Hello,
When I'm reading posts involving audio/music, it's common to find someone warning about the risk for some signals caring 3dB crest factor content in the low frequencies and once I got some knowledge I decided to investigate myself and I was never able to find anything close to this value. I'm opening this community request / database in order to collect and share information with mates.
Below some explanation from people that don't know much about this subject and also some guidance.
App to use (suggestion):
To edit audio file: Audacity (Free and open source available on Windows and Linux)
To calculate crest factor from audio file: GNU Octave (Free and open source available on Windows and Linux)
See attachment#1 for guidance: Crest Factor Thread.pdf
1) Find any music file witch is a good candidate to have low crest factor in the low frequency range.
2) Lets consider that this music in a PA will pass trough a DSP/Crossover going to Subwoofer channel that has Low-Pass Filter of 100Hz 24dB/oct and so apply this type of filter to your audio file.
3) Bennett Prescott, an Engineer from remarkable brand B&C said that a Sine Wave (Crest Factor of 3) takes 30s to heat the driver voice coil, so lets consider 60s (1min) of music file for this analysis. Source in the you tube video below:
4) Calculate the Crest Factor and other data using the the file in the attachment#2 Crest_Factor_Analysis.m (extract from zip)
5) Post the name of you music track, the artist, your music files (original + filtered), and your results (Crest Factor, Crest Factor dB, Peak and RMS).
Note: Narrow is the frequency range we evaluate, lower will be the Crest Factor, low frequencies (Subwoofers) are the worst cases.
Crest Factor is the ratio between the peak and and average level from a music signal and other signals. Sine wave is one of the must dangers signals to a transducer/driver once it carries a lot of energy. Crest Factor is 1/0.707 = 1,41 or in dB 20*log10(1/0.707) = 3dB
Note: attachment#3 has an 1kHz sine wave audio file to analyse. 1kHz.ogg (extract from zip)
Crest Factor per ser don't tell all the history, amplitude is also important, from the image above you can see 3 different sine wave with different amplitudes (1.5, 1.0 and 0.5), but all will have the same Cresft Facfor of 3 if analysed individually.
If the audio signal exceed the amplitude of amplifier input it can clip the signal increasing even more the energy sent to the transducer/driver. Lower the amplitude, lower is the energy.So the M-file will also calculate the peak value to the signal and its average value, if the Crest Factor is low and amplitude is also low, it might not be a problem.
When I'm reading posts involving audio/music, it's common to find someone warning about the risk for some signals caring 3dB crest factor content in the low frequencies and once I got some knowledge I decided to investigate myself and I was never able to find anything close to this value. I'm opening this community request / database in order to collect and share information with mates.
Below some explanation from people that don't know much about this subject and also some guidance.
App to use (suggestion):
To edit audio file: Audacity (Free and open source available on Windows and Linux)
To calculate crest factor from audio file: GNU Octave (Free and open source available on Windows and Linux)
See attachment#1 for guidance: Crest Factor Thread.pdf
1) Find any music file witch is a good candidate to have low crest factor in the low frequency range.
2) Lets consider that this music in a PA will pass trough a DSP/Crossover going to Subwoofer channel that has Low-Pass Filter of 100Hz 24dB/oct and so apply this type of filter to your audio file.
3) Bennett Prescott, an Engineer from remarkable brand B&C said that a Sine Wave (Crest Factor of 3) takes 30s to heat the driver voice coil, so lets consider 60s (1min) of music file for this analysis. Source in the you tube video below:
4) Calculate the Crest Factor and other data using the the file in the attachment#2 Crest_Factor_Analysis.m (extract from zip)
5) Post the name of you music track, the artist, your music files (original + filtered), and your results (Crest Factor, Crest Factor dB, Peak and RMS).
Note: Narrow is the frequency range we evaluate, lower will be the Crest Factor, low frequencies (Subwoofers) are the worst cases.
Crest Factor is the ratio between the peak and and average level from a music signal and other signals. Sine wave is one of the must dangers signals to a transducer/driver once it carries a lot of energy. Crest Factor is 1/0.707 = 1,41 or in dB 20*log10(1/0.707) = 3dB
Note: attachment#3 has an 1kHz sine wave audio file to analyse. 1kHz.ogg (extract from zip)
Crest Factor per ser don't tell all the history, amplitude is also important, from the image above you can see 3 different sine wave with different amplitudes (1.5, 1.0 and 0.5), but all will have the same Cresft Facfor of 3 if analysed individually.
