Amp power for music peak SPL?

[PENG]

Rich got me interested in measuring the SPL peaks in music. He asked in the other thread:

Is there an affordable device that could be attached to the speaker terminals to determine the average and peak power?

- Rich

Both Irvrobinson and I thought a multimeter alone could be used. Today I spent some time to do just that, using a Fluke 87 V and Diana Krall's The Look of Love track 5 (random pick).

The following are used in this experiment.

CD player: DVD 3910
Preamp: SC-7
Power amp: A21
Speakers: R900 (90 dB 2.83V/1M)
RS SPL meter with digital display, set to fast, 0.2 s response time,
Fluke 87V, response time 250 micro second, i.e. 0.2 ms

Results obtained using RS SPL meter:

Avg SPL at 1M: 79 dB (76 dB at 3M)
Pk SPL at 1M: 92 dB
DR = 13 dB

Results (Voltage) obtained using the Fluke meter:

V min.: 0.365
V avg.: 0.797
V pk.: 5.96

SPL calculated from above voltage readings:
Avg. SPL: 78.99 dB
Pk. SPL: 96.47 dB
DR = 17.48 dB

So the maximum peak SPL captured by the Fluke meter on that track is only 4.47 dB higher than that captured by the RS SPL meter.
The results also indicate that to get 76 dB from 3 meters (where I normally sit), I need about 0.08 W of average power and 4.44W of peak power base on the Fluke meter readings, or 0.16 W/8.88 W to allow for impedance dips coinciding with the peaks (worst case).

I am sure anyone will yield similar results using the 87V or equivalent meter. This experiment has not changed my mind, that a 100WPC amp will give me all the power I need so with 250W/8 ohms, 400W/4 ohms I do have enough power for peaks. Obviously, YMMV..

 

[avengineer]

...or you could have used this, and just plugged in the numbers:

Peak SPL Calculator

Perhaps a bit of conversion: the speaker sensitivity spec is given at 2.83 at 1M, and the nominal impedance of the speaker is 8 ohms, so that's the same as 1W@1meter.

In your measurements, the RS meter is never reading peak SPL, it does not have a peak detector, and the fastest ballistic is still an average. The two meters will never correlate because they use different detector types, the Fluke in Min/Max reads peak (hopefully both polarities!) the RS meter reads an average, similar to RMS, but NOT true RMS.

The ratio of peak:average is highly program material dependent. As a result, your ratios of peak to average are a bit misleading, and will be wildly different with different test material, and depend on musical content and instrumentation, and even more importantly the amount of dynamic processing applied to the recording.

No matter, though, since the only thing you really need to know is the maximum continuous SPL obtainable before the amp clips which will be fairly close to the maximum short-term SPL. You can get that from the calculator link above. Note that even though the calculator page is titled "Peak SPL Calculator", it is actually calculating the maximum RMS SPL, not peak, which is explained in the instruction section.

A Peak SPL meter is a specialized device usually only used for impulse noise measurement. Most affordable SPL meters are rough approximations of RMS, without a true RMS detector.

To be clear, a peak detector measures the highest peak of any waveform instantaneously and displays it accurately, but an RMS detector measures the square root of the arithmetic mean of samples of the entire wave for, and thus will read differently with different waveforms, though the peak voltage may be the same. To confuse things further, when measuring program material we must also take into account the meter integration time or response time, which is the time it takes to reach a stable and accurate reading of it's specific detector output, otherwise known as meter ballistics. All of this is also sort of academic since apparent loudness is what we are affected by, and that cannot be quantified by a peak detector at all, but requires an RMS detector with "history".

One last point, since the R900 has a varying impedance curve, the actual maximum SPL before amp clipping will be somewhat frequency dependent, though for practical purposes, with most low impedance output amps, you can ignore the variation. Related to this, the sensitivity of the speaker is also frequency dependent, so the voltage readings you take with the Fluke do not represent SPL accurately, and most certainly cannot be directly calculated to true power. This is a very important detail to keep in mind, especially when you make an assumption about what your average power is and what your media peak power is. But, again, this doesn't make all that much difference, as what you really want is the maximum capability of the system, average level is really not important.

 

[PENG]

...or you could have used this, and just plugged in the numbers:

Peak SPL Calculator

Perhaps a bit of conversion: the speaker sensitivity spec is given at 2.83 at 1M, and the nominal impedance of the speaker is 8 ohms, so that's the same as 1W@1meter.

I prefer to build my own spreadsheet, it is easy as abc using Excel. The online SPL calculator are too general, especially the Crown Audio ones. The one you link is better, and I had used it before but it still won't do everything I need in this experiment.

