Improving my OB midbass

I've been experimenting with OB for a while now. System is sealed 12" subs, 18" OB midbass, 8" OB midrange, AMT OB tweeters. Up until recently I wasn't super happy with OB midbass (mostly missing sense of impact) and was almost ready to go back to sealed, but added 2 additional 12" subs to the two I had, applied MSO, and now I'm happy with subs playing up to 80Hz.

System is active with digital xo on Acourate. When I run the correction routine, I see a sharp drop in raw response right before 300Hz that ends up being corrected by Acourate, but still bugs me that such a big drop is there. The 8PE21 midrange struggles to go much below 300Hz in the 30cm OB I have it on, so for now I xo at 300Hz and let Acourate correct.

20230710_054630.jpg 20230710_054720.jpg

I took some in-room measurements of the midbass to get a better feeling of what I have. Measurements are taken from the listening position, no IR windowing (500ms), speaker on its usual location. Far from ideal, but I haven't figured out how to measure an OB (in-room nearfield with windowing misses the backwave, and so does measurement outside plus it's a hassle in a large city like where I'm at).

2023-07-05 in-room with H without.jpg

Blue is how I'm running the 18". While it says H for the baffle, the structure where the mids/tweeter sits is separated from the floating midbass panel and the midbass was close to the front edge so it was more of a U open baffle than an H. With 1/12 smoothing it shows a 94.8dB at 231Hz, dropping to 71.1dB at 288Hz.

Orange is with the 18" baffle on the floor as usual, but I removed the midrange/tweeters and "H" (or U) structure around it. The dipole peak went up to 95.8dB at 242Hz - makes sense that with a smaller baffle the dipole peak raised the frequency at which it manifests. Interestingly, the gain I am getting on the lower end is less than I expected: about 2dB at 80Hz. The 18" Faital is quite efficient and a 400W class D driving it would have no issues compensating this with added power, so I'm thinking the H (or U) isn't helping much here.

The green line is with the speaker on the same location, but placed on top of a chair with minimal backrest (so it's acoustically transparent). I found this measurement quite interesting: It didn't drop below the orange (panel sitting on floor) on the lower range yet the dipole peak dropped by 8dB.

All measurement display a null around 210Hz and another at 139Hz. I wonder what might cause these?

I'm concluding:
  • I don't need the H/U baffle as there is minimal gain in the lower end. I'm not seeing evidence of the cavity creating nulls either, but there seems to be no upside to having the H/U.
  • As expected, a smaller baffle raises the dipole peak frequency.
  • Raising the panel from the floor seems to diminish the amplitude of the peak at the dipole peak.
A next step would be to make a smaller baffle (to increase the dipole peak frequency) and raise it from the floor (to reduce the amplitude of the peak).
And maybe when I'm done with improving the midbass I would try a larger midrange baffle to make it reach 250Hz.

Am I on the right track, or misleading myself from improper measurements, or other?
 
Member
Joined 2008
Paid Member
I’m interested to see what your solution will be.

I have minimal knowledge on the topic, but what I have read seems to say that nulls looking like this are due to the room.

The last article I read said they can’t really be gotten rid of but only minimized with tools such as multiple subs and room treatments to minimize reflections that cause the nulling?

It seems like raising your sub threw away some of its energy because you didn’t have the floor boundary reinforcement. It made your response curve look more even, but didn’t really affect the nulls you’re concerned with.
 
diyAudio Moderator
Joined 2008
Paid Member
I understand your wanting to measure the 18" anechoically, since the U is not front/back symmetrical.. however since it is low enough to be near room controlled you might see if you can get an idea by averaging a bunch of room responses.
 
  • Like
Reactions: 1 user
I find John Kreskovsky's method of measuring a dipole driver to be quite useful.

Measuring a dipole

The link to download the associated file is not very prominent, it's right above the first graph.

REW makes this method quite easy. Take a nearfield sweep of the driver, and switch to the All SPL tab. From the controls menu, open the Alignment tool. Select the same measurement for both drivers, then invert the second copy and add delay to match the path length as described in the method. For a nude driver, the delay = the radius of the driver.

I have tried this with a few nude drivers. In my experience the result matches the in-room measurement of the driver pretty well from the dipole peak down, but without room interference effects. Kreskovsky references a flat baffle in his example. I would think a symmetrical H-frame would be adequately modeled by increasing the path length / delay appropriately. I'm not sure how closely an asymmetrical baffle would be represented.

Bill
 
  • Like
Reactions: 1 user
I find John Kreskovsky's method of measuring a dipole driver to be quite useful.

Measuring a dipole

The link to download the associated file is not very prominent, it's right above the first graph.

REW makes this method quite easy. Take a nearfield sweep of the driver, and switch to the All SPL tab. From the controls menu, open the Alignment tool. Select the same measurement for both drivers, then invert the second copy and add delay to match the path length as described in the method. For a nude driver, the delay = the radius of the driver.

I have tried this with a few nude drivers. In my experience the result matches the in-room measurement of the driver pretty well from the dipole peak down, but without room interference effects. Kreskovsky references a flat baffle in his example. I would think a symmetrical H-frame would be adequately modeled by increasing the path length / delay appropriately. I'm not sure how closely an asymmetrical baffle would be represented.

