brightness/warmth measurement

[Alan Carruth]

Guitarplayer_PR wrote:

"There's a problem: I don't think brightness and warmth are opposites at all"

...which is the problem in a nutshell.

In the end I think all you can do is use measurements to look for things that correlate with 'brightness' and 'warmth' as you understand them, and then work from there.

The best example I know of that sort of study was published by Dunnewald back in May of '91 in the 'Journal' of the Catgut Acoustical Society, Vol 1, No. 7 (Series 2), pp 1-5. Subjective judgements of careful recordings were matched with objective measurements of the spectra of a large number of violins of varying quality. The qualities used were 'harsh vs clear' and 'nasal vs un-nasal', and these were defined at the outset using tonal examples. Evaluations of the spectra were done using a computer, which divided the frequency range from 190-6400 Hz into seven bands.

The only absolute quantity that was a good predictor of quality was the level of the Helmholtz resonance in the objective test: if it didn't reach an average of 18 dB in the anechoic chamber with a fixed drive power it was not a good fiddle, no matter what else it did. Otherwise the differences were relative; much more output in the 'B' band (650-1300) than the average of A,C, and D (low, and up to 2850 above the B band) resulted in an 'nasal' sound, and so on. The correlations proved to be in good agreement with the judgement of the market: 8.4% of 'factory' instruments met the objective criteria for a 'good' sound as defined in this study, while 30.7% of 'master instruments before 1800', and 92.5% of 'old Italian violins' did.

[KevWind]

Is there a tool that can be used to objectively measure a guitar's brightness/warmth? IMO No not really .
No more than you point to a particular spot in an MIR of the human brain, and say "that's where the mind is located "

[KevinH]

Originally Posted by printer2

As Alan said the time domain is important. What if one guitar has a bright attack and the other one more sustain in the treble? Which one is brighter, you could sum the frequencies and get an average number. Will that one tell you which is brighter. Please excuse the plot being of a bass. I would like to see a plot done on some popular guitars...

I ran some spectra on a couple of my guitars recently - a Taylor 522ce (all mahogany, X-braced), and a Waterloo WL-S Deluxe (spruce/cherry, ladder braced). I did that because those are the two guitars I own that sound the most different to me. I figured that would be the best chance of seeing differences in spectra. To my ears the Taylor is much warmer (yes, a warm Taylor) with a rich sound whereas the Waterloo is more woody and a little brighter.

The difference is really apparent on the bass strings. So to try and simplify things, I ran a spectrogram on the Taylor when just the open E string is plucked, all other strings muted:
TaylorEStringSpectra.jpg

And here is the same for the Waterloo:
WaterlooEStringSpectra.jpg

The numbers show the locations of the partials for the E string, 1 being the fundamental. Because the string was plucked at 1/4 the string length from the saddle (near the center of the sound hole), the 4th, 8th, 12th... partials are muted. The A and T are the locations of the main air resonance and the main top resonance.

To my eye, the main difference between the two is in the relative strength of the orange spike (2nd partial, 165 Hz) and the yellow spike (3rd partial). The Taylor's 2nd partial is more pronounced - because its main resonance of the top is fairly close to this frequency. It decays faster than on the Waterloo probably because, as Alan has mentioned, the resonance sucks the energy out of the string. For a similar reason, the Waterloo has a more pronounced and faster decaying 3rd partial (247 Hz, yellow spike) because of its proximity to its (higher) top resonance.

Does this explain the deeper/warmer sound I hear on the Taylor? Maybe it's the added bass from it's pronounced 2nd partial. And maybe the higher-numbered partials (the green spikes) that seem to contribute more on the Waterloo add to it's brightness. Or maybe not. I suspect timbre isn't so simple.

And, for the curious, here is an E Chord strummed on both guitars:
WLoo vs Taylor E Chord1.jpg

I've heard other people describe the Waterloo WL-S as having a woody tone, attributed to the ladder bracing. I'm not sure how to describe the sound, other than I know when it when I hear it. For a while I thought it was a fast decay, giving more of a "thunky" sound. But, other than the yellow spike, the Waterloo seems to decay at about the same rate as the Taylor.

