Hi everyone.
As I mentioned in my introduction post, I recently built two wireless speakers. Both sound great overall, but there's one major issue: port noise, especially from the subwoofer.
I built the subwoofer using a GRS 8SW-4HE-8 8" driver. I settled on a box volume of 40L, excluding port and woofer displacement. I chose a tuning frequency of 26Hz. The transfer function magnitude and SPL curve for this particular configuration looked quite good:
For the port, I went with a 12 by 5 centimeter rectangular port. At a tuning frequency of 26Hz, this port has to be 60cm long. To accomplish this, I had to bend the port once inside of the enclosure. I also added an angle to turn the 90° bend into two 45° ones. Here's a rough sketch of how it looks like from the side:
I accounted for all wood thicknesses during construction to make sure the port length was accurate. The internal length of the port came out to be exactly 60cm, just as planned.
In WinISD, the simulated peak port velocity looked perfectly acceptable for this size and power level. At 90W input, the velocity peaks at around 18.5 m/s at 24Hz, which is already below the tuning frequency of 26Hz, so I didn’t expect any major issues with port noise.
Despite the promising results in WinISD, there are two major issues in practice:
Now my questions are:
Where did I go wrong?
Do I have to simply flare the port?
Why else is there so much port noise?
As I mentioned in my introduction post, I recently built two wireless speakers. Both sound great overall, but there's one major issue: port noise, especially from the subwoofer.
I built the subwoofer using a GRS 8SW-4HE-8 8" driver. I settled on a box volume of 40L, excluding port and woofer displacement. I chose a tuning frequency of 26Hz. The transfer function magnitude and SPL curve for this particular configuration looked quite good:
For the port, I went with a 12 by 5 centimeter rectangular port. At a tuning frequency of 26Hz, this port has to be 60cm long. To accomplish this, I had to bend the port once inside of the enclosure. I also added an angle to turn the 90° bend into two 45° ones. Here's a rough sketch of how it looks like from the side:
I accounted for all wood thicknesses during construction to make sure the port length was accurate. The internal length of the port came out to be exactly 60cm, just as planned.
In WinISD, the simulated peak port velocity looked perfectly acceptable for this size and power level. At 90W input, the velocity peaks at around 18.5 m/s at 24Hz, which is already below the tuning frequency of 26Hz, so I didn’t expect any major issues with port noise.
Despite the promising results in WinISD, there are two major issues in practice:
- It doesn't seem to produce much sound at 26Hz. Audible output really starts around 32–34Hz when sitting directly in front of it, not the 26Hz I was aiming for.
- Port chuffing is excessive. From ~25Hz to ~40Hz, the port noise becomes very noticeable. At times, it seems louder than the speaker itself.
I’ve attached a video showing the port noise when playing a 28Hz sine wave for reference.
Now my questions are:
Where did I go wrong?
Do I have to simply flare the port?
Why else is there so much port noise?
The response at Fb should be at least -3dB from the 32Hz response.
You didn't mention if you had tested the Fb frequency is correct.
The cone looked to be moving quite a bit, it should move the least at Fb.
Because of the way our hearing works at low frequencies, as little as 4dB difference may seem like only about half as loud ~25Hz.
Port noise is high frequency, 40dB (around the noise level from a refrigerator) at 1kHz would sound as loud as 100dB at ~30Hz.
You may not have. Room modes can have a huge effect on SPL at any particular position.
Amp clipping or driver distortion may also be responsible for some of the noise, impossible to tell their contribution from the video.
Rounding the port's exit would reduce noise.
Single tone sine wave testing is the worst case for port noise.
Most musical low frequency is impulsive, and it's high frequency content masks a little wind noise.
That said, the port looks a bit on the small side for an 8" driver with 10.5mm Xmax.
Did you look at excursion in your simulation?
Art
I think the port is too small in terms of area, I bet if you were to measure output of the driver and port separately the port level would be significantly lower than the driver.
Although.. it might just be that the port is too narrow(slot like) and if it was more square turbulance would be lower.
Although.. it might just be that the port is too narrow(slot like) and if it was more square turbulance would be lower.
Thanks a lot for both of your responses!
---
I didn’t think the port was overly small. I chose the largest port I could fit with only one bend, since I've read that multiple bends increase turbulence. Simulated air velocity also looked acceptable. Is there something I might have overlooked?
----
---
I'm not entirely sure how to accurately test the Fb frequency, but I'll try playing different sine waves and focus on identifying the point where the cone moves the least.
Interesting. I'll try placing the sub closer to a corner or moving it to a different room to see if the response improves. The room I tested it in might be negatively affecting low-frequency performance, as another sub I tested in this room behaved strangely as well.
I don’t think it’s amp clipping or driver distortion in this case. I did try turning the volume higher, and while distortion did appear, it sounded very different from the chuffing I’m hearing at moderate levels.
I’ll give that a try by sanding down the port flares a bit. Should I do this on both the internal and external ends?
I've noticed this as well, the chuffing is less noticeable during music playback, but it’s still clearly audible in songs with lower bass notes.
Yes, I did take excursion into account. At 90W input, the driver doesn’t exceed its Xmax in WinISD:
I didn’t think the port was overly small. I chose the largest port I could fit with only one bend, since I've read that multiple bends increase turbulence. Simulated air velocity also looked acceptable. Is there something I might have overlooked?
----
I'm not sure how I’d measure port output separately. As I mentioned above, the simulated air velocity didn’t seem too high, but maybe there's more to it.
