Clean-sweep acoustic guitar recordings?

[-JK-]

Hi!

Would you sweep acoustic guitar tracks with a high Q (say 100) and look for resonances or would you go for a broader Q? And: is there an EQ that can detect resonances? A dynamic one maybe? ;-)

Regards, JK

[-JK-]

Does anyone know what I'm talking about? :-)

[-JK-]

No one? Where else could I post this then?

[Mark A. Jay]

Originally Posted by -JK-

Hi!

Would you sweep acoustic guitar tracks with a high Q (say 100) and look for resonances or would you go for a broader Q? And: is there an EQ that can detect resonances? A dynamic one maybe? ;-)

Regards, JK

It's a broad question...so my generic answer might not be what you seek per se. However, I can tell you that one thing I would recommend (when looking for resonance frequencies) would be to look at it as a sonogram. Because of the time-frequency paradox in an FFT, I would (assuming a sample rate of 44.1 or 48 kHz) perform the sonogram using a moderate line size (maybe 4096) but with a high percentage of overlap - maybe 75% or so.

To do this though (overlap) it has to be done as a post-process (from a recorded file) as you cannot, by definition, do overlap on a 'live' signal.

However, this begs a bigger question - are you looking for specific resonance frequencies from the body of the guitar? Are you looking in a specific location?

I'm assuming the answer here is 'yes', and so my advice would be to place several mics close to the physical location in which you expect to see resonance frequencies, and record all channels synchronously, then perform the sonogram on each.

At that point you could (in theory) slice out the resonance frequencies using an FIR and create time domain files from them, and look at their respective decay rates. However, the issue remains that to get repeatable and reliable data, you would have to ensure that the excitation (i.e. the strum or the string-pick) is very much the same from run to run.

If you are looking for resonance frequencies from the body proper, then you could do a sort of modal analysis (but have microphones available as well, which would allow you to compute the P/F (pressure/force) functions). By doing this, you could impact the guitar using a force hammer, which would then allow you to normalize the tests (because you would be dividing by the imparted force). This would allow you to have the complex FRF as well as coherence and thus give you an idea of where the resonances lie.

The bigger problem with things such as guitars though, at least from a measurement perspective, is that the modal density gets crazy-complex, and your modal accuracy (spatially) is defined by where you place the accelerometers (or as quasi-surrogates, microphones). So, if you happened to not choose wisely, the modal will not reveal that which you seek.

You could employ laser vibrometry (such as those from Polytec et al), but I suspect that this is outside the scope of your interest or will to employ such visualization tactics. Alternately, you could use visualization systems that emply eitehr beam-forming or acoustic holography. In any event, these things are pretty much relegated to a Noise and Vibration / Sound Quality laboratory as the associated analysis hardware and software are well beyond the economic means of anyhting but a large company.

Are you trying to do a comparative analysis of the guitar(s)? Are you attempting to see where the resonances lie so that you can apply some form of damping to alter the radiation?

Again, sorry that I cannot provide a direct answer to your query. Perhaps if you could shed a bit more light on the topic?

Lastly, if you want some expert advice / opinions on guitar resonance frequencies, you should contact Dr. Mark French at Purdue University. He's a good friend of mine and this sort of thing is his specialty - even teaches guitar-building courses and has worked with some of the heavy-hitters in the guitar industry. Check this link: http://www2.tech.purdue.edu/met/facstaff/rmfrench/

If you contact him, feel free to mention me by name.

[-JK-]

Hi Mark!

Thanks for your amazing reply! I never thought of getting down to the root of the trouble and analyze the guitar itself, but your scientific approach surely is interesting enough to give it a try someday! For now I found a way to deal with room (and probably guitar) modes quickly and easily with RND's Unique-lizer which has a built in spectrogram view and let's you notch out harmonics. Now I only gotta find a way how to notch them out just when the ringing occurs. Any side-chaining idea handy? :-)

Regards, JK

[Mark A. Jay]

Originally Posted by -JK-

Hi Mark!

Thanks for your amazing reply! I never thought of getting down to the root of the trouble and analyze the guitar itself, but your scientific approach surely is interesting enough to give it a try someday! For now I found a way to deal with room (and probably guitar) modes quickly and easily with RND's Unique-lizer which has a built in spectrogram view and let's you notch out harmonics. Now I only gotta find a way how to notch them out just when the ringing occurs. Any side-chaining idea handy? :-)

Regards, JK

...it sounds like (and correct me if I have mis-interpreted you) you want to make the filter(s) gain response a function of magnitude. That is, the filters should be transparent when no ringing occurs, but active when it does.

Is my interpretation of what you seek correct?

Bear in mind that filters are by their nature linear, that is, if enabled (switched on) they always do what you tell them to do, without regard to the magnitude of the signal wit which they are presented. However, there may be a means by which you can trigger the filter to turn on and off, with the trigger being a question of magnitude for the resonance frequency (ies).

One possibility that comes to mind would be to build this functionality out of pieces. I'm thinking that perhaps you could first use a 'send' channel and pass that through a filter (or filters) that are part of a noise gate that could then be used to trigger the filter on and off. In this way, I can see how you might be able to make the gate act as a switch that wouold turn the notch filter on or off.

I suppose another method would be to make a simple filter and then scale it by -1 (that is, 180 degrees of phase shift) and have that go active as a function of the noise gate level (or perhaps just define filters in the path of the noise gate send, and multiply that result by -1.

Keep in mind I'm just pulling this stuff out of the air (I'm kind of a minimalist when it comes to mixing, so I've not tried any of these concepts), and I suspect that there is a plug-in / plug-ins available that allow the user to do precisely what you seek. Again, I'm really not a big effects user, so there are bound to be otehrs out there reading this who are more accomplished at their DAW's capabilities and who might know of a better way, or a simpl plug in to do what you seek.

In the end though, in a block-wise way, what you seem to be asking is what I have described (at least conceptually). You could always resort to brute force; you could make the DAW learn to deepen the notch(es) as the signals are mixed down - this would be the simplest thing to do if the number of times this resonance issue occurs, but if it is frequent, then this approach is probably going to cause you to go blind, turn your hair grey, or both.

Anyway...I hope this helps.

[pkautzsch]

Originally Posted by -JK-

Now I only gotta find a way how to notch them out just when the ringing occurs. Any side-chaining idea handy? :-)

Regards, JK

Waves C4 with no makeup gain?

[-JK-]

Wow, took me over a month to reply on this... Didn't think I was THAT busy! :-D Well, anyway, you got the problem right, Mark, unfortunately neither a gate nor automation does the trick as the ringing has similar power as the other frequencies and it does occur on every second chord or so. ;-)

A multi-band compressor didn't suit my needs as well as even with a C6 the most narrow Q wasn't anywhere near as narrow as I needed it to be. So I guess I'll stick with a constant filter until I stumble upon a dynamic one with notching option. :-)