Guitars and Numbers

[pan_kara]

Good ideas guys, thanks. There are a lot of potential tests that I could do once I get the set-up more or less finalized..

I was thinking of a "scale-length" comparison stringing the guitar with strings say from 008 to 012 or so and tuning them such that they produce the high E on different frets. Taking a single string and just testing the behavior vs tension is another thing, though here a big factor of how it sounds, at least for lower tension, is the fact that the string goes sharp when plucked (and its easy to press it out of tune at the fret too).

The pickup excercise should be easy to do, if you get too close to the strings with a pickup the magnetic field "will" change how the string decays, sometimes in funky ways. I'll try to take a look at that.

[neotronic]

Good ideas guys, thanks. There are a lot of potential tests that I could do once I get the set-up more or less finalized..

I was thinking of a "scale-length" comparison stringing the guitar with strings say from 008 to 012 or so and tuning them such that they produce the high E on different frets. Taking a single string and just testing the behavior vs tension is another thing, though here a big factor of how it sounds, at least for lower tension, is the fact that the string goes sharp when plucked (and its easy to press it out of tune at the fret too).

The pickup excercise should be easy to do, if you get too close to the strings with a pickup the magnetic field "will" change how the string decays, sometimes in funky ways. I'll try to take a look at that.

Hi,

you might find this interesting: http://www.novaxguitars.com/info/technical.html

[pan_kara]

Interesting, like most people he doesn't look at the time dependence of stuff (just frequency spectra) but there is some nice info in there, thanks. He's also discussing some second-order effects, that I'm mostly trying to neglect here (different vibration modes etc).

I've been moving ahead a bit and analyzed the first set of measurements I made, with the two bridges I'm trying to compare for a start.

First of all its pretty clear by now that my method is immune to pickup choice. Here's a set of seven plots for the first harmonic of a G#3 note (G string, 1st fret) plucked seven times. The two lines are the neck and bridge pickup signal. The pattern seen here continues through all plots from different strings and harmonics - as expected the only difference is the amplitude (in some rare cases there is some noise polluting the singal, and then sometimes the two lines show more variation).

This is good since it means I don't need to use the same pickups to get comparable results. As expected.

Fast forward to my first actual result (sort of):

Here I have the same G# note, with the decay slopes averaged from 8 string plucks. The blue line "bands" represent the RMS widths of the averages, and the two lines are the Hannes in RED and the no-name in GREEN (I need to add a legend ....).

After the first 8 or so harmonics the measurement becomes unreliable (with some exceptions higher up). But below we can see a very clean effect of switching the bridge. The question is - how to interpret it. Does the no-name bridge dampen 800-1000 Hz or is this just something seen on this string at this fret.

I'll look at overlaying the results I have for the E A G and B strings, plus I need to do something in the direction of the scale-gauge comparisons - check the same note on different strings (normal tuning, just fretting to get the same pitch). There will be string-to-string differences of course (themselves interesting to look at) but what I really want to see is how the bridge-to-bridge differences will carry across.

Ok, more recording then.

EDIT: oh, and while I'm at it I'll verify the point that Psikot brought up since I now have to go back to the Hannes (the string set is still the same one)

Edited May 26, 2014 by pan_kara

[pan_kara]

Ok, time for some thread necromancy

wow, this really is from 2014... ? ... ? oops

So I'm bringing my "tone" investigation back to life. It got interrupted because I couldn't get consistent results and didn't understand why. At some point in time I think I did manage to figure out what was happening there.

In short:

My main idea was to try to see if I can find some numerical differences between the string overtone decay rates when comparing the Schaller Hannes bridge and some cheap hardtail. In the end my main problem appears to be caused by me having reused the same set of stings for all measurements - the idea was for that to ensure consistency, but it ended up messing everything up because the strings got deformed at the saddle break point and due to different mounting this kink then ended up in the "active" (vibrating) part of the string when swapping bridges, causing totally unpredictable results...

We'll I'll be more careful this time.