In your measurements, the RS meter is never reading peak SPL, it does not have a peak detector, and the fastest ballistic is still an average. The two meters will never correlate because they use different detector types, the Fluke in Min/Max reads peak (hopefully both polarities!) the RS meter reads an average, similar to RMS, but NOT true RMS.

Agree, and am fully aware of that. My Fluke did record both polarities. For that particular track, they were not the same, the + was quite a bit higher and that's the one I reported.

The ratio of peak:average is highly program material dependent. As a result, your ratios of peak to average are a bit misleading, and will be wildly different with different test material, and depend on musical content and instrumentation, and even more importantly the amount of dynamic processing applied to the recording.

Also agree, but that's why I specified the track used. I will do a few more using at least the SACDs that I think have the highest crest factors in my collection.

No matter, though, since the only thing you really need to know is the maximum continuous SPL obtainable before the amp clips which will be fairly close to the maximum short-term SPL. You can get that from the calculator link above. Note that even though the calculator page is titled "Peak SPL Calculator", it is actually calculating the maximum RMS SPL, not peak, which is explained in the instruction section.

Again, I just don't prefer to use online calculators that I cannot customize for my needs. I did use one in the linked below to check some of my findings and was happy to find that the results do correlate closely.

Audio calculations in English - acoustics calculator convert audio formulas sound calculations microphone formula sound recording studio useful stuff free audio calculator recording studio acoustic audio engineering site map online education sound ca

A Peak SPL meter is a specialized device usually only used for impulse noise measurement. Most affordable SPL meters are rough approximations of RMS, without a true RMS detector.

That's why I thought I should try the Fluke meter/electrical route instead. Again, thanks to Rich and Irv.

To be clear, a peak detector measures the highest peak of any waveform instantaneously and displays it accurately, but an RMS detector measures the square root of the arithmetic mean of samples of the entire wave for, and thus will read differently with different waveforms, though the peak voltage may be the same. To confuse things further, when measuring program material we must also take into account the meter integration time or response time, which is the time it takes to reach a stable and accurate reading of it's specific detector output, otherwise known as meter ballistics. All of this is also sort of academic since apparent loudness is what we are affected by, and that cannot be quantified by a peak detector at all, but requires an RMS detector with "history".

At 250 micro second, I think the 87V could capture the music wave peak voltage quite nicely. Being a true rms meter, it will likely still display the rms value, but for all intents and purposes, 250 microsecond peaks should be close enough to being "instantaneous". An individual (single)complex music waveform is not going to have crest factors much higher than that of a sine wave anyway. Again, my goal is simply to try and confirm that I won't running out of power (as in Rich case if you read his other thread) even during the rare extreme peak moments. In fact, the 5.96V happened just once and the more frequent peaks (say every 10 to 20 seconds or so) were well below 4V. So even if there were errors in the order of 200% I would still be well below the clipping threshold.

One last point, since the R900 has a varying impedance curve, the actual maximum SPL before amp clipping will be somewhat frequency dependent, though for practical purposes, with most low impedance output amps, you can ignore the variation. Related to this, the sensitivity of the speaker is also frequency dependent, so the voltage readings you take with the Fluke do not represent SPL accurately, and most certainly cannot be directly calculated to true power. This is a very important detail to keep in mind, especially when you make an assumption about what your average power is and what your media peak power is. But, again, this doesn't make all that much difference, as what you really want is the maximum capability of the system, average level is really not important.

Of course you are right but then as you said, that doesn't make all that much difference. I am still confident that the measurements do get me in the ball park, given that the 87V is true rms, and +/- 2% within a wide frequency range (5-20kHz).

 

[Irvrobinson]

I've done some speaker-terminal voltage measurements in the past, but I happen to have that same CD, so when I complete an urgent work assignment one of these days I'll run the same test on my system. It might be awhile, but I'm curious. By coincidence one of the tracks I have measured is also a rendition of The Look of Love, but it is on Herb Alpert's Colors album, and I measured levels in excess of 25v. Alpert's version is much more rambunctious, and the impressive voltage coincides with some impressive bass notes.

Your measurements beautifully illustrated the importance of sub-1w performance, which I personally find more interesting and intriguing than high power peaks.

 

[avengineer]

I prefer to build my own spreadsheet, it is easy as abc using Excel. The online SPL calculator are too general, especially the Crown Audio ones. The one you link is better, and I had used it before but it still won't do everything I need in this experiment.

Not sure what you are trying to customize, math is math...

Agree, and am fully aware of that. My Fluke did record both polarities. For that particular track, they were not the same, the + was quite a bit higher and that's the one I reported.

So long as you know the Fluke, even at the 250uS setting, is quasi-peak, not true peak.

