Description of an experiment to investigate the relationship between sustain and the construction of the neck joint of solid body electric guitars. This is a reprint of an article originally published in American Lutherie.
Last updated: January 23, 2020
Copyright © 2007 by R.M. Mottola
[This article was originally published in American Lutherie #91, Fall 2007]
Abstract- An experiment was conducted to ascertain what effect the type of the neck joint of an electric guitar has on sustain. A test instrument was built using neck through construction. Audio recordings were made of this instrument using a uniform picking mechanism. The neck was sawed off and then attached with screws (bolt-on configuration) and audio recordings were again made in the same manner. The neck was then glued in place (set neck configuration) and allowed to dry, and audio recordings were again made. Spectrographic analysis was performed on these recordings and averaged sound clips were produced for listening evaluation. Sustain for each iteration of the instrument was measured. Listening evaluation did not indicate any difference in sustain among the three instrument iterations. Measured sustain values indicated that the bolt-on neck iteration produced the greatest sustain.
A long standing bit of conventional wisdom has it that the type of construction used in the neck joint of an electric guitar has some influence on the duration that notes will sustain. A straw poll taken in preparation for the research detailed in this article suggested the perceived ordering of this influence is as follows:
I could find no formal research on this subject, and informal observation is problematic enough so that it is unlikely that reasonable conclusions could be made from it. Although in theory it would be possible to sit down to compare, say, a Gibson Les Paul (set neck) and a Fender Stratocaster (bolt-on neck), it would be difficult to accurately attribute any difference in perceived sustain to the construction of the neck joints of these instruments. This is due to the large number of other construction differences between these instruments that could conceivably affect sustain. These include scale length and string tension, neck shaft geometry, neck shaft material, body geometry, body material, bridge construction, pickup type, pickup placement, etc. From a research perspective, each one of these independent variables could conceivably have an independent effect on sustain. So it is not likely that a simple comparison between so differently constructed instruments would make for a meaningful analysis of neck joint type and sustain.
This article describes a series of experiments designed to provide some data which may be helpful in considering the relationship between neck joint type and sustain. The experiments made use of a series of three purpose-built test instruments (photos 1, 2). These were two string, solid body guitar analogs intended to control for as many independent variables as possible. The core of each instrument was a 25.5" scale length solid hardwood neck/body core with a fingerboard glued on. Glued to the sides of this core were hardwood slabs. The general appearance of these instruments was similar to that of very rough neck through solid body electric guitars. A single magnetic single coil pickup was set into a pickup cavity routed into the instrument top. The pickup was shielded with foil and connected with shielded wire to a jack in a small metal box screwed to the side of the instrument.
A conventional plastic nut and two standard tuning machines were used. The instrument was initially slated to use a Fender style surface mounted bridge, but preliminary experiments indicated that large sustain variations may be associated with small changes in string breakover angle at the saddle of that bridge style (a subject itself worthy of research), so a bridge/string anchor unit made of three blocks of solid aluminum was used instead (photo 3). The instruments were strung with the high E (0.009" plain) and low E (0.042" wound) strings from sets of new D'Addario EXL120 electric guitar strings. The strings were carefully wound on the posts of the tuning machines, as they would be removed and replaced twice during the course of the experiment and it was critical that they be able to survive this abuse.
Pickup height was adjusted relative to the underside of the strings. A number of preliminary tests were made to determine an optimal height for the pickup. Bringing the pickup close to the strings increases the gain of the signal, but if it is too close then magnetic "string capture" can damp string vibration. A relative string height was chosen that offered high gain but no obvious damping, then the pickup height was lowered a bit more to provide a margin for measurement error. This height, relative to the top of the pickup, was recorded for use in subsequent setup of the instrument.
