Jump to content

The ultimate small speaker - final design peer review thread


lowerFE

Recommended Posts

First off, I'd like to thank everyone for the incredible help given to me when I was doing research for my ultimate small speaker build. I call it, the Reference Mini. After too many hours of researching, modelling, thinking, etc, I think I got to a point where the design is mostly complete, and I want to share it here and get some comments on possible improvements from members here that knows much more than I do. 

 

For those interested, I talked about my previous attempt at such a speaker here. What I'm proposing here is a successor that improves upon every performance aspect after learning everything from the previous build. 

 

Design goal: Build the highest fidelity small speaker possible with as much bass output as possible in the 40-60Hz region to cover most music well. 

 

In more specific speaker design goals, I wanted to achieve the following, in no particular order:

 

1. Excellent clarity, deep bass response, and extremely high dB per liter ratio for bass.

2. Minimize/eliminate horizontal and vertical lobing errors

3. Matched directivity and smooth polar response

4. Minimal cabinet vibration

5. Excellent off axis response

6. Linear and minimum phase response 

7. Flat power response

8. Transient perfect response

9. Time aligned drivers

10. Minimize baffle diffraction

11. Ruler flat frequency response

 

Driver selection:

Tweeter: Scanspeak Illuminator D3004/6040-10 1'' Beryllium Dome Tweeter

 

d3004_6040-10.jpg

Midrange: Scanspeak 10F/4424G 3.5'' Neodymium midrange

10f_8424g.jpg

Woofer: 2x Wavecor WF152BD05 6'' Glass-fibre woofer

 

 

wf152bd05_06.jpg

 

Passive radiator: Either 2x Peerless 5.25'' passive radiator or 2x DIY PR from modifying the Tang Band W5-1138SMF

 

Enclosure design

 

The enclosure will be 14'' x 6'' x 6'' built with 1/4'' baltic birch plywood. In the horizontal configuration, the driver configuration looks like

W  P
WMTP

The enclosure is a pair of dual opposed configuration. The woofer and woofer will be back to back, and so will the PR's. The tweeter and midrange will be arranged vertically in the horizontal placement. This allows me to use a 1/4'' BB for the enclosure and still have a vibration free cabinet. While this is a convenient configuration, the rear firing woofer is not ideal because it can cause comb filtering. At 6'' deep, it will be 90 degree out of phase at 340/(4d) = 340/(4 * 0.15) = 566Hz. Even though the crossover is going to be 300-400Hz, I want a linear phase speaker, which won't happen. But then the fact that it is vented will screw up the phase anyways. I still would like to reduce this effect, maybe with a delay, but I don't know if it'll work. 

 

Another (big) issue is that it seems adding the PR's don't actually boost output much. It increases the output at tuning by 1dB with the same power vs sealed. It seems like all the PR is doing is reducing the excursion demand on the woofer. On a driver like the Tang Band W5-1138SMF or the Peerless SLS 5.25'', there is a 4dB increase in output vs sealed with the same power, and the PR excursion level is a lot higher. I'm not quite sure why, because the T/S parameter of the L16 and SLS are very similar. Only big difference is the Cms is double on the L16. 

 

DSP processing:

 

A miniDSP 2x4 HD will be used as the crossover and compressor. FIR filters will be used to achieve minimum phase. The speakers will be time aligned. The crossovers will be LR4, and things like baffle step correction, bass boost, high pass, and driver response correction will also be done after the speaker is built as I prefer to measure instead of relying on simulation graphs. 

 

A 2 band compressor will be used to protect the speaker. I'm thinking of leaving the >50Hz band uncompressed, and the 40-50Hz band will be compressed with a ratio of 50 and an attack of 1ms. This is extreme, but so far it has been working pretty well with the previous speaker. I need to learn more about compressors before I experiment to find better settings. Any tips here would be helpful. 

 

Amplifier:

 

I will be internally amplifying using the ICEpower 50ASX2 for the midrange and tweeter, which will provide 50W @ 4 ohms for each driver. The two woofers will be powered by the ICEpower 125ASX2, which will deliver 125W @ 4 ohms per channel to the woofers. The ICEpower amps are very high performance class D amplifiers that should sound much better than most of the class D amplifiers on eBay and Alibaba that may have questionable designs and parts quality. There will be no doubts on the excellent sound quality, power output, and build quality of ICEpower amps. They have integrated power supplies, the dimensions are incredibly small and they're not horribly expensive. 

