[DrWho]

If you want your graph to look like WinISD, then you'll need to take your sub outside to do the measurements since that's the environment that REW is predicting in. You might also see if your sub's frequency response doesn't change with output level outdoors as well....that will allow you to see what is room gain and what is the driver.

Btw, are these plots smoothed, or EQ'd? Can you post the raw responses?

 

[DrWho]

Yea, outdoor ground planes are how all the speaker companies achieve "anechoic" measurements at lower frequencies.

For what it's worth, I'm not seeing any of your spikes changing with output level. Instead, what I think you're seeing is non-linear room gain. Basically, as the lower frequencies get louder and mask the fact that you've still got a ~45Hz spike.

Could you provide more information about your setup? Like room size and the subwoofer you're measuring and where the microphone is located?

The 45Hz stuff looks like it might be two different frequencies and/or some of the polar response effects of having a subwoofer surrounded by walls. Looks like your sub is about 2ft from the walls? (or your microphone is). I would expect to see about a 6dB rise in response starting at around 60-70Hz....kinda like a shelf-filter until the driver response starts to naturally roll-off.

Are you opposed to plotting your WinISD predictions too? (preferably with the same scale).

 

[DrWho]

Are you setting your "microphone" on the couch when taking the measurement? And do you get the same results if you feed your signal directly to the sub amp instead of through your receiver?

If you use the all-measured tab, you could provide 10 measurements at 10 different SPL's...maybe start at 70dB and move up in 3dB steps? I think this will show the trend with SPL....I'm wondering if it's not voice coil temp rise / room gain / power compression? There are examples of this performed by Ilkka in the subwoofer measurement forum (it's down on the list somewhere).

Btw, are you more curious about the source of the behavior you're seeing, or are you trying to tweak things to sound better? or both?

Also, you can model VC temp rise in WinISD....basically, the resistance of the VC rises and causes a voltage-divider against the non-linear impedance response of the speaker.

 

[DrWho]

Well at first I was going to mention that the Bl isn't linear with excursion and that is a common source of compression...basically, as the voice coil starts to leave the magnetic gap, the amount of force it can generate goes down and thus the cone doesn't move as far. The suspension can also get stiffer as the excursion increases too.

However, if this was the problem, then you would see everything compressing below 40Hz especially in a sealed box because the excursion goes up with lower frequencies.

Instead, I think you might be clipping your SPL meter....or something in the measurement chain (and REW is probably not noticing). It looks like the compression starts happening near the same SPL at other frequencies once they get that high.

 

[DrWho]

Ah good, I'm glad to see things are starting to make sense :T

Btw, I was looking at your plots earlier....I bet you can get that 70Hz dip to move around by changing the height of your meter.

 

[DrWho]

In regards to 60Hz, check for boundaries that are about 4.7ft within the mic or the speaker...
http://www.peavey.com/support/technotes/soundsystems/boundarycancellation.cfm

You might try some outdoor measurements just to see what this looks like so that you can identify it in your room.

 

[DrWho]

Well you're getting proximity gain on the microphone by putting it right up next to the cone (in other words, changing the frequency response of the mic). I dunno how big of a deal it is with the ratshack meters. You'll also be getting some effects of room gain as well. Really, the only way to know is to do some outdoor measurements

Btw, the WinISD predictions are very accurate in terms of power response in half-space (middle of a field). Sometimes the shape of the cabinet and the arrangement of the driver(s) can affect the on-axis frequency response. Adding other boundaries (like putting the sub into a corner) will change the on-axis frequency response too. But in all these cases, the power response is essentially the same....you're just redirecting where it goes (and hopefully keeping it in phase).

If your subwoofer is "small" (relative to the wavelengths), then the polar response will be nearly omnidirectional - in which case, the power response and on-axis frequency response will have the same shape.

All that to say, you can take your WinISD prediction and then find a transfer function that matches your in-room measurement. I should confirm this before mentioning it, but room gain is typically about 12dB/octave, but looks more like a shelf filter because eventually the walls start getting more transparent as you go lower in frequency. But knowing the shape that room gain should be helps you in determining the transfer function that matches the difference between your measurement and the WinISD model. When you get it right, this transfer function can be applied to alternate designs and yield predictable results.

The nice thing about sealed cabinets is they also roll off at 12dB/octave....so if you can find a driver that has an F3 with a Q of 0.707 where your room gain kicks in, then you should expect to measure a frequency response that extends down to DC - but it'll roll off earlier at the point where the room starts getting more leaky.

