Subwoofer Tests - Fall 2007

[Chris Brunhaver]

WillyD,

Actually, that is purely coincidental. Unibox doesn't have a significant amount of nonlinear modeling capability and even if it did you'd need more information than just small signal parameters and a simplified excursion model.

My guess is that TC-2K does quite well down low due to a linear suspension and pretty good motor and not as good up high due to high inductance and big coil (i.e. modulation distortion and eddy currents in the steel etc.). I don't see how Unibox could be predicting this.

-Chris

[DLS]

Chris Brunhaver wrote:

I don't see how Unibox could be predicting this.

Actualy even far more sophisticated softwares can't predict it. Aside from Linearx's LEAP 5 wich seems to be the best modeling product, I don't see any other soft that can actualy ''see'' or predict nonlinear behavior/compresion at high power levels. TS data of course would be absolutly insufficient for that matter.

[WillyD]

I never said it wasn't coincidental. I was just making an observation. Of course I know that Unibox or any other simulation program can't accurate predict distortion limited output...who do you take me for?

I don't see how Unibox could be predicting this.

Unibox actually does have the ability to use a couple non-linear parameters, such as L2 and R2, which allow an "improved model of voice coil inductance simulating Eddy currents". So of course Unibox can help to predict how the inductance affects the output in the higher frequencies.

[Ilkka]

WillyD wrote:

Unibox actually does have the ability to use a couple non-linear parameters, such as L2 and R2, which allow an "improved model of voice coil inductance simulating Eddy currents". So of course Unibox can help to predict how the inductance affects the output in the higher frequencies.

L2 and R2 aren't non-linear parameters per se.

[Ilkka]

Chris Brunhaver wrote:

Hi Ilkka,

You didn't happen to measure the amplifier output voltage during the distortion test sweeps, did you? While you are already doing a massive amount of work, this would help settle the debate about how much power is required to really push these DIY sub drivers tot their CEA-2010 distortion limits etc.

Unfortunately not. That would require more sophisticated hardware than just a simple DMM due to burst signals.

[Ilkka]

SteveCallas wrote:

For sweeps, I would definitely agree. For 6 cycle bursts like were used (I believe?) for the CEA testing, I'm not so sure. Ilkka - was power the limiting factor for the TC2k CEA test above 25hz, or was it distortion?

For both TC2k and SDX15 it was power (but only at and above 50 Hz actually).

From sweeps or from bursts?

Both but it's more obvious when looking at the THD sweeps.

Isn't burst testing more representative of real world material reproduction?

Yes.

[Ilkka]

SteveCallas wrote:

I know the amp clipped during the sweep testing, but I don't believe it clipped during the burst testing. If that's the case, then more power wouldn't necessarily add much to the burst test.

Both the T.Amp TA2400 and the Crown CE4000 clipped during the burst testing at 50 Hz and above. I wasn't able to clip the CE4000 during the sweeps though.

[Ilkka]

Guys,

I'm not any more following some of the reasoning done by a few of you. That's maybe because I have seen all the data and you haven't.

But when it comes to TC2k/SDX15/LMS-5400 distortion, it's already posted here. Nothing actually changed during the GP/high power measurements. You can see the "pattern" already there.

The TC2k is more cleaner, has less compression and has more output than the SDX15, at every frequency (assuming same amp is used of course).

Now can you please stop fighting until you see the rest of the data?

[WillyD]

Ilkka wrote:

L2 and R2 aren't non-linear parameters per se.

I read this in a paper by Klippel.

To investigate the effect of each nonlinearity separately and to find
the dominant source of distortion the following nonlinearities
might be switched on and off during simulation:
• motor nonlinearity due to Bl(x)
• mechanical suspension nonlinearity due to Cms(x)
• inductance nonlinearity due to Le(x)
• para-inductance nonlinearity due to L2(x)
• losses from eddy currents due to R2(x)
• reluctance force (electromagnetic drive)
• adiabatic compression in enclosure CAB(Pbox)
• adiabatic compression of rear enclosure CR(Prear)
• radiation distortion (Doppler effect)

[WillyD]

Now can you please stop fighting until you see the rest of the data?