Also agree, but that's why I specified the track used. I will do a few more using at least the SACDs that I think have the highest crest factors in my collection.

Program stimulated measurements are interesting, but precision measurements require known, and controlled stimulus. For program stimulus, take a look for software or apps that can do a real Leq measurement over time.

I use these tools, but the require calibration which is nontrivial.
SPL Graph | Studio Six Digital

That's why I thought I should try the Fluke meter/electrical route instead. Again, thanks to Rich and Irv.



At 250 micro second, I think the 87V could capture the music wave peak voltage quite nicely. Being a true rms meter, it will likely still display the rms value, but for all intents and purposes, 250 microsecond peaks should be close enough to being "instantaneous". An individual (single)complex music waveform is not going to have crest factors much higher than that of a sine wave anyway. Again, my goal is simply to try and confirm that I won't running out of power (as in Rich case if you read his other thread) even during the rare extreme peak moments. In fact, the 5.96V happened just once and the more frequent peaks (say every 10 to 20 seconds or so) were well below 4V. So even if there were errors in the order of 200% I would still be well below the clipping threshold.

Again, the Fluke in the 250uS setting is quasi-peak, and is no longer using the RMS detector. However, it's too slow. 250uS is a half-cycle of 2kHz. There are far faster peaks in audio that it will miss. The manual states it captures peaks longer than 250uS so 2KHz is an absolute top end max to it's peak detection accuracy, it will be progressively less accurate above that.

The conversion from quasi-peak voltage to SPL is fraught with inaccuracies. However, the SPL calculator will put you within 1dB of reality. Also, the RS meters always require calibration, I've found them off by as much as 2dB, with steady state pink noise.

I'm not so much criticizing your work, but putting it in the context of reality. With the above mentioned errors and inaccuracies, and unknown program stimulus, I feel that the on-line calculator will give you an answer that is accurate enough to choose an appropriate amp for your system, where your measurements can and do mislead.

Of course you are right but then as you said, that doesn't make all that much difference. I am still confident that the measurements do get me in the ball park, given that the 87V is true rms, and +/- 2% within a wide frequency range (5-20kHz).

True RMS requires time to be true RMS. According to the manual, setting the min/max parameter to 250uS results in peak measurement, not RMS, as shown by the PEAK indicator. The manual also states that it is peak reading for all evens longer than 250uS which limits its useful accuracy to 2kHz in that mode. Remember, a peak, even quasi-peak measurement is waveform agnostic, RMS is not. The differential will be highly waveform dependent.

What you need to know is the highest SPL that can be obtained in your system before the amp clips, or choose an amp that can deliver the desired maximum SPL (not peak!) without clipping. The only accurate way to come up with the answer is to establish that clipping threshold as a function of power into a load, and calculate the SPL. If you wanted to measure that level, you'd drive your amp to clipping then back down a tad, calibrate your SPL meter, use pink noise with a know crest factor, do a spacial/temporal average, and blah blah blah, and possibly hurt speakers and ears in the process. That's why the calculator will get you a usable answer without the measurement errors and difficulties.

 

[Irvrobinson]

I'm not so much criticizing your work, but putting it in the context of reality. With the above mentioned errors and inaccuracies, and unknown program stimulus, I feel that the on-line calculator will give you an answer that is accurate enough to choose an appropriate amp for your system, where your measurements can and do mislead.

Actually, you're coming off in this thread like a troll. A multimeter, even the 87V, is far from a perfect way to determine the power being dissipated by a speaker system, but it gives us some idea of the current on the wire, opposed to sitting around on the internet and using calculators based on other uncorrelated data points. I think PENG's measurements illustrate nicely that most of the time speakers need less than a watt of power, and that peaks on commercial recordings aren't as bad as many people think they are. PENG wrapped his conclusions in disclaimers and caution, and measurements like this are just some data points, and he knows and I know and any reasonable person who knows what volts on a speaker cable represents knows, that these measurements are far from definitive and aren't intended to be.

 

[PENG]

Not sure what you are trying to customize

To my own listening environment.. Those online calculators, I mean the better ones including the one in your link, tend to limit you to a few options in terms of speaker placement. It does get you close if you pick the one nearest to your own configuration though. In this case I plugged pick in a corner.... and it gets within 15% of my measured/calcualted number, not bad.

math is math...

Not sure what your point is in saying that, as I thought it should be obvious I believe in math. I would say though, one reason albeit a minor one, is that I am not going to blinding trust the Googleable math. Again, it took me only a few minutes to build my own calculator and that I know will give me the correct numbers when plugged in the measured data with my Fluke meter. For example, I don't have to rely on some assumed numbers base on roughly described speaker placement.