Right now I'm trying to again establish the framework, including verifying various assumptions that I'm making. The main idea is (repeating some of the stuff I was mentioning earlier) that to study "tone" one should look not at the spectrogram (frequency content) of the sound of a plucked string, because this will depend strongly on a number of factors like type of pick, picking position, picking strength, picking motion etc etc

Instead we should look at the time dependence - how fast the individual overtones decay. This is largely independent on the initial conditions (picking) as the decay speeds are at the end of the day the result of properties of the guitar-string system itself. So is it really them that define the "tone of the guitar".

In such an approach where I decouple the initial conditions all experiments are much easier (no need so special jigs to replicate series of identical string plucks), plus I also take all electronics out of the equation (except for things like the magnetic field of the pickups influencing the string motion, something that we already discussed in this thread and something that I plan to address at some stage) - I'm really looking just at the string vibration itself, the source of the sound. Pickups come in later and shape it.

So what's the plan now? I'll try to release this as a series of youtube videos, and possibly also a write-up of some sort. Plus regular updates and hopefully discussion in this thread. I'm nearly done with the opening part, which is experimentally demonstrating that indeed the pick attack doesn't influence the results. Then I have a series of simple comparisons lined up - how the sound changes when changing fret material, nut material, string gauge .. maybe a few like that, before I'm confident I can attack everyone's favorite - wood choice.

More soon

 

[Prostheta]

I've spent a few hours thinking about this before replying. I keep typing a paragraph and then deleting it. I am sure that you have thought about the variables and potential for them causing exponential increases. You definitely go through a lot of strings!

[pan_kara]

I have thought about things and went through a bunch of literature that I could find online. I'm not sure if my method will prove fully correct, I guess we'll have to look and see. I'll wait until I have some solid conclusions before I start putting this series on youtube

And of course there are many variables here, I'll have to split them according to which ones I can control, neglect, or have to take into account. For example the string pluck does play a non-trivial role as I'm now discovering, most likely due to the delicate relation between the two perpendicular string vibration modes (up-down and side-to-side). But looks like my method is still holding up in general, I just have to be mindful of that. 

As for strings, well, here's my starter purchase - 10x sets of 3 strings for tests plus low E going from .042 to 0.054 in steps 0f .002. And a few nuts - brass, bone, graphite (I already have tusq and plastic). 

The guitar that I'll be using for most of these tests is my old nylon string build - because I want to do as much as possible with a piezo. Piezo gives me a much broaded palette of overtones - they reach higher and are not attenuated due to phase-cancellations (something I was trying to mitigate in the past by recording both bridge and neck pickup simultaneously in stereo and then picking the one with better amplitude for each overtone separately). As a bonus - I have no magnetic pickups here so any influence from those is automatically removed - but can be added in if I want to test it, by suspending a pickup above the strings in a location mirroring actual position in a guitar.

[pan_kara]

So here's a little sample from my current initial study that will probably form episode one.

I recorded a series of notes on an open A string, hitting it in different ways:

note 1: normal pick, normal position a little in front of bridge HB
2: back - above bridge HB
3: front - above neck HB
4: middle - above 12th fret
5-7: soft wood pick, normal position
8-9: hard wood pick
10-11: shell pick
12-13: finger pluck
14-15: metal wrench

Then I did my analysis on 40 (!) harmonic components of the sound. Yes, with the piezo I can still make out the 4400 Hz overtone from a 110 Hz low A - though it dies out completely in 0.5 seconds.

So for example here's the 440Hz overtone - aka 2 octaves of the 5th fret natural harmonic:

You can see that it can be of different strength initially (for example over 90dB when plucked with a hard shell pick or about 70dB when plucked with a finger), but the exponential decay is nearly identical across all notes. The only thing is the weird oscillations present in the 3rd note. Overall the overtone decays at 5.4 +/- 0.4 dB/s (average from the 15 measurements).

If we look higher, for example at 996 Hz, we have this:

still pretty consistent, though the oscillations appear to be confusing the straight-line fit a little so there's a bit more variation, overall we clock 10.6 +/- 1.2 dB/s.