 

What do you think of the design? Any comments or suggestions for improvement would be greatly appreciated. 

 

For those interested, these are the reasoning behind the design choices.

 

Driver selection:

 

Tweeter: Scanspeak D3004/6040-10

 
This is the finest small format tweeter available with exceptional distortion performance. While very expensive, at $250 (from Solen) it is actually very cheap for a Beryllium tweeter and it performs just as well as the large (and much more expensive) Scanspeak Beryllium tweeters. It has good sensitivity, dispersion, and can handle a 1500Hz LR4 crossover without too much strain. Best of all, all that performance is in a tiny 2.44'' face plate tweeter that can be further cut down to ~2'' to reduce size even further. 

 

Midrange: Scanspeak 10F/8424G

 

This is the best sounding ~3'' cone midrange available, and the largest midrange I can accommodate to achieve a 1/4'' wavelength crossover frequency with the tweeter. I will explain more of this significance later in the post. While there are 3'' domes, they are all bigger than the 10F, and it must be crossed much higher at around 600-1000Hz when I want the midrange to be crossed low closer to 300Hz to take advantage of the clarity of the midrange as much as possible. A woofer capable of the same clarity as the Scanspeak 10F would be extremely costly and may not be suitable for bass duties. Even if I did use a driver of that quality, the woofer compartment won't be fully stuffed like the midrange chamber to absorb the rear wave, which will result in an inferior sound. 

 

Woofer: 2x Wavecor WF152BD05

 

The woofer search gave me a lot of trouble due to the sheer number of choices and trying to find one suitable for bass. This is a classic compromise problem of bass and midrange clarity. Not only many of the high quality and high fidelity 5-6'' midwoofers do not reproduce bass very well at high excursion levels, most are not designed for small enclosures and have "high" Vas and low Fs, both are not "ideal" for small PR enclosures, which leads to very poor PR efficiency and less overall bass output of up to 3dB for the same input power. A driver that is ideal for high output bass reproduction, such as the Tang Band W5-1138SMF, does not have high fidelity midrange, which would be unacceptable given the best in class sound quality of the other components. 

 

I almost gave up and just used a high fidelity woofer and sacrifice bass output, but after some serious digging, I found a little known driver from Wavecor that has almost the best of both worlds. It had pretty much the ideal T/S parameters suitable for my design, while having extremely good sound quality comparable to Scanspeak Revelators. The result is a woofer that is 2-3dB more power efficient in the 40-55Hz range than nearly all high fidelity midwoofers, and just 1dB less than the W5-1138SMF, the ideal miniature subwoofer driver for this application.

 

Passive radiator: Either 2x Peerless 5.25'' passive radiator or 2x DIY PR from modifying the Tang Band W5-1138SMF. 

 

Using the Peerless would be the easiest solution, but there is only space for 1 PR per driver, and the PR's 6mm xmax is not enough excursion. I can use 3 of them, or 1.5 PR per active driver, which would be enough, but the mass needed to reach a 42Hz tuning, almost 2x the Mms, might be too much for the PR's suspension to handle. 

 

So a solution could be to make a DIY PR. The Tang Band W5-1138SMF is a very good candidate as it has a 9mm xmax, so the xmech must be a lot higher than 9mm. The driver has at least 2x more excursion than the Peerless PR, which should be enough for the woofer. The Mms of the driver, at 29g, is 3x higher than a typical 5'' woofer, which means the suspension has the greatest chance of handling the ~60g of Mms needed to hit the tuning. At $30 each, it is not expensive either. 

 

The DIY PR seems like the clear choice, but I'm a little haunted by the miserable failure the last time I tried that. Also, the Peerless PR has ideal PR parameters, while the TB woofer does not, and may be less efficient than the Peerless PR.

