RangerWalker71
Well-Known Member
You hear a lot of people say" Mil-Spec Meter testing standards".
When they are talking about testing suppressor.
Well here is what it is, and how it's done:
Link to article
Sound Testing Standards
Link to actual Mil-stands 147D - Mil- Std
http://www.silencertalk.com/docs/mil-std-1474d.pdf
MIL-STD-1474D is a military standard that was designed for generic hearing damage risk assessment for aircraft, shipboard, and community noise and does not address how loud a silencer would appear to a shooter or witness. Since most silencers on the market today are hearing-safe, it is much more interesting to compare silencers for how 'loud' or 'detectable' they are to the shooter or human observer.
There are three main reasons why the military hearing-risk standard is not applicable to perception of silenced gunshot noise. Let's address them one at a time. The problems are:
Problem 1: MIL-STD-1474D is a poor predictor, even of what it was designed to do.
The current hearing-risk standard is not even a good predictor of hearing risk. In the paper titled Weapon noise exposure of the human ear analyzed with the AHAAH model, the US Army has set out to compare MIL-STD-1474D to actual hearing loss and has found it to be accurate only 38% of the time. They have further studied A-weighted sound energy and found that to only be an accurate predictor 25% of the time. They concluded that "and in the case of impulses at very high levels and with a lot of low frequency energy, the errors with the A-weighed energy measure were very large."
The Army's modernized AHAAH system operates by processing a digital waveform much in the way that Silencer Tests used exclusively for its initial tests and now uses just for advanced research.
Problem 2: The hearing-risk standard specifies an 'A-weighted' filter on the measured audio spectrum.
In the paper titled The Difficulties in Evaluating A-Weighted Sound Level Measurement, it is explained that Fletcher and Munson generated equal-loudness curves to compare pure tones (not complex or impulse sounds) at different frequencies. Any spot on a given frequency curve should sound equally loud as a 1000 Hz tone of a given SPL.
A-weighting is an approximation of this curve at the 40 Phon level (40 dB SPL at 1000 Hz). A-Weighting is therefore an an easy-to-implement (with analog circuits) approximation of the Fletcher-Munson 40 Phon curve.
The problem with A-Weighing is that it does not match the ear's response at any sound pressure level. Even worse is that it comes closest to matching the ear's response for sounds in the 40-50 dB SPL range as described in the paper titled 'A' Weighting Filter For Audio Measurements. Suppressed gunfire is generally in the 120-140 dB SPL range and it would be much more appropriate to have selected the 120 Phon curve. The 40 Phon curve, and hence A-weighting, is entirely inappropriate for anything other than low-level (such as in an office) sound level measurement.
In fact, as shown in figure 3, unweighted measurements more closely approximate the way the ear hears most sounds at 120 dB SPL than the A-weighted curve. Why did the military go out of their way to impose a curve that is worse than no curve at all? Partly because they were not measuring gunshot noise perception but rather trying to predict generic hearing risk for a wide range of environmental sounds, and partly because A-weighting was already established as a standard for measuring lower-level environmental noise.
Problem 3: The hearing-risk standard measures peak noise and does not consider total sound output.
When taking measurements for the MIL-STD the meter is set to peak-hold and only the peak of the single-largest wave-crest is considered -- no matter that happens anywhere else in the audio spectrum. The other sounds might as well not exist, and their duration is not a factor in the results as it should be. For example, a 140 dB SPL impulse of 10 microseconds will sound about as loud as a 137 dB SPL impulse of 20 microseconds. Duration needs to be considered for gunshot perception.
The solution:
The solution is something new but based on solid research. ISO Standard 532 Acoustics - Method for Calculating Loudness Level references two methods developed by Zwicker and Stevens to calculate human perception of loudness by measuring the energy in 1 or 1/3 octave critical bands. These provide a loudness number in units called Sone by summing the noise in each band according to the formula St = Sm + F (S - Sm) where Sm is the maximum loudness index, S is the sum of all such indices and F is given the value 0.3 for octave bands, 0.2 for half-octave, and 0.15 for third-octave bands. One Sone was set to be 40 Phons at any frequency (at any point along the 40 Phon curve on the equal loudness graph) and twice as many Sones means a human would hear the sound as twice as loud. Here is a loudness calculator which works at the octave level, and another one for 1/3 octave. There is also an MS-DOS program to process files. Furthermore, Daniel A. Quinlan authored a paper comparing A-weighting to this Zwicker and Stevens method.
I propose an extension to this system for dealing with impulse noise because it is perceived differently than steady-state and the duration of the impulse needs to be considered. Such a change would involve determining an appropriate time-constant, which would need to be shorter than the standard 125 ms of a sound meter's 'FAST' setting. The 'Impulse' setting of a sound meter is a 35 ms time-constant and that still appears to be too slow. I suspect a time-constant between 0.0005 and 0.01 seconds would be best although the ultimate solution might be to have the time-constant be the same as the A-duration period.
In conclusion, the MIL-STD manufacturers commonly use for research, development, and marketing purposes is not appropriate for predicting human perception of suppressed gunshot noise even when perfectly and fairly implemented. The elimination of A-weighting and peak-voltage detection and replacing them with a better curve (C-weighting would be better) and a longer time-constant (such as the length of the entire A-duration) would be a good thing to test, but the ultimate solution would be a new standard based on a version of what is outlined in the ISO standard 532-B, modified for impulse sound signatures by choosing an ideal time-constant.
In the meantime, we will continue to post results in compliance with MIL-STD-1474D using calibrated and certified military-spec test equipment as well as unweighted results.
Hope this helps.
thanks
RLTW
Steve
When they are talking about testing suppressor.