Regarding the shape, I did some research before designing it and saw that a maximum ratio of about 1:5 is generally recommended. My port is 12 by 5 cm, giving a ratio of about 1:2.4, which should be well within that range. I've even seen builds with narrower ports.
From what I can tell, the lower the frequency I play, the more the cone moves, which seems different from what I expected.
I opened up WinISD and looked at the frequency where the cone should move the least (Fb), and compared it to where the cone should move the most.
- At 35Hz, the cone moved noticeably, but not excessively. I could also clearly hear the bass.
- At the tuned Fb of 26Hz, the cone moved much more than at 35Hz, and the audible output was significantly lower.
So now I’m not really sure what’s going on here…
The cone excursion should be minimum at the tuning frequency, and it should be fairly obvious by eye. If it's moving more as you lower the frequency with constant drive voltage, that indicates you are already below Fb. Perhaps the box is tuned higher than expected. Have you checked for air leaks? And the obvious, no bass boost in the amp or other processing affecting the signal during testing?
Your vent area seems reasonable at 9.3 in^2 , a bit better than 3" pipe which would be 7 in^2. I have a Dayton RSS 8" sub tuned Fb 25 Hz using 3" pipe vent, and it will chuff with 70W sinewaves but not noticeably with music.
Your vent area seems reasonable at 9.3 in^2 , a bit better than 3" pipe which would be 7 in^2. I have a Dayton RSS 8" sub tuned Fb 25 Hz using 3" pipe vent, and it will chuff with 70W sinewaves but not noticeably with music.
That would indicate the actual Fb (box tuning) is closer to 35Hz than 26Hz, more than 1/3 octave higher than you planned.
The excursion (and impedance) minima is at Fb.
Your port would need to be longer to drop the tuning frequency to your prediction.
Measuring the excursion curve accurately should give you an idea of the ratio the sim is off by.
At Fb, the output of the port lags the driver output by 90 degrees.
Below Fb, response drops at ~24dB per octave as the port output becomes 180 degrees out of phase with the driver.
The port output is basically useless and destructive less than 1/3 octave below Fb, and the excursion maxima is ~1/3 octave above Fb.
Putting a light colored dot near the cone surround makes it easier to see movement, I usually use a silver Sharpie marker pen. Look across the cone at an angle where you can just see the dot, holding a ruler near the dot.
Xmax is one way excursion, if you see 21mm peak to peak travel, your driver has reached it, which should take ~20 volts.
You don't need to test using that much voltage, around 10 volts should hit ~10mm peak to peak, should be easy to see the difference between 10mm and 4mm at Fb.
Or you can use this technique taping a paper ruler wedge with 2/1 increments marked on it:
If you don't have a sweepable sine wave generator, note the peak to peak excursion in 1Hz increments either side of Fb.
Sine waves contain ~eight times the average power of music reaching the same peak voltage, so keep the duty cycle low, a few seconds to measure, 10 seconds to cool while you chart the excursion.
Art
There should not be any air leaks. I’m pretty sure I’d hear them. There actually was a leak at the beginning that made a very distinctive noise, but I’ve since patched it up. As for signal processing, the only filtering happening is from the ADAU1452 DSP, which is cutting everything below ~22Hz and above ~90Hz.
Okay, that’s actually good to hear. It sounds like the amount of port noise I’m experiencing is normal for this port size.
That’s a good point. After reading these two threads:
https://www.diymobileaudio.com/threads/winisd-and-slot-ports.156355/
https://www.diyaudio.com/community/threads/first-diy-ported-sub-need-help-with-port.400541/
…I actually started wondering if my tuning frequency was too low, not too high, especially since WinISD doesn't seem to apply the correct end correction factor for slot ports. This can apparently lead to slot ports being longer than they should be. But that wouldn't quite be in line with the excursion values I saw today.
I'll have to accurately measure the excursion to check the actual tuning frequency of my box.
40 L is pretty big for an 8" driver. PE recommends 1 cu.ft. (28 L) tuned to 27 Hz. Me? If I had access to DSP I'd probably opt for a sealed design and use the DSP to extend the LF response, and / or opt for passively-assisted sealed design (think big capacitor added in series with the driver).
It's important to check to see if your build matches your sim, if you're noticing things that you didn't expect to notice. I usually start my troubleshooting by taking an impedance curve, as that can disclose a LOT of things (e.g. actual Fb, if the box is lossy, etc.). DATS is a good tool to do this, but if you want to spare some $$ you can build your own impedance jig to use with REW. Plans to do so are available via Google Search.
It's important to check to see if your build matches your sim, if you're noticing things that you didn't expect to notice. I usually start my troubleshooting by taking an impedance curve, as that can disclose a LOT of things (e.g. actual Fb, if the box is lossy, etc.). DATS is a good tool to do this, but if you want to spare some $$ you can build your own impedance jig to use with REW. Plans to do so are available via Google Search.
Just to clarify - Is that 90W into 8 ohms (Eg = 26.83 volts), 90W into 4 ohms (Eg = 18.97 volts) or a constant 90W at all frequencies?
This is worth checking. I had to shorten the vent in mine several inches vs the model in order to get the correct Fb, confirmed via impedance sweep.
You just close mic(couple centimeters away) the driver and port separately. This assumes you have a test mic and software like REW.. which is free. That will produce graphs that may look something like this where the contributions of each are clearly visible.
