Home Theater Forum and Systems banner

1 - 20 of 27 Posts

·
Banned
Joined
·
165 Posts
Discussion Starter #1
I have a question about PR break-in. I just put together my maelstrom sub with 2 opposed 2100g AE PR-18s. Using test tones (don't worry) I noticed that from about 12-14Hz (tuned to 16Hz) the box shakes side to side (box sitting on carpet). As I increase the volume it's fine until it some point where it begins to shake. Then looking at the PRs and sub the displacement is definitely not purely sinusoidal. Reducing the volume gets it back to normal but the point where it starts shaking isn't very consistent.

I assume this is just due to different resonant frequencies of the 2 PRs and will reduce one they break-in. Has anyone come across this issue before.

I'll be making a build thread soon with pictures
 

·
Registered
Joined
·
464 Posts
I have 2 lms's and 4 passives vertically opposed. 2 cabs. At the input I have been giving them, they do not budge or even vibrate--just the cones moving to the frequencies. I weighed each added mass to the hundred's of grams, so the masses are as close to matching as I could get them. I do not know about the PR's you have, but I would check the mass if I could and see if the mass is close to each other. Just a thought.

Good luck,

Robert
 

·
Banned
Joined
·
165 Posts
Discussion Starter #3
The PR's I have basically use 1" MDF as the diaphram and weight so I assume that they match pretty well. I can add some type of putty to the back as there is a circular tube for that but I guess I'd have to find a really accurate scale that can take a bit of weight and weight the whole PR including the frame.
I noticed that the thing that causes the shake is when they go into a visibly non-sinusoidal excursion mode at that small frequency range. And I'm sure it not the amp's clip limiter. I think only taking a short video of it will show it properly.
 

·
Registered
Joined
·
430 Posts
Is one of the PR's bottoming out before the other? Are either bottoming?

I was under the impression the AE PR's needed a minimal amount of break-in due to the very soft suspension that's used. I know my TD drivers from them needed no break-in at all according to John J. I ran them in anyway though.
 

·
Banned
Joined
·
165 Posts
Discussion Starter #7
update with measurements

The shaking effect is still present but it now occurs over a smaller frequency range and isn't as bad. I assume this is from the significant amount of break-in I've given the PRs.

I tried taking some nearfield measurements to see what I could figure out.
Here is a zoomed out view showing the shaking and pulsating nature of the PRs movement


And here is a zoomed in view. You can see how even to the eye this would not look like smooth sinusoidal movement.


I only had 1 microphone available at the moment but I will eventually try with both PR's mic'd at the same time. I swear I can feel some out of phase movement my resting my fingers on the PR surrounds

I also examined the signals frequency spectrum using matlab

They are normalized relative to the fundamental. For completeness, here are the correction factors that should be added to the graph to compensate for the microphone calibration: 6.3Hz +15dB, 12.5Hz +7dB, 20Hz +3dB

Note how the passive radiators have fractional harmonics. Instead of the normal 2nd, 3rd, 4th harmonics like the driver there were also 0.5, 1.5, 2.5, etc.

The box must be shaking side-to-side at 1/2 the frequency being played. This would cause modulation of output and generate additional peaks around each harmonic at +6.4 and -6.4Hz. This would also explain the out of phase movement I feel since the side-to-side shaking naturally modulates the PR's out of phase.
I verified this by using the spectrum analyzer in REW. At very low volume, before the box begins shaking, the extra fractional harmonics are not present for the PR NF. As I increase the volume it remains this way until the box begins shaking. The signals then grow out of the noise floor. Once it is shaking reducing the volume does not remove them until the volume gets very low.

I believe the effect may be exaggerated because it is sitting flat on carpet. Maybe it will disappear completely once it is on proper feet and a harder surface. I'm just really confused why this would be happening. The PR mass or parameters must be just slightly different to even initiate the shaking. I don't think its a fault of the box because it has double thick front and side baffles and decent internal bracing.
 

·
Registered
Joined
·
438 Posts
Interesting.

I wonder if maybe you didn't get one PR-2100 and one PR-1600? I believe the weight is controlled with the width of the MDF, so it should be rather easy to identify by comparing the two side by side. I'm curious if the change in tuning is around 6Hz? It's been a while since I modeled them so I'm not sure.