 

Enclosure design and driver placement

 

The midrange tweeter will be in a vertical configuration to save width. All drivers will have their frames cut off to reduce size and so they can fit closer together, especially the midrange and tweeter. When drivers have a crossover frequency that is the same frequency as the 1/4 wavelength of the center to center distance each other, the sound of the two sources couple as if it is coming from one source like a point source speaker. This eliminates any vertical or horizontal lobing errors, so the speaker can be placed horizontally like a sound dock or vertically when used as a pair with no performance degradation. Imaging and coherence is improved, and the directivity is matched between the woofer and tweeter, improving imaging further. 

 

Normally, this is impossible to do with a tweeter and midrange. With the standard 4'' tweeter frame and a 5.25'' midrange. The best case center to center distance is 5'', which corresponds to a 1/4 wavelength of 678Hz, clearly far too low for any 1'' tweeter. However, with the small format tweeter and small 3'' midrange, with the frames cut off I can achieve a CTC of just 2.2'', which is a 1/4 wavelength of 1500Hz, which is a doable crossover frequency for a capable 1'' tweeter. 

 

Why did I scrap the 4 woofer sealed design?

 

The speaker will be tuned to 42Hz to achieve useful output to about 38Hz or so. I analyzed many bass heavy pop songs, and most bass is above 40Hz, and mostly around 45-60Hz, so I want to concentrate to get as much output in this region. I'm sorry Josh, I won't be doing the 4 woofer design. The power efficiency of going from 2 woofers to 4 woofers is far too low to the point I can only achieve just a bit over half the xmax at the rated max power of the driver. Realistically I can't put that much power in the speaker to drive it anywhere near xmax. I would also need to use a smaller driver than the L16 because a speaker with 4 L16's would need to be a lot bigger than I would like. So a 2 woofer + 2 PR setup would have equal or greater max SPL than a 4 woofer sealed configuration everywhere except below tuning. The PR would contribute at least 2dB of extra output even an octave above tuning. The only benefit of a 4 driver design would be output to 25-30Hz instead of 38Hz, which would be very cool and brag worthy, but I have to use twice the driver, 4x the amplifier, bigger enclosure, the need for a 5 band compressor rather than just a 2 band for the PR version, which would also be extremely expensive and complicated to do. Over $1000 would be saved, and a lot less equipment. I do intend to actually use this speaker. I don't want to have to hide 6 stereo amplifiers (or an iNuke), 4 miniDSPs, and a 12 channel audio mixer for this little speaker. Plus, even with just 2 woofers I still cannot reach xmax at the 80W max rated power of the L16 (80% there though), further confirming that 4 woofers was wasteful and unnecessary. Josh, I pass the torch to you to build a crazy 4 woofer design to get the ultimate bass performance at any cost. Maybe use the Beryllium tweeter as well?  :D

post-1515-0-40082800-1470528822_thumb.png

post-1515-0-66500800-1470528827_thumb.png

post-1515-0-96726400-1470528840_thumb.png

Link to comment
Share on other sites

I just noticed this, and I think it is an important enough observation and question to deserve its own post. 

 

Another (big) issue is that it seems adding the PR's don't actually boost output much. It increases the output at tuning by 1dB with the same power vs sealed. It seems like all the PR is doing is reducing the excursion demand on the woofer. On a driver like the Tang Band W5-1138SMF or the Peerless SLS 5.25'', there is a 4dB increase in output vs sealed with the same power, and the PR excursion level is a lot higher. I'm not quite sure why, because the T/S parameter of the L16 and SLS are very similar. Only big difference is the Cms is double on the L16. 

 

After much modelling, it seems like there is definitely a correlation between Cms and the amount of boost PR gives vs sealed. Woofers with low Cms, around 0.5-0.7mm/N, will get a large 4dB boost while woofers around the 1.3mm/N get just ~1.5dB, and woofers with Cms close to 2mm/N gets almost no boost. Why does Cms affect PR efficiency so much?

Link to comment
Share on other sites

A 1500 crossover sounds optimistic to me. Have you looked at http://medleysmusings.com/scan-speak-illuminator-d3004602000-tweeter-testing/ ?