Bill
Thanks for pointing to this!

Looking into Kreskovsky's paper: "Thus, to obtain the low frequency response of the dipole we first invert the past of the near field data and then add the delay corresponding to the baffle delay and sum the result to the original near field measurement".

I believe he meant "phase" rather than "past". Can REW invert phase of a measurement? I see wrap and unwrap phase in the controls menu for a given measurement, but didn't find phase inversion. Can I add a delay in REW?
Otherwise I think Acourate can do this,
 
Kreskovsky's approach involves a nearfield measurement up to 330Hz (in his example in the paper) plus a 1m measurement for frequencies above 330Hz. Since the region I'm interested in is up to 300Hz or so, I could do just with a nearfield measurement, invert phase, add the baffle delay, and sum to the original nearfield.

Is it OK to perform the nearfield measurement on-axis and 3" away from the 18" dustcap, with the speaker standing on the floor as when in-use. Hopefully I can take these measurements tonight, and would like to make sure I perform them correctly.

Thank you!
 
REW makes this method quite easy. Take a nearfield sweep of the driver, and switch to the All SPL tab. From the controls menu, open the Alignment tool. Select the same measurement for both drivers, then invert the second copy and add delay to match the path length as described in the method. For a nude driver, the delay = the radius of the driver.

Thanks for the detailed explanation!! The first time I saw your post I went straight to the paper and I guess I didn't retain your explanation.

I took a nearfield measurement of the 18" mounted on the baffle, with U frame on. Mic at 1cm from dustcap, on-axis. Went into the alignment tool, selected invert on the second copy and added 1.75ms delay that corresponds with 600mm (U baffle is 600mm wide, so 300mm from center of driver to edge, plus 300mm U wing). Then clicked align sum.

Below are the plots: green is delayed (which is the same as the measured response), blue is the aligned sum (which I believe = measured + delayed responses).
2023-07-11 nearfield and simulation with U.jpg


What can I conclude from this?
I guess first thing is to verify I did this correctly.
If so, then the dipole peak is at 100Hz which is far away from the 300Hz I'm crossing over to the midrange. A small/minimal baffle would seem to make sense.
Interesting to note the steep upward slope for the simulated response (blue) up to the dipole peak. SPL at 100Hz is 23dB higher than at 50Hz. Steeper than I expected.
Below I overlaid the simulated response (aligned sum) vs what was measured at the listening position (was shown in blue in post #1) with 1/6 smoothing. The measured response showed the nasty peak around 240Hz and not as steep upwards slope from the lower end.
2023-07-11 simulation vs measured at listening position.jpg


Looking forward to you feedback!
 
Today I removed the 18" driver from the panel and set it up nude (without baffle) on a stool, and measured nearfield (1cm). I followed Kreskovsky's like above and got this:
2023-07-12 nearfield and simulation nude.jpg


When nude the dipole peak is at 400Hz. I'll device a stand to run the 18" nude and given them a listen. I'm expecting this to sound better towards the 300Hz xo but wonder about the sound closer to the 80Hz xo.
 
Hey Lewinski,
I have heard some great open baffle loudspeakers over the years and used a pair when I had a dedicated man cave, but now its all family home cinema and music in one system so sealed box all the way, but here are my thoughts on open baffle:
(1) Beyma TPL 200 AMT sounds superb in an open baffle, mine crossed over with a DSP/Eq crossover around 2KHz.
(2) Below 80 Hz (ish) best to go sealed box, certainly below 50 Hz sealed is best but must use min of two separate subs and three or four is ideal.
(3) 60Hz to around 400Hz is where open baffle cone drivers can really be astonishing with (IMO) the best overall compromise of life like dynamics, low distortion & low coloration with a natural, agile and textured sound that is magical.... Upright bass, cello and low end piano are simply believable and low male vocals are dense and rich without any trace of "cuppy" or boomy resonance.... This is vital to make the listening experience believable.
(4) Drivers... Size matters! Use big Pro orientated drivers with low Qts, low Mms/high Bl and design so that in your room and your highest comfortable SPL's the cones never exceed plus/minus 1.5 mm (the lower the cone excursion the better) between 80Hz and 400 Hz. For me that was easy, the Precision Devices 15 inch https://www.precision-devices.com/wp-content/uploads/2020/09/PD158-DATASHEET-310114.pdf was (sadly discontinued now) the perfect driver covering 80Hz to 400Hz.
(5) Baffle matters... Avoid straight sides, use curves... This 10 inch Volt midrange sounded gorgeous covering 100 Hz to 400Hz... Scroll down and check out the pics.... https://www.hifisentralen.no/forume...best-implementere-manger-i-diy-oppsett.83866/ and these curves in a large but very dense and heavy 25mm thick acrylic baffle made two little 8 inch drivers sound like a 15 or 18 inch driver with no baffle... (originally posted on Overkill Audio post) https://www.diyaudio.com/community/threads/acrylic-loudspeaker-cabinets-open-baffle.277692/
The curved sides eliminated the one frequency peak you get with "H" or "U" shaped baffles where the sound wraps around the front baffle and cancels the rear wave creating the side nulls and low point cut off.
(6) Having good DSP/Eq and room measurement / mic is pretty much essential to make it all blend and sing... Passive crossover is a non starter IMO.

