Calculate amp power for target SPL at a distance - assumptions/compromise/datasheet

[jmf11]

Hello,

I try to get some understanding of which power is needed in active speakers to achieve a SPL level at defined listening distance. I fist checked the datasheets of products like Neumann monitors or JBL 305P (and for KH80 and JBL 305P we know which amp chips are used).

I did some home work to learn the basis (https://audioxpress.com/article/target-spl-and-required-amplifier-size) and found an online calculator: http://www.mh-audio.nl/Calculators/REAP.html

I wonder:

• which amplifier headroom / crest factor to consider. I understand that 15 dB is a safe bet for music (uncompressed. What do you use?
• to which figure from amplifiers datasheets to consider for the power: Output power THD+N: 10% ? 1%? 0.1%?
• I imagine that with 15 dB Crest factor, the Crest values are rarely hit and high distorsion would apply to a minor part of the signal. But is this true?

Would 15dB crest factor x 0.1% be the target ? or overkill?

Which "combinations" would we expect in Active Monitors?

A STA350BW amp (the one inside the JBL) stays in 6R below 0.1% THD+N up to 2x10W (but is rated 2x50W). Using a Woofer with a 85 dB sensitivity, using the Calculator, I would achieve 74 db SPL at 2m (1 woofer - 10W). Using the 2x50W figure (10% THD+N), It would achieve 81 dB at 2m. Which is the more realistic assumption ?

Best regards,

JMF

 

[staticV3]

which amplifier headroom / crest factor to consider. I understand that 15 dB is a safe bet for music (uncompressed. What do you use?

The most compressed music I've come across is about 13dB between AVG and Peak. The least compressed track I have is 27dB.
15dB seems like a solid, best case assumption.

 

[Hatto]

The most compressed music I've come across is about 13dB between AVG and Peak. The least compressed track I have is 27dB.
15dB seems like a solid, best case assumption.

Are you sure 27dB is the crest factor (CF) and not the total dynamic range? That would approx. correspond to 54dB dynmaic range which sounds insane.

The highest CF I've ever come across was 17dB (and it was annoying to listen), now I'm extremely curious, care to share the file with 27dB CF?

 

[DonH56]

I use this one (below) and shoot for being able to hit the highest peaks I expect to ever need. In my case I just designed my system for 105 dB max peaks at the listening position to meet the THX spec (and 115 dB for the subs). I work backwards from the loudest volume I am likely to stand and figure the power needed from there. My actual levels tend to be significantly lower, rarely exceeding 90 dB peaks.

Peak SPL Calculator

For music, peak-to-average power is about 17 dB per an old, old paper I do not have at hand. See my usual post below. I go by rated max power which varies with the amp's spec, natch. But usually the 1% corner is pretty close to the spec max so I don't worry about it. I also figure a little clipping on the loudest peaks is not going to be noticed with distortion the speakers are likely generating, even assuming the loudest peaks are not cymbal crashes or explosions that sound like clipping anyway.

--- Old Post ---

Power Needs:

For many years I have used 17 dB peak-to-average power based on an old AES article I can no longer find. Anecdotally various audio fora report 20 dB or more for movies. IME most people use much less average power than they think, but require more peak power. An online calculator can help you estimate your average power; note 80 dB is very loud to me (YMMV). Here is one:

Peak SPL Calculator

Most people think of volume in dB and most modern AVR/AVP/etc. units list dB on the volume knob. Power in dB goes as 10log10(Power_ratio) so the change in power is 10^(dB/10). Here are some reference numbers in dB and power:

1 dB is barely noticeable and requires 1.26x the power
3 dB is what most people hear as "a little louder" and requires 2x the power
6 dB is significantly louder and requires 4x the power
10 dB sounds twice as loud and requires 10x the power
17 dB is the headroom for music and requires 50x the average power
20 dB for movies requires 100x the power

If you listen at around 1 W average, then you need 50~100 W to avoid clipping on most source material. You can figure out your estimated average power from the calculator knowing your speaker's sensitivity and distance from them. Note music may be more compressed and thus require less headroom, and the loudest sounds in movies tend to be things like gun shots and explosions where a little clipping is likely unnoticeable.

HTH - Don

 

[staticV3]

Are you sure 27dB is the crest factor (CF) and not the total dynamic range?

27dB is the difference between dBFS AVG and dBFS Peak. See here.

 

[charleski]

Are you sure 27dB is the crest factor (CF) and not the total dynamic range? That would approx. correspond to 54dB dynmaic range which sounds insane.