Not fighting, just discussing. I'll stop though. You've told me what I wanted to know anyways.

[SteveCallas]

Willy wrote:

Why don't you believe it did. Just have a lot of faith I guess?

Lol, no, I didn't mean "I don't believe" as in "that's not true", I meant it as in "if I'm not mistaken" or "I could be wrong but".

Ilkka wrote:

For both TC2k and SDX15 it was power (but only at and above 50 Hz actually).

Both the T.Amp TA2400 and the Crown CE4000 clipped during the burst testing at 50 Hz and above

Wow, I never would have expected the Crown to clip based on all the fuss made about it - at least not before the drivers got fried, even on a burst test. Good stuff.

Ilkka wrote:

The TC2k is more cleaner, has less compression and has more output than the SDX15, at every frequency (assuming same amp is used of course).

Now can you please stop fighting until you see the rest of the data?

I apologize Ilkka, in all honesty there was no ill intent behind this discourse, just trying to make heads and tails from some of this initial data. We've been tempted with the goods but won't see the full results for some time, so we gotta make due. TC2k is more capable at every frequency - got it.

[DevilDriver]

I'm not familiar with the non-linear calculations performed by Unibox, perse, but the addition of L2 and R2 does supply a certain unsophisticated non-linear prediction model. If you take your standard LR-2 model and identify that Le, Le2, and Re2 are functions of displacement x, then you can approximate values using a truncated power series expansion. I don't believe, however, that this model scales well with frequency, which is unfortunate because frequency is a very key component of inductance.

It's a shame that the Klippel analyzer does not support such measurement, but the model it does use does give good insight to inductance variation with coil position.

edit: I am extraordinarily surprised by the performance of the TC 2K. From posts in the other thread, it seems that distortion measurements at virtually every harmonic are, more or less, identical to the SDX15. Price is another story, but the performance is impressive.

[thxgoon]

ISLAND1000 wrote:

It's revealing that four out of the top five subs use TC Sounds drivers, that the top six subs are DIYers, and eight out of the top ten subs are DIYers.
This list contains some of the best commercial subs available and the DIY subs outperfomed them in these tests. We are at a point where DIY knowledge, available software modeling programs, and cooperation on these forums enable the hobbyest to make for themselves a sub that can outperform commercial units and probably do it for less money per performance dollar.
Congratulations all around.

Agreed!!!

Not that anyone else needs to say this but wow! Good job! These are great tests that just cannont be found anywhere else. Totally unbiased and honest raw data. If only the entire world of AV enthusiasts could see this... thanks!!!

[cjwhitehouse]

Isn't burst testing more representative of real world material reproduction?

Ilkka wrote:

Yes.

This is a very definite statement. Are there are scientific papers around to back it up?

While you might be able to make a case for this from music signals, I maintain that typical heavy HT material can contain bass energy that lasts for a lot longer than 6.5 cycles. What about the machines emerging scene from WOTW?

One of the factors affecting subwoofer performance is how it responds to continuous loading... voice-coil heating, power supply depletion etc. The amplifier is one of the most critical components in a powered subwoofer. When measuring amplifier power outputs, the hifi industry long ago learnt to eschew the use of "peak music power output" figures and adopted a more realistic "1kHz continuous RMS power" measure instead. Why are we now saying that a continuous test is less relevant to a subwoofer's performance?

As I see it, the CEA-2010 clean output limits are fine for giving you a relative scale by which to compare different subwoofers in a consistent manner. But it's just a number, derived from a wholly artificial signal. How does it relate to what you could expect a given subwoofer to deliver on real program material?

Actual program material is somewhere between tone-burst and continous sine-sweep. I would say that music is probably nearer the tone-burst end and HT material rather closer to the continuous end but this is debatable. You takes your pick as to which you think gives you a better handle on real-world program performance.

