Mini Statement Measurements

Sorry to draw this out I just find this all very interesting. So when I was running my first sweeps (weeks ago) I thought I had a polarity issue but the frequency response didn't seem to change very much when I would switch the polarity between the woofers and mid. I used REW (weeks ago) impulse to ensure that my polarity was correct, and I thought I had it right. I was just looking at my impulse graphs and it seems to be wrong again, I think.

Note: The two woofers should be positive polarity and the mid and tweeter are negative polarity.

The graph below go: left_woofer, right_woofer, left_mid-tweet, right_mid-tweet.

These make me believe that my left speaker has the woofer and mid-tweet on the same polarity where my right speaker has them opposite. Note, I am using the loopback connection during all these runs, although I don't think that a positive spike means positive polarity and vice versa. I don't think it matters as long as relative it is correct.

Do the woofer impulses look odd because there are two of them? Is there a way to determine the polarity of the woofers relative to each other based on this?

Wayne A. Pflughaupt said:

Seems like a glaring design defect, since the floor will be the same distance from the drivers in every room! You might try moving the speakers to other locations, even if it’s not a location you could actually use, just for the sake of experimenting to see if it’s being caused by proximity to other boundaries.

Regards,
Wayne

Click to expand...

Simple physics Wayne.
As I'm sure you already know, the floor bounce dip is caused by destructive interference. That is, the reflected sound from the floor has a longer path length than the direct sound and does not sum in phase relative to the direct sound. At certain frequency multiples the reflected sound will arrive at the listening position 1/2 wavelength behind the direct sound and cancel the the direct signal out. If you have found a method to EQ out a null, I'd sure like to hear about it...

Pete:
As I suggested to you in an email, a ground plane measurement of the speaker will eliminate the floor bounce as a potential cause for the dip, as the direct and floor 'bounce' will have the same path length. From what I've been able to determine, I can see no issues with the crossover construction, and I think we've got the speaker polarity thing gone through well enough. I suppose the next thing to do is to reverse the amps driving the woofers and mid tweeters to see if the 6 dB difference in response is caused by the amplification, or perhaps some setting in the receiver that is awry? Are you using separate amps or is this being driven by a HT receiver?

C

Wayne,

I'm most certain its a me/my room issue as these speakers have been well proven out. Wouldn't we have seen a large difference between the lp position and the 1m position if floor bounce was at play?

Curt,

I think the best I can do for a ground plane measurement would be to put the speaker much higher off the ground, that would certainly show a difference if the ground bounce was at play here. I thought I had the polarity thing down but the impulse graphs seem to show differently. That's a good call about switching the amps, this is a Pioneer SC05 receiver, I've never seen the issue with the old speakers. I'm thinking there shouldn't be any issue with switching the amps around right? The bass is a little heavy with these but I usually run it hot when I have my sub at play (although this is taking it to a higher frequency). I will play around with it more and get back to you guys. I'm just confused cause as Curt said we went through the crossovers, unless a value of one of the components is way off. It's gotta be something simple as we see the issue with both speakers.

Thanks,
Pete

Curt C said:

As I'm sure you already know, the floor bounce dip is caused by destructive interference. That is, the reflected sound from the floor has a longer path length than the direct sound and does not sum in phase relative to the direct sound. At certain frequency multiples the reflected sound will arrive at the listening position 1/2 wavelength behind the direct sound and cancel the the direct signal out. If you have found a method to EQ out a null, I'd sure like to hear about it...

Click to expand...

A bit off topic for this thread, but I find there is a tendency for any dip in a response to be labelled "a null", often accompanied by the "you can't EQ a null" comment. A null is what is says, zero, nothing, nada. It happens if, for example, a reflection has the same level as the direct sound - something that is quite rare in practice, since any reflection will have some reduction in level through absorption at the surface and diffusion from it. There is certainly no EQ solution for a true null, since whatever you multiply zero by you will still end up with zero. However, a dip in a response, such as those broad depressions in the posted plots, could certainly be EQ'd. Whether that is advisable would depend on the available headroom and whether the responses in other listening locations show the same characteristic and so would benefit from the same alteration. The difficulty with trying to EQ floor bounce is more usually that the location of the dip depends on the relative path lengths and so the listening distance, so any EQ could only be optimal at one listening distance. The location of the dip should move at different measuring distances, so if it stayed stubbornly the same there would seem to be more than floor bounce at work.

Pete, the main issue would appear to be the dramatic shelf in the mid-tweeter response below 500Hz, more than 10dB down, as shown in this plot:

The woofer response has no hope of filling that gap, does a near field measurement of the mid show the same characteristic?

Guys,

Appreciate all the responses, here are some more measurements that were taken per your requests.

Curt,

I flipped the bi-amp leads, still saw the same results, also found out that the pioneer puts a high and low pass which makes flipping the leads a bad idea for music listening (worked fine to find out if the amps were putting out the same gain).

