Home |
Search |
Today's Posts |
#1
Posted to rec.audio.pro
|
|||
|
|||
Getting accurate sound levels in spectrum analysis
I installed a freeware plug-in on Sound Forge that provides a spectrum
analysis. When I analyze something, it seems like there's a lot of sound energy at low frequencies, even though I don't seem to be hearing that much at the low end. There's a "slope" adjustment in the analyzer, but I'm not sure what it does--can someone explain it to me? I tried generating some white noise and then adjusting the slope so that the spectrum was relatively flat (since I presume that white noise contains equal amounts of sound energy at all frequencies), but I'm not sure that this accomplished what I want. I'd just like to see the actual sound levels for each frequency. If it makes a difference, the audio editing program is Sound Forge (the Audio Studio version) and the plug-in is VOXengo SPAN. |
#2
Posted to rec.audio.pro
|
|||
|
|||
Getting accurate sound levels in spectrum analysis
Mxsmanic wrote:
I installed a freeware plug-in on Sound Forge that provides a spectrum analysis. When I analyze something, it seems like there's a lot of sound energy at low frequencies, even though I don't seem to be hearing that much at the low end. There's a "slope" adjustment in the analyzer, but I'm not sure what it does--can someone explain it to me? I tried generating some white noise and then adjusting the slope so that the spectrum was relatively flat (since I presume that white noise contains equal amounts of sound energy at all frequencies), but I'm not sure that this accomplished what I want. I'd just like to see the actual sound levels for each frequency. If it makes a difference, the audio editing program is Sound Forge (the Audio Studio version) and the plug-in is VOXengo SPAN. Here are the Fletcher-Munson curves: http://blogs.msdn.com/blogfiles/audiofool/WindowsLiveWriter/LouderSoundsBetter_12855/FletcherMunson_EqualLoudness2.jpg -- Les Cargill |
#3
Posted to rec.audio.pro
|
|||
|
|||
Getting accurate sound levels in spectrum analysis
Les Cargill writes:
Here are the Fletcher-Munson curves: http://blogs.msdn.com/blogfiles/audiofool/WindowsLiveWriter/LouderSoundsBetter_12855/FletcherMunson_EqualLoudness2.jpg But I don't want the loudness measurement, I want the actual sound pressure level, since that is independent of anyone's hearing thresholds. |
#4
Posted to rec.audio.pro
|
|||
|
|||
Getting accurate sound levels in spectrum analysis
"Mxsmanic" wrote in message ... I installed a freeware plug-in on Sound Forge that provides a spectrum analysis. When I analyze something, it seems like there's a lot of sound energy at low frequencies, even though I don't seem to be hearing that much at the low end. What Les was trying to cue you into is the fact that your ears are increasingly insensitive to frequencies as they drop below 1,000 Hz. Your ears are 60 dB less sensitive at 20 Hz than 1 KHz. If it is less than 60 dB SPL you might not hear it at all! There's a "slope" adjustment in the analyzer, but I'm not sure what it does--can someone explain it to me? It's got markings, right? If they are like pink, white, etc., then they refer to whether the plot shows equal intensity per frequency or equal intensity per band. I tried generating some white noise and then adjusting the slope so that the spectrum was relatively flat (since I presume that white noise contains equal amounts of sound energy at all frequencies), but I'm not sure that this accomplished what I want. I'd just like to see the actual sound levels for each frequency. Then, don't apply any slope. That should make your FFT flat for a white noise input. If it makes a difference, the audio editing program is Sound Forge (the Audio Studio version) and the plug-in is VOXengo SPAN. I was kinda under the impression that SF had its own FFT facility. Most DAW software does! |
#5
Posted to rec.audio.pro
|
|||
|
|||
Getting accurate sound levels in spectrum analysis
Mxsmanic wrote:
I installed a freeware plug-in on Sound Forge that provides a spectrum analysis. When I analyze something, it seems like there's a lot of sound energy at low frequencies, even though I don't seem to be hearing that much at the low end. There's a "slope" adjustment in the analyzer, but I'm not sure what it does--can someone explain it to me? There are two possibilities. First, the lowest bins on an FFT system may not be accurate unless the window is made very, very large. You may want to play with that. Secondly, your speaker system probably just can't reproduce low end accurately. I tried generating some white noise and then adjusting the slope so that the spectrum was relatively flat (since I presume that white noise contains equal amounts of sound energy at all frequencies), but I'm not sure that this accomplished what I want. I'd just like to see the actual sound levels for each frequency. For that you'll need a sound level meter with some accurate calibration. --scott -- "C'est un Nagra. C'est suisse, et tres, tres precis." |
#6
Posted to rec.audio.pro
|
|||
|
|||
Getting accurate sound levels in spectrum analysis
In article ,
Mxsmanic wrote: Les Cargill writes: Here are the Fletcher-Munson curves: http://blogs.msdn.com/blogfiles/audiofool/WindowsLiveWriter/LouderSoundsBetter_12855/FletcherMunson_EqualLoudness2.jpg But I don't want the loudness measurement, I want the actual sound pressure level, since that is independent of anyone's hearing thresholds. Then you will have to measure it. However, you should be aware that it will change dramatically when you move the listening position by a foot. --scott -- "C'est un Nagra. C'est suisse, et tres, tres precis." |
#7
Posted to rec.audio.pro
|
|||
|
|||
Getting accurate sound levels in spectrum analysis
On Feb 7, 8:06*am, "Arny Krueger" wrote:
