The whole process in 10 images :)

For the fun of the eventual reader of this blog, here are the 10 most important pictures during this journey.

Credits

My heartfelt thank you goes out to:

• my family for driving me crazy enough to actually do this!
• Jim McCarthy for providing such a precise and readable DIY instruction
• The Unknown Sales Guy at Hornbach, who upon my description of what I wanted to with the bars pointed me to a router
• Herr Pfanz of Edelholzverkauf.de
• Sebi for getting me the paper of Ingolf Bork
• Jürgen for calling me nuts to do this
• Ingolf Bork for applying scientific curiosity to marimbas
• my neighbours for tolerating the screeching sound of the router on weekends
• my neighbour Manfred for lending me the chop saw
• myself, for holding on till the end and finishing it!!! 

It's done!

Today, the final steps for the marimba were completed. Due to a snazzling noise on some bars, a stripe of velcro was glued between the wooden groove and the insulation foam. This eliminated all snazzling, but it is a semi-permanent mounting, which can be removed relatively easy, should retuning be necessary.

So that's it, and this is what it looks like:

FINAL ASSEMBLY: The Xylorimba is ready!

Today the final assembly of the marimba so far has been completed, I call it a Xylorimba, since the resonators are not yet ready. However, the bars alone sound very nice, and with a final retuning, it is ready to play and practice with!

Here are some pics of the final stage.

This is my method of suspending the bars: Agroove was drilled into the rods, and pieces of insulation foam were cut to just under the bar's width. The groove is deep enough to hold half of the thickness of the foam piece.

Here is a close up. The piece of foam is tightly packed in the groove, it cannot move. The bars are glued on top of the foam.

Right before glueing.

Here it is: The router tuned xylorimba! Phew, what a task...


That's it for now. Later, I will add the resonator tubes, but for the next couple of months, there will be no work on the marimba (apart from practice, I hope!). If you have any questions, feel free place your questions here as a comment.

Tuning done!

Finally, the last two bars had been tuned, and 2 octaves and a fith, that is from C3 at 131.7Hz up to G5 at 789.34Hz, are now available. The last quart was lost in the tuning process, and will maybe be added later.

Here are the measurements. Deviation is given in cents when prefixed with + or -, or the actual versus the desired frequency is stated.

Note	Deviation 1st	Deviation 2nd	Frequency 3rd
G5	+12	-11
F#	+29	-36
F	+19	-12
E	+20	2349 (2655)
D#	+12	-12
D	+23	+11
C#	+10	+33
C	+16	-25
H	+3	+11
A#	+29	1856 (1877)
A4	+21	-7
G#	+3	+27
G	+17	+0
F#	+4	1532 (1490)
F	+2	1435 (1406)
E	+2	1387 (1327)
D#	+11	+48
D	+23	1230 (1182)
C#	+4	+11
C	+8	+44
H	+23	-11
A#	+15	+22
A	+23	-24
G#	+8	793 (836)
G	+12	+5
F#	-12	-7	1801 (1825)
F	+0	-20
E	-9	+1	1635 (1626)
D#	+10	-11	1541 (1534)
D	+22	-41	1447 (1448)
C#	+13	-18	1363 (1367)
C	+5	-	-

So all but 2 bars have a slightly too high fundamental. This is fine, going down is easy. The first quart is also triple tuned to some extent. All others are more or less double tuned.

When the bars have been up in the living room for a couple of weeks, I will fix the fundamental to pitch.

Next up:

• Finish the frame: Need to add a diagonal for stability.
• Suspension and mounting of bars: There will be a suprise!

Stay tuned!

Down at the C!

Hey, today I tuned the last 3 bars of the lowest octave, D, C# and C. And again, these bars needed a slightly different approach, this time making full use of all that Bork's paper had to say about tuning: How to get the f10 down without lowering any of the others.

Here are some pics of the D:

That is the D bar after full triple tuning. Note the deep groves at each end of the bar: This is the groove that lowers the f10 without interfering with f0 and f4. In addition, the final few Hz down for f0 were done with a scraper, see the very thin groove in the center. The rest of the bar followed the approach 6mm across the arc, then down between the areas for f4, and then f10/f1. Behind is the diagram that proved to be of great value while tuning the bars. Thanks, Mr Bork :).

Here are the spectra of the D-bar during tuning, with the final measurements of D:

Next: 2 bars have to be redone, the F# and G of the 2nd octave. No fix could bring freaky F# back to normal again....

Stay triple tuned! Soon, the marimba will be finished!

Triple tuning video

As promised in the last post, I've made a small video explaining how to triple tune a marimba bar. The bar, the lower D#, behaved very well, and could be triple tuned in approx 45 minutes, including the video setup time.

3 Bars to go ;) ;) ;) Stay tuned!

Down to 4 bars left!

