Official Luthiers Forum!

The idea behind this experiment is great. This is something I've always thought should be investigated. But the method of testing makes the results completely meaningless. Glue adhesion (and vibration transmission) on steel has no relevance to wood gluing. Many glues that adhere perfectly to wood do not adhere at all to steel. Sound waves are influenced by any change in density in the materials they travel through. A glue that had similar density to wood would be less than 1/15th the density of steel. Repeatability? With one sample of each glue, who knows how much variability might exist. Dry time? Will a 1x2 inch glue joint dry in 14 days in a completely non-porous material? The surface preparation used in this experiment is inadequate even for a test of steel gluing. Let's do this again with wood.

Eric I'm not an engineer but it's occurring to me that maybe Tim understood your very good points in advance and used steel anyway as a torture test of sorts for the glues where he knew in advance that adhesion would be poor but since all glues had to deal with the same materials it would still be telling?

Tim thanks for doing this! Todd thanks for posting this!

I have been wanting to post results of a recent glue test done by Darryl Keil from the Vacu systems veneer forum. He tested 6 different titebonds & his unibond urea glue which is exceptionally rigid. I can't figure out how to attatch the results but will continue to try.
Short answer, most glues were somewhat flexible after a week of drying. However Titebond
original "extend" which has an open time of 15 mins. & the clamp time of other PVA's came out almost as stiff as the urea. I think you would find his method interesting. I described it to titebonds tech support which is excellent by the way & they agreed it was an OK test & seemed to know, that is what the result would be for rigidity in comparing those glues.
Mike McNerney

_________________
Mike McNerney

With all due respect Eric, did you read the article in entirety or just skim it? I thought I made myself clear by pointing out in prefacing the article that my tests were NOT designed to test adhesion, shear, peel or any other physical property of glue. I was only interested to learn how vibration energy passed through the glue itself and if the data was even measurable or interpretable.

"Sound waves are influenced by any change in density in the materials they travel through." Precisely and one of the reasons why I chose steel. Tests on wood would be, in your words, "completely meaningless" too because wood has inherently more variation within so measurements would be all over the place.

"Repeatability? With one sample of each glue, who knows how much variability might exist." There was NOT just one sample but multiple samples of each glue tested. The data was averaged and then posted. This too was in the article!

"Let's do this again with wood." Feel free to step up to the plate and pick up the torch where I left off Eric. BTW, it will be interesting to see how [you] come up with wood samples that have less variation than steel that will make your data more accurate?

_________________
tim...
http://www.mcknightguitars.com

I haven't re read Tims test, so I'm going off of my memory. (which can sometimes be a problem LOL) I read it in it completely a while back.

First, Great experiment, and a lot of hard or tedious work. Great job Tim.
Second, what do we want from a glue joint? I want it to hold, and I don't want the joint to act as if it was a single piece, instead of 2 pieces glued together. Which glue had the same decay as a single piece of steel that was the same weight?

Adding decay, by adding a glue joint could, at least with metal could indicate a problem. A cast iron crankshaft (from an engine) will make a dull thud if you hit it with a hammer. If it has a hidden crack it will ring like a bell.


I've never use anything except HHG (in places where wood glue is used) and have no plans to change.


Added: I believe Tim's tests prove that different glues are.....different. If you're getting good results using whatever you're using, I'd recommend to keep using it.

_________________
I didn't mean to say it, but I meant what I said.
http://www.brackettinstruments.com/

I agree with good=bell; bad = thud, to a point. More bell than a single piece of the same mass probably isn't good. A cushion in a joint between two pieces of metal = thud. Two pieces of metal banging against each other because of a loose joint = more bell, but = bad. The decay from HHG was amost identical to a single piece of steel that was the same mass.

_________________
I didn't mean to say it, but I meant what I said.
http://www.brackettinstruments.com/

Tim's test was a good one as far as it went, I think. We're up against a lot of issues here. For example; if you want to do the test with wood, where will you find a lot of absolutely uniform samples? Also, suppose the damping factor of the wood is _higher_ than that of the glue: the effect of the glue will just be swamped.