That said, I disagree with the notion that "flat after room gain" is the goal, but that's getting into the realm of psychoacoustics instead of the realm of understanding why things measure the way they do. Knowing why things behave the way they do makes it possible to design for a target dictated by psychoacoustics. Anyways, I might suggest that we can perceive room gain without there being any music playing and therefore we establish our perceived "zero" as starting with the room's natural "frequency response". My current position is to be flat after boundary gain, but before room gain...which is kinda hard to quantify without getting extreme with the measurements. I've actually come across a recent article that talks along those lines too...I wonder where I put that one.

 

[DrWho]

I'm an electrical engineer myself and usually try to suppress the engineering topics because it ends up being more involved than people usually want to get (nothing wrong with that either)...so I enjoy an opportunity to talk engineering speak

Do you tune the simulation with equalization to represent the room's response?

Not sure what you're referring to here???

Also so can you explain the difference between boundary gain and room gain? Thanks.

Here's how I understand it:

Measure the subwoofer in the middle of a field - this will be half-space. Now measure the subwoofer sitting beside a really big building (basically a big wall in the middle of a field). This gets you a 1/4 space measurement and ideally you'll see +3dB at every frequency relative to the 1/2 space measurement. Then measure with a big corner out in the middle of a field....so a 1/8 space measurement. You'll ideally see +3dB at every frequency compared to the 1/4 space or +6dB from the 1/2 space.

What's happening is the energy that would normally be travelling rearward is reflecting off the wall and being redirected forward. Therefore, the intensity (amount of energy per area) increases in front of the speaker. The intensity at a single point in space is what is being shown in a normal frequency response plot. If you integrate the intensity over all space, then you end up with the power response (the total energy being delivered by the speaker at each frequency).

The +3dB from halving the space the driver is firing into is true when you assume that the speaker has an omnidirectional polar pattern and that the reflections take zero time to occur. So the only time this is true is for a point source located on the vertex of the boundaries.

In the real world, the driver has a physical size and the cabinet requires that the driver sit away from the boundaries. This prevents the existence of a perfectly omni-directional polar pattern and the reflections are delayed by a small amount. This seems pretty simple, but I've always been surprised by just how different an outdoor 1/8 space looks from an outdoor 1/2 space measurement. I wish I had some electronic copies to share as it is really quite surprising. Anyways, the point is that the difference in behavior is not at all trivial (we're talking like +-3dB at least).

When you put the sub in the corner of your room, you're always going to see the same 1/8 space behavior, but it gets a bit more complicated because you have a ceiling, and extra walls to deal with. Some of the reflection paths will be long, and some of the reflection paths will be short. When they're short, our ears perceive it as part of the direct sound and so we hear the dips/peaks that it introduces. When they're long, our ears perceive it as part of the natural sound of the room - so the dips/peaks aren't necessarily perceived.

So is the ceiling a part of the boundary or room response? Well I think it will depend on the frequency in question. I think the real question should be, "is the delay long enough to be perceived as part of the natural sound of the room, or is the delay short enough to be perceived as part of the direct sound from the driver?" I think answering that question in context of a specific situation should likely reveal the best path to a solution.

And then another behavior to throw out there...

When you have reflections in the room, there is a good chance that many of those reflections will find their way back to the driver itself. Increasing the air pressure in front of the driver will improve the coupling of the driver. Likewise, decreasing the pressure in front of the driver will reduce its coupling. Horns basically work on the principal of providing more pressure in front of the driver so that more power gets transferred. In fact, it is this behavior that creates the 12dB/octave room gain that we observe - basically, the wavelengths are long enough such that the reflections are in phase with the driver - thereby increasing the amplitude in front of the driver and increasing its coupling. As you go lower, you get more and more in phase (since the phase rotates slower and the time of the reflections are fixed). Things can get real complicated real fast when you consider all the reflection paths that might improve or reduce the coupling of the driver. So in a case of a horn, I'd call those "reflections" as happening early and therefore being perceived as part of the direct sound. In the case of the 12dB/octave room gain of the room, that behavior is going to be delayed by quite a bit since the sound has gotta slosh around the room for a bit to capture the full effect. We also naturally hear the power response of the room with no music playing either, which just reinforces our perception of it being more of a "reverb" than a "direct sound".

All that said, I don't think there is any way to easily determine what would constitute an ideal target response for a given measurement in a particular location. I might suggest that it's easier to think of it in terms of "boundary gain" vs "room gain", but in reality it's going to be a continuum of both - and the fact that it is a continuum makes it real hard to draw lines. But really, it may not be necessary since there is only so much we can, or are willing, to do....in which case it is important to be aware of how we perceive the sound so that we don't make unintentional sacrifices.