So what are these oscillations? For some frequencies they're there always (though can vary in size), and for some they come and go depending on the note. In general this type of modulation is a sign that the oscillator I'm looking at is coupled to another oscillator of a similar frequency, in which case energy will flow back-and-forth between them, causing this modulation on top of the normal exponential decay. The "other oscillator" can be several things. First of all it can be the other transverse component of the vibration - the piezo is mostly sensitive to up-down motion in the guitar plane (into-away from the bridge), but there is also the sideways motion that can in general have a slightly different frequency. The string is not anchored in the same way in these two directions, for example it might move a little side-to-side in the saddle, thus detuning the "sideways overtone" slightly and causing this "pumping". This I managed to observe in fact, I did a test with another guitar plucking the string with the same pick, in the same place, but with different motion, trying to have more of the sideways vibration one set of notes, and more up-down vibration in the other set. And indeed for some harmonics I see the modulation only appear (or be more pronounced) in one of the groups. For example:

and:

(first three notes are hit in one way and the subsequent four in a different way)

For some overtones it's less clear, sometimes the vibrations are there in all cases, sometimes the behavior is more erratic. But it does appear that their presence is in most cases correlated with the type of pick attack. 

What this means is that if I try to maintain a consistent style of picking for all studies I will probably get a more uniform set or results since the straight-line fit is a bit sensitive to these oscillations. On the other hand I don't see a systematic effect (like pulling the slope measurement in one direction when the oscillations appear), so in the end I should be fine whatever I do. 

Other things that could be causing these oscillations could be the interference with longitudinal vibrations in the string (there's also torsional vibrations i.e. the string core rotating around it's axis, but it appears that those can be ignored) or vibrations is the "rest of the guitar". Both are something I just have to live with - it will be interesting to see how that pattern changes when I change different components of the system. But that's for later. 

[pan_kara]

All right, I think I'm ready to show a first actual result. Or at least a result teaser.

I'm currently recording note samples with my selection of different nuts - so basically "how doest the guitar nut material impact tone".

Of course what I'm discovering first is how does the quality of the nut slotting job impact tone, but with that (hopefully) under control now, I'm starting to be able to actually observe stranger things.

Just two words about the setup: the fingerboard is flat so I'm just resting the nut somewhere behind the 1st fret  and letting string tension keep it in place. In fact now to make sure it's enough, I additionally clamp it down with a capo (of course taking care to note have the capo touch the strings). I make sure the vibrating lenght of the string is constant across all measurements (615mm IIRC) and all the strings are tuned to E. I'm measuring across 3 octaves with a .046, .026 and 0.010 gauge. Then there's a ton of felt and stuff to mute irrelevant parts of the string etc. The guitar is laying flat on some foam, connected to the sound card and to ground with a separate wire. Then I record a series of ~25 note plucks for each data point.

At some point I'll put out a video showing all of this, here's just a frame grab to illustrate the setup:

continued in next post

[pan_kara]

I promised some results, so here goes. I've been working on the actual signal analysis quite a bit and most stuff is not processed yet - it actually takes a while to run all the calculations for for development I was only working with the first 5 harmonic overtones on some first recordings. But at least I have enough stuff for first comparisons. So here is the low E string for a Tusq nut (RED color in all the following plots) and a brass nut (BLUE in all the plots).

First, the fundamental. Here is all the notes for both bridges, superimposed. Plotting the amplitude in dB vs time in seconds, but I'm normalizing all the lines such that they peak at 100 dB (I'm interested in the shape only anyway)

 

Pretty consistent decay between the two nuts, some interference patterns with very low frequency, in general looking good. So now for a clearer look, I'll average all the graphs for both notes. So only two lines remain:

I also zoomed in a little and smoothed the graphs. So now we start seeing some differences. In fact, let's zoom in more - looking at the note decay over one minute is interesting if I want to measure the sustain for example, but for just comparing the harmonic content difference between the two setups, I'd look at the first 5 or 10 seconds of the signal, which is how long notes get to sound in music normally.

So, zoom in more and remove the smoothing:

a yet different picture. So overall the fundamental doesn't differ much between the two nuts, the attack is similar (the first peak is in the same place), just that for the TUSQ after the initial attack the vibration actually increases over the first 1-2 seconds before going into the decay. So probably some energy first goes into some mode that we're not picking up and then comes back. 

