absolute sound level
absolute sound level
I am relatively knew to audacity. I am using it for a component of my masters thesis; I'm looking to capture the sound from stethoscopes' earpeaces to analyze differences in amplification heart sounds. However, as far as I can see, within each fft, the db reading is normalized to the 'loudest' sound in the file. I need something I can compare accross files. If I knew what the loudest intensity was in each file, I could calculate these values directly. It would be nice though to change the unit on the y axis to something comparable across multiple files. Can you help?
-
kozikowski
- Forum Staff
- Posts: 69374
- Joined: Thu Aug 02, 2007 5:57 pm
- Operating System: macOS 10.13 High Sierra
Re: absolute sound level
Might you be overthinking this?
The waveform on the timeline (when changed to dB instead of percentages) will tell you accurate dB levels of the peak of each sound. The wave is completely dependent on the input signals from the microphone and given one single stable microphone and multiple samples, should give you the numbers you wish -- relative to each other and not some arbitrary fixed reference. Convert to Average, RMS, or on a good day, dBSPL (C) for speaker presentation.
I can certainly see problems at the capture step. Audacity's frequency response will go down to DC. Battery voltage. You can edit a 3 volt watch battery into a show consisting of two 1.5 volt flashlight batteries. That's really bad in audio systems, but terrific for you.
What are you using for a microphone and how are you digitizing it? I believe blood pressure monitors don't use microphones. They use pressure transducers which won't work with violins and cellos particularly well, but they will produce DC voltages and work just fine at cardio frequencies.
http://kozco.com/tech/audacity/AudacityPanelFull.jpg
You will be in mono and not stereo, so your waveforms will be even bigger and more accurate.
Note that there are only level numbers on one half of the wave and even worse, the bouncing light sound meter only works on the upper half of the wave. And yes, I would be using Audacity 1.3 instead of 1.2. 1.3 has much better display tools.
Koz
The waveform on the timeline (when changed to dB instead of percentages) will tell you accurate dB levels of the peak of each sound. The wave is completely dependent on the input signals from the microphone and given one single stable microphone and multiple samples, should give you the numbers you wish -- relative to each other and not some arbitrary fixed reference. Convert to Average, RMS, or on a good day, dBSPL (C) for speaker presentation.
I can certainly see problems at the capture step. Audacity's frequency response will go down to DC. Battery voltage. You can edit a 3 volt watch battery into a show consisting of two 1.5 volt flashlight batteries. That's really bad in audio systems, but terrific for you.
What are you using for a microphone and how are you digitizing it? I believe blood pressure monitors don't use microphones. They use pressure transducers which won't work with violins and cellos particularly well, but they will produce DC voltages and work just fine at cardio frequencies.
http://kozco.com/tech/audacity/AudacityPanelFull.jpg
You will be in mono and not stereo, so your waveforms will be even bigger and more accurate.
Note that there are only level numbers on one half of the wave and even worse, the bouncing light sound meter only works on the upper half of the wave. And yes, I would be using Audacity 1.3 instead of 1.2. 1.3 has much better display tools.
Koz
Re: absolute sound level
Thank you for responding. I still have a problem though. I've been looking at other posts and doing some reading about how FFTs are calculated, but I am still a bit confused. Is there no way to get comparable values out of an FFT? db is a ratio...so what is the denominator for this measure of db (what's the reference)?
I was hoping to use audacity to compare the FFTs from different files; basically I want to be able to say the intensity of a particular frequency is different between files. I was thinking of inserting a "standard" sound into each file in the hopes that I could still generate a valid comparison between frequency intensities of different sounds using audacity. Does this seem valid?
Thanks for your input!
I was hoping to use audacity to compare the FFTs from different files; basically I want to be able to say the intensity of a particular frequency is different between files. I was thinking of inserting a "standard" sound into each file in the hopes that I could still generate a valid comparison between frequency intensities of different sounds using audacity. Does this seem valid?
Thanks for your input!
-
kozikowski
- Forum Staff
- Posts: 69374
- Joined: Thu Aug 02, 2007 5:57 pm
- Operating System: macOS 10.13 High Sierra
Re: absolute sound level
Gee, I was hoping someone else was going to jump in here.
The dB reading, once you're inside the digital system is very firm. 0dB is the maximum level possible, also known as 0dBFS -- Zero dB Full Scale. That's where you run out of numbers; in a fuzzy sense, that's where all the digital numbers (100010100111010) turn to "1".
So given, say, a video system, 48000, 16, Stereo, I can send a -20dB tone (standard tone reference in the US) from Los Angeles to New York Network Operations and they will derive exactly the same tone at the same level as I put it in. No variation and it doesn't matter how many satellite hops it took to get there.
It seems to me your real problem is an open ended capture system. For a true scientific comparison, you would fix the electronics -- and here we're assuming your stuff is up to the task -- and connect it to a standard human with a standard arm and a standard pulse. Capture a bit of that and then change arms to the test subject. Repeat the capture. What you do with the data later, Fast Fourier Transform, Peak Waveform Analysis, ANSI C16.5 Volume Meter, or wet fingers on a dry day is entirely up to you having already done everything you could to capture a correct reference and a test. You can't just do the test.
A conventional blood pressure test uses a Mercury Sphygmomanometer (or equivalent) and your ears (or equivalent). The test varies with human hearing, but it's usually close enough to go on. The fixed reference is the mercury column (very accurate) and the nurse's ears (somewhat less so). The generic philosophical reference is the knowledge that variations in blood pressure kill people, and you can draw graphs of blood pressure versus death over many people.
I'm reeeely interested in the capture step. How are you going to calibrate the pressure sensor or microphone? I glossed over the physical equipment, but blood pressure is an odd duck in that it has components of both pressure differential (atmospheric pressure is going down. It's going to rain unless you live in Los Angeles) and musical instruments (G Pedal, 16 foot Flute, any good pipe organ).
Koz
The dB reading, once you're inside the digital system is very firm. 0dB is the maximum level possible, also known as 0dBFS -- Zero dB Full Scale. That's where you run out of numbers; in a fuzzy sense, that's where all the digital numbers (100010100111010) turn to "1".
So given, say, a video system, 48000, 16, Stereo, I can send a -20dB tone (standard tone reference in the US) from Los Angeles to New York Network Operations and they will derive exactly the same tone at the same level as I put it in. No variation and it doesn't matter how many satellite hops it took to get there.
It seems to me your real problem is an open ended capture system. For a true scientific comparison, you would fix the electronics -- and here we're assuming your stuff is up to the task -- and connect it to a standard human with a standard arm and a standard pulse. Capture a bit of that and then change arms to the test subject. Repeat the capture. What you do with the data later, Fast Fourier Transform, Peak Waveform Analysis, ANSI C16.5 Volume Meter, or wet fingers on a dry day is entirely up to you having already done everything you could to capture a correct reference and a test. You can't just do the test.
A conventional blood pressure test uses a Mercury Sphygmomanometer (or equivalent) and your ears (or equivalent). The test varies with human hearing, but it's usually close enough to go on. The fixed reference is the mercury column (very accurate) and the nurse's ears (somewhat less so). The generic philosophical reference is the knowledge that variations in blood pressure kill people, and you can draw graphs of blood pressure versus death over many people.
I'm reeeely interested in the capture step. How are you going to calibrate the pressure sensor or microphone? I glossed over the physical equipment, but blood pressure is an odd duck in that it has components of both pressure differential (atmospheric pressure is going down. It's going to rain unless you live in Los Angeles) and musical instruments (G Pedal, 16 foot Flute, any good pipe organ).
Koz