SPL meter sample rate?

[pinebox]

Does anyone have knowledge as to how an SPL meter gets a reading? I had a discussion about firearm silencers today and was told that dB meters are now antiquated and a new system requiring a sample rate of ~1meg is required to catch the peaks of the blast concussion. After thinking about it, I started to wonder how SPL meters actually work?

 

[JohnRoberts]

Not sure that matters a lot.. life is a LPF.

Back last century I designed a micro based meter for a console... I found that I could reliably measure signal levels despite under-sampling the audio input. While the under-sampled signals generated aliases, the amplitude of those aliases tracked well with actual signal levels.

JR

 

[abbey road d enfer]

Does anyone have knowledge as to how an SPL meter gets a reading? I had a discussion about firearm silencers today and was told that dB meters are now antiquated and a new system requiring a sample rate of ~1meg is required to catch the peaks of the blast concussion.

1Meg sampling rate implies the need to detect 2 microsecon events. What are the physiological effects of a 2microsecond pulse? Or is there another reason? the person who told you that should know (or remain silent).

After thinking about it, I started to wonder how SPL meters actually work?

Analog or digital? Anyway, SPL meters are designed to provide measurements in the audio frequency domain, which is by convention 20Hz-20kHz.

 

[pinebox]

Thanks for the replies, they could not articulate the reasons why and I barely skimmed the beginning of the explanation they pointed me towards, which is where I got the super high sample rate from. All the info is HERE if you want to take a look. I will be reading it more in depth today.

 

[dmp]

In order to measure the peak accurately of a rapidly decaying noise event, the sampling rate needs to be much higher than the Nyquist criteria
For example, to measure the peak amplitude of a 20 kHz burst wave, you need to be sampling much higher than 20k S/sec, as the point on the wave you will randomly acquire will be randomly located. Most likely with only 1 acquired point on the first wave, the measured SPL will underpredict the peak. A data acquisition rate of 1M S/sec will give 1000/20 = 50 datapoints on the 'burst' wave first complete waveform and be much closer to the peak amplitude.
Acquiring the SPL of a burst is much different than acquiring the SPL of a steady tone, where lower sampling with random locations will be sufficient to get a accurate peak amplitude.
With lower dominant frequency the required sampling rate is lower, but the same principle applies.

 

[abbey road d enfer]

OK, read it.
He doesn't describe his test equipment, so I don't know what he used to measure pressure. Is it a mic, or some other device. If it's a mic, I doubt it has a response that goes much higher than 70 kHz, which requires only 150ksamples per second, not 1 Meg. I just surmise he had an acquisition card that does 1 MHz.
He didn't either demonstrates the nocivity of pressure pulses shorter than 1 or 2 microseconds.
The graphs he produced could have been done with a basic DS (96k SR) sound card. The main difficulty is getting a pressure sensor (microphone?) that provides reliable information at 170 dBspl. Not your standard SM57 or KM184...

EDIT: meantime I have seen the video of the test, where they use PCB soundmeters as front ends, and data recorded via USB.

 

[sr1200]

I was looking into something similar the other day. From what i read, sound in the MHz range can't really propagate too far from the source (i think it said something like 1-2Mhz can only travel about 1-2cm before becoming absorbed by the air itself). I would think amplitude would have some kind of effect on that as its been proven that freq. over 20K can be "heard" if the amplitude is high enough. Personally, dont see how having a higher sampling rate would make a difference, its not like you're going to have "inter-sample peaks" in air that would need to be detected for any kind of normal reason. This sounds very audiophool to me.

 

[JohnRoberts]

I was looking into something similar the other day. From what i read, sound in the MHz range can't really propagate too far from the source (i think it said something like 1-2Mhz can only travel about 1-2cm before becoming absorbed by the air itself). I would think amplitude would have some kind of effect on that as its been proven that freq. over 20K can be "heard" if the amplitude is high enough. Personally, dont see how having a higher sampling rate would make a difference, its not like you're going to have "inter-sample peaks" in air that would need to be detected for any kind of normal reason. This sounds very audiophool to me.

ding ding ding... The upper frequency range for human audition has been beat to death over at pukeslutz...

Presumably SPL meters are to provide some degree of human hearing protection. It seems conventional audio bandwidth (20-20k) should be sufficient.

JR

 

[sr1200]

Being that they're speaking about silencers (and firearms) I assume the inference is that the HIGH spl of the discharge of the weapon contains frequencies "above normal hearing", that they, (for some reason) want to take into consideration. I can't see how a frequency that high, unless at some ungodly amplitude would have any bearing on anything especially given how fast the transient is. I mean you can't even make the argument for some kind of detection (like the microphones we have setup on the "bad side" of town to try and triangulate gun fire), since freq that high aren't going to make it that far anyway.

 

[ccaudle]

It seems conventional audio bandwidth (20-20k) should be sufficient.