Hope this helps and happy experimenting!
Cheers
Alex.
 

Attachments

  • Finale by Overkill Audio.jpg
    Finale by Overkill Audio.jpg
    63.6 KB · Views: 83
  • Angel by Overkill Audio  2006.jpg
    Angel by Overkill Audio 2006.jpg
    91.3 KB · Views: 82
  • Angel side view.jpg
    Angel side view.jpg
    473 KB · Views: 80
@Hydrogen Alex
Thanks for the inputs!

I'm already onboard with several of your points:
1) Running my Beyma TPL-150H with open back. I'm due to try them without the waveguide (after I make my next improvement with midbass)
2) I have 4 sealed 12" subs below 80Hz, with MSO implemented. This was a significant improvement over 2 subs.
3) 80-300Hz (maybe 400?) is what I'm solving for now. I was missing slam until I added the 2 new subs, now I'm doing better there but I'm not fully satisfied in the upper range until the 300Hz. Hence this thread.
4) I'm using a pro 18" driver for 80-300Hz now. Not nearly as nice as your PD, but should work. At the end of the day I'm trying to figure out if OB midbass is the best solution in my living room (1.2m away from the absorbent wall) and I would move to a better driver like AE LO15, or go sealed midbass.
5) Point taken about curved sides for baffles! Those deep acrylic baffles are meant to reach as low as possible (no subs). Overkill's Finale has a sub with the midbass using curved sides (interestingly those are very close to the wall). I'm going to try a small flat baffle next (or no baffle), and if I don't like the lower end (close to 80Hz) I will look into narrow baffle with curved shallow wings.
8) I'm using Acourate for DSP.
 
Nice work Lewinski!
I should have read the whole thread before I jumped in but hey, it sure looks we are on the same page with OB!
If you look really closely at the photo, the big Finale speakers are actually in front of big custom made room treatment / absorbers, they were made by a studio treatment company and tuned to absorb a lot of the 100Hz to 400 Hz frequencies from the rear of the OB driver. Apparently they worked really well and allowed the owner to place the speakers much closer to the walls.
Re your open baffle 18 inch driver,what is the Mms/Bl?
I chose the PD 158 because of its light weight (85g) Mms and good Bl of 25, so the ratio was a really fast 3.4. Big 18 inch drivers tend to be around 200g and 30 Bl ie the Precision Devices 1850/3 is 215g and 31 Bl giving a much "slower" figure of 6.9.
When I say "slower" I refer to the stopping time of a cone in motion rather than the acceleration of a stationary body at rest ie Those who sell high Mms low Bl drivers only focus on a one note impulse response and claim its all just a matter of frequency response... I dont believe that is the whole story...
Once real music is playing the cone is oscillating back and forth and attempting to respond to a never ending stream of individual pulses ie Compression (cone pumping out) or Rarefaction (cone pumping back in) and the real world damping factor / back EMF force exerted by most amplifiers is too weak to offer effective cone control.... So it falls to the driver magnet/suspension to control the cone and hence I believe a low Mms/high Bl is very important...YMMV!
Cheers
A.
 
Indeed my Faital 18HP1010 don't have very strong motors: Mms: 186g, BL: 18.4, son Mms/BL=10.1...far from your PD.

If I decided to stay on OB I would consider AE LO15 acknowledged as avery good performer. It has similar Mms and BL to the 18hp1010: Mms: 74g, BL: 7.5, Mms/BL=9.9.

A new option I'm looking forward to read from early adopters is Purifi 10". Much smaller, but huge displacement with very low distortion. These parameters don't look much different, though: Mms: 99g, BL: 12.6, Mms/BL=8.
 
I'm confused now.

I simulated the 18" nude in dipole in Edge and shows the dipole peak around 500Hz:
Edge 18hp1010 nude dipole.png


I built a quick 3-legged stool to hold the driver in place. Driver is now 20 degrees looking up:
20230715_193435.jpg

Took the measurement and applied the method above to sum a copy with inverted polarity and delayed by the radius of the driver. The measurement at 1cm from dustcap shows a peak at 120Hz, and the simulated dipole response doesn't really show a dipole peak:
2023-07-15 nearfield and simulation nude.jpg


I tried a number of different things to make sure I wasn't making a mistake when measuring. I couldn't find it. Looks too good to be true.

I will build another stool for the right speaker, run the measurement/correction routine and listen. Will probably take me a week to complete though.
 
There are at least 2 factors what cause this results.
1. Edge simulation is not very accurate and is made for distance 3 m in current case.
2. Nearfield measurement of OB speakers did not give precise general speaker response with calculations "sum a copy with inverted polarity and delayed" as distance of measurement point affects on FR a lot. Dipole peak 120 Hz for this nude speaker is definitely not correct, it must be at least 3 times higher.
 
Last edited: