting to note that the glue was never a point of failure in this
trial.** The samples would often
break near the glue line, but they always left a fine “beard” of
wood fibers on top of the glue boundary.
Most Frequent
Left: Redwood
splitting along the grain, just above the liner near
the glue line. Most glued (only)
samples failed in this mode
Right: Rupture across
the grain at the bending point near the string rib. Many of the samples that were glued and
screwed failed in this mode.
Less Frequent
Left: Screwed Cedar
Sample failed when the liner cracked away from oak bass Plate
Right: Screwed
Many
of the glued and screwed samples had an interesting dual failure mode. At 25-45 loads the sound board would
audibly crack just above the liners but the sample would not fall apart as the
screws held the wood being tested to the fixture. After an addition 10-30 lb load had been
added, the sample would then rupture or break off the liner below the tip of
the screw.
Tabular Data:
Deflection
Tables with detail on failure modes.
|
|
|
|
|
|
|
|
|
|
Load |
#1 |
#2 |
#3 |
#4 |
#5 |
#6 |
Average |
|
0 |
0.000 |
0.000 |
0.000 |
0.000 |
0.000 |
0.000 |
0.000 |
|
5.8 |
0.056 |
0.061 |
0.062 |
0.067 |
0.065 |
0.064 |
0.063 |
|
15.8 |
0.151 |
0.153 |
0.158 |
0.166 |
0.160 |
0.153 |
0.157 |
|
25.8 |
0.231 |
0.250 |
0.245 |
0.250 |
0.260 |
0.231 |
0.245 |
|
35.8 |
0.300 |
0.326 |
0.322 |
0.320 |
0.335 |
0.299 |
0.317 |
|
45.8 |
0.365 |
0.400 |
0.388 |
0.394 |
0.405 |
0.348 |
0.383 |
|
55.8 |
0.430 |
|
|
0.460 |
0.467 |
0.419 |
0.444 |
|
65.8 |
0.489 |
|
|
0.520 |
0.535 |
0.494 |
0.510 |
#1
#2 Failed as #1,
but at 55.8 lbs load
#3 The Sitka
failed above the liner on both sides.
#4 Failed as
#1
#5 Failed in
the liner on both sides, liner cracking just below screw tips
#6 Failed as
#5
|
Redwood |
#1 |
#2 |
#3 |
#4 |
#5 |
#6 |
Average |
|
0 |
0.000 |
0.000 |
0.000 |
0.000 |
0.000 |
0.000 |
0.000 |
|
5.8 |
0.088 |
0.133 |
0.193 |
0.128 |
0.089 |
0.115 |
0.124 |
|
15.8 |
0.347 |
0.288 |
0.373 |
0.274 |
0.369 |
0.264 |
0.319 |
|
25.8 |
0.459 |
0.397 |
0.505 |
0.388 |
0.498 |
0.380 |
0.438 |
|
35.8 |
|
|
|
0.503 |
0.600 |
0.472 |
0.525 |
|
45.8 |
|
|
|
0.583 |
0.695 |
0.554 |
0.611 |
|
55.8 |
|
|
|
0.653 |
0.807 |
0.664 |
0.708 |
|
65.8 |
|
|
|
0.778 |
|
0.754 |
0.766 |
|
75.8 |
|
|
|
0.857 |
|
0.854 |
0.856 |
#1 Redwood failed
at liner at both sides.