"Mxsmanic" wrote in message ... I installed a freeware plug-in on Sound Forge that provides a spectrum analysis. When I analyze something, it seems like there's a lot of sound energy at low frequencies, even though I don't seem to be hearing that much at the low end. What Les was trying to cue you into is the fact that your ears are increasingly insensitive to frequencies as they drop below 1,000 Hz. *Your ears are *60 dB less sensitive at 20 Hz than 1 KHz. *If it is less than 60 dB SPL you might not hear it at all! There's a "slope" adjustment in the analyzer, but I'm not sure what it does--can someone explain it to me? It's got markings, right? If they are like pink, white, etc., then they refer to whether the plot shows equal intensity per frequency or equal intensity per band. I tried generating some white noise and then adjusting the slope so that the spectrum was relatively flat (since I presume that white noise contains equal amounts of sound energy at all frequencies), but I'm not sure that this accomplished what I want. I'd just like to see the actual sound levels for each frequency. Then, don't apply any slope. That should make your FFT flat for a white noise input. If it makes a difference, the audio editing program is Sound Forge (the Audio Studio version) and the plug-in is VOXengo SPAN. I was kinda under the impression that SF had its own FFT facility. Most DAW software does! It is also interesting to note that I have seen two styles of FFT analyzers. One style has a constant resolution BW (standard FFT ) and therefore will display a flat spectral density with WHITE noise input and a decreasing density with a PINK noise input. The other type sometimes called a real time audio analyzer has a BW that increases with frequency i.e. it has say 5 bands per octave for example. This type will display a flat spectral density with a PINK noise input and a rising density with a WHITE noise input. You need to know what kind of ruler you are using. Mark |
#8
Posted to rec.audio.pro
|
|||
|
|||
Getting accurate sound levels in spectrum analysis
Scott Dorsey writes:
Then you will have to measure it. That's what I'm trying to do, but it doesn't seem that my spectrum-analysis plug-in provides a clear indication of actual intensity in each frequency band. What does the slope adjustment do? |
#9
Posted to rec.audio.pro
|
|||
|
|||
Getting accurate sound levels in spectrum analysis
Arny Krueger writes:
What Les was trying to cue you into is the fact that your ears are increasingly insensitive to frequencies as they drop below 1,000 Hz. Your ears are 60 dB less sensitive at 20 Hz than 1 KHz. If it is less than 60 dB SPL you might not hear it at all! But there is a certain intensity recorded in the recorded material itself. I want to see what that intensity is at each frequency, whether I can hear it or not. It's got markings, right? If they are like pink, white, etc., then they refer to whether the plot shows equal intensity per frequency or equal intensity per band. Not markings, just a number, from like -9 to +9 dB/octave. Then, don't apply any slope. That should make your FFT flat for a white noise input. It seems to ... so does that mean that slope = 0 shows the actual distribution of sound at different frequencies? I was kinda under the impression that SF had its own FFT facility. Most DAW software does! This is the consumer version. Either it doesn't have it or I'm not sure where to look for it, but someone pointed me to the VOXengo free plug-in, so I've been trying to use that. |
#10
Posted to rec.audio.pro
|
|||
|
|||
Getting accurate sound levels in spectrum analysis
Scott Dorsey writes:
There are two possibilities. First, the lowest bins on an FFT system may not be accurate unless the window is made very, very large. Which window? Secondly, your speaker system probably just can't reproduce low end accurately. I'm not listening to the sound so much as looking at it. For that you'll need a sound level meter with some accurate calibration. Just to look at the recording? That shouldn't be necessary, since the sound levels in the recording are fixed by definition. I set the slope to zero, which seemed to flatten out the spectrum for white noise. So far, so good. But then I looked at a recording I made on the street, and it seems that the bulk of the sound in the recording is at very low frequencies, below 100 Hz. Is this an accurate reflection of what the sound is actually like outside, or is it an artifact of the microphone or the recording process, or what? When I was actually recording it and monitoring it, the sound in the headphones seemed virtually identical to what I heard without the headphones, but it puzzles me that I see such an enormous proportion of the sound at such low frequencies. Where could all that low-frequency sound be coming from? There wasn't any wind, so it wasn't that. |
#11
Posted to rec.audio.pro
|
|||
|
|||
Getting accurate sound levels in spectrum analysis
Mxsmanic wrote:
Scott Dorsey writes: Then you will have to measure it. That's what I'm trying to do, but it doesn't seem that my spectrum-analysis plug-in provides a clear indication of actual intensity in each frequency band. Then get one that does, like SpectraFOO. SpectraFOO will also allow you to adjust the window width and shape, so you can see what this does to the bottom bins. What does the slope adjustment do? It's an EQ that lets you go from white to pink, etc? --scott -- "C'est un Nagra. C'est suisse, et tres, tres precis." |
#12
Posted to rec.audio.pro
|
|||
|
|||
Getting accurate sound levels in spectrum analysis
Mxsmanic wrote:
Scott Dorsey writes: There are two possibilities. First, the lowest bins on an FFT system may not be accurate unless the window is made very, very large. Which window? The window over which you're collecting samples to put into the FFT. Secondly, your speaker system probably just can't reproduce low end accurately. I'm not listening to the sound so much as looking at it. So you're looking at the signal coming out of the noise generator, without actually putting them through the speaker system and reference microphone? In that case, what you see should be constant over whatever interval the noise is constant over. Except the lowest FFT bins may contain trash because that's how it is. The longer you set the window on the FFT, the lower frequency you'll see accurately, and the longer it will take to update the display. For that you'll need a sound level meter with some accurate calibration. Just to look at the recording? That shouldn't be necessary, since the sound levels in the recording are fixed by definition. I thought you said you wanted to measure what you were HEARING. The vast majority of response aberrations in what you're hearing are due to the speaker. I set the slope to zero, which seemed to flatten out the spectrum for white noise. So far, so good. But then I looked at a recording I made on the street, and it seems that the bulk of the sound in the recording is at very low frequencies, below 100 Hz. Is this an accurate reflection of what the sound is actually like outside, or is it an artifact of the microphone or the recording process, or what? When I was actually recording it and monitoring it, the sound in the headphones seemed virtually identical to what I heard without the headphones, but it puzzles me that I see such an enormous proportion of the sound at such low frequencies. Where could all that low-frequency sound be coming from? There wasn't any wind, so it wasn't that. Set the window so it's as wide as the whole recording, and see what the low end looks like. If you're trying to look at the spectrum in realtime, you are taking periodic chunks of it, and how low the FFT is valid at will depend on how long a chunk you are taking. --scott -- "C'est un Nagra. C'est suisse, et tres, tres precis." |