 

Yes I did, and I realized the speaker's max volume will be limited with this low of a crossover. However, the speaker won't be playing all that loud for the tweeter to struggle. Eventually I want to find a way to make a limiter that changes the tweeter's crossover frequency from 1500Hz to 3000Hz above a level where the tweeter will start distorting. One forum member from another forum says the DSP tools he's working on could help me do this. But one step at a time. 

Link to comment
Share on other sites

Yes I did, and I realized the speaker's max volume will be limited with this low of a crossover. However, the speaker won't be playing all that loud for the tweeter to struggle. Eventually I want to find a way to make a limiter that changes the tweeter's crossover frequency from 1500Hz to 3000Hz above a level where the tweeter will start distorting. One forum member from another forum says the DSP tools he's working on could help me do this. But one step at a time. 

I don't understand why you'd want to design something that requires you to deliberately introduce distortion in a critical part of the frequency range, it seems completely contrary to some of your goals too. At what distance will you listening btw?

 

FWIW I made a speaker recently for which the 1st iteration of the crossover pushed the tweeter too low, I found this introduced a harshness/roughness that was really obvious with certain content even at lower levels. It was quite unpleasant, almost unlistenable in fact, but only on that specific content. I didn't think I'd even pushed it that far beyond and the measurements didn't look bad but it was still rough (this was roughly an LR6 at 1500Hz vs an LR4 at 1900Hz). Different speaker obviously so ymmv. 

  • Like 1
Link to comment
Share on other sites

I don't understand why you'd want to design something that requires you to deliberately introduce distortion in a critical part of the frequency range, it seems completely contrary to some of your goals too. At what distance will you listening btw?

 

FWIW I made a speaker recently for which the 1st iteration of the crossover pushed the tweeter too low, I found this introduced a harshness/roughness that was really obvious with certain content even at lower levels. It was quite unpleasant, almost unlistenable in fact, but only on that specific content. I didn't think I'd even pushed it that far beyond and the measurements didn't look bad but it was still rough (this was roughly an LR6 at 1500Hz vs an LR4 at 1900Hz). Different speaker obviously so ymmv. 

 

I'm pushing the crossover at 1500Hz because it is emulate a coaxial speaker's dispersion characteristics. When drivers have a crossover frequency that is the same frequency as the 1/4 wavelength of the center to center distance each other, the sound of the two sources couple as if it is coming from one source like a point source speaker. This eliminates any vertical or horizontal lobing errors, so the speaker can be placed horizontally like a sound dock or vertically when used as a pair with no performance degradation. Imaging and coherence is improved, and the directivity is matched between the woofer and tweeter, improving imaging further.

 

I'm gonna go change the crossover to 1500Hz on my previous speaker with the same mid and tweeter and see if I hear problems like you experienced. This speaker is gonna be used in a bedroom, so 10 ft away or so. I would use these as my main speakers, as they do sound significantly better, but being active and all doesn't play well with receivers.

 

Also, after some thought, it might not be too bad that the PR efficiency is bad with this soft suspension driver. Because of this, it "alleviates" the problem of having only 1 PR per driver. The PR's unfortunately won't boost max SPL much, but it will just serve to reduce distortion near tuning and active driver excursion levels. 

Link to comment
Share on other sites

I get the reason why you want to push the XO lower, it's a good reason for sure. I'm just saying it might be a risky strategy and that dynamically compressing the content to deal with it is not, IMV, a good trade off. I'm not saying what you're doing won't work btw or that it's an inherently bad idea, testing (using your ears) is the best and quickest way to verify.

 

FWIW I have found there are a few tracks from Adele 21 to be perfect for stress testing this aspect of a crossover, she can really belt it out and her voice has a broad range (which you can see if you watch a live spectrum analyser alongside listening).

