Collecting Spectrographic data

[Dominic]

Hi Trevor, How do you go about holding the guitar when you collect spectrographic data on unfinished guitar bodies. In the appendix you say hold the guitar as if you are playing it and by this I assume you are referring to a finished guitar. Do you hang it from the sound hole when tapping or do you place it on foam blocks or some other method?

Again, loving the books and loving all the technical details and models. Great fun.

Also in the appendix on creating Chladni patterns, you say to house your speaker in a tube with the end closed off. Does this produce better vibrations when creating Chladni patterns at lower sound levels? I am currently mucking around with a open speaker and am not getting much apart from the T(11)2 mode.

Cheers
Dom

[Trevor Gore]

Dominic,

I don't do much tapping until the guitar is "boxed". I've never been able to make much of free plate tapping or tapping without the back, because so much changes when the box is closed. Once closed, you can hold the guitar pretty much as normal, albeit without a neck (or with the neck if you've got that far).

"Boxing" a speaker reduces the out of phase sound coming off the back of the speaker cone. Generally, it makes the speaker less efficient in terms of sound output per watt of electrical power, but the sound you get is more directional. For my set up (YMMV) the "boxed" speaker gives somewhat more guitar top movement for the same amount of noise. How well things work also depends on what you're using as an indicator. I use old dried herbs. Others use poppy seeds, glitter (difficult to remove), saw dust (you need to work on getting the right particle size) tea leaves and then there's the old original lycopodium powder (if you want a bit of fun). You can help the process by tapping first and seeing what peaks come up. Then tune your oscillator to the frequency of one of the peaks (or there abouts) and you should find the resonances quite easily. Just as in indicator of the sound levels - if you're not wearing hearing protection, you should be.

[Dominic]

Thanks Trevor, I am not having any trouble getting the T(1,1)2 mode but the others are harder. (Pleasingly the T(1,1)2 modes don't seem too high at this stage and they don't have bridges yet.) But I can see the other modes are there because there is lots of activity with the tea leaves I am using but I am not getting clean patterns. I am wearing hearing protection and it seems loud so i'll try boxing in my speaker and direct more onto the soundboard. I am using a decent 60w speaker but the amp is a little 40 w kit amp from jaycar. Might have to look out for an old stereo amp. Also, I am using a program called SweepGen which seems OK but only rises in 2 Hz increments. Do you use software signal generator or a dedicated one? I've seen little digital test ones on ebay starting about $50.
Thanks again
Dom

[Trevor Gore]

Dominic,

I'm using 10w RMS and have plenty of power! I think your problem is the 2Hz increment on the sig gen. I use an analogue one and read the frequency with a mic and computer (or you could just use your mobile and a suitable app)

You could try Visual Analyser (Wave tab). VA is a free download. Just Google it. In theory it has 0.01Hz resolution on its sig gen. Feed the output from your computer to your amp/speaker and you should be right, but adjusting the frequency is not as convenient as twiddling an old analogue knob.

[Dominic]

Others who are thinking about setting up to create Chladni patterns might find my experience helpful. As I mentioned below, I am having trouble getting enough output to generate all the modes. I've reading up about software signal generators and some people are saying their output signal is too weak to be useful. I've tried a few and some you have to go into the settings to change the frequency each step which is a real pain. You really want something you can adjust simply with your keyboard or mouse. Not having much luck so far. If anyone is using a good software signal generator let me know.

Since that I have been looking on ebay for machines. You can pay up to $60,000 for one but there are much cheaper ones around for about $100. This one is a low frequency generator from an Australian dealer for $115. We don't need the really high frequencies so I thought this might be useful. http://www.ebay.com.au/itm/Low-Frequenc ... 19c8335685. Given I don't really know what I need I wouldn't mind any comments or suggestions. I suspect many others will be doing this soon as they read through the books and want to start testing so this info will be useful to others.
Also keeping my eye on a couple of old heavy looking ones for less then $100. There seems plenty on ebay but many are overseas. Anyway, I’ll keep people posted about my progress.

Cheers
Dom

[peter.coombe]

I have been measuring Chladni patterns on my mandolin free plates for many years. I use a signal generator I made from a Jaycar kit, a 140W amp from a Jaykar kit, and a 120W speaker also from Jaycar. I added speaker protection to the amp so I don't blow the speraker. It is very easy to blow speakers doing this stuff, I have triggored the speaker protection a few times and the amp does get hot sometimes. Ear muffs are essential. I think most of what I use is unfortunately no longer available from Jaycar. The results have been published in JAAMIM.

