Fast transient in music recordings - tweeter load - amp response

[pma]

It is generally acknowledged that music signals do not contain much of high frequencies / fast transients of high amplitude and that the music signal looks like a pink noise. It is also suggested that tweeters do not need to handle much power and that amplifier power capabilities at higher frequencies are unimportant, because of the suggestion here above. However, this is often not true and especially in case of some recording of rock music.

Please let me show time record at the amplifier output, playing Secret World Live album by Peter Gabriel, Red Rain song. I could demonstrate a plenty of similar plots. The graphs are calibrated, sampled at amplifier output with 390 kHz sampling frequency. The trigger level was set at 30Vpeak, this would make 225Wpeak at 4ohm. The time needed from 0V to 30V is only 38 microseconds. This is to be handled exclusively by the tweeter, so see what is the tweeter peak power. And also the amplifier is asked to deliver this in full amplitude and not clipped. If clipping occurs, the tweeter power load is much higher.

Yes it was playing very loud.

Looking forward to averaged "pink noise spectrum" suggestions

 

[pma]

Now, the simulation of amplitudes that a mid-woofer section and tweeter section of one of my speakers (with the real crossover schematics used) see with the Red Rain song used as a source. 30s sample is used and 10us is the shortest sampling. Crossover frequency is 2.4kHz.

Top trace is the song data, middle trace is the mid-woofer section and bottom trace is the tweeter section. One can see that tweeter section amplitudes are up to 0.6 FS (full scale), mid-woofer section is up to 0.8 FS and the maximum data sample is about 0.9 FS. This clearly busts the myth that the tweeter section does not need to be rated at similar peak power as mid-woofer section. It all depends on music used and the worst case must be always considered.

 

[KSTR]

Thanks, Pavel.
In a passive speaker, the tweeter is usally padded down so it will not actually see that much peak power but still substantial, of course.

 

[KSTR]

This clearly busts the myth that the tweeter section does not need to be rated at similar peak power as mid-woofer section. It all depends on music used and the worst case must be always considered.

As I've written recently here, 300W of amp power on a 10W tweeter in an active speaker is not overkill for transients. Needs a proper thermal limiter of course.

 

[ctrl]

The trigger level was set at 30Vpeak, this would make 225Wpeak at 4ohm.

As I've written recently here, 300W of amp power on a 10W tweeter in an active speaker is not overkill for transients. Needs a proper thermal limiter of course.

How loud should the "transients" be reproduced? An amplifier with large power reserves is certainly a fine thing, especially for enthusiasts, but for the normal average listener significantly less power should be sufficient - or not?

A tweeter, depending on quality and impedance, will reach about 87-92dB@1m sound pressure at 2.83V - let's assume at 5 ohms. If you seriously want to achieve high sound pressure levels, you will use a compression driver (plus horn) anyway, and then you will be well above 100dB@1m.

At 100W amplifier power, we would be around 22V voltage in the above example (at 5 ohm impedance).
The possible achievable sound pressure level would then be 20*log(22/2.83) ~ 18dB above [email protected] - if my calculation is correct

For the tweeter this would be 105 - 110dB@1m and for the compression driver well above 118dB@1m. At 300W power, you would be able to achieve another 4-5dB higher sound pressure level.

These would be the sound pressure levels in the high frequency range, if we then also consider the room response slope, then we are at "insane" sound pressure levels in the low frequency range.

 

[KSTR]

It all depends on the program material. With loudness-war crushed-to-death final masters you don't need much power reserves. In a studio monitor used to monitor raw signal it's another story. 20dB SPL headroom above continuous thermal limit is not completely over the top. That means 1kW for a 10W tweeter ;-)
Ever wondered why active monitors, as well as hifi/studio amps in general, almost never have instantaneous clipping indicators? You clip them all the time if you give them (semi-)raw program material.

 

[Matias]

38 us = 2,6 kHz, right? This is the bottom range of most tweeters, maybe even crossover range. And hardly high frequency for the amplifiers range.

Edit: No, wait, 38 us was from 0 to 30 V, so 1/4 wave. Total wave would be 4 x 38 = 152 us = 650 Hz, right?

 

[KSTR]

38 us = 2,6 kHz, right?

No, missed by a factor of 10.

 

[Matias]

No, missed by a factor of 10.

1 / (38 x 10^-6) = 0.026 x 10^6 = 26 x 10^3 = 26 kHz, you are right.
And the entire wave would be 6.5 kHz, right in the middle of tweeter range.
Thanks.

 

[DVDdoug]

This clearly busts the myth that the tweeter section does not need to be rated at similar peak power as mid-woofer section.

If it was a myth we'd be swimming in blown tweeters...