The structure of this attack is something that I might still look into, maybe my normalization is messing with it a little. Clearly the string takes a moment to start properly vibrating after being struck with the pick. For the higher harmonics this is less messy.

Here is the first octave overtone (I'm only showing the zoomed-in plots):

 

Nice match, TUSQ clearly decays a little bit faster, we quickly have a 2-3 dB dip at 167 Hz.  

Next overtone:

here the difference is bigger, 250 Hz dies out even quicker with a TUSQ nut.

 

AAt 330 Hz we pretty much have a match. There's some oscillations visible for the BRASS nut, but no big deal.

 

Finally, at 415 Hz again the TUSQ nut decays faster, but not dramatically.

That's all I have for now. I should be able to look at the higher harmonics soon, and also at the other strings. Of course higher up is where interesting stuff should be happening - we get into how "bright" the tone is and how the higher frequency content is shaped. I'm also wondering to what extent the results will be compatible across the three strings - if I see an effect at 2 kHz for example, will that be visible for all three, indicating that this is really a propery of the nut that I'm probing. We'll see.

BTW @Prostheta I keep wondering what was that paragraph that you kept typing in and deleting  I would be very much interested in hearing your and others' thoughts on this. I hope it will get more interesting soon!  

[Prostheta]

I think the biggest challenge is to factor out or quantify every single variable. In terms of the "nut slotting job" being a potential for adding in variability, perhaps a string run that is pulled over a straight course via an unslotted nut would allow for the material to be observed in experiments rather than the slot. Hold up, you just replied....

[Prostheta]

Wow, lots of data. I'm particularly interested in the TUSQ nut, since I use those myself. Albeit, in a different context. Generally they are only used as guiding nuts behind a zero fret, which is great since they're pre-slotted. Brass also. I'm a fan of the material by its acoustic quality in bridges and nuts....do you think that you can summarise any findings from your initial tests, or is this going to take a LOT of data point gathering?

[pan_kara]

1 hour ago, Prostheta said:

I think the biggest challenge is to factor out or quantify every single variable. In terms of the "nut slotting job" being a potential for adding in variability, perhaps a string run that is pulled over a straight course via an unslotted nut would allow for the material to be observed in experiments rather than the slot. Hold up, you just replied....

Actually now that "nut slotting" is in writing I gave it some more thought and indeed a dedicated test would make sense (the list is growing endlessly...). The only thing is the actual break angle - I might need to redo the brass low E measurements, when I was doing them initially I heard some faint buzzing that I couldn't locate no matter how hard I tried - I looked at the tuners, other parts of the string, loose felt pieces, loose nut, loose frets?? - (the guitar has no truss rod so that's out of the question). Finally I realised that the string is rattling in the slot in the nut. Looks like I made it to flat (parallel to the fingerboard surface) to the string wasn't resting hard enough against the nut edge, causing buzzing. I ended up clamping it down and got rid of all the buzzes but I'm not 100% convinced that was "legal". I already see that there is a similar problem with the slots that I cut in the bone nut, so I have to postpone the recording of the bone samples until I have my nut slots back (in a few days).

But I could compare a nut that has an angled surface with the string running over the flat edge, against having a slot cut for the string. Actually if I rest the string between the slots in the graphite nut that is installed at the moment, that might work.. hmm...

Anyway, in general I agree with your comment, Carl, my first thing to check was the validity of the method. Whether I can take out of the equation the things that I want to take out - the electronics, the initial conditions (string pluck) etc. Not having the influence of those things removed is for me the biggest flaw of all the other approaches to this problem that I came across. Strangely, I haven't found anybody doing this my way, which continues to puzzle me since for me this is the most obviously correct approach (unless it's in fact incorrect, for some reasons that I'm still missing).

That was step zero, from what I learned up to now it is a success.

Step one is checking the precision of the method - how repeatable are the results if I hit the string many times, possibly in different ways. What kind of effect I can consider an actual phenomenon and what falls below my precision or statistical uncertainty. The plost that I just posted illustrate that to some extent - the differences from changing the nut material are clearly bigger than the variations between the individual notes (measurements).