#2 Failed as
#1
#3 Redwood
ruptured, cracking near center at edge of “string rib”
#4 Failed when
the screw pulled out of liner on one side
#5 Sample
ruptured into three pieces, failing ¼” in from each liner and at
Center near “string rib”
#6 Sample
ruptured into two pieces, failing ¼” in from one liner and at
Center near “string rib”
|
Cedar |
#1 |
#2 |
#3 |
#4 |
#5 |
#6 |
Average |
|
0 |
0.000 |
0.000 |
0.000 |
0.000 |
0.000 |
0.000 |
0.000 |
|
5.8 |
0.105 |
0.099 |
0.142 |
0.103 |
0.146 |
0.103 |
0.116 |
|
15.8 |
0.243 |
0.230 |
0.307 |
0.251 |
0.319 |
0.235 |
0.264 |
|
25.8 |
0.355 |
0.335 |
0.442 |
0.378 |
0.438 |
0.347 |
0.383 |
|
35.8 |
0.454 |
0.427 |
0.544 |
|
0.531 |
0.438 |
0.479 |
|
45.8 |
0.537 |
0.504 |
0.657 |
|
0.613 |
0.537 |
0.570 |
|
55.8 |
|
|
|
|
0.718 |
0.619 |
0.669 |
|
65.8 |
|
|
|
|
0.806 |
0.728 |
0.767 |
|
75.8 |
|
|
|
|
0.874 |
0.798 |
0.836 |
#1 Failed
Cedar at liner, both sides
#2 Failed Cedar at liner, one side
#3 As did #1
#4 Failed when
liner separated from oak base plate
#5 Ruptured at
center (near string rib), then within each of the liners pulling each of the
screws out
#6 As did #5
The
data tables shown in stress/strain graphical format
Comparative
Averages for the different kinds of woods:
Breaking load for
samples tested (in lbs)
|
Glued Only |
Screwed
and Glued |
||||
|
|
|
Specie average |
|
|
Specie average |
|
SS1 |
75.8 |
62 |
SS4 |
75.8 |
76 |
|
SS2 |
55.8 |
SS5 |
75.8 |
||
|
SS3 |
55.8 |
SS6 |
75.8 |
||
|
RW1 |
35.8 |
36 |
RW4 |
85.8 |
79 |
|
RW2 |
35.8 |
RW5 |
65.8 |
||
|
RW3 |
35.8 |
RW6 |
85.8 |
||
|
WC1 |
55.8 |
56 |
WC4 |
35.8 |
69 |
|
WC2 |
55.8 |
WC5 |
85.8 |
||
|
WC3 |
55.8 |
WC6 |
85.8 |
||
|
|
|
|
|
|
|
|
Average |
51 |
|
Average |
75 |
|
|
Deviation |
13.3 |
|
Deviation |
16.2 |
|
SS –
Some useful findings
1. The amount of
deflection for a given load is fairly consistent within each species, regardless
of whether the sample was screwed down or not. Sitka Spruce had the most tightly
grouped (consistent) load/deflection data points. The standards deviation values for any
given load averaged 5%. Cedar and
redwood demonstrated greater variations of 11% and 12% respectively.
2. When the
3. For glued-only
construction,
4. Reinforcing the
attachment to the fixture with screws only increased the load capacity of the
Sitka Samples by 21%. Screwing the
redwood samples to the fixture dramatically increased their load bearing capacity
(up 121%).
5. The glued and screwed
6. The batten/screw
reinforcement increases the samples’ load bearing capacity by about 45%
on average.
7. The samples and pine
liners in the fixtures had significant variations in their breaking strength,
even though these samples were sawn from top quality boards, adjacent to each
other. Wood is a natural product
and even when one selects top grade spruce for the sound board, there is
significant variation. As harp
builders we should not be surprised when some soundboards break sooner (and
later) than the average.
Shortcomings of this
Experimental Approach:
Critics
would note that string loads are not perpendicular to the plane of the
soundboard. String-to-sound board
angles typically range from 25-35 degrees, so the perpendicular component of the
string loads are going to be significantly lower than they were in this
experiment. That is
true. In many designs, the string
rib will take much of the load along the length of the sound board, and the
sound board itself must resist the normal component of the string’s
tension, especially where the belly is the greatest, about 2/3rd of
the way towards the bass end.
In
these tests, loads were rapidly increased from 0-85 lbs in a minute or
two. Most REAL soundboards fail at
sustained loads several days or years after they have a sub-critical load
imposed on them. Since wood deforms plastically over large time spans, the
abbreviated nature of the test may not accurately reflect what is actually
happening.
The
thick oak base plate used in the fixture was quite stiff. This imposed end
conditions on the sample that replicate a very stiffly braced sound box. Many harp designs have sound boxes that
are much more flexible, they are built lightly with little or no bracing.
Statisticians
will point out that the number of samples are small,
and 3-5 times more trials would need to be completed to produce reliable
(statistically significant) results.
To
all these critics, I say phooey! If
you think you can do better, go do your own tests and publish YOUR resul. . . . (Editors note: Please excuse Rick, he is
taking a short break and has promised us he will resume his medications post
hoc. He will also complete two
weeks of mandatory sensitivity training).
A
model is only useful if it can accurately reflect (even better, predict)
real world conditions. Do the results on the samples correspond
to real soundboard failures?
Based on two dozen of my own soundboard failures and repair jobs brought
in by my clients, I would estimate that about half the boards fail when the
soundboard splits at the edge, just above the glue line at the liner. Most of these have been old pedal
harps or boards that were secured only with glue. The other half fail in rupture, usually
near the string rib, about 2/3rds of the way down at the bass end. In 15 of the 18 trials, the samples
broke in one of these two failure modes.