#13
Posted to rec.audio.pro
|
|||
|
|||
Getting accurate sound levels in spectrum analysis
Mxsmanic wrote:
Les Cargill writes: Here are the Fletcher-Munson curves: http://blogs.msdn.com/blogfiles/audiofool/WindowsLiveWriter/LouderSoundsBetter_12855/FletcherMunson_EqualLoudness2.jpg But I don't want the loudness measurement, I want the actual sound pressure level, since that is independent of anyone's hearing thresholds. Well, when I do a spec an on generated white noise in CoolEdit, the result is ruler flat. -- Les Cargill |
#14
Posted to rec.audio.pro
|
|||
|
|||
Getting accurate sound levels in spectrum analysis
Scott Dorsey writes:
So you're looking at the signal coming out of the noise generator, without actually putting them through the speaker system and reference microphone? Yes. I'm just trying to get the spectrum to match what the actual signal should be like. For white noise, I presume it should be flat over the entire frequency range. In that case, what you see should be constant over whatever interval the noise is constant over. Except the lowest FFT bins may contain trash because that's how it is. That's how it is? Why is it that way? I thought you said you wanted to measure what you were HEARING. Oh no, I'm just trying to get a visual image on the screen that accurately reflects the frequency distribution of the sound in the recording. That way I'll be able to correctly assess parts of the sound spectrum that I may not be able to hear. For example, if there's a great deal of high- or low-frequency noise in the recording, beyond my range of hearing, I'd like to see it on the screen, so that I can take steps to remove it if necessary. Set the window so it's as wide as the whole recording, and see what the low end looks like. If you're trying to look at the spectrum in realtime, you are taking periodic chunks of it, and how low the FFT is valid at will depend on how long a chunk you are taking. OK, I will try that, if this plug-in allows it. |
#15
Posted to rec.audio.pro
|
|||
|
|||
Getting accurate sound levels in spectrum analysis
"Mxsmanic" wrote in message
... Scott Dorsey writes: So you're looking at the signal coming out of the noise generator, without actually putting them through the speaker system and reference microphone? Yes. I'm just trying to get the spectrum to match what the actual signal should be like. For white noise, I presume it should be flat over the entire frequency range. Sorry, Charlie. The power spectrum of white noise rises at 3dB/8ve. This is because it has equal energy "per frequency". You need pink noise for a flat response. Pink noise has equal energy per octave. |
#16
Posted to rec.audio.pro
|
|||
|
|||
Getting accurate sound levels in spectrum analysis
William Sommerwerck writes:
Sorry, Charlie. The power spectrum of white noise rises at 3dB/8ve. This is because it has equal energy "per frequency". You need pink noise for a flat response. Pink noise has equal energy per octave. I'm looking for equal energy per linear increment in frequency, but I see your point. I guess you have to integrate over some non-zero interval just to get the spectrum in the first place. I still don't understand what the "slope" control actually changes, though. |
#17
Posted to rec.audio.pro
|
|||
|
|||
Getting accurate sound levels in spectrum analysis
"Mxsmanic" wrote in message ... Arny Krueger writes: What Les was trying to cue you into is the fact that your ears are increasingly insensitive to frequencies as they drop below 1,000 Hz. Your ears are 60 dB less sensitive at 20 Hz than 1 KHz. If it is less than 60 dB SPL you might not hear it at all! But there is a certain intensity recorded in the recorded material itself. I want to see what that intensity is at each frequency, whether I can hear it or not. It's got markings, right? If they are like pink, white, etc., then they refer to whether the plot shows equal intensity per frequency or equal intensity per band. Not markings, just a number, from like -9 to +9 dB/octave. Then, don't apply any slope. That should make your FFT flat for a white noise input. It seems to ... so does that mean that slope = 0 shows the actual distribution of sound at different frequencies? 0 probably means no adjustment. FFTs by default show a flat spectral response for equal energy per frequency. White noise should show as being flat. -3 probably corresponds to equal energy per octave, so +3 should make pink noise show as being flat. +6 should make red or brown noise show as being flat. Kinda a neat option. |
#18
Posted to rec.audio.pro
|
|||
|
|||
Getting accurate sound levels in spectrum analysis
Mxsmanic wrote:
Scott Dorsey writes: So you're looking at the signal coming out of the noise generator, without actually putting them through the speaker system and reference microphone? Yes. I'm just trying to get the spectrum to match what the actual signal should be like. For white noise, I presume it should be flat over the entire frequency range. Depends on how your display is set up. White noise has equal power for a given bandwidth (ie. equal power per hertz). Pink noise has equal power in bands that are proportionally wide (ie. equal power per octave). Your spectrum display will have settings that allow you to look at the frequency plot in several different ways. In that case, what you see should be constant over whatever interval the noise is constant over. Except the lowest FFT bins may contain trash because that's how it is. That's how it is? Why is it that way? Because you're not going the FFT over the whole dataset, you're doing the FFT on small chunks of the dataset so you can see the display in realtime. How big a chunk you pick and what algorithm you use to overlap the chunks will affect the low frequency display. --scott -- "C'est un Nagra. C'est suisse, et tres, tres precis." |
#19
Posted to rec.audio.pro
|
|||
|
|||
Getting accurate sound levels in spectrum analysis
Arny Krueger writes:
-3 probably corresponds to equal energy per octave, so +3 should make pink noise show as being flat. So -3 means reduce the height of the curve by 3 dB per octave of increasing frequency? Or am I still misunderstanding it? |
#20
Posted to rec.audio.pro
|
|||
|
|||
Getting accurate sound levels in spectrum analysis
Scott Dorsey writes:
Because you're not going the FFT over the whole dataset, you're doing the FFT on small chunks of the dataset so you can see the display in realtime. How big a chunk you pick and what algorithm you use to overlap the chunks will affect the low frequency display. Hmm. Well, if you do it for 10-millisecond chunks, then I suppose that would distort the spectrum for sounds down around 100 Hz and below. But intuitively I'd expect the low end of the spectrum to diminish, since low frequencies would be harder and harder to discern with smaller and smaller sampling intervals. I had not previously thought of the influence of chunk size. My ultimate purpose is mainly to make sure that there are no weird sounds intruding on the recording, particularly sounds that I can't hear. If there's some sort of loud whistling at 30 kHz in the recording that I can't hear but that could still cause a problem, I'd like to be able to see it represented in correct proportions on the spectrum so that I can do something about it. |
#21
Posted to rec.audio.pro
|
|||
|
|||
Getting accurate sound levels in spectrum analysis
Mxsmanic wrote:
Hmm. Well, if you do it for 10-millisecond chunks, then I suppose that would distort the spectrum for sounds down around 100 Hz and below. But intuitively I'd expect the low end of the spectrum to diminish, since low frequencies would be harder and harder to discern with smaller and smaller sampling intervals. I had not previously thought of the influence of chunk size. This is very fundamental to using any FFT displays. There are a large number of parameters that you can set which will totally change the way the signal is displayed. You need to read the manuals. My ultimate purpose is mainly to make sure that there are no weird sounds intruding on the recording, particularly sounds that I can't hear. If there's some sort of loud whistling at 30 kHz in the recording that I can't hear but that could still cause a problem, I'd like to be able to see it represented in correct proportions on the spectrum so that I can do something about it. I think you'll find that in the long run the spectrum analyzer is a poor substitute for having accurate monitoring. --scott -- "C'est un Nagra. C'est suisse, et tres, tres precis." |
#22
Posted to rec.audio.pro
|
|||
|
|||
Getting accurate sound levels in spectrum analysis
"Mxsmanic" wrote in message ... Arny Krueger writes: -3 probably corresponds to equal energy per octave, so +3 should make pink noise show as being flat. So -3 means reduce the height of the curve by 3 dB per octave of increasing frequency? I would think so. What happens when you apply this adjustment to a source that should display as flat, such as white noise? |
#23
Posted to rec.audio.pro
|
|||
|
|||
Getting accurate sound levels in spectrum analysis
"Mark" wrote in message ... It is also interesting to note that I have seen two styles of FFT analyzers. One style has a constant resolution BW (standard FFT ) and therefore will display a flat spectral density with WHITE noise input and a decreasing density with a PINK noise input. The other type sometimes called a real time audio analyzer has a BW that increases with frequency i.e. it has say 5 bands per octave for example. This type will display a flat spectral density with a PINK noise input and a rising density with a WHITE noise input. You need to know what kind of ruler you are using. Both could be the same FFT only with different ways of displaying the data. The RTA display is also known as (fractional) octave display. Here's an article that might help: http://www.dspdimension.com/admin/dft-a-pied/ |
#24
Posted to rec.audio.pro
|
|||
|
|||
Getting accurate sound levels in spectrum analysis
Arny Krueger writes:
I would think so. What happens when you apply this adjustment to a source that should display as flat, such as white noise? If I set it to zero, the spectrum looks flat. If I set it to -9, the low end rises and the high end falls, and vice versa for +9. |
#25
Posted to rec.audio.pro
|
|||
|
|||
Getting accurate sound levels in spectrum analysis
On 2/8/2012 1:59 PM, Mxsmanic wrote:
My ultimate purpose is mainly to make sure that there are no weird sounds intruding on the recording, particularly sounds that I can't hear. If there's some sort of loud whistling at 30 kHz in the recording that I can't hear but that could still cause a problem, I'd like to be able to see it represented in correct proportions on the spectrum so that I can do something about it. If it's there and you can't hear it, you probably shouldn't be working in this field, or you should have better monitors so you CAN hear it. I remember in the early days of home recording, people were sending projects in for mastering and replication with way too much low end because they couldn't hear it on their home speakers. My local tape duplicator would put a 50 Hz high pass filter between the tape deck and console on everything that came in just to prevent damaging their monitors when first playing through the tape. As for 30 kHz, if you're working at or mastering for standard sample rate, you won't record anything at that frequency anyway. -- "Today's production equipment is IT based and cannot be operated without a passing knowledge of computing, although it seems that it can be operated without a passing knowledge of audio." - John Watkinson http://mikeriversaudio.wordpress.com - useful and interesting audio stuff |
#26
Posted to rec.audio.pro
|
|||
|
|||
Getting accurate sound levels in spectrum analysis
You need to know what kind of ruler you are using. Both could be the same FFT only with different ways of displaying the data. The RTA display is also known as (fractional) octave display. Here's an article that might help: http://www.dspdimension.com/admin/dft-a-pied/ thanks for quoting me Arnie, :-) I was beginning to think I was invisible.. (which actually may be true for some since I'm posting via Google) Mark |
#27
Posted to rec.audio.pro
|
|||
|
|||
Getting accurate sound levels in spectrum analysis
Mike Rivers writes:
If it's there and you can't hear it, you probably shouldn't be working in this field, or you should have better monitors so you CAN hear it. I can't agree. A person might be an expert in audio systems and yet not have perfect hearing. A recording might contain loud noise at 18 kHz that a sound engineer might not be able to hear, but the intended audience might be able to hear. I don't think the engineer should change careers just because he can't hear it, but I also think that he should take steps to ensure that even things he can't hear are at least noticed and handled appropriately, since there may be other people who can hear those sounds. Of course, if a person is entirely deaf, working with audio is going to be difficult, but writing off a career simply because of a hearing impairment seems extreme, insofar as many other skills are involved besides simply hearing things alone. I remember in the early days of home recording, people were sending projects in for mastering and replication with way too much low end because they couldn't hear it on their home speakers. My local tape duplicator would put a 50 Hz high pass filter between the tape deck and console on everything that came in just to prevent damaging their monitors when first playing through the tape. If those people had looked at the spectrum of the sound before sending in their projects, perhaps they would have seen and removed the low end. |
#28
Posted to rec.audio.pro
|
|||
|
|||
Getting accurate sound levels in spectrum analysis
|
#29
Posted to rec.audio.pro
|
|||
|
|||
Getting accurate sound levels in spectrum analysis
"Mxsmanic" wrote in message ... Arny Krueger writes: I would think so. What happens when you apply this adjustment to a source that should display as flat, such as white noise? If I set it to zero, the spectrum looks flat. If I set it to -9, the low end rises and the high end falls, and vice versa for +9. That's what I would expect. |
#30
Posted to rec.audio.pro
|
|||
|
|||
Getting accurate sound levels in spectrum analysis
"Mike Rivers" wrote in message ... On 2/8/2012 1:59 PM, Mxsmanic wrote: My ultimate purpose is mainly to make sure that there are no weird sounds intruding on the recording, particularly sounds that I can't hear. If there's some sort of loud whistling at 30 kHz in the recording that I can't hear but that could still cause a problem, I'd like to be able to see it represented in correct proportions on the spectrum so that I can do something about it. If it's there and you can't hear it, you probably shouldn't be working in this field, or you should have better monitors so you CAN hear it. I don't know about people can hear 30 KHz and monitors that can reproduce 30 KHz (particularly off axis). I would never fault a person or a loudspeaker only because of inability to respond to 30 KHz. I remember in the early days of home recording, people were sending projects in for mastering and replication with way too much low end because they couldn't hear it on their home speakers. My local tape duplicator would put a 50 Hz high pass filter between the tape deck and console on everything that came in just to prevent damaging their monitors when first playing through the tape. I've been known to do that. As for 30 kHz, if you're working at or mastering for standard sample rate, you won't record anything at that frequency anyway. Exactly. |
#31
Posted to rec.audio.pro
|
|||
|
|||
Getting accurate sound levels in spectrum analysis
"Mxsmanic" wrote in message ... Mike Rivers writes: If it's there and you can't hear it, you probably shouldn't be working in this field, or you should have better monitors so you CAN hear it. I can't agree. A person might be an expert in audio systems and yet not have perfect hearing. In fact there are a lot of experts in audio who have far from perfect hearing. At 65 years, my hearing isn't the best, but recently I was sitting in a listening room next to a well-known audio expert who is a lot younger than I am. I was having a hard time enjoying the music due to a very strong 60 Hz noise that was probably due to a ground loop. He didn't notice it. Caveat: this expert does design, development and production of equipment, not recordings. I don't know how someone could be that deaf and do live sound or recording. IME live sound probably requires the best hearing because the opportunities to work through live sound situations with FFTs and the like are pretty limited. With track-per-mic recordings and good DAW software, the right tools are more available and you have the time to use them. A lot of older audio workers have younger assistants that they rely on to double check their work at least occasionally. Let's face it, most of what really matters is in frequency ranges and at levels that are easy enough to hear. A lot of audio production is about knowing which knobs to turn to get things to sound a certain way which does not require hearing really subtle details. A lot of people can hear those subtle details but never master knowing which knobs to turn. A recording might contain loud noise at 18 kHz that a sound engineer might not be able to hear, but the intended audience might be able to hear. To be heard by anybody, that sound at 18 KHz would have to be very large, because of masking. For example, the world is full of MP3s that are brick-walled at 16 KHz, and still manage to sound crisp and even hot. In general almost nobody hears brick wall filters very easily until their corner frequency is 16 Khz or even somewhat lower. I don't think the engineer should change careers just because he can't hear it, but I also think that he should take steps to ensure that even things he can't hear are at least noticed and handled appropriately, since there may be other people who can hear those sounds. Being able to read waveforms and use a FFT can be very useful for everybody, but for folks whose hearing isn't the best, they can be survival tools. DAWs make using these tools far easier. Of course, if a person is entirely deaf, working with audio is going to be difficult, but writing off a career simply because of a hearing impairment seems extreme, insofar as many other skills are involved besides simply hearing things alone. I remember in the early days of home recording, people were sending projects in for mastering and replication with way too much low end because they couldn't hear it on their home speakers. My local tape duplicator would put a 50 Hz high pass filter between the tape deck and console on everything that came in just to prevent damaging their monitors when first playing through the tape. If those people had looked at the spectrum of the sound before sending in their projects, perhaps they would have seen and removed the low end. If they'd looked at the cones of their woofers... |