So step two is to learn how big is the influence of various things. Right now I'm looking at nut material, for most other tests I want to use a fretted note as baseline (when comparing scale lenghts, string gauges, bridges, finally woods, bolt-ons etc). Then the nut goes out of the equation. I'll compare different fret materials to see if I can measure any difference caused by that. And so on.

1 hour ago, Prostheta said:

Wow, lots of data. I'm particularly interested in the TUSQ nut, since I use those myself. Albeit, in a different context. Generally they are only used as guiding nuts behind a zero fret, which is great since they're pre-slotted. Brass also. I'm a fan of the material by its acoustic quality in bridges and nuts....do you think that you can summarise any findings from your initial tests, or is this going to take a LOT of data point gathering?

I use TUSQ for pretty much everything. Except when putting in a locking nut. Zero frets not yet but in general I like the idea and might switch to that. Ok, another idea since you mentioned zero fret - I can add that to the comparison. If I fret the strings at the first fret. The scale will be a few mm off, but I hope that won't affect the result much. (will know later when I do the big scale length set of measurements....)

For now I there's not a lot to summarize, you can see all that I have in the plots - brass generally has a bit more sustain than Tusq, at least in the low mids. Should be able to look past 1-2kHz soon, that's going to be interesting. And then all the other materials - plastic, bone, fretted note - when I plot all of these together we'll see how they compare. 

[Prostheta]

Setting as level a playing field for all of the test subjects is the only way to prevent other factors affecting what you're attempting to examine and quantify. I shouldn't need to tell you this of course, since you're the physics scientist and I'm a wood butcher. 

It would be interesting to compare the results from the materials, ordering by mass, etc. I almost wrote that "brass rings like a bell", whilst that should be "bell bronze". hahaha

TUSQ does have an interesting tap tone when several nuts are shaken in a tray, that's for sure. Not sure how - or if - this translates to desirable properties. Do you have any Corian to test? I could send you some if you want.

[pan_kara]

hmm nope, I don't have any Corian. Before we get into mass ordering or any other ordering I guess I have to still prove that I'm seeing a difference between the materials (BTW speaking of mass, I could also see if adding mass to the headstock makes a measurable difference). I'm starting to get there. I still need to work out some loose ends, the slotted vs unslotted test needs to be done properly (I only did a quick one). Plus I'm starting to wonder if string aging can be impacting my results (I still have the same set of strings on since I started the nut tests, it's been like a month or two - but I practically don't touch the strings ...)

But while I cross-check those things - I got my analysis to a point where I can do comparisons between several recordings, and I have the full set of measurements for four nuts now. So here's the fundamental frequency for about 25 notes recorded for brass, tusq, graphite and  an old roller nut that I found. Each nut is represented by a different color.

it's not super pretty, but the colors do appear to form different structures. It's easier to see what's happening if I make a new plot by averaging all the notes from a given nut:

 

By the way - I'm normalizing all the plots to have the peak at the same height. This is in line with my concept of measuring the decay properties (that are a parameter of the system) rather than the initial conditions (which depend on how the string was hit). So, anyway, There appears to be some distinction between the different colors, though I have to visualize also the uncertainty/precision of the averaged graph so that we can see if the differences are within statistical error.

But as we go to higher harmonics, the picture quicky clears itself out quite a bit:

Here the separation is clearly visible. 

So how to summarize these results without looking at 50 plots showing the decay of different sting vibration modes (which is what I anyway did)? We can show the plot vs frequency - we know the frequency of each overtone. So what de we plot? I went away from fitting straight lines to these plots for the moment, end I'm doing a conceptually simpler thing: since these plots are all normalized, all the overtones begin decaying from 0 dB. So one could ask a question - how much will they have decayed after - for example - 2 seconds. Well we can plot this vs frequency:

So this is already starting to tell us something - looks like the brass nut is sustaining best, closely followed by the tusq.  The roller nut is worse, and the graphite nut is even worse. This is most apparent in the 1-1.5kHz region. For the lower overtones the differences are more subtle. 

In any case all of these follow a similar shape, that determined by the rest of the elements of the test system (body, neck, bridge etc).