My feeling is that with a proper glue line thickness of .002"-.006" the actual nature of the glue is next to irrelevant in terms of tone. Unless it's glass or rubber cement it's not going to make a whole lot of difference, and even then... At that thickness you're talking about the transmission of _forces_, not _sound_, since the wave length of any sound you can hear will be orders of magnitude larger than the thickness of the glue line. That means that the entire thickness of the glue line will either be in tension or compression at any one instant, and the transition between states will be slow enough that the glue line itself will always be fairly uniformly stressed: all in tension, or all in compression

There's also a lot of 'discussion' as to the nature of the glue bond itself. If it's structural, with the glue interpenetrating the wood and bonding it mechanically, that has different consequenses than if it's a chemical bond. The current weight of opinion as I understand it is that the bond is primarily chemical, which would suggest that the nature of the subtrate (wood, glass, whatever) is less important than the thickness of the bond. We know that HHG bonds very well to glass if the surface is clean and smooth, and it would not surprise me to find that other types do too. If that's the case, then it won't matter what substrate we use for the tests, just so long as all of the glues used will bond well to it.

So, as always, I think we need to do some more testing. Tim's work helps set some limits for the future, and that's worthwhile, but I don't think the final word is in, if it ever will be.

Woody's correct. Tim's testing was well planned, quite thorough and well executed, but the interpretation of the results is wrong and misleading.

hey Grumpy! Just wanted to say it is good to hear from ya

seconded!! Hey how cold is it outside in your neck of the woods Mario?

I totally agree Alan. My testing was only a starting point and I do hope someone will pick up where I left off. In the onset I wasn't sure that I could isolate and measure each glues effect on vibration transmission through it as a medium or even see any difference between samples. It was encouraging to see that the particular testing method chosen could show some graphical differences between some glue. We may very well be splitting hairs but I still hold to the belief that an instrument is affected by the sum of its parts. How much the glue part contributes to the whole will remain a mystery ... at least for the time being.

Mario, please share your thoughts on how you interpreted the data.

_________________
tim...
http://www.mcknightguitars.com

Studies like this make me wonder about some things. It's one thing to prove something with scientific methods and mathematical equations on paper and it's another entirely different thing as to whether you can actually hear the difference or not. Can you build two identical guitars say from the same flitch to the same dimensions one with TB and one with HHG and hear a difference? And further more can you determine that if there is a difference in tone that it's not because of other factors like the stress imparted on the top by the side bends etc...

I don't know really, just thinking out loud and I think it's great to see experiments like this.

It was done and I related the experience on the OLF a couple of years ago. In the production shop where I was working 2 sister guitars were built, one with TB1, one with HHG. The selected sets were as similar as sets can be (red spruce and EIR), in stiffness, weight, tone and appearance. The builds were obviously carried in a parallel manner, the guitars were built at the same time, the plates tuned to sound and feel identical and so on. No effort was spared, I know, I was part of it, and it's a shop that has built over 5000 guitars.
8 weeks later the guitars were set up and ready to play, and everybody gathered to play them (a dozen people). Nobody but the set up person and the head of the shop knew which was which, the guitars really looked identical. We all played the guitars a few times, listened as others played them etc.
General consensus was that the TB1 guitar sounded marginally better. A hair more openness, dynamic range and power. Marginally.
IMHO the difference is obviously not due because TB1 sounds better than HHG, but simply because of the natural and unavoidable variations between builds and sets.
Of course this is one test, carried as best as it could I believe, and may not prove anything. On the other hand, if HHG had such an impact on tone I believe a couple dozen of highly trained ears would have heard something.

Now the funny thing is, both guitars were sent to the same dealer and they absolutely loved the HHG guitar, finding it unmistakably superior in every way to the TB1 one. Of course they knew which was which before opening the cases.

The experiment did nothing to convince me of the tonal superiority of HHG, but everything to prove the ineffable reality of psychoacoustics.

Now, before somebody attacks me, I regularly use HHG and find it has a lot of advantages, I just do not hear a recognizable difference due to using HHG.
I also believe that using HHG probably represents a step forward for a lot of builders as it is a little bit more demanding to use than TB1.