#32
Posted to rec.audio.pro
|
|||
|
|||
Getting accurate sound levels in spectrum analysis
|
#33
Posted to rec.audio.pro
|
|||
|
|||
Getting accurate sound levels in spectrum analysis
Arny Krueger writes:
A lot of people can hear those subtle details but never master knowing which knobs to turn. That would be me. My hearing is still pretty good, but I hear something wrong, I don't necessarily know how to fix it. I have the same problem with music. If someone is even very slightly out of tune, I hear it, but I'm unable to describe what's wrong because I just don't know enough about music theory. Of course, I could learn these things, I guess. But there is so much to learn in so many domains, and there are only 24 hours in a day. To be heard by anybody, that sound at 18 KHz would have to be very large, because of masking. I've read about extremely low sounds that make people very irritable, anxious, and restless, even though they might not hear the sounds. I've already mentioned somewhere here the concert at which the audience was made very restless by an extremely high whistle that (presumably) nobody could actually here. I would want to eliminate those from my recordings if I found them. For example, the world is full of MP3s that are brick-walled at 16 KHz, and still manage to sound crisp and even hot. I've noticed some MP3s that seemed to stop at exactly 16 kHz, and was wondering about that very thing. So it's a deliberate choice, not a limitation of MP3? Being able to read waveforms and use a FFT can be very useful for everybody, but for folks whose hearing isn't the best, they can be survival tools. DAWs make using these tools far easier. Since vision provides much more bandwidth than hearing, in theory it should be possible to represent sounds in a completely visual format with no loss, allowing even a deaf person to perceive sound accurately. It would take training but it should work. I thought up a system like this for computers and deaf people once, but nobody was interested. Unfortunately, the opposite is not true: there's no way to represent visual information to blind people completely with a simple audio system. The bandwidth difference is 1000 to 1. If they'd looked at the cones of their woofers... I've seen commercials and movies where a close-up of a speaker is shown with the cone bouncing in and out, and I've always wondered if that's just a special effect for purposes of exaggeration, or if it's a real sound being played over the speaker. Which in turn reminds me of a certain scene in _Back to the Future_ where Marty McFly turns up a huge amplifier to play his guitar through a huge speaker. Which in turn reminds me of a certain print advertisement for a stereo system with a guy sitting in his chair, his wind blown backwards by the sound blast from his stereo (but I can't remember the brand!). |
#34
Posted to rec.audio.pro
|
|||
|
|||
Getting accurate sound levels in spectrum analysis
Arny Krueger writes:
I don't know about people can hear 30 KHz and monitors that can reproduce 30 KHz (particularly off axis). A loud ultrasonic noise could still affect listeners, even if they couldn't actually hear the sound. |
#35
Posted to rec.audio.pro
|
|||
|
|||
Getting accurate sound levels in spectrum analysis
Mxsmanic writes:
Arny Krueger writes: A lot of people can hear those subtle details but never master knowing which knobs to turn. That would be me. My hearing is still pretty good, but I hear something wrong, I don't necessarily know how to fix it. A spectrum analyzer is handy tool, just like a VU or peak meter. A VU and peak meter might be like your doctor's stethascope and blood pressure cuff, where the analyzer might be more like an ultrasound of your heart (or maybe one of those micro-cams snaking through your blood vessles!) You can simply "see" certain things with one tool that the other tools might only hint at. And ears can be limited at times; it's useful, IMO, to occasionally "look" at our work in other ways. Because I'm always curious, I listen to CDs in general with a spectrum analyzer running on one of the side video monitors. I'm amazed at the number of recordings showing a 15,750 hz spike from a CRT monitor. (Heck, even one of my own pipe organ recordings managed to acoustically pick up 15,200 hz or so from a closed-circuit 9" video monitor that was up on the organ console. The organ was a tracker, which meant the console was part of the organ, and the organ was way up in a loft. The monitor allowed the organist to watch a conductor's position down on the main floor of the church. The monitor was probably 1970s vintage; seems high-volume flyback noise was pretty common in those days. First video production room I walked into -- with all those various CRTs -- made my eyes water and my teeth hurt because of all the flyback noise.) But a few things to consider: I don't think very many folks hear a flyback spike embedded in a recording -- they're typically 20-40 dB down, and usually well-masked by lots of energy at surrounding frequencies. Even with the narrowest Q you might have on an EQ, it's damn near impossible to notch out a single frequency like that. A 1/12 octave notch at 15K is going to put quite a dip in your spectrum up there -- and you *will* hear that, whereas the single offending freq you well might not. The analyzer can also help you find system problems. I had a noisy full-sine UPS that was putting out spikes at 19.5K and 26K and sometimes (oh the horror) 10.5K. Worse, it was injecting this crap into the building ground! Zooming in I saw additional "sidebands" on these spikes at 60 and 120 Hz intervals. The spectrum analyzer was a very useful tool to investigate this. The manufacturer thought I was crazy, until I showed them short video screen caps from the analyzer. "Oh, yeah. Well... Those are our oscillators starting up on the UPS..." To their discredit, the issue was never resolved and the UPS was abandoned in favor of lighter, cheaper systems that were dead quiet, even when fully powering the loads. (I've also seen that 19.5K spike in a few other commercial recordings; I'm guessing they had the same defective UPS running during the recording!) Bottom line is that while you can't (and shouldn't) "mix" on an analyzer, it's a useful piece of kit to have. Nowadays we have fairly comprehensive ones that can be done cheap or even "free" in software, whereas a few decades back such measurement devices were bulky hardware, their measurements coarse by comparison, and they cost a bloody fortune. Frank Mobile Audio -- |
