Feature request - please help save our tweeters!

[lovingdvd]

Matthew and all - just a quick line to let you know that John graciously added the feature request to 5.19 beta 9, released today, that I originally suggested at the beginning of this thread, to help protect tweeters during sweeps should the master volume inadvertently be set too high. I think this will be VERY helpful! Thanks John!!

 

[Matthew J Poes]

Glad to see this was addressed!

 

[dc2bluelight]

One factor to remember when considering tweeter power handling: while tweeters generally do not handle much more than 20W RMS, and while a crossover will adjust for tweeter efficiency so it matches the other drivers, nearly all program material has a spectral RMS energy distribution that puts energy at 20kHz between 20dB and 30dB below energy at 1kHz. That means if you hit peaks at say 200W mid band, program energy will be between 2W and .2W at 20kHz. An REW sweep changes that 20kHz to a level equal across the entire band. Nothing about the tweeter or crossover could save a tweeter with a full power full bandwidth chirp.

 

[John Mulcahy]

An REW sweep changes that 20kHz to a level equal across the entire band.

That's not correct. It would be right for a linear sweep, but REW uses a log sweep which has a pink spectrum, so the energy at HF is dropping at 10 dB/decade. The 10 kHz level is 30 dB below the 10 Hz level.Here is the energy content of a full range sweep:

 

[dc2bluelight]

Does this mean that feeding the REW sweep into a broadband RMS volt meter would show voltage dropping 10dB with every decade of frequency increase?

 

[John Mulcahy]

No, not at all, an rms meter has no indication of spectral content and the sweep has constant peak to peak amplitude. The energy reduction is a result of the increasingly rapid change of frequency as the sweep progresses, meaning it spends ever shorter time at any given frequency as frequency increases. A log sweep spends as much time traversing the octave from 10 to 20 Hz as it does on the octave from 10 kHz to 20 kHz.

 

[dc2bluelight]

That was my understanding as well. So perhaps I used the wrong terminology before. My point was that the sine sweep could apply unusually high RMS power to a tweeter as compared to program material.

 

[John Mulcahy]

It isn't really meaningful to talk about rms power handling for signals that are not continuous. The main concern is the ability of a driver to dissipate energy. The energy content of a log sweep at high frequencies is low, as the spectrum shows, and not dissimilar to program content. On a full range 256k log sweep at 48 kHz the entire period from 2.5 kHz to 24 kHz is less than a second.

 

[dc2bluelight]

Agreed, but wouldn't total heating (resulting in tweeter destruction) be a function of RMS energy, time and dissipation? So the more time the sweep spends above crossover (and assuming a flat tweeter impedance curve, which may not be right) the more heat would be applied to the tweeter. When time and RMS energy result in more heat than the tweeter can safely dissipate, out she goes.

Considering the issue of why the OP toasted tweeters with an REW sweep, but not program... With dissipation held constant, either RMS or time or both had to change. I don't see in his data that he used a 256k log sweep, it might have been longer, and level seems undocumented other than an AVR volume setting, so we don't know. But for a given RMS level and a log 10-24kHz sweep, wouldn't tweeter heating be a function how much time that sweep spend above the crossover where the energy would be applied to the tweeter, along with the specific level of RMS power applied?

Yes, I understand RMS is difficult for signals that vary over time, but depending on the thermal time constant of the device being heated (and tweeters are actually fairly short), we might be able to integrate over that time constant, perhaps a bit longer, and get a feel for what kind of heating a given signal causes. For example, take the pink noise from REW as compared to the portion of the 10-24k 256k log sweep that lands above 2.5kHz (which lasts about .35 seconds), and analyze that for heating capability (RMS). We land at an arbitrary -12dBFS for the sweep because that's the level of the signal and it's amplitude doesn't change with frequency. So that's .35 seconds of -12dBFS RMS. Now take a sample of pink noise .35 seconds long, generated also at -12, but filtered at 18dB/octave below 2.5kHz to represent the portion of the signal applied to the tweeter. We get -17dBFS RMS. So -12dBFS pink noise applied to the speaker for the same time as the sweep in this example applies 5dB less RMS energy to the tweeter, and correspondingly less heating for the same amount of time. The sweep's spectral distribution may match pink, but I don't think tweeter heating is particularly spectrally sensitive, it just gets heated by any RMS energy applied with much less regard for frequency, up to some practical limit of course.

Now we would have to consider the dissipation of the tweeter and its thermal time constant, we could then generate 3 axis graph showing time vs level vs heating and perhaps even superimpose a fatal threshold for some given tweeter. I wouldn't bother to attempt generate the graph, because it would be very specific to a given tweeter and full data may not even be available. but clearly the combination of level and time (as applied with respect to the tweeter's thermal time constant) would affect where along the curve any signal would land with respect to tweeter fatality. And the REW sweep signal likely contained more than one chirp, with less than a second of cooling time between high frequency heating energy applications, possibly pushing up higher on the thermal time constant toward the destruction point. Of course, we don't have the actual time constant, so I admit to this being conjecture.

Music, particularly above 2.5kHz, is less constant resulting in far less RMS energy over a typical tweeter thermal time constant interval than a continuous sine wave, swept or not, for the same interval, likely less even than pink noise. My theory here is that the extra energy from a sine wave constant amplitude sweep, with the possible Audyssey 8dB boost as a factor, is what cased his tweeters to blow with the high power sweep when program material didn't blow them.

I also recognize there isn't enough data here to be sure. We don't know specific power levels, we don't know what overall gain offset Audyssey put in place to meet reference cal, and we don't even know what sweep rate was used for certain, though likely the default. And .35 seconds does seem to short to blow a tweeter, but it certainly did. All I'm trying to tune into here is the delta between the sweep energy and program or pink. Seems like there may be quite a bit more in the sweep than we think. And then apply significant power, clearly too much, with burn out as the result.

And yes, I'm overthinking all of this.