  • Like 1
Link to comment
Share on other sites

It's important to be sure the tweeter can handle the same peak SPL as the rest of the range.  Even though the tweeter will never see as much power, transient SPL peaks can still be very high up there.  It's not just a question of whether you can hear distortion but whether you cause premature wear of the driver.  With my horns, I did successively higher sine sweeps to look for signs of both power compression and THD rise.  You may need to use something like CEA bursts instead to test THD at high SPL without cooking the voice coil.  I don't know of any guidelines for safe max THD, but I tend to think 10% for the highest peaks is pushing it for a high frequency driver.  I want my stuff to last a while.  :)

 

I think you could compromise somewhat on the 1/4 wave rule between the mid and tweet.  If they are a half-wave length apart (double the XO), you'll likely end up with a single central lobe and nulls only all the way out at +/- 90 degrees.  You could have the tweeter and mid mount to a separate piece of the baffle that can be rotated 90 degrees to keep the tweeter and mid vertical, depending on what orientation you use the speaker in.  This would get you most of the benefits of 1/4 wavelength separation without pushing the tweeter too hard.  The trade-off is that there will be more directivity peak at the XO if the drivers are further apart.  Still, I think you can find an excellent compromise somewhere in the middle, like at 2 kHz.

 

I would avoid any kind of non-linear crossover that shifts the XO point up on higher peaks.  Imaging depends a lot on transients, and having that XO doing funny things hurt the imaging at higher playback levels.

 

Unfortunately, from the data posted, it looks like that tweeter has rather low sensitivity, like 82 dB/1m@2.83V and 4 ohm in the 2-3kHz range.  Peak output with 50W will be about 96 dB @ 1 m, which is closer to 86 dB at 10 feet.  If you ask me, that's pretty whimpy, and this tweet may struggle with transient peaks in well-recorded live acoustic music.  Another issue is that it starts to heat up with any input higher than about 4V or about 4W of power.  This only gets you to about 85 dB continuous @ 1 m in the 2-3 kHz range, or closer to 75 dB continuous at 10 feet away.  This could be a problem with some content at lively playback levels.  For a bullet proof design, you should consider a long term limiter on the tweeter because the 50W amp can probably cook it.  Overall, the tweeter looks to me like the weak point of the design.

 

The mid driver is a bit more efficient, but it is 8 ohm and won't be able to get as much power out of the amp?  The sensitivity (from manufacturers data) appears to be about 86 dB @ 2.83V, and with BSC of +3 dB, you can expect peaks of 101 dB @ 1 m or 91 dB at the listening position.  I'd say that's still a bit weak for playback of dynamic content, but not as bad as the tweet.  Presumably it has much better power handling than the tweet, so compression is a lot less likely to be an issue.

 

While it would be helpful to see sims of the PR alignment, the bass section looks pretty good on paper at least.  Given that you can only do one 1 PR per driver, I think the frequency response behavior you are seeing is a good thing.  Even if the PRs don't increase sensitivity much, they do improve power handling.  Depending on their thermal limits, you may be able to get away with a lot more power on the bass drivers than you would otherwise.  This can improve your extension in turn if you do a bit of signal shaping with the DSP.  Another thing is that your response likely has a wider transition region between flat and 4th order roll-off, and this will keep group delay in check leading to better sound quality without DSP and less aggressive DSP needed to make it do what you want.  And of course, most usage situations will involve room gain that might make up for whatever gradual roll-off there is before the tune, so you might want to just leave it that way.

 

And yes, it makes sense that this happens because of the high compliance of the bass drivers.  As I understand it, you have two impedance peaks in a ported response.  The higher of the two peaks is essentially analogous to the peak in a similar sealed system.  It involves resonance of the moving driver vs. the combination of suspension and air spring.  Driver excursion typically peaks around there if a HPF is used to protect the driver from unloading.  The lower peak represents the resonance of the driver and port air mass vs. the driver suspension in an I.B. as that is essentially what you get when the port unloads.  I believe the further apart they are, the more gradual the transition between flat and roll-off, and the woofers with high compliance are likely helping to move that lower peak (as well as the tuning frequency) lower.

 

My understanding is that PR designs work very similar to ported designs with the PR Sd being analogous to the port area and PR mass being analogous to port length.  The air spring stiffness in the ported box is analogous to the combination of air spring and PR stiffness in the PR system.  In many cases, the PR stiffness is low enough to neglect.  In any case, your challenge will be to find or build PRs with sufficient mass to achieve the tuning you want.