#36
Posted to rec.audio.pro
|
|||
|
|||
Getting accurate sound levels in spectrum analysis
On 2/9/2012 4:20 AM, Mxsmanic wrote:
Mike Rivers writes: If it's there and you can't hear it, you probably shouldn't be working in this field, or you should have better monitors so you CAN hear it. I can't agree. A person might be an expert in audio systems and yet not have perfect hearing. A recording might contain loud noise at 18 kHz that a sound engineer might not be able to hear, but the intended audience might be able to hear. I know many audio engineers, both live and studio, that have hearing loss and still manage to do good work. They don't look at a spectrum analyzer to see what they can't here, though. I don't think the engineer should change careers just because he can't hear it, but I also think that he should take steps to ensure that even things he can't hear are at least noticed and handled appropriately, since there may be other people who can hear those sounds. That's true. He needs to recognize his limitations and figure out how to deal with them. This might be as simple as asking someone to listen to the program material and ask if he hears any booms or whistles. But the best way to handle problems like "loud noise at 18 kHz" is to avoid them in the first place. Where might such a noise enter the recording process? It isn't likely to come from anyone's vocal cords or an instrument. It likely means something is broken. I only suggested a career change because of a clear misunderstanding of what spectrum analysis is good for. I remember in the early days of home recording, people were sending projects in for mastering and replication with way too much low end because they couldn't hear it on their home speakers. If those people had looked at the spectrum of the sound before sending in their projects, perhaps they would have seen and removed the low end. In those days, a spectrum analyzer cost $25,000. Mastering (which includes removing things that clearly shouldn't be there, if possible) only cost $100, and you had a real human to make the judgment, not a piece of hardware that you probably shouldn't trust with a critical decision that affects the sound of your production. -- "Today's production equipment is IT based and cannot be operated without a passing knowledge of computing, although it seems that it can be operated without a passing knowledge of audio." - John Watkinson http://mikeriversaudio.wordpress.com - useful and interesting audio stuff |
#37
Posted to rec.audio.pro
|
|||
|
|||
Getting accurate sound levels in spectrum analysis
On 2/9/2012 10:02 AM, Mxsmanic wrote:
A lot of people can hear those subtle details but never master knowing which knobs to turn. That would be me. My hearing is still pretty good, but I hear something wrong, I don't necessarily know how to fix it. This is what engineers learn (or they get out of the business). I I have the same problem with music. If someone is even very slightly out of tune, I hear it, but I'm unable to describe what's wrong because I just don't know enough about music theory. You can't say "the guitar is out of tune? Of course, I could learn these things, I guess. But there is so much to learn in so many domains, and there are only 24 hours in a day. Well, you see, there are tools for people like you who want to make music but don't have the skills or experience. It's easy to assemble a system that doesn't have any fundamental flaws. And if you know what you're recording, you can be reasonably safe that the only thing that will make your recordings unpleasant for someone (or you) to listen to is your own bad taste. But this is something that you have to develop as an artistic skill, not by buying a tool. I've read about extremely low sounds that make people very irritable, anxious, and restless, even though they might not hear the sounds. This is true. That principle is even used in warfare. Make the enemy want to **** and he'll have something else on his mind than fighting a war. But think: How is such a sound going to creep into your recording? the concert at which the audience was made very restless by an extremely high whistle that (presumably) nobody could actually here. Somebody could hear it, probably a lot of people could. Maybe there was no engineer manning the controls. Maybe there was and he couldn't hear it. Maybe he could hear it, did everything he could to eliminate or reduce it and still couldn't get rid of it. I used to do shows in an auditorium in a Government building where, every hour, something would be sent along the power lines to correct all the clocks. It caused a whistle in the PA system for about 30 seconds and then it was gone. I noticed it, as did some people in the audience. They probably thought I did something to cause it. I would want to eliminate those from my recordings if I found them. And the way to do that is to know that your equipment is working properly and simply not make noises like that. I've noticed some MP3s that seemed to stop at exactly 16 kHz, and was wondering about that very thing. So it's a deliberate choice, not a limitation of MP3? It's a choice. There are a lot of options for creating an MP3 file. Properly conducted tests have shown that it's possible to make an MP3 file and a CD of the same material and not be able to tell them apart. And it's not just for selected material, it was with a wide range of material. Thing is that MP3 files grew out of a lack of storage space so the original goal was to provide something distinguishable as music which could be stored in as little space as possible. Now that iPods have 160 GB of storage, if someone were to give you 320 kbps 44.1 kHz MP3 files to load on to it, you wouldn't be able to tell that you weren't listening to a CD. But then there are people who still think that CDs sound unacceptable. Some do, some don't. But what you're talking about is a problem that you can eliminate at the source. Using a spectrum analyzer to find it after the fact is only helping you to put a Band Aid on it, not fix the problem. Since vision provides much more bandwidth than hearing, in theory it should be possible to represent sounds in a completely visual format with no loss, allowing even a deaf person to perceive sound accurately. It would take training but it should work. Yeah, but can you dance to it? For a realistic perspective, check out Flanders & Swann's "A Song of Reproduction." http://youtu.be/7fJmmDkvQyc -- "Today's production equipment is IT based and cannot be operated without a passing knowledge of computing, although it seems that it can be operated without a passing knowledge of audio." - John Watkinson http://mikeriversaudio.wordpress.com - useful and interesting audio stuff |