You can learn where
the soundboard is going to break through sad experience or through predictive
methods like Finite Element Analysis (FEA). These pictures are courtesy of Patrick
Jordan, a fellow Uilleann piper who a genuine rocket scientist for his Day
job. The light green areas in the
second panel (showing Von Mises Stress) are the areas most likely to break as
predicted by the same kind of fancy pants computer program they use at
NASA. Pretty slick, eh?
In
the other 3 trials, the pine liners failed. Each of those three failures
happened in a sample with screw reinforcements. Failure in the liner is not something I
have encountered on my own harps or through repair jobs. I have not seen a sound boards
that failed in the top half of the
sound board or at the very bass end.
I
have never had seen a harp soundboard that failed when the string ribs cracked
either. There are a few reasons
this happened in the trial. The
short 1¼” grain in the pine “liner” used in the
fixture is much more likely to split than a full length liner is would be. Also, sound boards are glued over the
sound box edge, and in most of the designs I have examined, this is a composite
component, made from a strip pine on the iside (the
liner) and the hardwood or plywood of the sound box side. The composite structure is much
less likely to split too.
Conclusions:
Returning
to our original goals for this experiment:
Q: How does
soundboard deflection correspond to load?
It
is a fairly shallow concave curve, with notable differences between
species. Within the small set of
samples tested,
Q. Does a batten/screw reinforcement measurably improve the sound
board’s load bearing capacity?
Yes, by about 45% on average, but the limited
number of tests indicate that that figure is likely to vary significantly
depending on the species used.
Q. What are
the relative breaking strengths and failure modes of Spruce, Western Cedar and
Redwood?
Species tended to fail in similar modes on
similar fixture conditions. Glued
samples tended to fail when cracks formed along the grain in the sample just
above the glue line that attaches them to the liner. Samples that were glued and screwed
usually failed with the sample material rupturing at a bending point near the
string rib, with a smaller group rupturing near the edge of the sound
board.
In samples that were only glued to the
fixture, the
Q. Do these
samples fail the same way as real sound boards do?
In general, yes. There were a small number of samples that
failed in the liner or when the liner cracked away from the oak base
plate. This is an unusual failure,
but is likely a result of the fixture not replicating the composite
construction (typically pine and hardwood) where the sound box is joined to the
sound board.
End
Note
In
writing up this experiment, I tried to avoid getting preachy about how other
builders should build their harps.
I’m not saying
·
A
builder is having trouble with a design where the soundboards are splitting
away at the edge of the sound box, the screw batten edge treatment may be a
good way to address that weakness.
·
If
he believes a lighter, thinner soundboard will improve the harp's performance, and
isn't screwing his boards down, he can reduce the thickness of his boards,
secure the edge with battens and screws and the soundboard is quite likely to
stay together.
·
If
he has been insisting on using hard to find (expensive)
Hopefully,
harp repairpersons will also take a closer look at blown boards to see how they
fail and what went wrong. They may
be able to take some proactive steps to affect a longer lasting repair.
Next
Steps
When
I have the time and funds, I plan to conduct an expanded battery of tests that
would incorporate many of these improvements;
·
Double
the sample sizes (6-8 samples instead of three)
·
Test
tapered soundboards that are profiled to 70, 55 and 40% of their thickness at
each edge.
·
Test
the stiffness and breaking strength of the various aircraft birch plywoods used
for sound boards
·
Test
some common domestic hardwoods samples – cherry, tulip poplar, and aspen
·
Use
a composite (pine and hardwood, split resistant) liners on the fixture
·
Verify
whether the same trends hold when the fixture holding the sample is not as
rigid
·
Test
string ribs with feathered to see if it mitigate bending fractures
·
Add
oriented strand composites at center and edge of the sound board samples to see
if they improve the breaking strength of the sound board.
Are
you still reading!? Nobody reads
the footnotes anymore do they?
*
Wood selection: any impoverished harp builder would be crazy to use his best
stock for destructive testing, right?
I was sorely tempted to skimp.
I am as likely to give in temptation as the next guy, but in this case I
remembered, hey, this is for Science, my little contribution to the harp
building brain trust! For once I
let my conscience prevail over selfish, base instincts. The
**
Yellow PVA glue (Titebond II) was the sole adhesive used in the fixture and to
attach the soundboards samples.
Many builders have had significant problems with yellow glue for
attaching soundboards because it can slowly creep over long periods of time when
subject to high sheer loads. This
tendency is exacerbated by elevated temperatures or prolonged humidity. The Titebond II was not a point of failure in any of the
samples tested, but the duration of these loads was very short. Titebond is a great
glue for many other applications, but I cannot recommend it for the critical
soundboard to sound box joint. I
use Raka 1:3 epoxy for that and would recommend new builders use something more
creep resistant than the common PVA glues like Titebond for their sound boards.