  • Like 1
Link to comment
Share on other sites

I don't understand why you'd want to design something that requires you to deliberately introduce distortion in a critical part of the frequency range, it seems completely contrary to some of your goals too. At what distance will you listening btw?

 

FWIW I made a speaker recently for which the 1st iteration of the crossover pushed the tweeter too low, I found this introduced a harshness/roughness that was really obvious with certain content even at lower levels. It was quite unpleasant, almost unlistenable in fact, but only on that specific content. I didn't think I'd even pushed it that far beyond and the measurements didn't look bad but it was still rough (this was roughly an LR6 at 1500Hz vs an LR4 at 1900Hz). Different speaker obviously so ymmv. 

 

Hearing is definitely quite sensitive to THD in the mid/high frequencies, depending on the content.  If I listen carefully I can definitely hear the distortion (about 0.5% IIRC) in my DNA-360s with sine sweeps of only 75 dB at the MLP.  Room acoustics can attenuate or amplify THD, depending on the situation, so it's always good to keep it as low as possible.

  • Like 1
Link to comment
Share on other sites

I think you could compromise somewhat on the 1/4 wave rule between the mid and tweet.  If they are a half-wave length apart (double the XO), you'll likely end up with a single central lobe and nulls only all the way out at +/- 90 degrees.  You could have the tweeter and mid mount to a separate piece of the baffle that can be rotated 90 degrees to keep the tweeter and mid vertical, depending on what orientation you use the speaker in.  This would get you most of the benefits of 1/4 wavelength separation without pushing the tweeter too hard.  The trade-off is that there will be more directivity peak at the XO if the drivers are further apart.  Still, I think you can find an excellent compromise somewhere in the middle, like at 2 kHz.

 

I would avoid any kind of non-linear crossover that shifts the XO point up on higher peaks.  Imaging depends a lot on transients, and having that XO doing funny things hurt the imaging at higher playback levels.

 

Unfortunately, from the data posted, it looks like that tweeter has rather low sensitivity, like 82 dB/1m@2.83V and 4 ohm in the 2-3kHz range.  Peak output with 50W will be about 96 dB @ 1 m, which is closer to 86 dB at 10 feet.  If you ask me, that's pretty whimpy, and this tweet may struggle with transient peaks in well-recorded live acoustic music.  Another issue is that it starts to heat up with any input higher than about 4V or about 4W of power.  This only gets you to about 85 dB continuous @ 1 m in the 2-3 kHz range, or closer to 75 dB continuous at 10 feet away.  This could be a problem with some content at lively playback levels.  For a bullet proof design, you should consider a long term limiter on the tweeter because the 50W amp can probably cook it.  Overall, the tweeter looks to me like the weak point of the design.

 

 

I thought about the 1/2 wavelength crossover as well, as 1500Hz is pushing things pretty far. I probably will just switch between the two settings. my normal listening volume is very low, like ~60-65dB. If I need to play louder I'll just switch the setting. On the miniDSP 2x4 HD there are 4 preset configurations that can be switched through a remote, so a settings change is very easy and quick. However, when you presented your analysis with numbers, I never realized how inefficient the tweeter is. I thought the 85.9dB figure was the 1W/1M figure since Scanspeak rates them at 89.2dB @ 2.83V. But I see now I can actually strain the tweeter quite easily. Unfortunately I am very limited on the choice of tweeters because there are very few high quality small format tweeters, and the Scanspeak stands out as having the best sound quality. Maybe I should get the beryllium version? It's got 1dB higher sensitivity!  :D

 

1/2 wavelength is not a bad compromise, but apparently the magic really happens at 1/4 wavelength as that's when the drivers sound coherent. I'll just try it and see. I love active speakers!

 

The mid driver is a bit more efficient, but it is 8 ohm and won't be able to get as much power out of the amp?  The sensitivity (from manufacturers data) appears to be about 86 dB @ 2.83V, and with BSC of +3 dB, you can expect peaks of 101 dB @ 1 m or 91 dB at the listening position.  I'd say that's still a bit weak for playback of dynamic content, but not as bad as the tweet.  Presumably it has much better power handling than the tweet, so compression is a lot less likely to be an issue

 

I will get the 4 ohm version instead. Your analysis makes sense. 