[curtisa]

2 hours ago, pan_kara said:

So this is already starting to tell us something - looks like the brass nut is sustaining best, closely followed by the tusq.  The roller nut is worse, and the graphite nut is even worse. This is most apparent in the 1-1.5kHz region. For the lower overtones the differences are more subtle. 

So sustain on open strings is a function of rigidity and hardness of the nut material. Which is why a softer material (graphite) or a mechanically sloppy material (roller nut) will sustain less than solid brass, If you could mill a nut from stainless steel I'd expect it to perform as good as, if not better than brass.

The roller nut may also be hamstrung in that there is also less material to transfer the strings vibrations into the surrounding structure of the neck and body. So despite being made from similar materials as the solid brass nut, each roller has to spin freely on a tiny pin, the string being in contact with only a tiny portion of that roller as it passes over, and the roller itself having contact with only a tiny portion of the pin.  Whereas the solid nuts have all their mass directly seated into the neck and the string (ideally) in contact with the full length of the slot as it travels from tuner to neck.

Have you tried bone yet as well? My gut feel is that it'd be similar to graphite in terms of overall sustain.

Thus, at the other end of the guitar, I'd expect to see a similar correlation in sustain when comparing between roller saddles, brass, steel, or graphite, inlcuding the way the saddles are mechanically coupled to the body - fully floating trem, trem blocked off, hardtail, tune-o-matic etc.

Mind you, I'd also expect the differences between nuts to vanish as soon as you fret a note. Comparing differences in sustain between nut materials is fine, but at the end of the day that's only six notes out of a possible 144, and not many players will find an application for just plucking the open strings at 40 second intervals...except for perhaps bands like Sun O))) Perhaps you can investigate using fretted notes as control tests to verify that the age of the string isn't gradually skewing your results as you go?

[Prostheta]

Awesome! I figured few people here would be aware of Sunn O))). Somehow it doesn't surprise me that you would be one of them Andrew! Haha

I always had the impression that brass is more sonorous than stainless. That would be an interesting match.

[curtisa]

Thread derail!

Knew the name Sunn O))) for years but wasn't familiar with their music. Was only just recently reminded to check them out on Youtube via another unrelated video I was watching. Drone doom metal \m/

[pan_kara]

4 hours ago, curtisa said:

Thread derail!

Knew the name Sunn O))) for years but wasn't familiar with their music. Was only just recently reminded to check them out on Youtube via another unrelated video I was watching. Drone doom metal \m/

lol now I'll need to check them out.

perheaps they would be interested in my results

[pan_kara]

8 hours ago, curtisa said:

So sustain on open strings is a function of rigidity and hardness of the nut material. Which is why a softer material (graphite) or a mechanically sloppy material (roller nut) will sustain less than solid brass, If you could mill a nut from stainless steel I'd expect it to perform as good as, if not better than brass.

Clearly this is in line with the expectations. Though for the higher strings I'm having slightly mixed results. In general I've been testing materials that I've seen nuts made out of so didn't consider steel. I do consider it for a fret test, which is probably next in line - currently I have EVO gold in that guitar, so I plan to record a sample with that and then pull the first fret and put a stainless one in, then normal nickel silver, then back to EVO gold. I expect the differences to be much smaller (if measurable at all - we'll see).

8 hours ago, curtisa said:

Have you tried bone yet as well? My gut feel is that it'd be similar to graphite in terms of overall sustain.

Thus, at the other end of the guitar, I'd expect to see a similar correlation in sustain when comparing between roller saddles, brass, steel, or graphite, inlcuding the way the saddles are mechanically coupled to the body - fully floating trem, trem blocked off, hardtail, tune-o-matic etc.

Bone is in the queue. As is a piece of rosewood that I wanted to have a slighlt more extreme case. And a plastic nut.

The bridge side is another thing that I'm thinking of, though replacing bridges on the guitar is not something that's on my shortlist at the moment (especially after the previous failure). Swapping out saddles in a hardtail is something that I could imagine. Also I do plan to do a floyd rose floating vs blocked test. (I'd do a replacement "tone block" test also if I had one laying around).