Just a quick, text only note: There are only 4 bars left to tune, the 4 lowest ones! Tuning the last octave, I've found a really quick way to double/triple tune a bar, within 30 minutes maximum! I'm thinking about making a movie of one of the last four bars...so you can believe me.

Here are the steps:

• Remove 6mm material between calculated beginning and end of arc.
• Remove another 3-4mm in the center right away.
• Measure: The bar should have 3 nice spikes. f0 and f10 already quite close, f4 way to high.
• Tune down f4 by removing material in 0.5mm steps from the position of f4 towards the center. Keep an eye on f0 and especially on f10. This creates a step of approx 3mm down from the first 6mm. If f10 gets close, stop extending the f4 area towards the center, keep it closer to its core area.
• When f4 is almost right (say 1-3 cent up), tune down f10 and f0 in the center. Measure often, watch f10 closely, and f0.
• When f10 is right, and f0 is still too high, scrape at approx 2cm left and right to the center. This brings down f0 only, and does almost nothing to f4 and f10.
• Here you are: You've just triple tuned a bar in no time at all, typically 30 minutes.

You'll see that in the upcoming video. The locations are all taken from Bork's paper and the graph of the regions of influence for the f0, f4, f10 and f20. The graph really is like a map, which provides orientation on the unknown bar.

Next up: Video, sound samples! Stay tuned!

Fixing broken bars (or trying to tune UP!) - The Experiment

Three bars so far have fundamental frequencies that are too low by approx 30 cents. Now tuning up is really hard, it is possible to do maybe 10 cent by removing quite a bit of material at the end of the bar, but that's about it.

While contemplating the effort involved in redoing the whole bar, the following idea came to my mind: Removing material reduces the stiffness of the bar in the area, and the frequency function is proportional to the stiffness^3. That is why removing only fractions of a mm in the last few steps drops it to the final frequency - or below, if you're not careful. So why not try to give back a little stiffness directly at the node's area, I was asking myself?

Here's how:

 The high B is too low...

 ..as is the high Bb.

 This F has a perfect fundamental, but two nasty peaks at f4:...

...one at E directly below the desired f4 and one at F# directly above; this 7th and 9th interval really sounds nasty, even when the bar gets hit in the center.

Here are the patients: Low B and Bb to the rear, double-f4 in front. Notice the uneven surface of the bar in front, I think that is the reason for the 2 close-to-f4 frequencies.

Now here is the idea: Let's take some of the sawdust, mix it with wood glue and apply this goo directly to the node!

All three patients resting now. I am very keen on measuring the bars again when the glue has dried. I'll give it a week or so, to let it thoroughly dry. If this works out well, then the fear of tuning too low can be somewhat reduced ;).

Stay tuned!

Fairly acceptable measurements

With the first couple of bars ready, it was time to measure them against the target frequencies based on A4=443Hz. IMHO, the results are promising in terms of the method and fairly acceptable for the first try at building a marimba.

Here are the details. The measurements were made in the basement, at 19.7 °C.

	f0	2 Octaves	3 Oct + 4 semi	f0	f4	DeltaF0 %	DeltaF1%	Ratio
C6	1053.64	4214.55		1055	4183	0.13%	-0.75%	3.96
H	994.50	3978.01		978	4054	-1.66%	1.91%	4.15
A#	938.68	3754.74		919	3801	-2.10%	1.23%	4.14
A	886.00	3544.00				-100.00%	-100.00%	#DIV/0!
G#	836.27	3345.09				-100.00%	-100.00%	#DIV/0!
G	789.34	3157.35		791	3118	0.21%	-1.25%	3.94
F#	745.03	2980.14		754	2913	1.20%	-2.25%	3.86
F	703.22	2812.87		708	2779	0.68%	-1.20%	3.93
E	663.75	2655.00		666	2331	0.34%	-12.20%	3.50
D#	626.50	2505.99		629	2480	0.40%	-1.04%	3.94
D	591.33	2365.34		597	2367	0.96%	0.07%	3.96
C#	558.15	2232.58		560	2266	0.33%	1.50%	4.05

So the deviations are quite small, with the exception of the E, which has a f4 that is too low (it's on the D, actually). A and G# have to be completely redone, due to tuning mishaps.

Finally, this is what the spectra look like:

The peaks of the fundamental and the f4 are clearly visible here. Several other peaks are present; the higher the bar, the more peaks are in between.

The bars now "rest" in the first floor, where it has approx 25°C. I'll measure them again in a couple of days, to see the influence of the temperature on the bars, and if I need to correct the target frequencies for the basement temperature. Target operating temperature for the marimba is 21°C.

Next up: More tuning, taking into account Jim's advice that the shape of the arch may change with the length of the bars. Then, some experiments for alternative bar suspension need to be done.

Stay tuned for more tuning results!

Marimbas have an arch, and for a reason!

Finally, there is progress, in understanding the arch as well as in tuning the bars. After much frustration with the upper A, G# and G (which are currently lost), I finally arrived at a scheme that works consistently. If it works, it works like a charm: double tuning the bars takes approx 20 minutes now (not kidding!), and the process is very predictable and repeatable.