That would be correct if you were talking about hearing something, but this use case is about possibility of damage by conditions far outside normal conditions, in some cases approaching the point where air itself becomes nonlinear. And the point about rapid absorption of high frequencies by air may be pertinent, but the distances are sub-meter, and in many cases are supersonic shock waves, not typical sound wave transmission, so again rules-of-thumb are not useful.

All that to say, normal limits which have implicit assumptions of linearity do not apply and need to be explicitly checked for these conditions (i.e. approaching 180dB SPL). In such a case it is definitely appropriate to use tools capable of measuring far beyond normal acoustic events until you can determine what physiological limits apply in those cases.
The original link to PEW Science is multiple pages, I have not finished reading through them yet, but so far (in the first two pages) is definitely in the careful science category, and nothing related to "audiophool" nonsense.

 

[JohnRoberts]

do you mean like the concussion grenades they used during basic training... ?

JR

 

[abbey road d enfer]

The original link to PEW Science is multiple pages, I have not finished reading through them yet, but so far (in the first two pages) is definitely in the careful science category, and nothing related to "audiophool" nonsense.

I agree, but nowhere do they give a justification for the 1 Ms/second SR. I suspect it's just because they have it.
It's extremely difficult in that type of measurement to extract the actual peak value, because the stimulus is not too far from a Dirac, which peak value, as you know, increases as the measurement BW increases.

 

[john12ax7]

There are definitely detrimental physiological effects due to high power supersonic events. Given that they are also transient you need much higher sampling rates than typically used for audio. 1M seems very reasonable.

 

[JohnRoberts]

fire arm silencers are pretty much passive, mechanical LPF...

I somehow doubt higher sample rates will reveal much new information.

JR

 

[moamps]

I read that article and it seems very interesting and in a way instructive to me.
Their conclusion "At lower sample rates, gunshot waveform peaks are missed. In those cases, reported" peak "values may actually be values occurring elsewhere in the waveform, away from the real peak altogether." it seems correct to me.
The question is whether a better system with a larger sample rate would give even better and more accurate results. Abbey suggested 140kHz as a sufficient sample rate. This would IMO mean that approximately only every eighth sample is taken in the graph below and the rest are deleted. Such a graph would certainly be inaccurate.



When it comes to a microphone that could be used for measurement, I would first suggest 1/4" capsules B&K 4135 and 36, or 4138 1/8" capsules with a bandwidth of 150kHz and a maximum SPL close to 170dB.

 

[gyraf]

..why the sample rate discussion - why not just peak-hold in analogue and call it a day..?

/Jakob E.

 

[abbey road d enfer]

..why the sample rate discussion - why not just peak-hold in analogue and call it a day..?

I think they need more than just the peak value. The actual evaluation of ARUs (Audio Risk Units) takes into account an integration of the convoluted waveform.

 

[abbey road d enfer]

Abbey suggested 140kHz as a sufficient sample rate. This would IMO mean that approximately only every eighth sample is taken in the graph below and the rest are deleted. Such a graph would certainly be inaccurate.

View attachment 92458

I suggested that in view of the HF response of the mic.
I saw the video of the experiment -which unfortunately I can't retrieve. The PCB soundmeters and capsule they used has a 70kHz HF response, which suggests a rise time of about 5us. The graph shows a little more than that, so the mic's HF response is probably the dominant restricting factor.
I submit the idea that, had they used a mic with an even more extended HF response, they may have measured a higher max pressure.
How it affects possible hearing or physiological damages is another story.
I agree with you that this is a very interesting and challenging thread.

 

[JohnRoberts]

As usual when discussions wander off into the weeds like this one, it is worth revisiting the OP..

This thread is about SPL measurements in the context of firearm silencers. I can imagine much higher rise time requirements for direct measurements near an un-silenced weapon discharge. Even a crude crappy silencer (soda bottles work in a pinch) will dramatically diffuse the rise time of discharge impulses. We may be over-thinking this.

JR

 

[user 37518]

I once took an acoustics instrumentation class in college, there are several ways of measuring SPL and its all standardized, just look at the standard. What I recall is that there are predefined integration times for slow integration, fast integration, it has an Leq reading, percentiles, etc.. weighting is usually A-weighted but can be varied in some cases, most noise level applications require an A-weighted curve so most SPL meters are like that.

Here are some articles that will give you an idea

Fast, Slow, Impulse Time Weighting - What do they mean?

What are LAeq and LAFmax?

What I also remember is that good SPL meters, like B&K cost thousands and thousands of dollars, and they can do much more than just measure SPL. If you are looking for an instantaneous SPL level meter, I don't know any that does that, it might be a specialized unit. Most meters integrate because what is important is the sound level over a significant sound period, transients can have a high level but little energy, and for noise standardization, an ultra fast reading is mostly useless, but rather the average level over a period of time.