The highest CF I've ever come across was 17dB (and it was annoying to listen), now I'm extremely curious, care to share the file with 27dB CF?

It's not unknown, especially with classical music. Here's a histogram of a track from Aida:

Peak -0.54dB, RMS Average -29.31dB

 

[Hatto]

27dB is the difference between dBFS AVG and dBFS Peak. See here.

And @charleski

I don't know how those sound amplitude values correlate to voltage amplitude values at the amplifier output (or speaker input).

However, 27dB voltage crest factor corresponds to 65Vpeak for a mere 1Wrms power (assuming 2.83V into 8ohms), which then corresponds to 530W peak power!!!

530W peak power requirement for just 1W RMS sounds pretty unreasonable, which makes me wonder if sound CF can be used directly as a baseline for amplifier power headroom requirements.

As another example, M-Noise Audio Sample (https://m-noise.org/procedure/) is advertised as having "a crest factor of at least 17.5 dB in every 10 second interval", which corresponds to 7.5x amplitude ratio, however the voltage crest factor is only 4.5x in absence of any clipping or compression (measured here: https://audiosciencereview.com/foru...te-new-schiit-speaker-amps.42329/post-1527503).

 

[NTK]

As another example, M-Noise Audio Sample (https://m-noise.org/procedure/) is advertised as having "a crest factor of at least 17.5 dB in every 10 second interval", which corresponds to 7.5x amplitude ratio, however the voltage crest factor is only 4.5x in absence of any clipping or compression (measured here: https://audiosciencereview.com/foru...te-new-schiit-speaker-amps.42329/post-1527503).

I ran some analyses on M-Noise. You'll need to test pretty long samples to get the correct CF. The highest peaks are few and far in between.

Here are my numbers for the 53 seconds signal. I calculated the CF's for entire 53 seconds, and by splitting the 53 seconds signal into 2, 4, 8, 18, 32, and 64 equal length slices.

And the RMS values.

 

[Hatto]

I ran some analyses on M-Noise. You'll need to test pretty long samples to get the correct CF. The highest peaks are few and far in between.

Here are my numbers for the 53 seconds signal. I calculated the CF's for entire 53 seconds, and by splitting the 53 seconds signal into 2, 4, 8, 18, 32, and 64 equal length slices.

View attachment 283359
And the RMS values.

View attachment 283362

Your results (7.75x CF) seem consistent with M-Noise specs (17.5dB). My sampling rate for the voltage crest factor was way much higher at 20kSa/s (above 1 million slices for 53s as you would put it) and the voltage crest factor was nowhere near 7.75x.

Which only counts more towards demonstrating that digital amplitude CF for the sound file is not the same as analog voltage CF output for the amplifier.

 

[NTK]

Your results (7.75x CF) seem consistent with M-Noise specs (17.5dB). My sampling rate for the voltage crest factor was way much higher at 20kSa/s (above 1 million slices for 53s as you would put it) and the voltage crest factor was nowhere near 7.75x.

Which only counts more towards demonstrating that digital amplitude CF for the sound file is not the same as analog voltage CF output for the amplifier.

I did measure it with a scope too, and it came pretty close (17.1 dB) for a 10+ second slice. I think in this case the minimum phase reconstruction filter in my Apple headphone adapter reduced the CF a little bit. Normally it should be the reverse (minimum phase reconstruction filter increases CF, e.g. with square waves).

 

[Hatto]

I did measure it with a scope too, and it came pretty close (17.1 dB) for a 10+ second slice. I think in this case the minimum phase reconstruction filter in my Apple headphone adapter reduced the CF a little bit. Normally it should be the reverse (minimum phase reconstruction filter increases CF, e.g. with square waves).

View attachment 283363

This is interesting. What device did you measure wotht the scope?

 

[NTK]

This is interesting. What device did you measure wotht the scope?

I used a Digilent Analog Discover 2 to measure an USB-C Apple headphone adapter and the headphone output of my laptop (results below). My laptop headphone output gave a CF of 18 dB. My guess is that it uses a linear phase reconstruction filter and can maintain the CF better.

 

[DonH56]

Not sure if I'm happy or sad that my 40-50 year old 17 dB number is still reasonable...

 

[DVDdoug]

I don't know how those sound amplitude values correlate to voltage amplitude values at the amplifier output (or speaker input).

A 27dB difference in the digital level makes a 27dB difference in amplifier output (assuming no clipping and everything remaining linear, etc.).