Do you have any means for taking an impedance measurement? That would be a good way to verify what is going on with the tuning points. Another thing you might try is removing one passive radiator, seal the hole, and then take a FR measurement. Then swap the PR's and measure again. The FR should line up if they're tuned the same.

Btw, that frequency spectrum looks very much like clipping. Do you know if your microphone is getting overloaded?

I also wonder what the CMS for those PR's looks like...I wonder if maybe you're pushing into a non-linear realm of the spider. Might wanna really crank it hard and loosen up them spiders as much as you can. It's my understanding that the spider dominates the behavior more than the surrounds.

And one last note....can you measure the frequency response of the input signal to the subwoofer? Or at least show a time domain response? I don't know what you're using to generate your sine waves, but the common look-up tables aren't very sine-like....just something to keep in mind.

Ok, one more thought...do you have any air leaks?
 

·
Banned
Joined
·
165 Posts
Discussion Starter #10
Thanks for giving this problem some thought.

I checked that both of them were 1" MDF (2100g) when I initially received them. Also, I attempted to measure their tuning frequency. I'll describe my method if you promise not to laugh. In free air I held a mic near them and hit them. I then filtered the recording to only show content below 10Hz and measure the frequency. This was before any breakin and they were pretty stiff. My method did show slightly different resonant frequencies. About 6 and 7.5Hz. They're spec'd for 3.65Hz.

I've been playing them pretty hard with test tones and extremely low-bass heavy music so I assume they're probably at least 75% broken in.

Do you have any means for taking an impedance measurement?
I have taken an impedance measurement. The tuning looks fine and the phase is nice and smooth. It can be seen in my build thread.
http://www.hometheatershack.com/forums/exodus-audio/18068-maelstrom-x-2x18-page-ranking-25in-cube-16hz-build-thread.html

Btw, that frequency spectrum looks very much like clipping. Do you know if your microphone is getting overloaded?
The mic was fine, the level was pretty low and also the mic preamp gain was set to a moderate level. Yes, the harmonics resemble clipping because they are so high relative to the fundamental. If you calculate the THD from those figures its on the order of 20%. But when you take into account the microphone calibration the fundamental has a larger amplitude so its actually closer to a couple % THD

And one last note....can you measure the frequency response of the input signal to the subwoofer? Or at least show a time domain response? I don't know what you're using to generate your sine waves, but the common look-up tables aren't very sine-like....just something to keep in mind.
I was using tone generator by tolvan. I find it to be really good and have taken numerous nearfield THD measurements using it and REW's spectrum analyzer. I also use it to run 10Hz sine bursts at night to break in the PR's.

Ok, one more thought...do you have any air leaks?
I know there are some minor air leaks. I currently am not using the rubber gaskets. I will be once I get the box painted. It leaks enough that pushing in both PR's moves the cone but then holding them in the cone returns to it's center point after about 2 seconds.

Maybe uneven air leaks between the two sides is contributing to the problem.
 

·
Registered
Joined
·
438 Posts
If the Fs of your PR's is actually closer to 6Hz, then perhaps at larger excursions it's getting just non-linear enough to trigger the resonance (much like tapping your finger on the driver) and then modulating the response with the audio signal.

I also wonder if the two PR's aren't beating with each other.

Btw, I totally dig all the measurements and proof of performance you've been doing. Great writeup in that other thread.
 

·
Banned
Joined
·
165 Posts
Discussion Starter #12 (Edited)
If the Fs of your PR's is actually closer to 6Hz, then perhaps at larger excursions it's getting just non-linear enough to trigger the resonance (much like tapping your finger on the driver) and then modulating the response with the audio signal.

I also wonder if the two PR's aren't beating with each other.
I like what you're thinking here. I did some more testing and have come up with interesting results. First, I moved the sub to another room and placed it near a corner. I then took measurements at different listening positions within the room instead of nearfield.

I am using a microphone calibrated to 10Hz, but I have extrapolated the calibration to lower frequencies by curve-fitting a first order HPF to it; the fit was very good. Also, I made the REW soundcard calibration of my entire signal chain - E-MU 0404, BFD, EP2500. By including this it gives true "speaker in room" measurements. These two factors were important because they help shape the graph into recognizable form by removing the two sharp rolloffs of the mic and signal chain at single-digit frequencies.

Here is the average plot with the mic and sound card calibration curves.


On a side note I was truly amazed when I saw this plot. There is tremendous room gain at about 35Hz because of the corner, but also the low frequency response down to 8Hz is even with the upper bass region.

