Official Luthiers Forum!

[QUOTE=crazymanmichael] perhaps our huge industry could utilize its great economic and politcal influence get congress to finally adopt the metric system as mandatory instead of the ridiculaous optional situation which has applied for over a hundred years.
[/QUOTE]

The problem isn't Congress. Industry has a few hundred $billion invested in machine tools and tooling.

Even so, MSC and Enco offer a lot of metric gauges.Jerry Hossom38556.5975578704

I need a sign made that says, "THINK METRIC" and I need it big so make the sign about 2 feet by 4 feet

having lived with the metric system for so long i have a dreadful time trying to think in imperial measure, even after being"home" for almost ten years. anything i design, which admittedly is not a lot, from scratch i do in metric. it is just so much easier. why the country did not get rid of the illogical, disconnected imperial system at the same time they got rid of pounds, shillings, and pence over 200 years ago is still a mystery.

There are formulas and tables for converting measurements of any given type from one unit to another. The ultimate that I know of is 'Measure for Measure', a little book by Young and Glover, Sequoia Publishing, ISBN 1-889796-00-X. Over 800 pages, and they even have a list of things like 'bodge' and 'fors' that they _don't_ give conversions for. I'd suggest that any listing we set up be in metric (say, kilogram-meter-second) units, since that is the international standard. If you really need in it English it's easy enough to convert.

Dynamic testing gets you a couple of things that you can't get from static tests. One is the Q factor: a measure of how much loss (or dissipation) there is in the material. There is some debate over how important this is in the real world, but nobody denies that the tradtionally preffered woods, especially for the back and sides of guitars, are those with low losses. Another eason for using dynamic tests is that they often give slightly different results, particularly for the Youngs modulus of wood, which is generally found to be higher in dynamic than static tests. Again, there is some debate about this, but it is known that wood 'cold creeps' when stressed over time, which could give at least slightly lower Youngs modulus values.

There are, of course, problems with any sort of test. I've recently been measuring the stiffness and Q values of my top and back wood stocks, and I'm getting rather lower Qs than many people report for similar woods. This is most likely a problem with my test setup, which puts the plates in too close proximity to the table top, thus adding damping.

There are two ways around these sorts of problems. One is to specify test methods exactly, and only accept those results that were done by standard methods. This gives the greatest possible accuracy and the most comperable data, but it does restrict the number of people who can participate. Another is to accept all data so long as we know the test method used. If you know how I do things you can always discount my data: "With his plates right on the table top old Carruth always gets Q values that are too low, so we'll just fudge them up a bit". I suspect that this would actually be adequate to our purposes, at least at first.

What data we decide to post will depend a lot on what we mean to do with it, and how we think the guitar works. There will be differences of opinion on both of those things. For me what's important is the way things correlate at various stages of the game. Particularly, i've been interested in whether things like the mass and damping of the plates, and the mode shapes and frequencies of the 'free' plates, have any strong relationship to the final tone. Thus I keep track of those things. I have only recently started measuring wood on ore than an occasional basis, to see how things like the stiffness ratio relate to the mode shapes. I find one of Dave Hurd's measurements, the deflection of the top below the soundhole under string load, to be an interesting and potentially useful one. Although I keep records of things like 'impulse spectra', and find them useful, Ithink it would be just too hard to standardise on those things enough so that we coul compare data directly.    

I agree with Michael.

I was educated during the change over from Imperial to Metric, and for me Metric is much more logical, or though even now I describe peoples Height in Feet and inches.

Guess it's a matter of what you are used to.

bugger,mydaughterspilledadrinkonmykeyboard.nomoreposting
forme'tilibuyanotherone.

just take it apart and clean it out!

[QUOTE=Alan Carruth]
There are, of course, problems with any sort of test. I've recently been measuring the stiffness and Q values of my top and back wood stocks, and I'm getting rather lower Qs than many people report for similar woods. This is most likely a problem with my test setup, which puts the plates in too close proximity to the table top, thus adding damping.

There are two ways around these sorts of problems. One is to specify test methods exactly, and only accept those results that were done by standard methods. This gives the greatest possible accuracy and the most comperable data, but it does restrict the number of people who can participate. Another is to accept all data so long as we know the test method used. If you know how I do things you can always discount my data: "With his plates right on the table top old Carruth always gets Q values that are too low, so we'll just fudge them up a bit". I suspect that this would actually be adequate to our purposes, at least at first.

What data we decide to post will depend a lot on what we mean to do with it, and how we think the guitar works. There will be differences of opinion on both of those things. For me what's important is the way things correlate at various stages of the game. Particularly, i've been interested in whether things like the mass and damping of the plates, and the mode shapes and frequencies of the 'free' plates, have any strong relationship to the final tone. Thus I keep track of those things. I have only recently started measuring wood on ore than an occasional basis, to see how things like the stiffness ratio relate to the mode shapes. I find one of Dave Hurd's measurements, the deflection of the top below the soundhole under string load, to be an interesting and potentially useful one. Although I keep records of things like 'impulse spectra', and find them useful, Ithink it would be just too hard to standardise on those things enough so that we coul compare data directly.    [/QUOTE]

Like you said Al, if we could devise a standard procedure that everyone agrees on then at least we have overcome one hurdle in data communication.

For starters, I am 100% in favor of using the metric system all the way through. Imagine if we bought Ohause triple beam balances, which measures in grams and then used gallon jugs of water, (Whose mass in my opinion is too spread out over a wide area), measured in ounces (and not liquid ounces, but avoirdupois) and then had the density in grams per cc and the deflection in Inches per pound. Messy.

Ah, that's better, at least I have a computer at work! The wet keyboard doesn't seem to want to come apart, even with all the screws out, it's a wireless keyboard which means it had power running through it when the incident occured - so I suspect somethings shorted. Can't blame my daughter she's only two after all (maybe I should blame the knucklehead who left a drink next to it, which would be, er, me).

What I was going to say before the interuption; is that while it's fairly easy to convert between units for someone who does it often, it's still non-ideal as it can be a source of error (see also: Beagle II & the European Space Agency). While a crashing spaceship is a good indicator that something has gone wrong with the calcs, any error produced in our area is more likely to go unnoticed, leading to dodgy conclusions. Maybe better to just use your gallon jugs or whatever you're familiar with, and leave us metric types out of it.

When we went to the metric system in '73 I'd just been taught how to get my little 7 year old head around the imperial system, then spent the next year getting my little 8 year old head around the metric system. Worked out well for me, I'll use whatever units happen to be facing up on the rule when I'm working at home. At work it's all SI units, just like every other lab.

If I might make a suggestion. One of the problems with using some base units, like meters and kilograms, is that you can end up with answers like 0.00000123. Some thought should be given to using units which will result in the most useful numbers for comparison. 12.3 is a much handier answer, and grams and/or millimeters are pretty easy to measure and use.

When I test woods the numbers I get are densities that range from maybe 300 to 1000 kg/cubic meter, Young's modulus from, say, 2000 to 20000 megaPascals, and Q values from maybe 30-250. The only one that might throw you is the Young's modulus: if you forget to divide by a million someplace the numbers run off the calculator. It's pretty obvious. :)

I'm in, lets go metric.

I think I may have been misunderstood in a previous post by some. I'm not question the value of testing. I done it for quite a while and get very good information from it. Just the type of testing.

[QUOTE=John Kinnaird]

For starters, I am 100% in favor of using the metric system all the way through. Imagine if we bought Ohause triple beam balances, which measures in grams and then used gallon jugs of water, (Whose mass in my opinion is too spread out over a wide area), measured in ounces (and not liquid ounces, but avoirdupois) and then had the density in grams per cc and the deflection in Inches per pound. Messy.[/QUOTE]

John, The gallon of water might or might not be too spread out over a wide area if you sat it on the wood directly. But if you hang it from the center of the brace you get a very small area of contact.

Another point I"ll bring up is 3 point vs. 4 point deflection testing. I've been doing 3 point but in reality 4 point give better numbers as the stres is distributed better throughout the beam. 4 point is the standard in the composites industry.

Al, yes you won't get the Q value from static testing. But the dynamic testing always seemed to me to have too many variables and I wasn't sure I could ever be really repeatable without a very expensive set up, which is why I choose static testing. Of course we wouldn't need a weight then would we!

I also am for the meteric system. Units are alway messy!



Jim_W38558.4968981482

_________________
Jim Watts
http://jameswattsguitars.com

Jim:
The setup is actually fairly simple. I learned about doing it this way from Mort Hutchins, who had picked it up from Dan Haines, who used strip testing to do one of the most complete studies on tone wood out there. One reason he used this methodology is that it can be more accurate than static testing, which is always plagued by the unknown amount of cold creep. As usual, though, there are some caveats.

The most important thing is that your results are only as accurate as your measuremnts: garbage in, garbage out. Because the strips are small Mort used a lab balance and weighed them to .01 gram, which most of us can't do. This limits our accuracy to within a few percent, instead of the 1% or smaller errors he was getting. It's easier to get lots of significant digits on bigger pieces, and halves of backs and tops are really inviting. I've been measuring those of late, and am finding that, as usual, there's no free lunch.