[jakeman]

Excellent post CJ. I've thought for some time now that the burst testing advocated by the CEA standard has become outdated with current demands on sub performance. How a sub responds to continuous loading is one of the defining tests which identifies flaws in build and design. Many of the torture test discs I use in listening sessions are ones which contain intense long cycle bass for detecting audible signs of stress. Loud deep sustained organ music as well as WOTW have been known to bring out the worst in a few subs which otherwise have acceptable looking graphs. I'd suggest its not a fine line between tone bursts for music and continuous sine waves for HT. Testing for both conditions would be ideal.

Peak long demands for energy stress all the system components and are much more revealing of how well the amplifier and driver can handle sustained loads in real world listening environments. As the sub stresses under such demands the ability of the sub components to dissipitate heat not only causes compression but audible swings in FR and distortion. The remedy one reads often is to add more subs but that just skirts the real problem IMO.

The downside to this kind of testing is greater risk of damaging the DUT. On the other hand if a sub cannot handle near reference levels of WOTW type bass, that is invaluable information indeed.

[Ilkka]

WillyD wrote:

I read this in a paper by Klippel.

Sorry, I meant that they are not non-linear parameters in Unibox. Meaning that Unibox doesn't do any non-linear modeling. If you add input power, all simulated variables stay linear. So the effect of L2 and R2 is only valid at low power levels.

[ISLAND1000]

It seems these deep honest discussions regarding sub testing needs a final resolution.
If the CEA-2010 standard is not a true representation of sub duty because it s too light weight and
continuous sine-sweeps is too strenuous what then would be the "perfect" subwoofer test? Who will set the standard?
I think this forum should be the responsible party. We have our scientists, engineers, builders, tinkerers, and DIYers. What better diverse group of enthusiasts exists?

For my first suggestion for an improved standard I propose:

PINK NOISE 10Hz to 100Hz @ 105 DB for 60 seconds or until compression or stress shows.

Next . . . . . ?

Discussion . . . .?

Oh, I know . . . . . we need a new thread for this . . . . . Sonnie?

[Ilkka]

SteveCallas wrote:

Wow, I never would have expected the Crown to clip based on all the fuss made about it - at least not before the drivers got fried, even on a burst test. Good stuff.

It has nothing to do with CE4000 being bad or anything like that. It's just that these drivers can take HUGE amounts of power at those frequencies when using short/burst signal. I'm assuming you are not familiar with car audio SPL contests? There it is quite the norm to use several tens of thousand of watts per driver. And they are using even longer signal than 6.5 cycle burst.

There is an old test made by Richard Clark where he tested an AA Brahma and a JL Audio 13W7. They both could take 32 kW at 80 Hz when using burst signal.

The peak spec for the LMS-5400 is 8000 watts, but I'm guessing that's way way underrated. It should be able to take 40-50 kW (above ~65-70 Hz) or more in a small sealed enclosure when using burst signal (when distortion is not limited naturally).

I apologize Ilkka, in all honesty there was no ill intent behind this discourse, just trying to make heads and tails from some of this initial data. We've been tempted with the goods but won't see the full results for some time, so we gotta make due.

I understand it but making wrong conclusions is often worse than no conclusions at all.

[Ilkka]

cjwhitehouse wrote:

This is a very definite statement. Are there are scientific papers around to back it up?

While you might be able to make a case for this from music signals, I maintain that typical heavy HT material can contain bass energy that lasts for a lot longer than 6.5 cycles. What about the machines emerging scene from WOTW?

One of the factors affecting subwoofer performance is how it responds to continuous loading... voice-coil heating, power supply depletion etc. The amplifier is one of the most critical components in a powered subwoofer. When measuring amplifier power outputs, the hifi industry long ago learnt to eschew the use of "peak music power output" figures and adopted a more realistic "1kHz continuous RMS power" measure instead. Why are we now saying that a continuous test is less relevant to a subwoofer's performance?

As I see it, the CEA-2010 clean output limits are fine for giving you a relative scale by which to compare different subwoofers in a consistent manner. But it's just a number, derived from a wholly artificial signal. How does it relate to what you could expect a given subwoofer to deliver on real program material?