Here are several measurements taken


Now the interesting parts. PR designs should have a dip in their output at the free-air resonant frequencies of the passive radiator. For mine this should be 3.65 Hz. But the measurements are consistently showing nulls at about 3.5 and 6.3Hz. Could it be that the two PR's have tunings that drastically different? They were ordered at the same time and the look and feel identical.

Also, I made sure that during the REW sweep the PR's excursion was larger than the box shake excursion (again it was sitting on carpet with no feet). So the shake was not able to cancel out the output. The shake is not that prevalent during sweeps. I can only get it when signals in the 12-14Hz region last more than about 1 second.

Seeing that there is a dip at 6Hz does make one think. As DrWho suggests, hypothetically, if one of the PR's has an fs at 6Hz this would mean that it naturally has larger excursions at that frequency, causing the box to shake at that frequency. This would then modulate the near-field output spectrum results in the previous posts and also explain the out of phase-ness I detected by touching both surrounds at the same time. Since the two PRs are linked by the enclosed air volume, this might explain why the spectrum of the two PRs is the same even though only one of them is having the problem. The test results done with 12.8Hz tone suggest that maybe even just the 2nd harmonic of this resonance is enough to excite it.

I won't be able to do any more testing until next weekend. But hopefully I'll work out a way to do accurate fs measurements of the PRs. Maybe something like mounting one PR in the box by itself then putting the mic inside the box and hitting the PR. This will give much better results than the last time I did something similar, but without the box. Plus they are more broken in now.
 

·
Registered
Joined
·
165 Posts
It will be interesting to see what is causing this shake. I am sure you will figure it out. These graphs are very impressive! Getting that level of output to 8Hz is amazing. Maybe this is the type of design I should use. Do you plan to EQ this sub?

Mike
 

·
Banned
Joined
·
165 Posts
Discussion Starter #14
It will be interesting to see what is causing this shake. I am sure you will figure it out. These graphs are very impressive! Getting that level of output to 8Hz is amazing. Maybe this is the type of design I should use. Do you plan to EQ this sub?

Mike
Thanks Mike. I have a BFD that I'll eventuallly use to eq.
 

·
Registered
Joined
·
2,431 Posts
SturmMD,

Did you take a nearfield FR of each PR and the driver. You should see a broad peak like a mountain profile for the PR's corresponding to the tuning and the driver should have a notch at the tuning. Is one of the PR's close to a wall or other boundary? Your impedance plot looked like you a tune in the 14-16hz range.

Ignore the stuff below about 7 or 8hz in your graphs. It is just artifacts from the huge boosts in the cal files. If that double peak is there it's a dead give away. You have to get truly massive output levels to get any sort of reading however compromised down there.
 

·
Banned
Joined
·
165 Posts
Discussion Starter #16
Thanks for the ideas and discussion. I'm sure we'll be able to solve this eventually.

SturmMD,
Did you take a nearfield FR of each PR and the driver. You should see a broad peak like a mountain profile for the PR's corresponding to the tuning and the driver should have a notch at the tuning. Is one of the PR's close to a wall or other boundary? Your impedance plot looked like you a tune in the 14-16hz range.
Yes, my NF measurements of each PR did show the broad tuning peak. I had also attempted to observe the null at the resonant frequency. For one PR I was able to get a null in the response but it took several measurements. For the other side I couldn't measure a null. I assumed it had to do with uneven room effects.

After trying that I remembered that it is the combined response that theory predicts the null at the resonant frequency. So I took additional measurements in a different room.

Ignore the stuff below about 7 or 8hz in your graphs. It is just artifacts from the huge boosts in the cal files. If that double peak is there it's a dead give away. You have to get truly massive output levels to get any sort of reading however compromised down there.
For the in-room measurments I didn't expect to be able to get any meaningful results at such low frequencies. Nevertheless, the dips were consistent over many sweeps and I tried 4 different listening positions, both high and low in the room. I also had the level set to about 90dB rather than 75dB (I offset the plots). I'm not trying to argue that the measurements are accurate, just that I believe those two measured dips aren't coincidental or just noise. Also, the room isn't large enough to support a 6.5Hz mode :)
 

·
Registered
Joined
·
2,431 Posts
Thanks for the ideas and discussion. I'm sure we'll be able to solve this eventually.