One issue is keeping the thickness uniform over the whole plate. Since thickness comes in as a cubed value in figuring out the stiffness, a 1% error in the reading comes out as a 3% error in the result. If the thing is only 3-4mm thick to begin with it's awful easy to lose that last percentage or three!.

Bars or strips have crosswise and 'torquewise' vibration modes that are well above the lengthwise bending modes that we're trying to measure. That's not true with plates. In some cases there have been 'other' modes close enough to the one I'm trying to read that they become 'mixed'. You can almost always tell that this is happening because the node lines get curved. The resuts in these cases are a sort of average of those of the two modes that are interacting, and the only way to sort them out is to re-cut the piece so as to move things out of range. If the objective is to test a back or top that you will then use to build something out of that might not be possible.

The equations that give the lengthwise and crosswise Young's modulus are set up assuming that the plates are rectangular. I have a number of oddball shaped tops (most notably red spruce ones) that I simply will not be able to test as plates, and may not even get big enough strips from.

The large surfaces of the plates tend to move a fair amount of air, which shows up as increased damping. I'm getting consistently lower Q values than I expect in my tests because of this.

With all of the problems, plate testing has a lot to recommend it; most particularly that you can _use_ the wood you tested to make instruments, and thus get correlations. I will try to do strip tests of the off cuts from plates to compare with those tests of the whole plates. These have their own drawbacks: most particularly that you usually cut off the worst part of the plates, so the strips often don't look as good as the body of the wood. Still, with both plate and strip tests, and the instruments made from them, we should be able to figure out if any of the measurements we do mean anything.    

Aside from the lab balance the setup Mort used was pretty basic. His original (Eico?) signal generator burned out, and I set him up with one based on an 8038 with a frequency counter that went to .1 Hz. The sine wave was not particularly 'clean', but it was good enough. Output was about 2W, iirc. The units Don Bradley is making now are delux for this, with plenty of power, good resolution, and a built-in meter if you want one for measuring the output of a mic. Mort had a little rig with threads stretched between parallel rods to support the strips. The threads could be moved to the node points easily. There was a coil for a driver, with a permanent magnet near it to magnetize the armature, which was a small piece of iron glued to the end of the wood with Duco cement. He cut the armatures from transformer core stock, which has low magnetization and thus low hysteresis, but I'm not sure how much of an issue that is. The Duco could be cleaned off the strip later with acetone or toluene. A dB meter reads the amplitude: I like the RadShack analog unit, which is still available for about $40. Evan Davis, of Boeing, says they're surprisingly accurate. Set it on the C scale. In my setup the triple beam balance is probably the most expensive thing after the signal generator. In a pinch you could use a program like 'Sweepgen' on your computer, and some of the new digital scales are getting pretty accurate and low in cost. Other than that most of the stuff, like coils from doorbells or old relays and so forth, is cobbled up or salvaged.

For Starters lets adopt a procedure that anybody can easily do in their shop. Lets begin by establishing a static protocol for testing the stiffness and density of the top and back material.

I propose we use grams per cubic centimeter as the units for reporting density. That's pretty standard in most labs and I think would fall within the range of the materials handily enough so that our values would range somewhere between .1 to 1 grams per cc. We might get a little lower than 0.1 but I doubt it and probably won't get higher than 1 gram per cc unless we find some wood that is so dense that it does not float.

For testing stiffness I propose we test plates that are actually glued together and still rectangular, dressed down to 1/8 of an inch. That thickness would probably be the max thickness of any top and would allow for eventual thinning if necessary. I propose that we test top plates that are 16 inches wide and we do that by placing them between two parallel dowels whose centers are 19 inches apart (convert all those inches to cm, it is so hard to be consistently metric) and we place an appropriate weight on a thin stick that spans the entire width of the plate, 16 inches long and about .75 inches wide. This will spread the load all along the top of the plate, engaging all the grain lines and eliminate the effect that cross grain stiffness has when centering the weight in the middle of the plate.

The total mass that is to be put in the plate needs to be determined, but the mass of the spreader stick needs to be included. I don't want to suggest a mass that is so heavy that it snaps the top, but we do need a good deflection, and I think we need a force that reasonably reflects the tension placed on the top by the strings. At least it should be in the same order of magnitude.

I also suggest we test the plate across the grain using a smilar testing apparatus but of course the dowels would have to be closer and the load spreading stick narrower and the weight perhaps lighter.