Actual program material is somewhere between tone-burst and continous sine-sweep. I would say that music is probably nearer the tone-burst end and HT material rather closer to the continuous end but this is debatable. You takes your pick as to which you think gives you a better handle on real-world program performance.

Hi Roger,

Sorry for being so vague, I was just making some quick replies last night.

I'm not familiar with any scientific papers around this subject. Dr. Floyd Keele's AES papers regarding his 6.5 cycle tone burst might have some, but unfortunately I have no access to them.

Of course there is music and HT material that lasts longer than 6.5 cycles. No one has ever claimed otherwise. But it is a whole different case to say that the max SPL tests performed using this short signal represent better what happens with real program material. The majority of the most demanding HT material nowadays contains rumble similar to narrow-band pink noise, and shorter thumps/explosions/hits/bursts alone or more usually on top of this rumble. Signals similar to straight sine waves are VERY rare. Especially the ones that last longer than one second or more. I can not think of more than a couple movie sound tracks that contain this kind of signals. Music actually has more of them; organ and some modern synthesized type of kind comes to mind.

So IMO it's fair to say that these burst SPL tests represent better the real world situation. Also if you measure the sound pressure levels with an SPL meter in your room during a movie scene, it's these short thumps and explosions that create the highest numbers, not this constant rumble or straight sine waves.

Notwithstanding the above, I don't try to "sell" the CEA-2010 as a perfect and only solution when it comes to subwoofer performance testing, as the original purpose of this standard probably was. I have always taken and will always take longer sine sweeps additionally to these burst tests. They provide a lot more information about compression and power handling then the CEA-2010. But the sine sweep method isn't good nor safe way to measure maximum SPL capability, especially at higher frequencies.

As you can see from majority of my own and Roger's measurements, the 5 dB sweep level increment used by both of us isn't small enough to show the absolute maximum SPL at all frequencies. Usually it's pretty useful at low-end due to higher compression but up higher the highest sweep doesn't usually compress that much and the true max SPL remains as a mystery. Using a smaller sweep level increment doesn't help either because then it becomes very unsafe to squeeze that last bit of output at those critical frequencies (low impedance means high current through the VC, but not much cooling due to low excursion).

This isn't a problem with CEA-2010's burst signals and individual test frequencies. I can safely and accurately measure the maximum clean (remember the stepped distortion threshold) SPL at each 1/3 octave wide test frequency, which basically covers the whole passband of the subwoofer. One number per each test frequency covers two variables. With regular sine sweep one has to look at two graphs: SPL and distortion.

IMO the CEA-2010 is a valuable addition to my battery of tests. It doesn't tell the full story but helps to understand how the subwoofer reacts to shorter, more program material type of, signals. It is really interesting to compare the maximum sound pressure levels/compression measured with these two different methods/test signals. One more variable to keep on eye on. Currently it is also the only safe and accurate method for measuring the absolute maximum SPL at every frequency in the subwoofer passband. Of course it doesn't bother me that also (additionally to Floyd Toole and Don Keele) two of the most highly regarded and noted subwoofer reviewers, Keith Yates and Tom Nousaine, use this same or very similar burst test as a part of their subwoofer reviews.

[Ilkka]

ISLAND1000 wrote:

For my first suggestion for an improved standard I propose:

PINK NOISE 10Hz to 100Hz @ 105 DB for 60 seconds or until compression or stress shows.

So are you trying to find at what level and after what period of time the subwoofer breaks? Please see my explanation above.

[SteveCallas]

Ilkka wrote:

It has nothing to do with CE4000 being bad or anything like that. It's just that these drivers can take HUGE amounts of power at those frequencies when using short/burst signal. I'm assuming you are not familiar with car audio SPL contests? There it is quite the norm to use several tens of thousand of watts per driver. And they are using even longer signal than 6.5 cycle burst.

But at the same time I read cases of frying voice coils with a few thousand watts. I believe it was Brian Bunge's Tumult that practically blew up with less than 4000 watts. I guess other issues were at play.

cj wrote:

While you might be able to make a case for this from music signals, I maintain that typical heavy HT material can contain bass energy that