 

 

While it would be helpful to see sims of the PR alignment, the bass section looks pretty good on paper at least.  Given that you can only do one 1 PR per driver, I think the frequency response behavior you are seeing is a good thing.  Even if the PRs don't increase sensitivity much, they do improve power handling.  Depending on their thermal limits, you may be able to get away with a lot more power on the bass drivers than you would otherwise.  This can improve your extension in turn if you do a bit of signal shaping with the DSP.  Another thing is that your response likely has a wider transition region between flat and 4th order roll-off, and this will keep group delay in check leading to better sound quality without DSP and less aggressive DSP needed to make it do what you want.  And of course, most usage situations will involve room gain that might make up for whatever gradual roll-off there is before the tune, so you might want to just leave it that way.

 

I have updated the first post with screenshots of the simulation.

 

Regarding PR's increasing power handling, they only increase power handling if the driver is excursion limited before thermally limited right? 

 

No matter what alignment, there's going to be a ton of DSP shaping to get this speaker flat. I want this speaker to be completely flat (not F3) to 40Hz outdoors. On top of that I will add an additional house curve. With my current Reference Mini I have the bass set about 8dB hot, and I'll do the same with this speaker, maybe even more since it has ~10dB more headroom. 

Link to comment
Share on other sites

I thought about the 1/2 wavelength crossover as well, as 1500Hz is pushing things pretty far. I probably will just switch between the two settings. my normal listening volume is very low, like ~60-65dB.  If I need to play louder I'll just switch the setting. On the miniDSP 2x4 HD there are 4 preset configurations that can be switched through a remote, so a settings change is very easy and quick. However, when you presented your analysis with numbers, I never realized how inefficient the tweeter is. I thought the 85.9dB figure was the 1W/1M figure since Scanspeak rates them at 89.2dB @ 2.83V. But I see now I can actually strain the tweeter quite easily. Unfortunately I am very limited on the choice of tweeters because there are very few high quality small format tweeters, and the Scanspeak stands out as having the best sound quality. Maybe I should get the beryllium version? It's got 1dB higher sensitivity!  :D

 

My analysis of the tweeter is based on the 3rd party data that 3ll3d00d posted.  I just looked at the manufacturer's data, and while the response shape shown by the manufacturer is similar, the sensitivities are like 5 dB apart.  That's a big discrepancy, and I can't say who's more right than the other.  If you already have one of these tweeters, maybe you should measure its sensitivity and impedance yourself.  I think the sensitivity of the driver at all frequencies is a pretty crucial point with this design.  Either way, the dip in the 2-5 kHz range is a bit of a drag.  I bet a lot of people cross it much higher.

 

Admittedly, I know very little about the merits of different dome tweeters as I prefer a good waveguide + CD.  Too bad I don't know of any that would fit that form factor, and anyway, it's hard to really do a good horn crossover with 1/4 wavelength separation.  For that matter, it's hard to get the 1/4 wavelength separation in the crossover with any speaker.  Unless you do something like Danley with everything in one large horn or use a coaxial, you are going to have to make some serious compromises.  Small format and low extension in a driver tend to correlate with low sensitivity, so this may be a compromise you have to accept to achieve your other goals.  In the end you decide what's best.  That's the best part about DIY.

 

At least with an active XO and multiple presets, you can experiment with XOs and figure out where that point source magic reveals itself (if it does at all).  You know, 1/4 wave length isn't true point source, it's just a lot closer to point source than 1/2 wave length.

 

I have updated the first post with screenshots of the simulation.

 

Regarding PR's increasing power handling, they only increase power handling if the driver is excursion limited before thermally limited right? 

 

No matter what alignment, there's going to be a ton of DSP shaping to get this speaker flat. I want this speaker to be completely flat (not F3) to 40Hz outdoors. On top of that I will add an additional house curve. With my current Reference Mini I have the bass set about 8dB hot, and I'll do the same with this speaker, maybe even more since it has ~10dB more headroom. 