Voice coils are burned-out by heat and it takes some time to heat-up anything. How much time I don't know, but RMS power is a better indication of the energy dissipated by the driver.

If you were to re-wire/miswire your crossover, sending the lows to the tweeter you'd probably kill it.

It might be interesting to see if the tweeter is linear on those peaks...

 

[spacevector]

Yes it was playing very loud.

Can you estimate dB SPL at 1m?

Are you able to also probe the tweeter current and plot along with voltage? If you have a good impedance model, may be this current can also be simulated?

 

[NTK]

... Voice coils are burned-out by heat and it takes some time to heat-up anything. How much time I don't know ...

Back of the envelope calculations ...

Tweeter voicecoil geometry assumptions:

Wire diameter = 38 gauge = 0.1 mm = 0.0001 m​

Coil diameter = 25 mm = 0.025 m​

Number of turns = 24​

Cross sectional area of voicecoil = 0.25 * π * 0.0001^2 = 7.854e-9 m^2
length of wire = π * 0.025 * 24 = 1.885 m
Volume of wire = 1.48e-8 m^3

Density of copper = 8940 kg/m^3
Specific heat capacity of copper = 385 J/(kg °C)

Therefore, the amount of energy to raise the voicecoil temperature by 1 °C, assuming no heat loss, is:

Volume of wire * Density of copper * specific heat capacity = 0.05 Joule/°C​

Melting temperature of copper is 1085 °C. To raise the temperature of this voicecoil by 1000 °C will require 0.05 * 1000 = 50 Joule. Thus, 50 W for 1 second (or 100 W for 1/2 second, etc.) will do it for this hypothetical tweeter.

 

[bigjacko]

Back of the envelope calculations ...

Tweeter voicecoil geometry assumptions:

Wire diameter = 38 gauge = 0.1 mm = 0.0001 m​

Coil diameter = 25 mm = 0.025 m​

Number of turns = 24​

Cross sectional area of voicecoil = 0.25 * π * 0.0001^2 = 7.854e-9 m^2
length of wire = π * 0.025 * 24 = 1.885 m
Volume of wire = 1.48e-8 m^3

Density of copper = 8940 kg/m^3
Specific heat capacity of copper = 385 J/(kg °C)

Therefore, the amount of energy to raise the voicecoil temperature by 1 °C, assuming no heat loss, is:

Volume of wire * Density of copper * specific heat capacity = 0.05 Joule/°C​

Melting temperature of copper is 1085 °C. To raise the temperature of this voicecoil by 1000 °C will require 0.05 * 1000 = 50 Joule. Thus, 50 W for 1 second (or 100 W for 1/2 second, etc.) will do it for this hypothetical tweeter.

Are all the power towards the tweeter ends up being heat? Will some power ends up being movement for the tweeter so we can hear it playing? If it is true then how much power goes to heat and sound?

 

[dc655321]

Are all the power towards the tweeter ends up being heat? Will some power ends up being movement for the tweeter so we can hear it playing? If it is true then how much power goes to heat and sound?

Drivers are about 1% efficient in converting electrical to acoustic energy. So, almost negligible...

 

[bigjacko]

Drivers are about 1% efficient in converting electrical to acoustic energy. So, almost negligible...

Is the low efficiency because of high impedance and hard to drive phase? Is hard to drive phase due to voice coil, and can it be avoided somehow?

 

[dc655321]

Is the low efficiency because of high impedance and hard to drive phase? Is hard to drive phase due to voice coil, and can it be avoided somehow?

It's due to impedance mismatching - acoustic impedance, not electrical impedance.

 

[Kvalsvoll]

and can it be avoided somehow?

Yes, if you don't need bass, and can live with very large speakers. But, why..

1% is actually quite good for efficiency, most hifi speakers are well below that. There is at least one commercial speaker I know of with 30% efficiency - downside is, that is about the only thing it gets right.

 

[pma]

Thanks, Pavel.
In a passive speaker, the tweeter is usally padded down so it will not actually see that much peak power but still substantial, of course.

Exactly, it all depends on driver's sensitivity and if the series resistor is/is not used.

As I've written recently here, 300W of amp power on a 10W tweeter in an active speaker is not overkill for transients. Needs a proper thermal limiter of course.

That's a great point because the requirements on peak amplitude applies to the amp as well.

It all depends on the program material. With loudness-war crushed-to-death final masters you don't need much power reserves. In a studio monitor used to monitor raw signal it's another story. 20dB SPL headroom above continuous thermal limit is not completely over the top. That means 1kW for a 10W tweeter ;-)
Ever wondered why active monitors, as well as hifi/studio amps in general, almost never have instantaneous clipping indicators? You clip them all the time if you give them (semi-)raw program material.