_________________
Laurent Brondel
West Paris, Maine - USA
http://www.laurentbrondel.com/

Thanks for relating your experience Laurent. Many years ago I made the switch from TB1 to HHG and I could have sworn I heard a mean shift in the tone of my guitars. That psycoacoustic perception was one of the reasons I abandoned use of TB1. There are so many blasted variables in the process its really hard to put your finger on any one of them as the major contributor. It was from that point that piqued my interest to test glues. However, the testing didn't materialize until many years later.

The other major reason that I abandoned TB1 is its propensity to cold creep. Having used many gallons of it in carpentry work with first hand experiences of glue joint telegraphing in the harsh winter months was reason enough to make the switch. I still have a guitar hidden away with a top center seam joint that was joined with TB1 that shows an obvious cold creep joint under the finish.

_________________
tim...
http://www.mcknightguitars.com

Tim, I think you did a thorough job in your test, great idea. I wonder how the sustain would change on the baseline sample if it was made of two pieces of steel the same size as the others, properly welded together with an overlap joint? I think it would be different than adding washers to make the massas equal, since the stiffness would be different.

PS, when I say properly welded, I mean that the steel is returned to it's original hardness and relieved of the stress form welding,, not so easily done.

Chuck

_________________
_________________

Alan, it's my understanding also that glue bonding is primarily chemical rather than mechanical. Because it is chemical, it behaves very differently on different materials of the same surface texture. This explains why silane coupling agents are required to bond epoxy resins to glass fibers, and why a phosphoric acid anodizing is needed for a durable bond of epoxy to aluminum. Because the bond is a chemical interaction between the glue and the substrate, some glues bond much better to one material than another.
As Tim makes clear, we're not concerned with making a durable glue joint in the steel samples. But is a glue line that is not adhering to the substrate going to transmit sound in the same way that it would if it adhered? Is the vibration of the test enough to degrade the bond, and reduce transmission? It seems to me that we don't know, and we wouldn't need to know if the samples were wood. Similarly, the tendency of sound waves to reflect off of layers of different density creates a very different picture when using a test material that is 10-20 times as dense as wood.
If the damping of wood is high enough and variable enough that no effect of the glue line could be detected measuring multiple samples under ideal conditions, then I think we can safely ignore glue selection in trying to perfect the sound of guitars. It may well be, as Laurent suggests, that glue selection doesn't affect the sound of a guitar. It also seems possible that there would be an effect that was measurable (and maybe even audible) in a well conducted test of wooden samples, that would be difficult to isolate by listening to completed guitars.
Tim, you've got us all thinking, and talking which is one of the best things a test can do. And you're right, if I think a better test could be done with wooden samples, then I should get off my ass and do it.

I design experiments all the time as an engineer. Tim's test are great for what he was doing. The following is not to be critical, as I am not and the results are good and confirm what was expected. To do a great experiment, more replicates are needed to be more confident in the results. All that takes is a lot more time and money!




This is why I brought up the replicate issue above. As you know, as an experiment, the 2 guitar build did nothing at all to confirm anything. To be done correctly, you need to make about 30 guitars from each glue and then do a large number of blind tests. Obviously, no one is going to do that. Even if someone did, it is not likely that there would be a clear winner because the results would be subjective. Ultimately it would be a waste of time and money.
The debate will continue because all we will ever have is opinions.


One thing interesting to me about Tim's data is that you could chose adhesives based on the sound you want. HHG and LMI white would be better for guitars when longer sustain is important and TB for guitar that you did not want as much sustain.

_________________
"An adventure is only an inconvenience rightly considered. An inconvenience is an adventure wrongly considered." G. K. Chesterton.

But you probably did. Maybe other factors, or a convergence of factors as switching to HHG probably changed a couple of things in your routine.


Tim, I will politely disagree. TB1 has more tendency to creep when heated, as in a car trunk. HHG will simply let go, but at a much, much higher temp. I think Frank Ford did some pretty explicit tests. Otherwise the cold creep TB1 is accused of has no bearing in lutherie, our applications are just not that demanding. Even with 192lbs of string tension. Wood itself cold creeps.
I know 1000s of guitars that have tops joined with TB1 and still show no visible glue line. Maybe your top was baked in a trunk?