For those interested here some links
http://www.petercoombe.com/Chladni_patterns.htm
http://www.petercoombe.com/jaamim7.html
http://www.petercoombe.com/jaamim8.html

Scroll down and you will see a picture of my setup here
http://www.petercoombe.com/Workshop_and_tools.htm

Peter

[Dominic]

Hi Trevor, I wonder if you could explain how you tap the frequency response into Visual Analyser. I have a little pdf that Jeff sent me the link to from your web site which has set up instructions and I have them right so I can tap 10 times around once every 1.6 seconds. But I am getting inconsistent results. Wide or rounded peaks and moving modes. So I figure it is the way I am tapping. Can I tap slower than this to get a good reading or do I need to stick close to the 1.6? Could you explain your process a bit and what II should be doing to get consistent results?
Cheers
Dom

[Trevor Gore]

If you have VA set up correctly and you are tapping once per buffer (~1.6 seconds) it's hard to see what you're doing wrong.

Try tapping in the same place 10 times. As the live screen updates you should get a "flash" picture that looks more or less the same every time. If that is not happening for you I'd be making basic checks. For example, record a tuning fork and see if you get what you expect.

I get fairly nasty looking results if I use my laptop's internal mic. I normally use an external dynamic mic (Shure PG 57) which gives a much better result.

If you can't find the problem, post a shot of your results (use File, Save from the averaging window) and I'll have a look.

[Dominic]

Thanks Trevor, I had better results on a finished guitar by tapping on the bridge wings. But I had the trouble with a classical with no bridge so perhaps I was tapping too much on the centre line. In the book you say to "tap systematically in a consistent pattern", where would you tap out this pattern? Is it best to tap along a line where the bridge would be? I want to take measurements before and after i glue on the bridge so I only want to be measuring the impact of the bridge and not different techniques.

Thanks again Trevor, your books are so interesting and I am having so much fun with this stuff. I hope you get the recognition you deserve for the great service you have done for the luthier community. And I also hope you make good money from them.

Cheers
Dom

[Trevor Gore]

How and where you tap is entirely up to you!

Things to remember:

1) Frequencies of peaks will correlate well to Chladni frequencies
2) The frequencies of the peaks should remain constant until you change something in the "structure"
3) The magnitude of the peaks depends where and how hard you tap, so unless you devise a fairly elaborate piece of kit to systematically tap, your peak heights are fairly random
4) That said, it's possible to be amazingly consistent. Probably to do with log plots!
5) It's possible to identify modes by tapping. For example, tap on the centreline only and if you had a cross dipole peak it will tend to disappear (as you are driving it on a node). It's possible (with a bit of experience) to be reasonably certain of the low order modes just by tapping nodes/antinodes and seeing which peaks grow/decay
6) If you want an approximation to string excitation, just tap along the saddle position.

If you want information more specific than this, you need to invest in piezo hammers, accelerometers etc., but it's a case of diminishing returns.

[luke_lee]

Can someone help with the settings for VA, Im not sure what settings I have to adjust. I have done the Sample size(16384) and sampling frequency (11.025kHz), anything else?

Thanks Luke

[Trevor Gore]

Luke, you can download some more detailed instructions from here:

http://www.goreguitars.com.au/main/page ... nical.html (right at the bottom of the page).

With that and what's in the book you should be right. I still use version 9.0.1, an older version which was still available last time I looked. The later versions are similar but take more clicks to do what I want, which is why I've stuck with V9.

[luke_lee]

Here is the Frequency responece for a guitar I make several years ago. X braced, 14 fret neck joint. It is my best one to date but after read the book I think there is room for improvement.

Luke

[Dominic]

No back (T(1,1)3) mode. So I assume you have a very stiff back on this one. And quite peaky in the lower frequency range with a sharp drop off in the higher frequency range.
Body resonance looks to be between G and G# which is good
Top main freq looks to be around F# and G but its hard to read the scale from the pic. So these 2 modes seem to be around the right place.
Can you describe the sound for us. Punchy, shimmery, bassy, projection etc etc. With this freq chart it should be giving a particular sound.
Cheers
Dom

[woodrat]

Could that be the back at 360ish hertz? Or would a very stiff back not show up as a peak because it hardly moves? Its great that we have some conversations like this going on. I have found that my backs are too loose and it seems that they have split the T(1,1)2 into two peaks as Trevor suggested in Design at 2.3-8. They are ladder braced with 4 braces and I am going to add two shorter braces in between the lowest and the next to lowest brace to stiffen the back up. They will have flat tops so I can plane them down to drop the frequency if I am a bit high. The symptoms sound wise was a clunky 6th string so I hope that it will liven it up when I separate the T(1,1)2 and the T(1,1)3.