The basic procedure for the data collection part of the experiment was as follows. As initially constructed, each instrument was an analog of a neck-through electric guitar. The strings were tuned to pitch using an electronic tuner (high E at 329.628 Hz, low E at 82.407 Hz). The test instrument was positioned on its back on a foam pad on the workbench, with an additional foam pad under the neck at the nut. Fifteen long duration digital recordings (16 bit, 44.1 kHz sampling rate) were made using the audio input and sound card in a laptop computer from notes struck on each of the two strings for later analysis. The recording software was Total Recorder (http://www.highcriteria.com. The instrument was un-strung and then the neck was sawed off from the body. The cut was made to approximate the dimensions of a typical bolt-on neck and neck pocket. The neck was then screwed to the body using standard Fender-style neck attachment hardware. The instrument was then re-strung using the original strings and brought up to pitch. Pickup height relative to the strings was set to the originally specified value. Digital recordings were made of this bolt-on iteration of the instrument. Then the instrument was un-strung again and the neck was glued to the body using woodworking glue. The glue was allowed to dry overnight and then the instrument was restrung using the original strings and brought back up to pitch. Pickup height was checked and reset as necessary. Again, digital recordings were made of this glued/set neck iteration of the instrument.
As can be seen, most of the possible influences on sustain were well controlled by using identical materials and construction for the iterations of the test instrument. This included the use of the neck attachment plate and screws in all iterations of the instrument, not just the bolt-on version. Photo 4 shows the neck hardware on the neck-through iteration. One thing that was not identical among the iterations of the test instrument was a result of cutting the neck off. Although small, the saw kerf eliminated some mass from the structure and also resulted in slightly different geometry, as the neck sat a bit lower in the bolt-on and set neck iterations than it did in the neck-through version. To compensate for this and to check to see if it was significant, the first test instrument was also subjected to a fourth iteration. After the bolt-on iteration was readied and recorded, a second bolt-on iteration was made. This one included a veneer spacer sandwiched between the neck and the neck pocket. The spacer was made from the same wood species as the neck and body core and it was dimensioned to the same thickness as the saw kerf made when the neck was sawed off (photo 5). Subsequent analysis revealed that there was no substantial difference whether the spacer was used or not.
One possible source of influence not controlled by the design of the experiment was the sequential nature of the sampling. The neck through variation was always sampled first, followed by the bolt on, then by the glued neck iteration. About the only possible influential independent variables I could think of that might be a consequence of this sequencing were related to changes to the physical properties of the strings themselves as a result of repeated picking and repeated un-installation and re-installation and then being brought back up to pitch. A preliminary test which subjected the strings to the same treatment they would receive during the actual experiment but which did not involve changes to the neck joint of the test instrument did not reveal any changes to sustain values however.
Although the methods used to analyze the recordings did not require that each note be struck using identical picking force, a simple excitation method was used that maintained a good degree of similarity across trials. Each note was struck using the wire pull method. A small diameter (44 AWG, 0.00198") length of shielding wire taken from the shield of a piece of audio cable was looped around the string to be picked and positioned at the end of the fingerboard (18.5" from the nut) (photos 6,7). The wire loop was pulled horizontally until it broke, releasing the string to vibrate. This technique has been used in a number of experiments and it seems to yield quite consistent results. In the experiment detailed here, this picking method yielded a mean power value of 51.24 dB for the first 10 milliseconds of signal, with a standard deviation (an indication of variability among trials) of 1.39. These mean and standard deviation values indicate that 95% of all picks performed by this method yielded initial signal levels within 3 dB of the average value. Roughly speaking, 3 dB represents the smallest change in power that results in a reliably perceived change in audio volume.
|High E String||Low E String|
|Mean||Std. Dev.||Mean||Std. Dev.|
|Bolt-on wo/ spacer||3.35||0.35||15.80||1.99|
|Bolt-on w/ spacer||3.58||0.17||17.08||1.52|
Table 1 - Average sustain values in seconds for four different types of neck joint.
The collected data were analyzed by three different methods. The first analysis compared power plot data from the trials recorded. A power over time plot was generated for each trial using the facilities provided for this in Wavesurfer sound manipulation software (http://www.speech.kth.se/wavesurfer). The time of initiation of the tone was determined for each trial as was the power level at initiation. The time at which the power level decayed -21dB from the initial power level was determined and the duration of the note was calculated as the time difference between initiation and the time at which power decayed -21dB. Using duration values from signal onset to a fixed relative power level made it possible to meaningfully compare trials that were not initiated with identical excitation (picking) force. The -21dB drop was selected arbitrarily - the note was certainly still audible for some time after its level had decayed -21dB. But at -21dB it had not dropped too near the noise floor, even for relatively fast decaying tones. These duration values were considered as representative of sustain for purposes of comparison. Duration values for all 15 trials for each string/test instrument iteration were averaged. The results appear in table 1 and figure 1.