_________________
Laurent Brondel
West Paris, Maine - USA
http://www.laurentbrondel.com/

On the contrary, it proved that even if HHG has a positive impact tonally (and I am not denying that it could), it is easily overwhelmed by other factors.
And that's when all other things are more or less equal. And I emphasize "more or less", because the build variations in the experiment were obviously minimized as much as they could. Yet they were enough to make the TB1 guitar the better one, by consensus.
I am not arguing that different glues have no impact on tone. I do not know and certainly can not hear it, or hear what difference it makes. I switched to HHG a dozen guitars ago, and I noticed no difference in my guitars performance that I can attribute to the glue, nor in the tone of the plates when carving braces.
And I'd like to add that I can hear many other minute things, and am pretty well trained as a musician and sound engineer. Just so I do not appear like the Plebeian raining on the parade…

_________________
Laurent Brondel
West Paris, Maine - USA
http://www.laurentbrondel.com/

Mario, please share your thoughts on how you interpreted the data.

Quite simply what Woody has already said; the HHG, followed by the polyurethane glue(if i remember correctly) were the two closest the the solid steel baseline. That the others had more sustain is a -bad- thing, as it shows that they are isolating the two steel halves. You interpreted the increased sustain as being better, because, I imagine(but open to be corrected), we always think of increased sustain as being good in an instrument. But here, it showed exactly what we don't want, and that is that the glue line of those glues is acting to isolate, or de-couple, each half. Same as Woody's example of the cracked crankshaft(old hot rodder and engine builder here, BTW <bg>). Stone masons will do the same when they tap on stone, also, to listen for internal fractures before working with a slab that may fail.

I suspected this from the first time I read the test, but couldn't put my finger exactly on the why of it, but your test was being discussed on a guitar list(it's been discussed on many, many forums and lists) and one doctor type explained the why of it clearly. If I can ever find his response again, I'll post it or mail it to you. If I recall, he was impressed with the test, also, but pointed out that the interpretation of the results was incorrect.

As you know, as an experiment, the 2 guitar build did nothing at all to confirm anything. To be done correctly, you need to make about 30 guitars from each glue and then do a large number of blind tests.

Well, how about hundreds of thousands of examples of each, then? We know the exact date and model where Martin switched to PVA from HHG. Martin has been very consistent in their builds and build qualities over the years. Yes, some things change here and there, but those changes are also well documented, and quite minor, with the bad changes mostly reversed back to the good specs(like their return to small maple bridge plates from the large EIR ones, for example). There's your test. No blindfold needed. And not just tone and response, but if we could gather neck reset numbers, and the amount of reset needed, I'm sure it would be clear which ones have proven more stable; I don't have hard numbers, but I see more 70's, 80's and 90's Martins needing a lot of reset than I do pre-'65. For sure, my own instruments are much more stable since switching to HHG. There are many long term adhesive tests going on in Europe in laminated beam structures, and early-on, they ruled out ALL PVA type glues as unacceptable due to excessive creep and temperature sensitivity(as were all epoxies). Last time I found that article, they were now at 20+ years with something like 10 tons of constant load on the various joints; these folks were serious!

It's not just about tone.

Laurent:
I, too, have tried to build 'identical' guitars. In my case I used redwood tops cut in flitch, and mahogany B&S ditto. The necks were from the same plank, and the fingerboards and bridges were cut in flitch. I matched thickensses, masses and the pitches of the first ten modes of the tops and backs as closely as possible: iirc, the modes were all within 3 Hz, and the plate masses within 3 grams. The bridges were 'identical' within the limits of my measurements. I strung up the two guitars on a Friday, and due to my class on Saturday morming, I only had the opportunity to keep them up to pitch. In the afternoon they were submitted to a Boston Classical Guitar Society performance party for judging. THe upshot was that, though they were very similar, one was clearly prefered. I _think_ i know what was different in that case, but as usual, more testsing is in order.