John,

Got a closeup (4-6") of the mid, lower woofer and upper woofer. When I did the mid I unhooked the woofers (so it was the mid-tweeter playing and vice versa).

Graphs go: Right Mid closeup (4-6"), Right Mid closeup vs Right 1m full range, Right bi-amped flipped 1m vs Right bi-amped normal 1m, Right closeup (4-6") lower woofer, Right closeup (4-6") upper woofer, Right both woofers, Right individual woofers vs 1m full range.

Looks like to me that the woofers are destructively interfering with each other.

Thoughts/Feelings?

Also,

Here is a graph of the mid closeup (4-6") vs mid at LP.

There definitely does seem to be a large difference here, not sure what the issue could be though.

JohnM said:

A bit off topic for this thread, but I find there is a tendency for any dip in a response to be labelled "a null", often accompanied by the "you can't EQ a null" comment. A null is what is says, zero, nothing, nada. It happens if, for example, a reflection has the same level as the direct sound - something that is quite rare in practice, since any reflection will have some reduction in level through absorption at the surface and diffusion from it. There is certainly no EQ solution for a true null, since whatever you multiply zero by you will still end up with zero. However, a dip in a response, such as those broad depressions in the posted plots, could certainly be EQ'd. Whether that is advisable would depend on the available headroom and whether the responses in other listening locations show the same characteristic and so would benefit from the same alteration. The difficulty with trying to EQ floor bounce is more usually that the location of the dip depends on the relative path lengths and so the listening distance, so any EQ could only be optimal at one listening distance. The location of the dip should move at different measuring distances, so if it stayed stubbornly the same there would seem to be more than floor bounce at work.

Click to expand...

Hi John, nice to make your acquaintance.
I agree the null analogy is a bit simplistic, but still essentially correct. As you state, the advisability of EQing it may be problematic due to the multiplicity of room resonances and their variation with listening position. Although I agree it is possible, and may even be desirable to EQ the dip if the same response is representative of the entire listening area, it is also possible that EQing out a 10 dB dip will likely result in a 10 dB peak a scant few feet away. Since peaks are subjectively more audible than dips, I generally recommend a cautious approach to active equalization.

Pete:
I agree with John that the response attenuation above the crossover ‘dip’ is of more concern to me than the actual dip itself. Perhaps you should verify the midrange and tweeter resistor values are in the correct places, as I could not see this from the pics. I also note in your farfield measurements almost the same dip appears in the mid response as it does in the woofer. Furthermore, the dip is not evident in your nearfield measurements. Both these observations suggest a room boundary reflection issue rather than a crossover or design issue. Are you measuring the speakers well out away from the walls, or in their intended listening positions?

C

Curt,

I haven't moved the speakers from their intended position throughout all these measurements, these things are heavvvy, ha. I was hoping that through close measurements and gating I could get most of the info I needed. I see what you are saying about the dip in the mid going away, that's why i posted that last graph showing the differences. Not sure what could be causing it, seems to be pretty broad band, ~1 octave. I have been using the right speaker for these latest measurements because it is further away from walls. The picture below shows my setup. The picture was take from the back of the LP seat, and off to the right of the right speaker is another couch. The thing that gets me is that both the left and right speakers are showing this large dip and yet the boundaries around them seem to be far different (left speaker close to a wall). When I took the crossovers out a couple weeks ago I checked the labeled values on the components and had a buddy double check them so I don't believe the components are in the wrong spot unless the measured values are way off (hard to believe since the left and right seem to be pretty close).

Do these impulse graphs of the lower woofer and upper woofer make you believe they are the same polarity or opposite?

Note: These were taken with the mic ~4-6" away from the lower then ~4-6" away from the upper with no mid-tweeter playing. The lower was playing while taking the upper measurement and vice versa.

Curt,

Forgot you had asked about a gated measurement in one of your e-mails. Here is the right speaker at the LP gated for 4ms before the first large peak (as seen in the impulse) to 5ms after that peak.

FYI after another battery check on the woofers, although the impulse graphs I posted earlier make me think they are out of phase, the battery check shows them in phase.

Appreciate all the help so far.

vettett15 said:

Do these impulse graphs of the lower woofer and upper woofer make you believe they are the same polarity or opposite?

Note: These were taken with the mic ~4-6" away from the lower then ~4-6" away from the upper with no mid-tweeter playing. The lower was playing while taking the upper measurement and vice versa.

Click to expand...

vettett15 said:

Curt,

Forgot you had asked about a gated measurement in one of your e-mails. Here is the right speaker at the LP gated for 4ms before the first large peak (as seen in the impulse) to 5ms after that peak.

FYI after another battery check on the woofers, although the impulse graphs I posted earlier make me think they are out of phase, the battery check shows them in phase.