While looking at these spectra is fun, I agree that interpreting them in terms of warmth/brightness/etc isn't so easy.

[Alan Carruth]

I like to use 'impulse' spectra to look at guitars. One thing I have noticed is that guitars that strike me as 'warm' sounding have a broad output peak around the 'main top' pitch. This can be helped if the 'main back' mode is fairly close in frequency. In that case the two acting together produce a double peak, like a 'Butterworth filter' or a bass reflex cab.

Another thing that helps with this is to get the top 'cross dipole' up out of range. On classical guitars the cross dipole is fairly low, and on Flamencos it's even lower. The dipole 'steals' energy from the top monopole, which is by far the most effective radiator of sound on the guitar, and cuts into the band width. I believe it was Meyers who noted that guitars with a narrow monopole peak tended to sound 'cutting' or even 'harsh' (I have been unable to find my copy of the paper). On a classical this can help, but since there's a lot more power in the high frequencies in steel strings it's not necessary, and having a broad monopole peak helps give 'fullness' to what can be a 'bright' sound.

And, yes, that supports the notion that it's high frequencies that make the sound 'bright', at least to some extent. Note, here, that the narrow monopole band near the fundamental of the G string on a Flamenco can produce a 'sharp' sound, even though there may be no more energy in the 2-4 kHz range than in a similar classical with a higher pitched dipole. There can be several things that contribute; we haven't even gotten to the 'time domain' yet.

[KevinH]

Originally Posted by Alan Carruth

I like to use 'impulse' spectra to look at guitars. One thing I have noticed is that guitars that strike me as 'warm' sounding have a broad output peak around the 'main top' pitch...

Where/how do you generally apply the impulse?

[Alan Carruth]

I have a setup in a small 'semi-anechoic closet' that I've been using for some time now. You can do this sort of thing in a room, but the room resonances can be hard to account for. The closet isn't perfect, but it's always the same.

The guitar is hung 'freely' from the ceiling, with loops of string under the E string tuner buttons. I put thin strips of foam between the strings and the fretboard, at the 4th and 11th frets, which does a pretty good job of damping them. The guitar is hung at a spot about 1/3 of the length of the closet from one end, more or less on the center line. The microphone is one meter out in front, at a height between the bridge and the sound hole, and it's a little off the center line; the guitar is turned slightly to face it.

I use a hard plastic bead, about 5/8" diameter, as the 'hammer'. It swings from the ceiling a little way in front of the guitar on a string; rigid arms introduce resonances. There is a small magnet on the bottom of the ball that can be captured by a steel bolt that is on a solenoid operated piston out in front. Triggering the solenoid releases the ball, which then swings in and hits the saddle. I try to get it to hit between the 4th and 4th strings, but it doesn't always. I may shorten up the string the ball is on to try to improve that.

I'm using an old analog dB meter from Radio Shack as the mic. It's enclosed in chicken wire as a 'Faraday cage' to cut down on 60 cycle hum. The response is pretty flat up to 1 kHz, which is the range I'm most interested in, and I can use the different gain settings to get the best signal without clipping. It's plugged into my old desk top, running some ancient version of Windows, that supports the various sound analysis and editing programs I've been using. I record the tap, erase as much noise as I can, pad out the signal with silence to more than 16384 bits in the 'Cool Wave' program I'm using, and save it to the directory of the ancient FFT4WAV3 program. The FFT crashes if it doesn't get along enough file. That's balanced out by the fact that it can save the 'real' and 'imaginary' parts of the transform as comma delimited text files, that I can open in a spreadsheet. I can calculate the amplitude, convert it to dB if I want, and format the chart the way I like it.

I've been using pretty much the same system for a long time, with incremental updates, but the charts I have from 20 years ago are reasonably comparable. I think the closet goes back about 6-8 years. It's easy to argue for or against various features of this setup; but IMO consistency is a real virtue. The absolute values of any measurement may be suspect, but they're pretty reproducible, so I can compare the most recent instruments with ones from a few years ago with some confidence.