So long as you know the Fluke, even at the 250uS setting, is quasi-peak, not true peak.

I thought I told you I was fully aware of that, or maybe not.

Program stimulated measurements are interesting, but precision measurements require known, and controlled stimulus. For program stimulus, take a look for software or apps that can do a real Leq measurement over time.

I use these tools, but the require calibration which is nontrivial.
SPL Graph | Studio Six Digital

Thanks, I will certainly take a look.

Again, the Fluke in the 250uS setting is quasi-peak, and is no longer using the RMS detector. However, it's too slow. 250uS is a half-cycle of 2kHz. There are far faster peaks in audio that it will miss. The manual states it captures peaks longer than 250uS so 2KHz is an absolute top end max to it's peak detection accuracy, it will be progressively less accurate above that.

Of course there are far faster peaks but I really don't care about those because they will be high frequencies that hardly stress the amplifier. 0.25 ms represents one full cycle of a 4000 Hz signal, and it a 5000Hz signal last longer than one cycle, the Fluke should be able to catch it. It may well miss a cymbal crash but again would catch everything after the first cycle and may even catch the first impact of a large one use in symphony orchestras.

The conversion from quasi-peak voltage to SPL is fraught with inaccuracies. However, the SPL calculator will put you within 1dB of reality. Also, the RS meters always require calibration, I've found them off by as much as 2dB, with steady state pink noise.

That I fully disagree, that SPL calculator cannot not put you within 1 dB of reality, simply no way even if you only consider the assumption of speaker placement, and there are other factors that it does not allow/account for.

I'm not so much criticizing your work, but putting it in the context of reality. With the above mentioned errors and inaccuracies, and unknown program stimulus, I feel that the on-line calculator will give you an answer that is accurate enough to choose an appropriate amp for your system, where your measurements can and do mislead.

Now that I check my results with that calculator, I would have no problem using it as I am fine with even +/- 50% accuracy, but then now I have done my own I still won't need it. I am not sure how you figure my measurements can and do mislead when you have not seen my calculation sheets and not asked any questions. You can go ahead and criticize my experiement so long as you have necessary information to do so.

True RMS requires time to be true RMS. According to the manual, setting the min/max parameter to 250uS results in peak measurement, not RMS, as shown by the PEAK indicator.

It can capture the peak and then take the time needed to calculate the RMS, I don't see that a being problem for a $500 multimeter. I actually wish you were right, that it could actually capture and display the peak as is but I am not sure you are right as I read the manual more than once. When it says peak, I believe it still simply means the highest (hence peak) rms V captured at 250 microsecond response time. Again, I do wish you were right and it would even make sense to me but going by what I read, I highly doubt that is the case. This is what the maunal does say:

"Peak records the signal extremeslasting longer than 250 μs"<o></oThe fact that it says "lasting longer than 250μs tells me it is not the literal peak of the waveform. To me, to capture literally the peak point of the waveform it means instantaneous, that's way too fast for that Fluke meter. Anyway, you got me interested enough I may give Fluke a call tomorrow.

The manual also states that it is peak reading for all evens longer than 250uS which limits its useful accuracy to 2kHz in that mode. Remember, a peak, even quasi-peak measurement is waveform agnostic, RMS is not. The differential will be highly waveform dependent.

Which page did you find that information? I use the search function to search for peak, frequency, accuracy and could not find anything that talks about limiting its useful accuracy to 2 kHz in the peak mode.

What you need to know is the highest SPL that can be obtained in your system before the amp clips, or choose an amp that can deliver the desired maximum SPL (not peak!) without clipping. The only accurate way to come up with the answer is to establish that clipping threshold as a function of power into a load, and calculate the SPL. If you wanted to measure that level, you'd drive your amp to clipping then back down a tad, calibrate your SPL meter, use pink noise with a know crest factor, do a spacial/temporal average, and blah blah blah, and possibly hurt speakers and ears in the process. That's why the calculator will get you a usable answer without the measurement errors and difficulties.

Okay, let's be clear now, I agree with what you are saying. My experiment is inspired by the point brought up by Rich in his thread:

http://forums.audioholics.com/forums/amps-pre-pros-receivers/87598-you-running-out-amplifer-power-i-am.html

So I wanted to find out even if I agreed with him (that I really didn't) that he might be running out of power for the reasons he cited, I still would not have to worry about running out of power in my own set up. In doing so, I also can confirm that he is correct about the RS SPL meter not able to capture the music peak SPLs. It is not surprising and RS never claims such capability anyway, but our measurements put some hard numbers on the table, mine seems not as extreme as his, that's all.