Perhaps surprisingly, the neck through configuration yielded the lowest average sustain for both the high E and low E strings. Sustain values for the high E string trials were similar for both of the bolt-on neck configurations and the glued neck configuration. Comparing averages of the low E string trials, the two bolt-on neck configurations show sustain values higher than both the neck through and glued neck configurations.
All low E string sustain values were large enough to yield musically sufficient sustain values for all but the most extreme musical material. The lowest average value observed (12.13 seconds for the neck through configuration) is the equivalent of more than three tied whole notes at a slow 60 beats per minute tempo. For this reason subsequent analysis focused on comparing recorded tones for the high E string only. Since the two bolt-on configurations yielded such close results, subsequent analysis involved only the bolt-on with spacer neck configuration.
The second method of analysis compared spectrograms of the first two seconds of recordings of the high E tones for three of the test instrument configurations. Although the power based comparison above is useful the results would not necessarily correlate with human perception of sustain for these test instrument iterations. For example, rapid decay of high frequency partials could make sustain appear to be lower in a sample that had overall higher power. Likewise a sample in which power predominated at frequencies to which human hearing is most sensitive could appear to sustain longer than a sample that had overall higher power.
Since sustain values differed from trial to trial using the same test instrument configuration it was not likely that a reasonable comparison could have been made between individual trials from different test instrument configurations. For this reason ensemble averaged tones were used for the spectrographic comparisons. Each of the fifteen trial recordings for each instrument configuration was trimmed to start at signal onset using the Audacity audio editor (http://audacity.sourceforge.net). The aligned signals for each instrument configuration were then ensemble averaged (mixed down) using the Audacity editor. Spectrograms of each signal average for the three instrument configurations were generated using Wavesurfer software (Hamming window, 512 point FFT window length, analysis bandwidth = 689 Hz with a window width of 64 points), and are compared in figure 2.
Overall power around the fundamental (329.628 Hz) as well as the second and third harmonics is highest and sustains longest for the averaged signal from the bolt-on neck configuration of the test instrument, followed by that of the glued neck configuration. In the band centered around 3500 Hz sustain is highest for the bolt-on configuration as well, but second highest for the neck through configuration. Differences in the bands centered around 5500 Hz and 7500 Hz can also be seen, with sustain in these bands again being greatest for the bolt-on configuration.
The third method of analysis was to attempt to ascertain relative sustain values by listening to the ensemble averaged audio files generated for spectrographic analysis. Although this analytical method may seem the most obvious and certainly the most meaningful for practical applications of instrument construction, it turned out to be difficult to arrange an appropriate presentation for the recordings. When played back sequentially and in random order using the Audacity audio editor software through high quality closed ear headphones it was not possible for five different listeners to discern any difference in sustain among the three recordings. The most optimal method of presentation involved the simultaneous playing of two files at once, one played through the left headphone and the other played through the right. Here too the five listeners could not consistently identify any one recording as sustaining longer than the other two. Since the pitch of the note was the same in both recordings presented simultaneously, with instructions to determine the position in the left-to-right stereo field that the sound appeared to be coming from it was possible for the five listeners to reliably identify the recording from the bolt-on neck iteration of the test instrument as appearing the loudest. But even with this information the listeners could not reliably identify one of the three recordings as sustaining longer than the other two.
Although results are not shown here, the entire experiment as described was repeated for three different test instruments. Results were similar for the power and spectrographic analyses and identical for the listening evaluation. Although experiments involving a larger population of test instruments would be desirable, the experiment as conducted does suggest that there is little practical difference in sustain of notes based solely on the construction type of the neck joint. This may be a fruitful area for additional study.