Here's how:

In the pic above, today's result is present. The arch is now cut starting at the position influencing f4 the most down to the middle. The terasse is formed by stepping the router approx 0.7mm down in each step, resulting in a final depth of approx 5mm. With this initial arch, the f4 is almost spot on (typically +100Hz), and by increasing the steps by 0.5mm again, f4 can be brought exactly to match. The fundamental in that stage is high by 3-4 tones, and can be brought to pitch by deepening the centre groove alone. Since this does not really affect f4, the tuning is stable.

And here it is, my 5 semitone-marimba:

And it sounds like this :)

My understanding of the arch is that it serves 2 purposes:

• bringing the frequencies 
• down
• to specific pitch
• prolonging the decay time

The decay time gets longer with an arch present, probably due to the overall reduced stiffness of the bar. The frequencies of the fundamental f0 and the f4 are influenced at regions along the arch:

• f4 at approx 30% of length
• f0 all the way from 30% to the middle

This means that when the arch starts at approx 30% of the length, both the f4 and f0 are lowered. Most of the arch between 30% and 40% lowers f4, and all the way from 30% to the middle lowers f0. So when the first arch is cut, both f0 and f4 drop. Then, by deepening the grooves around 30-40% of length, the f4 can be brought to pitch, also further deepening f0. However, f0 is still high enough. The final tuning step of lowering f0 can be achieved by cutting right in the middle, where the influence on f4 is practically zero (it isn't but I think it does not matter). That is why you need and arch and grooves ;).

I wonder how things work out when I triple and quadruple tune the lower bars...much head-scraping, I think!

Stay tuned!

Influence of the arch on decay time

It's been a while since the last post, now there should be again some time to work on the marimba. In this post, 2 topics will be discussed: A final tuning experiment, and the planned procedure for tuning the bars.

Influence of the Arch on Decay Time

The arch depicted in Bork's paper extends towards the edges beyond the areas of influence for f4. I was wondering whether removing this material in an arch like shape would influence decay time or pitch.

Before working on the arch, the decay time was measured in the following way: Strike the bar 3 times hard, and measure the time from maximum to 30dB below, as displayed by the current baudline setup. I am only interested in detecting any influence, and not in determining the actual T60 of the bars, so this should be ok. Measured this way, the average decay of the original bar was 731ms on the fundamental.

The straight-edge shows the how the arch would extend towards the end of the bar.

 Material was removed in steps with a 6mm router drill.

On both sides, the typical marimba bar arch is now present.

Then, the decay time was measured again, as shown in the images above. It was found to be 772ms after removing the material. Even after flattening the whole area to the level of the innermost step, the decay time remained approximately the same (745ms).

Results

By removing the material in an arch like shape lead to

• a small increase in decay time of the fundamental (+5%)
• a small decrease of f10
• no other effects noticed

So, I don't consider it worth the effort to finish the arch from the position of f4 outwards.

Tuning strategy

I am going to tune the bars in the following way:

• start at the C6 bar, work from highest pitch downwards
• double tune bars C6 (994Hz)  to A4 (443Hz)
• triple tune bars G#4 (418Hz) to A3 (221Hz)
• quadruple tune bars G#3 (209Hz) to C3 (131Hz)

Inspecting the archs of  a commercial marimba, I found that the thickness at the middle of the arch decreases with decreasing pitch. So starting at the top pitch, I can use the higher bar as a guide to set the router, this should speed up things a bit.

Let's see!

Frame check: A-ok!

Construction of the frame makes progress, all the parts are ready now to be joined together to the complete frame. The Ivar-System poles form a very rigid A-shaped form, for receiving the bars. Here are the pics:

Here are the two A-shapes aligned on the floor, for a basic length check. Towards my feet, this A-frame holds the "white" keys, the one above the "black" keys in piano terms. To the left and right, the Ivar-system racks are visible. And yes, the calculator on the floor is a HP-48G!

 Left, the angled strut needs to be aligned flush with the spacing pieces.

Right, after cutting the spacer pieces.

 

Here, the top spacer pieces align perfectly flat with the angled strut!

The frame components set up roughly as complete frame. The A-sections were screwed&glued together at their spacer pieces, resulting in a rigid structure.

The end pieces get feet: a bevelled piece of Ivar-pole is glued&screwed to the bottom of each side rack.

This is the small end rack.

 Update from the garden shed: A second goat of line oil glazing was applied today. Now, the bars really shine in their golden brown colour!

With the coating still wet, the structure of the wood can clearly be seen. I hope that the second coating will dry to a smooth enough surface that the structure still can be seen when the coating is dry and hard.


Next up: Joining the frame together, in a way that allows for relatively easy deconstruction. A lower stabilizing strut must be added. And then, when the coating is dry, it's time for tuning! Stay tuned...;)