Eric Reid wrote:
"Alan, it's my understanding also that glue bonding is primarily chemical rather than mechanical. Because it is chemical, it behaves very differently on different materials of the same surface texture."

That I don't know. I'm under the impression that a hydrogen bond is a hydrogen bond, but I'll have to ask my son the chemist. In a test like this, I'd be more concerned that the different glues would bond differently to steel or whatever, than I would be with the differences between their bonding to wood and steel.

I must confess, it's been a long time since I read Tim's report, and there are details I'm forgetting. Let me float my own proposal for a test, and see what people think.

First: what are we looking for? I think there's no argument over the relative ranking of HHG and PVA based glues in terms of cold creep. Performance under sustained shear loads is not something we need to test.

We're discussing the acoustic effects of these glues. In terms of transmission of sound through the thickness of a joint I think I've made my views clear: we're looking at 'forces' there, not 'sound' traveling through these thin layers. There could be effects due to the difference in impedance at the glue line, and that would be interesting to look at, if we could figure out how. But, IMO, the thing that's going to make the most difference is the loss factor in shear loads. This would occur, for example, when the top bends as it is vibrating, and produces a shearing stress in the joint between the braces and the top. In this case, if the glue has any flexibilty it is likely also to add damping. On the surface this would tell against the PVAs, but we can't be sure of how a properly thin glue line would act until we try it.

As much as it would be a wonderful thing to do these tests with wood samples, there are some drawbacks. Most notably, it's very hard to find really uniform wood, so we could never be sure we were comparing apples with apples, so to speak. Also, wood has higher damping than many other materials, and if the differences in damping between the glues were small, they would be swamped by the damping of the wood. Finally, if it's hard to find poor glue lines in steel joints, it's probably harder in wood. Thus, I'd opt for glass. There are very few easily available things in this world as uniform as float glass. It also has very low damping, so you'd be pretty sure that any differences you'd see would be because of the glue. We already know it glues well with HHG, and the others can be tested. Note that I'm not proposing that the bond with glass would be the same as that with wood, just that each glue would (I hope) bond to glass in the same way as every other. This would allow us to isolate variables, and talk about the glue line itself.

I'd make up bars of a convenient size, of two strips of glass glued together for their whole length. It would be simple to monitor the thickness of the glue line and it would be easy to see any gaps. These bars could be vibrated simply in the 'free-free' condition (like a xylophone bar). Assuming the glass was uniform, and we got glue lines of uniform thickness, any differences in pitch and damping would be due to the shear modulus and damping of the glue line. Since we're not transmitting sound through the glue line, we don't need to be concerned with differences in impedance leading to reflection/refraction effects.

Obviously, it would take a while to make up the samples: I imagine the glue would be very slow to dry. Very narrow bars would help.

I can see lots of problems with a test like this. OTOH, it's hard for me to think of a better one.

The two things that keep shouting at me about this test whether using steel plates or glass plates are;

1. As I stated in another post, the baseline model is not consistent with those tested. It would need to have an overlap joint also, since it would effect the stiffness. How do you connect the two pieces without using glue? Answer, weld them together, but the piece will have to be returned to its original hardness and also have the stress from the weld released. Now using glass, I am not sure how to connect the two pieces without glue? But regardless, how can you compare results to a baseline model that is not consistent with the other test samples?

2. Should we really be testing glue with a material that all of the different glues are not designed to adhere to?

Chuck

_________________
_________________

Alan,
How about using microscope slides? They would all be uniform in size and mass. Are you going to excite them ala Chladni style? How would you measure the slides?

_________________
tim...
http://www.mcknightguitars.com

Chuck, I did consider welding but it would add mass and also change the physical property of the steel and likely aneal it. Since I was using cold rolled steel there is no way to return it to its un-anealed state, is that a word?

_________________
tim...
http://www.mcknightguitars.com

Without re-doing your entire test, Tim, could you find a machine shop to mill a solid billet of the same cold rolled steel(I mean the same composition) to the shape of the lap jointed glue pieces?

Or not I think the test stands well, just the interpretation of the results. Even without the solid "standard", the results hold up and show how some glues isolate and de-couple the two halves.