#38
Posted to rec.audio.pro
|
|||
|
|||
Getting accurate sound levels in spectrum analysis
"Mxsmanic" wrote in message ... Arny Krueger writes: I don't know about people can hear 30 KHz and monitors that can reproduce 30 KHz (particularly off axis). A loud ultrasonic noise could still affect listeners, even if they couldn't actually hear the sound. Been there done that, but we're talking *really* loud, like over 120 dB SPL. That doesn't happen with recordings, as a rule. For openers, they rarely get played that loud, and in many cases if they were played that loud, the 30 KHz tone that created 120 dB SPL would quickly fry the tweeter, which has happened in the real world. |
#39
Posted to rec.audio.pro
|
|||
|
|||
Getting accurate sound levels in spectrum analysis
On Thu 2012-Feb-09 04:20, Mxsmanic writes:
If it's there and you can't hear it, you probably shouldn't be working in this field, or you should have better monitors so you CAN hear it. I remember in the early days of home recording, people were sending projects in for mastering and replication with way too much low end because they couldn't hear it on their home speakers. My local tape duplicator would put a 50 Hz high pass filter between the tape deck and console on everything that came in just to prevent damaging their monitors when first playing through the tape. If those people had looked at the spectrum of the sound before sending in their projects, perhaps they would have seen and removed the low end. Whether you agree or not, you don't have the experience to make that judgment, as is obvious by that remark. Many folks didn't have those tools readily available in real time before the advent of the daw. Have you ever mixed live for broadcast or for paying customers? Did you just decide to dabble in audio once you found low cost software? I'd guess the later. IN the era that Mike's referencing most folks didn't have spectrum analysis tools readily available as I said, ears and the monitoring chain were what was used to make those decisions. Mike is one of us in this group who's made his daily bread working in audio, and we got some neophyte basement tinkerer going to tell him he's full of ****? Those are indeed beneficial tools, but but folks have been making production decisions without them for the better part of a century. Such tools are especially useful as a sanity check, or to make quicker decisions. Still, if you're going to cause money to change hands you need a better monitoring chain it sounds like. if you're playing and tinkering for your own good time, have at it, but expect the professionals in this group to challenge such assertions when you make them here. After all, the neophytes know no better, and they're trying to do work that is intended for public consumption. Regards, Richard .... 10% of everything isn't crap, watch closely or you'll miss it! -- | Remove .my.foot for email | via Waldo's Place USA Fidonet-Internet Gateway Site | Standard disclaimer: The views of this user are strictly his own. |
#40
Posted to rec.audio.pro
|
|||
|
|||
Getting accurate sound levels in spectrum analysis
Mike Rivers writes: snip But the best way to handle problems like "loud noise at 18 kHz" is to avoid them in the first place. Where might such a noise enter the recording process? It isn't likely to come from anyone's vocal cords or an instrument. It likely means something is broken. Indeed, and one can figure out those things before that happens, if one has some knowledge. I only suggested a career change because of a clear misunderstanding of what spectrum analysis is good for. A post I made which didn't propagate stated much the same thing to mxmanic. Spectrum analysis is a valuable tool, but not the panacea he thinks it is. snip If those people had looked at the spectrum of the sound before sending in their projects, perhaps they would have seen and removed the low end. In those days, a spectrum analyzer cost $25,000. Mastering (which includes removing things that clearly shouldn't be there, if possible) only cost $100, and you had a real human to make the judgment, not a piece of hardware that you probably shouldn't trust with a critical decision that affects the sound of your production. As I also stated, and for some of us, who had to do production for right here right now, the trick was resolving such issues by applying some knowledge to make sure they didn't creep in. Often those problems can be audible as harmonics of the fundamental. Mxmanic is doing this as a hobby, no doubt because he found low cost software which permits him to dabble, and now thinks it qualifies him to tell professionals how they should be doing their jobs. It's good that he's interested in learning, but I still think the faq for this newsgroup and other reading material would help him gain a much better understanding than stumbling around with a daw then coming into a production oriented newsgroup and telling those of us who actually do this to earn a dollar we're full of whatever it may be, that he has all the answers. Just for mxmanic, I record using digital not analog these days, adn still don't bother to deal with spectrum analysis, as old blind man uses a recorder that acts like the analog recorders he built his skills using, and these don't offer spectrum analysis tools. During later phases of production I've had folks look at something with such tools as spectrum analysis, but i endeavor to find other fixes for such problems before they're necessary in mastering. But, if i miss, that's why tehre's mastering g. Finally for Mxmanic: Folks such as MIke and i aren't just trying to urniate in your cornflakes just to have something to do. Neophytes lurk here for useful information, and they're endeavoring to put their product out before the public, even if self produced and self engineered. Hence, we have to do whatever we can to dispel myths and misinformation. I'm not the gentlest at doing so. Sorry 'bout that. Richard webb, replace anything before at with elspider |
Reply |
Thread Tools | |
Display Modes | |
|
|
Similar Threads | ||||
Thread | Forum | |||
Panoramic spectrum analysis | High End Audio | |||
Software for sound spectrum analysis | Pro Audio | |||
Turntable Spectrum Analysis | Pro Audio | |||
good web tutorial on spectrum analysis? | Pro Audio | |||
Spectrum Analysis in SF . hmm??? | Pro Audio |