8 hours ago, curtisa said:

Mind you, I'd also expect the differences between nuts to vanish as soon as you fret a note. Comparing differences in sustain between nut materials is fine, but at the end of the day that's only six notes out of a possible 144, and not many players will find an application for just plucking the open strings at 40 second intervals...except for perhaps bands like Sun O))) Perhaps you can investigate using fretted notes as control tests to verify that the age of the string isn't gradually skewing your results as you go?

I agree that "nut tone" doesn't immediately impact most of what you hear from the guitar as any impact vanishes the moment you fret a note. And in fact I forsee nearly all of my other measurements to be done fretted - to have the nut out of the equation.

Here's why I think this study is useful

• I'm still figuring out the method, so this is a good test case, allowing me to keep most things constant
• I can see how big the effect from the nut is, I can compare this later to other things I'll be measuring - does changing the nut make more difference than steel vs nickel frets? Than mahogany vs alder body? Than old vs new strings?
• This will to some extent translate into the differences caused by different bridge materials

The idea about using fretted notes as a control sample vs string aging is a very good one, sadly I didn't consider string aging when I started the current recording batch so I can only test from now into the future. I have one other option though: some of these have been re-recorded since in the first batch I was mostly doing short 6-10 second notes, and for the low E to really see the decay you need closer to a minute. But in fact with the way I'm looking at this now (average RMS after 1-2 seconds or so) the old recordings would be perfectly fine to compare. So I need to make some plots for those, maybe I'll be able to see if there is any effect going.

I hope not, otherwise I'd have either to keep chaging the strings, or do everything as quick as possible which would mean having to set aside quite some time for sitting down and plucking a string once every 30-60 seconds for ... I don't even want to cound how many times ..

[ScottR]

Now you've got me wondering what (if any) impact string gauge has on sustain.....

SR

[pan_kara]

It's coming

On 11/17/2017 at 11:40 PM, pan_kara said:

As for strings, well, here's my starter purchase - 10x sets of 3 strings for tests plus low E going from .042 to 0.054 in steps 0f .002. And a few nuts - brass, bone, graphite (I already have tusq and plastic). 

 

[pan_kara]

I'm constantly learning things with this project. Turns out my method is super-sensitive to various things. Which echoes @Prostheta's comment about having all variables under control. I keep discovering new ones. I already mentioned some of them - nut slotting and sting age. String age started being an issue since I was on the same set of strings from the moment I started the nut comparison recordings - they have been taking a while to complete, I think ~2 months or so. Every once in a while I had some time to do some recordings, once or twice I figured I need to redo everything because of some new discovery...

To put things in perspective - my current test sample is about 25 notes in sequence letting the string fully ring out. For the low E that's about a minute. So every recording is about 20 minutes of basically just sitting there and hitting the string every 40-60 seconds. 

Anyhow

The difference between a 2-month old almost-not-played string is small but definitely measurable, as I verified over the weekend but simply putting in a brand new string.

I was also having repeatability issues, doing seemingly the same measurement after a day or two was giving me slightly different results. Well guess what. It depends on how exactly the body is supported. I made a series of tests, just changing how the guitar is positioned on the pink foam that it's laying on (see pic 10 or so posts back). There's a clear trend. It's not very strong, but it is there, especially for the lower frequencies. Which makes sense, I'm damping body resonances in different ways when I change how the guitar is supported.

So I had to start from scratch again. I made a new setup with the guitar supported by 4 bench cookies on very rigid support and then I put on the fresh E string and recorded all the nuts in one day, so I'm pretty confident the only thing that's changing is the nut. I checked the tuning of all strings after every nut swap. I checked that the guitar is in the same spot each time...

I think the results are much more consistent this time. I need to work on the data analysis a bit and then I'll post something.

Also, please welcome the new guest star in the nut shootout:

[Prostheta]

Hopefully you don't get excellent results with a pencil and end up breaking the Internet. We need that thing!

[pan_kara]

That was my secret plan but looks like the numbers speak otherwise. Funnily enough, one nut from my test collection sounds (and measures) almost as bad as the pencil ... 

[Prostheta]

Yeah yeah, guitarists blaming their nut again. Heard this story before! hahaha