Appreciate all the help so far.

Click to expand...

Pete,
I’d be concerned if they did look identical. Remember this is a 3.5 way design, and an extra 4.7 mH inductor is used on the lower woofer adding additional phase rotation. I don’t see any phase concerns…

With respect to your gated plot: The dip around 350 Hz is much less noticeable and more importantly, the mid/tweeter response no longer looks attenuated with respect to the woofer response. Response irregularities from adjacent boundaries often make it difficult to obtain ‘clean’ plots as the impulse is corrupted by the reflections. What is surprising is how much the room modal issues can affect measurements, yet the speakers will not sound compromised.

To make my measurements, I generally move the speaker in near the center of the room, and angle it so it is not orthogonal with the walls. I also throw a stack of eggcrate foam on the floor between the speaker and the mic to ameliorate the floor dip. Try this at a 1 meter distance to the mic. Put your start gate immediately prior to the impulse and your stop gate 5mS later.

Nearfield measurements are wonderfully free of boundary issues as the acoustic energy of the direct signal is many times larger than the reflected energy. Unfortunately they also do not show the diffraction effects of the baffle, and therefore are of limited use for accurate room measurements or crossover design.

Wayne A. Pflughaupt said:

Welcome to the Forum, Curt!

Understood, but theoretically all speakers should have the same problem – a dip in response like Pete’s, caused by floor bounce - but obviously they don’t. So either it’s typically compensated for in the design (especially with a tower speaker, where the distance between the drivers and the floor is a know factor that will obviously be the same in every room), or else Pete’s problem is either because of a deficient design, or else it’s being caused by something else.

Nulls are typically narrow and deep, not like what Pete’s seeing, which is more accurately a trough or depression, not a null. So Pete’s problem could probably be addressed to some extent by equalization.

Regards,
Wayne

Click to expand...

Hi Wayne! Thanks for the welcome!
Theoretically all speakers do have the same problem. One only needs to look a one of John Atkinson’s in room plots of a speaker response in Stereophile to see the boundary mode issue is prevalent in most traditional designs. –And this is after he averages a grid of 12 separate measurement positions to mitigate the effects of those very same room modes on the response.

In a passively crossed design we have precious few tools to address response dips, but some attention to details can minimize its effects: The floor dip frequency of each driver is easily determined by its given driver height above the floor and a given listening distance. In a 3 way, a common method is to use crossover frequency that equates to the mean dip frequency between the woofer and the mid, thus partially compensating for the issue by manipulation of the transfer function of the individual driver’s transition bands. In a 2 way, this option is obviously not available to us, and in the case of a bookshelf size speaker, even the height above the floor is not a given, denying any chance of compensation. When you consider the effects of the wall reflections will be equal to or greater than the floor bounce, and how arbitrary those distances will be, the issue seems hopeless. -Yet we’ve nonetheless managed to enjoy our speakers regardless of room position. How can this be? While I’m certainly no expert in this area, I presume there is some psychoacoustic signal processing done in our hearing centers that allow us to hear ‘through’ these response aberrations. -Of course, these boundary aberrations are best controlled by careful placement of the speakers in the room and room treatments. After all, it isn’t the speaker that is broken, it’s the room that’s causing the issue.

My LspCAD software contains a relatively simple room sim program. Based on the computation time, it likely only takes into account axial modes, but not tangential or oblique modes, although to its credit the relative attenuation of these modes generally make the axial modes predominant. The program is interesting in that you can select/deselect each boundary surface and see the effects each has on the summed response. Selecting more than one boundary makes those sharp nulls disappear, and the resultant summation of the 6 boundary modes make for a response that is both unsettling to look at and difficult to interpret. –And similar to what Pete has posted.

At this juncture I’d like to reiterate that we are basing our presumptions of floor dip issue on speculation. This is one possible explanation for Pete’s measurements, but certainly not the only possible explanation. As I’ve suggested before, measurements are a 3D slice of a 4D world, and these measurements themselves are fraught with numerous potential pitfalls in interpretation and execution: Last week I was running harmonic distortion sweeps on a group of tweeters and found they all exhibited high 3rd order distortion between 3K and 5K, but otherwise apparently measured normal. Disappointed in the tweeters, I went so far that evening as to spec out a different tweeter for the design. However, I did a retest the next day and in checking the equipment over, found the errant HD plots caused by a low battery in the mic preamp.

Sorry for the loquacious response. I’ll try to be more succinct from now on…


Edit:
John, my apologies. I just saw this is the REV forum, so we’re not very on topic, -but could be if it turns out to be a procedural issue with measurement. I did take a few minutes to look over the REV overview: It looks quite impressive. It appears full featured and robust with all the tools necessary. Kudos for giving away your intellectual property for the good of audiophiles everywhere!

C