Sure, and the program material with fast transients cannot be excluded from considerations. The "average music" is no measure.

38 us = 2,6 kHz, right? This is the bottom range of most tweeters, maybe even crossover range. And hardly high frequency for the amplifiers range.
Edit: No, wait, 38 us was from 0 to 30 V, so 1/4 wave. Total wave would be 4 x 38 = 152 us = 650 Hz, right?

10 times more, 6500Hz as you have already corrected. Yes but as always, we need a deeper view which is often not clear from a simple consideration. The right way is to see which portion of the program material travels through the crossover to the tweeter. Simulators are able to tell us this exactly, if we feed them with crossover circuit used and take a song wav file as an input. It is not difficult and gives exact results and answers. Below is a part of the transient analysis of input song data and tweeter section voltage, behind the crossover circuit. Please see the frequencies up to 14.5 kHz at high amplitude, but also please consider that the frequencies are of secondary importance, what we care about are the peaks because we see them in their real amplitude at the tweeter input. And there is not a thermal issue as a primary one, but the tweeter has to handle the high peaks without high distortion (membrane break-up) and mechanical damage due to high excursion.

 

[Kvalsvoll]

Now, the simulation of amplitudes that a mid-woofer section and tweeter section of one of my speakers (with the real crossover schematics used) see with the Red Rain song used as a source. 30s sample is used and 10us is the shortest sampling. Crossover frequency is 2.4kHz.

Top trace is the song data, middle trace is the mid-woofer section and bottom trace is the tweeter section. One can see that tweeter section amplitudes are up to 0.6 FS (full scale), mid-woofer section is up to 0.8 FS and the maximum data sample is about 0.9 FS. This clearly busts the myth that the tweeter section does not need to be rated at similar peak power as mid-woofer section. It all depends on music used and the worst case must be always considered.

View attachment 133118

This is a picture of a typical situation playing music material - energy levels fall of at higher frequencies, but peak level remains high. This means peak spl capacity should be dimensioned equal for high frequencies, and for all speakers not utilizing compression drivers and horn this means high frequency amplifiers must have same peak power capacity as the amplifiers for lower frequencies.

In practical situations in small rooms (that is all our rooms) power handling will not be an issue even for dome tweeters, as they will be too loud for comfortable listening before they burn out. But sound quality will be compromised, as there is not enough peak spl capacity.

Now some people enjoy discomfort and screaming highs, so they will eventually burn out both their tweeters as well as their hearing.

 

[restorer-john]

Back of the envelope calculations ...

Tweeter voicecoil geometry assumptions:

Wire diameter = 38 gauge = 0.1 mm = 0.0001 m​

Coil diameter = 25 mm = 0.025 m​

Number of turns = 24​

Cross sectional area of voicecoil = 0.25 * π * 0.0001^2 = 7.854e-9 m^2
length of wire = π * 0.025 * 24 = 1.885 m
Volume of wire = 1.48e-8 m^3

Density of copper = 8940 kg/m^3
Specific heat capacity of copper = 385 J/(kg °C)

Therefore, the amount of energy to raise the voicecoil temperature by 1 °C, assuming no heat loss, is:

Volume of wire * Density of copper * specific heat capacity = 0.05 Joule/°C​

Melting temperature of copper is 1085 °C. To raise the temperature of this voicecoil by 1000 °C will require 0.05 * 1000 = 50 Joule. Thus, 50 W for 1 second (or 100 W for 1/2 second, etc.) will do it for this hypothetical tweeter.

None of this is useful whatsoever for determining the failure modes for typical tweeters when hit with transients.

It is not the voicecoil that fails in transient situtations, it is the lead-in wires and the interface to the voice coil. They fail like a flash bulb. The voice coils themselves are fine. I have seen (and repaired) literally hundreds like that. Usually the transients are not a clean transients that causes the damage, they are amplifiers momentarily bursting into oscillation for a number of cycles.

This was routinely tested for back in the day with amplifiers. Drive the amplifier to 50% rated power (-3dB) and then overload it with an assymetrical (diode clipped) toneburst 10dB higher, putting it 7dB into overload. See what happens in terms of overload recovery.

Tweeters, woofers and midranges can absorb huge single cycle or low cyle count clean transients from capable amplifiers without damage. An example from a review of my current main speakers:

The voice coil windings adjacent to one another and layered means they take a long time (relatively) to heat up to the point where the enamel/adhesive melts/moves/bubbles and/or becomes deformed enough to foul either the pole piece or outer.

Tweeters in typical 100-200 rated speakers are usually rated 5-10 watts. There is no way most domestic tweeters would survive even 5 watts continuous for very long without failure.