Originally Posted by Eyleron
At some percentage of its power handling these will appear. (Still waiting to hear others confirm a generalization of where this occurs).
With speakers whose power handling is only 150 watts, I think you'll have distortion. (but I'd like to hear from others what they think the extent would be. Is the speaker-loafing distortion of 5% going to rise to 10%? 20? Will you still get the 95db peaks?)
I still didn't know how compression affects peaks, because most of the stuff I've read about compression was with regard to continuous or program signals.
I started a thread about this topic here
Essentially, what I've learned so far is that thermal compression doesn't occur when you stay within the speaker's peak power handling. There may be other forms of compression such as magnetic and physical at play, and I'm still trying to find this out.Measuring Peak Output and Distortion
This is more difficult to measure. I read that special sound meters are required to read these very fast peaks. I don't know if a typical < $200 meter or a measurement mic, hooked up to Real EQ Wizard, counts as being able to measure those transient peaks? What tone should we use? I couldn't get REW to put out a 500ms or 1 second burst tone. Linkwitz sells a CD of shaped burst tones. THX uses burst tones in its tests. CEA 2010 subwoofer tests uses such, I think?
It's hard to find info on this, either because it's hard to measure, it's not important (if you stay under max peak watts you're okay?), or a combination of both.Thermal Compression
With enough continual (or frequent peak) high power signal being fed to the speaker, the voice coil heats up (speakers take electricity and convert a small fraction to acoustic energy; the rest is dissipated as heat).
As the VC heats up, its resistance changes. The frequency response of the speaker is now out of balance, because eg. the woofer will be producing less sound with the same watts but more resistance. You won't be getting the same bass peak level dynamics, and the speaker will sound more "bright." This is distortion, because what comes out is not what went in.
In that first post I linked to a Google spreadsheet showing four speakers and their manufacturer-stated compression specs. With bass pink noise, after only 5 minutes, the speakers compress a fraction of a dB at only 1/10 continuous power rating. At half that rating, they compress 1-2dB. At full power rating, they compresses 3-4dB.
Peaks, though, are by definition not continual. The VC has a chance to cool down.Magnetic Compression
More typically in bass drivers, at high levels, stuff happens to the magnetic flux fields that make the power response nonlinear.
Also, the woofer can be driven so hard that it's out of the linear portion of the magnetic gap, and the field weakens.
I still don't know if that typically happens if you stay within the speaker's max peak power rating.Physical Causes of Compression
With enough power, the driver can bottom out.
Also, with enough excursion, the spider itself is exerting a resistant force that's no longer linear, so the output from an increase in power is not what's expected: distortion.How can you tell?
Again, there's distortion that sounds horrible and harsh, that we can easily discern, and we turn it down. This is more on continuous bass/drum sounds, and not necessarily the more infrequent peaks.
But if the peaks are just not hitting quite as loud...it's difficult to tell, except if one is cringing and wincing, it may be due to distortion on those peaks. If it's just softer, we may be missing that last icing of realism, and we would have to A/B test two speakers rapidly and tell the difference, but in judging our own speakers we may not realize it.
Also, just turning it down isn't always an option, because if the softest sounds are now beneath the noise floor of the room, then you've still lost dynamic range. Instead of losing it at the top, now you've lost it at the bottom.
MKTheater suggested the technique of playing passages of a soundtrack and using REW to measure the peaks, and I tried this yesterday with some good effect. I was testing the subwoofers, but one could do the same on main speakers, using A-weighting. It might help to know what the recorded level of a certain passage is, at what frequency, and in what channel, so that you could narrow it down better. But I don't know for sure at what point REW can pick up what short peaks at what frequency. It might be time/frequency/sample-rate dependent?Changes to List
For now, I've removed the columns in the spreadsheet that showed what dB output you'd get on peaks, but using only 1/8th the stated power (I had thought that the rule-of-thumb of where compression starts affected peaks as well as continuous signals).
As long as typically one can avoid magnetic and physical compression by staying within the stated peak power handling, that's good news, as one can just use the simple calculator to roughly judge what output one will get.
This does deserve more investigation, however.