Transient peaks with REW and RS meter

[nelmr]

A couple years ago on the S&V forums, I remember saying that the 1812 overture was only peaking at this and that amount on my RS meter. Flint, mentioned that it really isn't capable of showing the transient output as even the fast setting is more of an average, just faster than slow. Looks like we can capture them after all using REW. This is a hard hand clap about 4 feet from the mic:

It's interesting that a hand clap is about 105dB from that distance (the LZpeak value). Whereas the RS Meter on SLOW setting is showing about 70dB. On the fast setting it peaks past 80dB.

Pretty cool if you ask me.

 

[soundhound]

Those numbers are only valid if the microphone and preamplifier have been calibrated using a calibration standard. They are pretty expensive to buy.

http://www.bhphotovideo.com/c/produ...Audio_CM_C200_CM_C200_Calibrator_for_SPL.html

 

[nelmr]

Those numbers are only valid if the microphone and preamplifier have been calibrated using a calibration standard. They are pretty expensive to buy.

http://www.bhphotovideo.com/c/produ...Audio_CM_C200_CM_C200_Calibrator_for_SPL.html

Soundhound, I realize the values are not really precise. But doesn't it give a better approximation than just using the RS meter by itself? For example, the RS meter would show my hand clap at best would be around 85dB or less. 100+ sounds closer to reality for transients.

 

[soundhound]

If the meter is not calibrated, the numbers are almost useless. The ballistics of the RS meter do NOT conform to SPL meter standard specifications, and can be off by many dBs relative to a real SPL meter. Using it as a standard for your other program is not very useful. That's why real SPL meters are expensive: they are accurate and conform to known specs. What you have are really relative numbers which may or may not be truthful.

 

[nelmr]

If the meter is not calibrated, the numbers are almost useless. The ballistics of the RS meter do NOT conform to SPL meter standard specifications, and can be off by many dBs relative to a real SPL meter. Using it as a standard for your other program is not very useful. That's why real SPL meters are expensive: they are accurate and conform to known specs. What you have are really relative numbers which may or may not be truthful.

Yes, I understand that. I do use a calibration file to help offset it's inaccuracies that someone put together over at home theater shack averaged from several meters. In my post about the 15 configurations for the HSU sub, it should at least be able to show the relative differences between the settings and the overall effects.

But I do understand that without true calibration, the meter could be way off (even with an adjusted calibration file). For example, here is a sample of more accurate meters (85 mics) and there relative variances:

I'm not trying to say the RS meter is accurate, or that the hand clap is actually 105dB. I just thought that it is more useful in knowing the SPL difference between average and peak levels. With the RS meter alone, there isn't any feedback regarding peak transients at all.

 

[soundhound]

A calibration file is better than nothing and will at least get you in the ballpark.

There are a few definitions worth knowing about:

"Average" is what most analog and digital meters use, and this is an intregral approximation of true RMS. The problem is that "average" can be almost anything depending on the manufacturer. Cheap meters like the RS just hang a capacitor across the signal line from the mic preamp to slow down the meter movement. "Fast" and "slow" settings just use a different value capacitor (and that capacitor usually is electrolytic with +/- 20% or more tolerance).

"True RMS" is an absolute value which represents the energy in a waveform as defined as the heating ability of the waveform. In early meters, this was done by actually heating a filament within the meter and reading the temperature with a thermocouple. In modern meters the value is calculated mathematically. The RMS value of a sine wave is very close to .707 of its peak value (or another way of looking at it is the peak value is 1.414 of the RMS value). The true RMS value of an an AC wave has the same energy as DC of the same voltage. The RMS value of complex waveforms is much more difficult to measure, even with modern instruments. True RMS meters measuring transient waveforms are still subject to the variances of the ballistics of the meter.

"Peak" is also an absolute value, and as its name suggests is the highest instantaneous value of the waveform at any given moment. It has no relation whatsoever to "average" or "true RMS" values. It is difficult to measure the true peak value of a waveform with an analog meter as there will always be a ballistics lag unless sample and hold circuits are incorporated to hold the peak value for enough time for the ballistics of the meter to catch up. With purely electronic meters (such as your software), true peaks can be more easily read. Even in these instances however the software incorporates a sample and hold in order to freeze the actual peak value so that your eye can register it (or the peak number is held until a higher peak value is sampled). It is also important to know if the software algorithm is performing full wave rectification of the waveform so that both positive and negative peaks of as little as a single cycle (Hz) can be captured and held. The human ear does not really respond to "peak" SPL; the way the ear responds is more like the "average" value.

 

[nelmr]

Thanks soundhound! That is very informative.

 

[Aaron German]

Good discussion, guys.

 

[nelmr]

Those numbers are only valid if the microphone and preamplifier have been calibrated using a calibration standard.
...
If the meter is not calibrated, the numbers are almost useless.

Okay, now that I'm using a proper mic, couldn't it and REW be used to determine how well a system can handle dynamic (transient) peaks? For example take a look at this SPL logging of my office setup playing the 1812 overture (telarc 2001 CD track of the SACD).

The SPL being measured is zero-weight fast (125ms average). The average level of the music in this section is about 77dB. This one minute capture starts around 14:25. There are 11 cannon blasts in this section (they are easy to see - the big long peaks in the middle of the chart). The maximum 125ms average is the 10th cannon blast, just under 90dB. However, the "peak" SPL captured is about 101dB, also on the 10th cannon. So from 14:25-15:25 the dynamic peaks are about 24dB louder than the average level. The peaks don't really sound that loud, so I'm guessing that this is a pretty good measure of the dynamics, as the 13dB louder 125ms average peaks are more what one notices.

 

[soundhound]

The best test for how well a speaker handles transients is either a single cycle pulse wave, or a square wave, both of which have a leading edge slope which is infinitely steep. Most of the energy in a cannon shot is in the bass, which by definition is a slow event. Plus the cannon shots have been run through microphones (which do not have infinitely steep transient response), plus the recording medium which also introduces errors from an infinitely steep wavefront.