John

[luke_lee]

The back is not live, it's a conventional ladder braced. As for the Body resonance it is at 103Hz, Top main is 185Hz and the other peak is 358Hz. I am concerned with frequency falling very close to the G# and F#. Not to sure what to do but I would like to drop the main by 5Hz. The sound is ok, I think it is balanced, the volume and sustain is good. I would say it is lacking responsiveness and tonal contrast. It's hard to evaluate without a direct comparisons to a really good guitar.

Luke

[Somerby]

Trevor - I am trying to work out the kinks in getting VA to work properly. I am getting some results, but I can not be sure that the program is capturing the data correctly - My expectation is that the program will wait for a signal on the input to begin the capture. Instead, I see the capture move around in the display, some times centered, some times at the ends. It is not starting an acquisition based on input. Do I have this wrong?

Thanks!

[charangohabsburg]

Somerby, using VA's menu do this: Settings - Capture - Start capture spectrum

More info and more adjustments here:

Luke, you can download some more detailed instructions from here:

http://www.goreguitars.com.au/main/page ... nical.html (right at the bottom of the page).
[...]

[Dominic]

Trevor - I am trying to work out the kinks in getting VA to work properly. I am getting some results, but I can not be sure that the program is capturing the data correctly - My expectation is that the program will wait for a signal on the input to begin the capture. Instead, I see the capture move around in the display, some times centered, some times at the ends. It is not starting an acquisition based on input. Do I have this wrong?

Thanks!

Somerby, try these instructions I posted recently. Works for me.
Cheers
Dom
viewtopic.php?f=1&t=3696

[Trevor Gore]

For those how haven't spotted it already, there's a "Capture Spectrum" button on the main screen:

[Trevor Gore]

Hmmm, I thought I added a screen shot there. Not sure where that went!

[jeffhigh]

Anyone know how to save your settings in VA, preferably as the default?

[Trevor Gore]

Settings, Save As (on the right)

[Dominic]

Hey everyone, not having any trouble collecting data but I always seem to get a much higher body resonance than the main in my charts. That is T(1,1)1 is the highest peak by about 20%. Everything is in the right place its just the order of the peaks. I have not glued the bridge on yet. Curiously, it does not look like this on the original chart that VA generates, only when I resize/zoom that chart or copy the data into excel. I assume it is either some kind of VA bug, has something to do with the way I hold the guitar to the mic or the way I tap or all three. Perhaps it is a setup issue.
I sit with the guitar soundhole level with the mic and about 30cm away. I tap with a small rubber ball on a stick. Should I be pointing the mic at a different place on the body? Or should it be pointing straight into the sound hole.
Anyone got any ideas?
Can't work out how to copy an excel chart into a thread or i would put one up and show you.
Cheers
Dom

[Trevor Gore]

Hi Dom,

Hey everyone, not having any trouble collecting data but I always seem to get a much higher body resonance than the main in my charts. That is T(1,1)1 is the highest peak by about 20%......I sit with the guitar soundhole level with the mic and about 30cm away.

Yep, that's typical of what you get if you mic that close in front of the sound hole. Try backing off a bit and "pointing" the mic at one of the bridge wings (or where the bridge wing is going to be). If you "point" to the centreline you may find that some of the symmetric modes don't show up well.

Curiously, it does not look like this on the original chart that VA generates

That's definitely curious. Before you scale the chart, save it to disc, find it, then double click on the text file. It should open in Notepad or something. Check down the list of data and find the "peaks" and see if they correspond to the graphic. There will be some minor differences due to the way the data is interpolated for the display, but the numbers are the real data and you should easily be able to match the peaks for amplitude and frequency. If they don't match, there has to be a bug in the graphical display. I use version 9 and have not seen this scaling issue. If you're using a more recent version, maybe try one of the earlier versions. Some are still available from the VA site.