Well here is what it is, and how it's done:
Link to article
Sound Testing Standards
Link to actual Mil-stands 147D - Mil- Std
http://www.silencertalk.com/docs/mil-std-1474d.pdf
Sound Testing Standards
MIL-STD-1474D is a military standard that was designed for generic hearing damage risk assessment for aircraft, shipboard, and community noise and does not address how loud a silencer would appear to a shooter or witness. Since most silencers on the market today are hearing-safe, it is much more interesting to compare silencers for how 'loud' or 'detectable' they are to the shooter or human observer.
There are three main reasons why the military hearing-risk standard is not applicable to perception of silenced gunshot noise. Let's address them one at a time. The problems are:
Problem 1: MIL-STD-1474D is a poor predictor, even of what it was designed to do.
The current hearing-risk standard is not even a good predictor of hearing risk. In the paper titled Weapon noise exposure of the human ear analyzed with the AHAAH model, the US Army has set out to compare MIL-STD-1474D to actual hearing loss and has found it to be accurate only 38% of the time. They have further studied A-weighted sound energy and found that to only be an accurate predictor 25% of the time. They concluded that "and in the case of impulses at very high levels and with a lot of low frequency energy, the errors with the A-weighed energy measure were very large."
The Army's modernized AHAAH system operates by processing a digital waveform much in the way that Silencer Tests used exclusively for its initial tests and now uses just for advanced research.
Problem 2: The hearing-risk standard specifies an 'A-weighted' filter on the measured audio spectrum.
In the paper titled The Difficulties in Evaluating A-Weighted Sound Level Measurement, it is explained that Fletcher and Munson generated equal-loudness curves to compare pure tones (not complex or impulse sounds) at different frequencies. Any spot on a given frequency curve should sound equally loud as a 1000 Hz tone of a given SPL.
Fig 1, Equal loudness curves.
A-weighting is an approximation of this curve at the 40 Phon level (40 dB SPL at 1000 Hz). A-Weighting is therefore an an easy-to-implement (with analog circuits) approximation of the Fletcher-Munson 40 Phon curve.
Fig 2, Transfer function defining A-weighted curve.
The problem with A-Weighing is that it does not match the ear's response at any sound pressure level. Even worse is that it comes closest to matching the ear's response for sounds in the 40-50 dB SPL range as described in the paper titled 'A' Weighting Filter For Audio Measurements. Suppressed gunfire is generally in the 120-140 dB SPL range and it would be much more appropriate to have selected the 120 Phon curve. The 40 Phon curve, and hence A-weighting, is entirely inappropriate for anything other than low-level (such as in an office) sound level measurement.
Fig 3, Actual 120 Phon equal loudness curve (green) compared to A-weighting (blue) and unweighted (red).
In fact, as shown in figure 3, unweighted measurements more closely approximate the way the ear hears most sounds at 120 dB SPL than the A-weighted curve. Why did the military go out of their way to impose a curve that is worse than no curve at all? Partly because they were not measuring gunshot noise perception but rather trying to predict generic hearing risk for a wide range of environmental sounds, and partly because A-weighting was already established as a standard for measuring lower-level environmental noise.
Problem 3: The hearing-risk standard measures peak noise and does not consider total sound output.
When taking measurements for the MIL-STD the meter is set to peak-hold and only the peak of the single-largest wave-crest is considered -- no matter that happens anywhere else in the audio spectrum. The other sounds might as well not exist, and their duration is not a factor in the results as it should be. For example, a 140 dB SPL impulse of 10 microseconds will sound about as loud as a 137 dB SPL impulse of 20 microseconds. Duration needs to be considered for gunshot perception.
The solution:
The solution is something new but based on solid research. ISO Standard 532 Acoustics - Method for Calculating Loudness Level references two methods developed by Zwicker and Stevens to calculate human perception of loudness by measuring the energy in 1 or 1/3 octave critical bands. These provide a loudness number in units called Sone by summing the noise in each band according to the formula St = Sm + F (S - Sm) where Sm is the maximum loudness index, S is the sum of all such indices and F is given the value 0.3 for octave bands, 0.2 for half-octave, and 0.15 for third-octave bands. One Sone was set to be 40 Phons at any frequency (at any point along the 40 Phon curve on the equal loudness graph) and twice as many Sones means a human would hear the sound as twice as loud. Here is a loudness calculator which works at the octave level, and another one for 1/3 octave. There is also an MS-DOS program to process files. Furthermore, Daniel A. Quinlan authored a paper comparing A-weighting to this Zwicker and Stevens method.
I propose an extension to this system for dealing with impulse noise because it is perceived differently than steady-state and the duration of the impulse needs to be considered. Such a change would involve determining an appropriate time-constant, which would need to be shorter than the standard 125 ms of a sound meter's 'FAST' setting. The 'Impulse' setting of a sound meter is a 35 ms time-constant and that still appears to be too slow. I suspect a time-constant between 0.0005 and 0.01 seconds would be best although the ultimate solution might be to have the time-constant be the same as the A-duration period.
In conclusion, the MIL-STD manufacturers commonly use for research, development, and marketing purposes is not appropriate for predicting human perception of suppressed gunshot noise even when perfectly and fairly implemented. The elimination of A-weighting and peak-voltage detection and replacing them with a better curve (C-weighting would be better) and a longer time-constant (such as the length of the entire A-duration) would be a good thing to test, but the ultimate solution would be a new standard based on a version of what is outlined in the ISO standard 532-B, modified for impulse sound signatures by choosing an ideal time-constant.
In the meantime, we will continue to post results in compliance with MIL-STD-1474D using calibrated and certified military-spec test equipment as well as unweighted results.
Hope this helps.
thanks
RLTW
Steve