If the audio signal exceed the amplitude of amplifier input it can clip the signal increasing even more the energy sent to the transducer/driver. Lower the amplitude, lower is the energy.So the M-file will also calculate the peak value to the signal and its average value, if the Crest Factor is low and amplitude is also low, it might not be a problem.
Last edited:
Below you can find my example, it isn't the best case but the music is well recorded with some synthesizer in the low frequencies but the Crest Factor is a little high.
Crest Factor Data (two columns means two channels - Left and Right)
Artist: Combichrist
Music: All Pain is Gone
CF =
3.7803 3.8881
CF_dB =
11.551 11.795
Peak_x =
0.9149 0.9876
RMS_x =
0.2420 0.2540
Crest Factor Data (two columns means two channels - Left and Right)
Artist: Combichrist
Music: All Pain is Gone
CF =
3.7803 3.8881
CF_dB =
11.551 11.795
Peak_x =
0.9149 0.9876
RMS_x =
0.2420 0.2540
@LORDSANSUI It would be preferred if it was better explained what is meant by a statement such as "a Sine Wave (Crest Factor of 3) takes 30s to heat the driver voice coil". After all, isn't it just as true that any signal will "heat a driver voice coil"? It would be nice to be on the same playing field when discussing this topic.
I would guess that what was meant was that it takes a sine-wave signal 30 seconds to heat up a voice coil so that its temperature reaches a steady state condition. I'd be grateful for some further confirmation or some attendant clarification!
That clipped waveform looks suspiciously like the textbook square wave. Hence there is the fundamental sine-wave component at the original frequency of the square wave. We also have higher odd-order components at 3x, 5x, 7x, etc., of the original signal frequency. Interestingly, their amplitudes are 1/3, 1/5, 1/7, etc., of the amplitude of the square wave.
Hence, when feeding an amplifier with a square wave signal, getting it to the maximum amplitude that the amplifier can reproduce does require a greater power output than for a pure sine wave alone. If we take the fundamental plus the first three harmonics, we see that the total signal contains about 17% more power than a pure sine wave at the fundamental frequency that is of the same peak amplitude as the square wave.
Last edited:
Sure. Even one pW of dissipated power will increase the temperature of the voice coil by some minuscule amount, regardless of waveform. I agree that precision in language is often lacking.
Tom
I would expect that it needs infinite time to reach steady state, and that it is within some unspecified error percentage from its steady-state value after 30 seconds.
Regarding sloppy language, the average value of a sine wave without DC offset is 0, not 0.707 times the peak value. Its RMS value is (1/2) √2 ~= 0.707 times the peak value.
I've been told that some of the music of Armin van Buuren is quite close to square waves. I haven't checked if it is true, but the person saying that designed audio power amplifiers for a living.
Last edited:
That´s a nice lil trainwrack we have here, good late morning! :-D
Where physics and math hits reality. The issue is easy. With synthetic "disco kind of music" it is very easy to burn down the speaker. Even high end setups have about 10dB limiter in place in order for it to not happen. So I am not with Bennett at one in this. He says "get all the power you can to feed the speaker", and that does not work in practise. I have some (parts of) tracks with dynamic range of about 4-4,5dB. Could that be crest factor of 7-7,5dB? Don´t know. What I know is that I cooked few coils a bit, not even trying to abuse em. Might participate some more in the topic, but at the moment I cannot be swayed by any data presented, showing that it´s all okay and music is not that demanding. 🙂
Where physics and math hits reality. The issue is easy. With synthetic "disco kind of music" it is very easy to burn down the speaker. Even high end setups have about 10dB limiter in place in order for it to not happen. So I am not with Bennett at one in this. He says "get all the power you can to feed the speaker", and that does not work in practise. I have some (parts of) tracks with dynamic range of about 4-4,5dB. Could that be crest factor of 7-7,5dB? Don´t know. What I know is that I cooked few coils a bit, not even trying to abuse em. Might participate some more in the topic, but at the moment I cannot be swayed by any data presented, showing that it´s all okay and music is not that demanding. 🙂
I'm focused on measure music file and check it's Crest Factor at low frequency range. That sentence was just a background, if you watch entirely the video from B&C engineer you find your answer there with more in deep analysis.
Average value 0 is valid only for voltage, but I was speaking about power average and it's not 0, you probably pay a bill for electric consumption for your house.