27dB does "seem like" a lot, but I'm not a classical music fan. I know the 1812 Overture has cannon shots at the end! But if I'm remembering it correctly it's not otherwise a "quiet" piece so although the peak is probably louder than anything else (uncompressed) the average probably isn't that quiet.

 

[RayDunzl]

A STA350BW amp (the one inside the JBL) stays in 6R below 0.1% THD+N up to 2x10W (but is rated 2x50W). Using a Woofer with a 85 dB sensitivity, using the Calculator, I would achieve 74 db SPL at 2m (1 woofer - 10W). Using the 2x50W figure (10% THD+N), It would achieve 81 dB at 2m. Which is the more realistic assumption ?

JBL LSR 308 multiple sweeps.

Two speakers playing, measured at 19 feet.

The lower frequencies start to fall apart above 95dB,

Single speaker would be 89dB (at 10 feet in a "normal" room)

They seem to distort readily, the rise starts around 80dB at 10 feet.

I can't tell you if it is the amp or the driver that is providing the high distortion levels.

Probably the flexi-plastic woofer, for the most part.

 

[Deleted member 48726]

I use this one (below) and shoot for being able to hit the highest peaks I expect to ever need. In my case I just designed my system for 105 dB max peaks at the listening position to meet the THX spec (and 115 dB for the subs). I work backwards from the loudest volume I am likely to stand and figure the power needed from there. My actual levels tend to be significantly lower, rarely exceeding 90 dB peaks.

Peak SPL Calculator

For music, peak-to-average power is about 17 dB per an old, old paper I do not haver at hand. See my usual post below. I go by rated max power which varies with the amp's spec, natch. But usually the 1% corner is pretty close to the spec max so I don't worry about it. I also figure a little clipping on the loudest peaks is not going to be noticed with distortion the speakers are likely generating, even assuming the loudest peaks are not cymbal crashes or explosions that sound like clipping anyway.

--- Old Post ---

Power Needs:

For many years I have used 17 dB peak-to-average power based on an old AES article I can no longer find. Anecdotally various audio fora report 20 dB or more for movies. IME most people use much less average power than they think, but require more peak power. An online calculator can help you estimate your average power; note 80 dB is very loud to me (YMMV). Here is one:

Peak SPL Calculator

Most people think of volume in dB and most modern AVR/AVP/etc. units list dB on the volume knob. Power in dB goes as 10log10(Power_ratio) so the change in power is 10^(dB/10). Here are some reference numbers in dB and power:

1 dB is barely noticeable and requires 1.26x the power
3 dB is what most people hear as "a little louder" and requires 2x the power
6 dB is significantly louder and requires 4x the power
10 dB sounds twice as loud and requires 10x the power
17 dB is the headroom for music and requires 50x the average power
20 dB for movies requires 100x the power

If you listen at around 1 W average, then you need 50~100 W to avoid clipping on most source material. You can figure out your estimated average power from the calculator knowing your speaker's sensitivity and distance from them. Note music may be more compressed and thus require less headroom, and the loudest sounds in movies tend to be things like gun shots and explosions where a little clipping is likely unnoticeable.

HTH - Don

There should be a "mark as best answer" button.

I would add that the current capability of the amp should be considered as an important feat. as well. Look at the ratings of the amp at that calculated power and see if the manufacturer have specs. into 4 / 2 or 1 ohm.

 

[jmf11]

Thanks to all about your answers, and especially DonH56 for the very practical and actionable elements and the confirmation of the orders of magnitude. Much appreciated.

Thanks also to Raydunzl for the JBL 308 use case and "trade-off values" for that product. As a comparison, from the datasheets:
- the JBL 308 with its 8" relies on 2x56W
- the Neumann KH150 with a 6.6" implements 100+150X (10% THD).

Making the assumption that the drivers sensitivities are similar (to be checked), one of the 2 designs has taken more margins that the other ;-)

And as a side note, I have to pay attention that all this is log scale, when we are used to manipulate amp power in linear scale.

Best regards,

JMF

 

[Sokel]

I don't know if I learned well what @staticV3 taught me but that's what I get from his 27db song:

Average-peak:




Their Delta:

 

[jmf11]

Second step question that popped to my mind. I understand the principle for rull range operation. What happens if the Frequency range is split in two: low / high?

How to factor the allocated frequency range in the process?

JMF

 

[Koeitje]

IME most people use much less average power than they think, but require more peak power.

This 100%. You need both more and less power than you expect.