Yes, my NF measurements of each PR did show the broad tuning peak. I had also attempted to observe the null at the resonant frequency. For one PR I was able to get a null in the response but it took several measurements. For the other side I couldn't measure a null. I assumed it had to do with uneven room effects.

After trying that I remembered that it is the combined response that theory predicts the null at the resonant frequency. So I took additional measurements in a different room.



For the in-room measurments I didn't expect to be able to get any meaningful results at such low frequencies. Nevertheless, the dips were consistent over many sweeps and I tried 4 different listening positions, both high and low in the room. I also had the level set to about 90dB rather than 75dB (I offset the plots). I'm not trying to argue that the measurements are accurate, just that I believe those two measured dips aren't coincidental or just noise. Also, the room isn't large enough to support a 6.5Hz mode :)
You won't get a null in the response of a close miked PR. It will be in the active driver. Is this what you meant?

The fact that you get the same dips below 10hz no matter where in room is not coincidence. This should be a clue that points to it not being real output, noise or otherwise. It's the huge boosts in the cal files applied inside REW to compensate for the electronics and mics IMO. I've taken hundreds of sweeps in multiple rooms, houses, outdoors, in vehicles, with different subs in different alignments with different signal chains and I can show you a response that looks eerily similar to what you are showing below 10hz in many of those.

Are you using an ECM8000? I'm not saying that I'm right and you are wrong, just that I've looked at this stuff a lot and that I'd disregard the stuff below approximately 6.5hz in your graphs shown above personally. In my testing if you raise the output level you'll get better accuracy in the ULF range and it can be seen as the response changes outdoors. Again just my experience.


Back to the PR's. Can you post a low volume 0-200hz close mic (2" or so from the dust cap) of each PR and the active driver?
 

·
Banned
Joined
·
165 Posts
Discussion Starter #18
Yes, I am using an ECM8000. I was initially taking NF measurements of the PRs to observe the tuning and see I could observe the null until I realized that it would only be on the combined driver/pr response.

I agree with the skepticism because of the large calibration boost. If you say in your experience, which is much more extensive than mine, the single-digit measurements are junk I'll have to try to work out some way to measure it. I did make sure the house was as quiet as possible - no vents, fans, etc. Maybe trying many different locations and measurement positions and turning the volume up higher.
 

·
Banned
Joined
·
165 Posts
Discussion Starter #19
I tried a new method to measure the resonant frequencies. I removed the driver and left only 1 PR in the box. I placed the microphone inside the box, sitting on polyfill. I then struck the PR from the outside several times and saved the recording. Looking at the filtered recording, this method provided very strong and clear low frequency components in the recording. Much better than NF from the outside.

Here is the spectral content of the recording for each PR. The microphone calibration is included in this plot.


So, it looks like still after about 7 days of torturously-loud break-in including subsonic test tones and sine bursts at night the PRs still need A LOT of breakin to be able to meet spec (fs = 3.65Hz)

These more accurate fs measurements do show that the shake observed and the NF pr output modulation and harmonics shown in the previous post should eventually change.
I think it is safe to hypothesize that playing a tone that is the second harmonic of the PR resonant frequency can cause the box to shake at the pr resonant frequency. This is the case that was observed in previous posts.

Modeling with this higher fs doesn't affect output much until about 12Hz and below but it does raise the tuning frequency by about 1Hz. It also explains why the lower impedance peak was at a higher frequency than expected.

I still have to figure out what is causing the odd PR behavior at very large excursions in the 12-14Hz range. Perhaps it is some really bad non-linearity in the suspension that still need to smooth out. The only problem is that I can't drive the PR as hard as I want because the house rattles too much. I have to settle for < 10Hz frequency tones at night for break-in sessions but the PR excursion isn't as great there.
 

·
Registered
Joined
·
438 Posts
I don't think your test is accurately identifying the Fs of the PR's...you've got the phase angle to deal with and the effects of the enclosure around the PR. And who knows what your microphone is doing at those lower frequencies too...there's a lot of physics going on here.

Btw, the resonant frequencies you're looking at are well below the Nyquist rate of typical video cameras. I would recommend just filming the PR movement and then dumping it into windows movie maker where you'll be able to calculate the frequency with the timestamps.

Also, you need to measure the PR Fs in the same manner that it was spec'd. I dunno if this is a half space measurement or calculated from a fixed cabinet volume or what. Changing the environment around the PR will changes the resonance you measure.
 
1 - 20 of 27 Posts
Top