What do you guys think so far?

dowels would have to be rigid; 3/4" steel or so else you would have your "dowel" deflection measured with you plate deflection. the dowel should have 0" deflection or be a standard dowel with known consistent deflection under the specified test mass which can be factored out.

"Dowles" could be cf. LMI has them in specific sizes.

Just a thought and John Kinnaird touched on this. Keep it simple. I mean start with basic tests specifically defined and then from there increase the complexity of the testing. This way everyone can get involved at the level they desire and can benifit/contribute from/to the info collected. Some people like me care little about glitter patterns or modes and nodes. Instead we lean on a more intuitive approach. Having said that I am interested in collecting certain information that I feel will help my building.

In my own works I switch between metric and US (decimal/fraction) whichever is more appropriate/accurate. Measurements taken with non-metric dial indicators or other devices can be converted easily enough.

16" wide plates... A seperate catagory for classical builders? Maybe just an * near the results.

I might be wrong but I believe the volume of water changes with temprature but not the mass.

You can buy 1/4" x 1+" mild steel bars at Home Depot. Held on edge they don't flex. Measure from inside edge to inside edge for the distance.

John,
I have pretty much the same set up you just decriped for testing tops. I'll try to get a picture of it and post it tonight. It's not perfect and perhaps we can redisn it somewhat. But it does work pretty well as it provides repeatable results.

_________________
Jim Watts
http://jameswattsguitars.com

Great Jim! I would love to see it.

[QUOTE=RCoates]
I might be wrong but I believe the volume of water changes with temprature but not the mass.[/QUOTE]

Yeah, fair enough. Actually it's the density that changes with temp, waters a very weird substance in that it's density decreases when it's cooled or heated (sort of a U shaped curve if it was represented graphically).

My argument was that just filling a gallon jug by volume and calling it good was not accurate. I should have read the original post a little more thouroughly, my appologies. You're right weighing it shouldn't be a problem (man, it's a lot harder to gloss over being wrong, when you've got written evidence in plain sight). But, if you're going to weigh it out, why would you chose a number like 3780g? Anyway, I'll just shut up now...

John,
heres a couple of pictures of my test set up. The wooden dowels that the top rest on are a bad idea as dowels aren't straight. PVC or steel pipe/tube would be much better. Also more thought needs to be given to the mounting of the dial indicator. I just clamp it to my bench, which works ok, but I'm sure you could come up with a better way.
The first pic shows the whole set up in use. There are 2 sets of grooves that the dowels can set in allowing me to rotate the top and test it in each direction.
The second is a picture showing the bottom of the beam that the load is applied with. It has a 1/4 steel rod that contacts the top, which spans the whole width/length.
anyway,food for thought, maybe a good jumping off point, maybe not.

_________________
Jim Watts
http://jameswattsguitars.com

Jim
Great pictures. Thanks.
   From the bottom picture it looks like the block of wood with the X on it is a pedistal that must have a rod that goes down and transmits the weight of the water to the bar below. Very clever. That keeps the weight centered over the bar and balanced at the same time.   If a gallon of water weighs about 3.8 kilos, or a little over 8 pounds (if my morning math brain is working) that seems to be a lot of weight. How much deflection are you getting with that amount of weight? Are you testing a joined top or just a half of a top?
     Is the dial micrometer in the center of the plate or near one edge?
      
Thanks for the pic
John

John,
The block of wood with the X on it is the weight platform.
I test the unjoined halves. I sand them just enough to clean them up and get them an even thickness.
A gallon of water + the platform + the beam and all the weight is approximately 10.3 lbs. This is much more than I thought I would originally need. The example I just looked up in my records shows a top that was .158 thick had a deflection of .109.
interestingly enough the deflection is in a very similar range when testing each direction, this is due to the differences in modulus and span lengths for each axis. Just worked out that way, nothing I actually planed for!
The dial indicator sits directly under the 1/4" rod at the end of the plate.

As a side note one thing that I've found and not surprisingly is that the denser spruce is almost always the stiffer wood. There are exceptions however. You can also easily hear the difference in the wood for the least stiff to the stiffest, assuming they are all the same dimensions. Jim_W38560.6522337963

_________________
Jim Watts
http://jameswattsguitars.com

I don't have my copy of Hurd handy, but he gives all the formulas for calculating the Young's modulus from these sorts of deflection tests. If we could post the moduli, with a short note about test methods(static deflection, dynamic free-free, whatever)that would give us the information we need without the need for everybody to convert the other guy's results.