Power = Voltage^2/resistance
any negative value ^2 will became positive.
Lets check it, the proposal for this thread is to measure real music file, Armin van Buuren is a good candidate.
Sometimes what people say can be or not aligned with real data. If you not able to measure I can do it for you, just let me know what music.
Post your tracks here, with it's data.
Unfortunately, that remark about Armin van Buuren was just a short remark someone made during a meeting over a decade ago. He didn't give any details about which Armin van Buuren tracks he referred to.
@LORDSANSUI
I can be wrong. I am willing to be wrong. Let´s this be a lesson for me. I don´t know. Let´s see... It depends...
25:18-27:40 where this is, track name in the app header.
What strucks me this is quite slow chill music. Finding something harder, I believe it can be outdone.
I can be wrong. I am willing to be wrong. Let´s this be a lesson for me. I don´t know. Let´s see... It depends...
25:18-27:40 where this is, track name in the app header.
What strucks me this is quite slow chill music. Finding something harder, I believe it can be outdone.
Last edited:
Here, better.
I know, 6dB, but what basshedad DJ dark psy whatever doesn´t clip a little bit of the signal. Seeing the difference between peak RMS and average, that´s just 2.35dB. It is not full dynamic range, but when playing hot, one CAN get close to it. And that´s still music. Like, something consumable, half commerce. Give me some real underground....
I know, 6dB, but what basshedad DJ dark psy whatever doesn´t clip a little bit of the signal. Seeing the difference between peak RMS and average, that´s just 2.35dB. It is not full dynamic range, but when playing hot, one CAN get close to it. And that´s still music. Like, something consumable, half commerce. Give me some real underground....
Post the music file, just compress it as zip file and you can attach.
by the graphic in your app, it looks very dynamic, so the Crest Factor will be high.
I don´t know if it is fair to hog the server with these files. Tried to upload it somewhere online:
https://limewire.com/d/ccc37e2c-c62...2#H06CxvoC9Q27TlK3h7TNsKNo7aNBrBcU4TYLJtrQh5o
The whole track with all pauses and different beats will of course have more crest factor. This is cutout from about 7-8 minute....
https://limewire.com/d/ccc37e2c-c62...2#H06CxvoC9Q27TlK3h7TNsKNo7aNBrBcU4TYLJtrQh5o
The whole track with all pauses and different beats will of course have more crest factor. This is cutout from about 7-8 minute....
I wonder if it would even be legal to upload a complete minute or even more of a copyrighted music recording.
As sugested by @Crashpc
Artist: Space Noise
Music: Musical Hallucinations (Progressive Psytrance Mix Okt 2017)
CF =
3.0870 2.8667
CF_dB =
9.7909 9.1476
Peak_x =
0.8917 0.8294
RMS_x =
0.2888 0.2893
Note: It was possible to post only the 60s sample with LP Filter. The original sound is too big and not accepted by the diyaudio server.
Artist: Space Noise
Music: Musical Hallucinations (Progressive Psytrance Mix Okt 2017)
CF =
3.0870 2.8667
CF_dB =
9.7909 9.1476
Peak_x =
0.8917 0.8294
RMS_x =
0.2888 0.2893
Note: It was possible to post only the 60s sample with LP Filter. The original sound is too big and not accepted by the diyaudio server.
Good question, maybe moderator can help, if not legal we can post only the results to avoid issues.
@AllenB do you mind to check
@LORDSANSUI
Yes, but arguably, that's not what is often finally fed to the speaker. As you wrote, filtered, and somewhat compressed by slight output clipping or DSP limiter that ďoesn't only limit big averages but also peaks. More investigation would be needed.
Yes, but arguably, that's not what is often finally fed to the speaker. As you wrote, filtered, and somewhat compressed by slight output clipping or DSP limiter that ďoesn't only limit big averages but also peaks. More investigation would be needed.
That Crest Factor is what in fact feeds the driver/subwoofer once I already applied the Low-Pass Filter. But what we don't know is the amplitude, the driver power handling long term, amplifier voltage output, gain chain, etc.
But the main point is, people mention that there are music with Crest Factor close to sine wave, Where are they? Are they a myth?
Yes, we don´t know the chain. It is an argument we cannot take on, but "otherwise there would not be so many burned speakers in the field". It is a nice attempt to tackle this though. I lightly cooked my speakers in front of my eyes playing normal "hard dance disco music". Nothing specifically ill going to the speakers.