 

Wow, you were definitely serious about the lack of gain from the PRs.  To be honest, the frequency response looks like a sealed box with a big leak.  It's not clear that you're gaining anything over just using the pair of woofers sealed.  Indeed, the extra excursion control doesn't seem to be needed here unless you are willing to step up again in amp power.  I'd say you reconsider the 4 sealed woofers and double the amp, but if you have to switch to less suitable drivers, I guess that kinda breaks the deal.

 

And yes, you do still have to worry about thermal limits.  If you're going to turn on heavy bass music, turn it up "as loud as it will go", and then let it play all night, you're going to want to have some kind of long term thermal limiting.  Or else you'll want to keep the amps real small enough to avoid failure in a worst case thermal scenario.

 

Also, your plots say "2 parallel" and "at 160W".  I thought you were going to put 160 W on each, or am I just confused by the labelling here?

 

Anyway, good luck.  It'll be interesting to hear how this turns out.

Link to comment
Share on other sites

My analysis of the tweeter is based on the 3rd party data that 3ll3d00d posted.  I just looked at the manufacturer's data, and while the response shape shown by the manufacturer is similar, the sensitivities are like 5 dB apart.  That's a big discrepancy, and I can't say who's more right than the other.  If you already have one of these tweeters, maybe you should measure its sensitivity and impedance yourself.  I think the sensitivity of the driver at all frequencies is a pretty crucial point with this design.  Either way, the dip in the 2-5 kHz range is a bit of a drag.  I bet a lot of people cross it much higher.

 

Admittedly, I know very little about the merits of different dome tweeters as I prefer a good waveguide + CD.  Too bad I don't know of any that would fit that form factor, and anyway, it's hard to really do a good horn crossover with 1/4 wavelength separation.  For that matter, it's hard to get the 1/4 wavelength separation in the crossover with any speaker.  Unless you do something like Danley with everything in one large horn or use a coaxial, you are going to have to make some serious compromises.  Small format and low extension in a driver tend to correlate with low sensitivity, so this may be a compromise you have to accept to achieve your other goals.  In the end you decide what's best.  That's the best part about DIY.

 

At least with an active XO and multiple presets, you can experiment with XOs and figure out where that point source magic reveals itself (if it does at all).  You know, 1/4 wave length isn't true point source, it's just a lot closer to point source than 1/2 wave length.

 

 

I agree this is a large discrepancy. I don't know how to measure the sensitivity of a speaker. I have no way of measuring the voltage from the amplifier. This is not really designed for high SPL, so I'm willing to trade a bit of max volume for cleaner sounding highs. I think for mono use (when I take this on trips and stuff), the crossover will be 1/2 wavelength at 3000Hz. I don't really need the benefits of 1/4 wavelength in mono mode. With only one, the speaker has to work harder as well.  In stereo use it may or may not be at 1500Hz, depending on the audibility of the benefits/drawbacks of 1/4 wavelength crossover. 

 

It is unfortunately Hoffman's Iron Law is applying to tweeters as well... A CD + waveguide would be sweet, but to put it in some perspective of what I'm dealing with, the darn SEOS 12 waveguide itself is nearly as big than this entire speaker! When designing a speaker this small, I'm surprised I'm not facing more compromises! 

 

What about adding a second tweeter? I can low pass it at 5000Hz or so to avoid comb filtering, and I'll get 6dB more output in the lower treble region. Expensive way though. Need an extra amp and miniDSP. Originally I planned on not having a budget limitation for this project, but this has gotten incredibly expensive (and fun). I can build like a dozen SI HT18's with the same money if they were still available.  :o

 

 

Wow, you were definitely serious about the lack of gain from the PRs.  To be honest, the frequency response looks like a sealed box with a big leak.  It's not clear that you're gaining anything over just using the pair of woofers sealed.  Indeed, the extra excursion control doesn't seem to be needed here unless you are willing to step up again in amp power.  I'd say you reconsider the 4 sealed woofers and double the amp, but if you have to switch to less suitable drivers, I guess that kinda breaks the deal.

 

And yes, you do still have to worry about thermal limits.  If you're going to turn on heavy bass music, turn it up "as loud as it will go", and then let it play all night, you're going to want to have some kind of long term thermal limiting.  Or else you'll want to keep the amps real small enough to avoid failure in a worst case thermal scenario.

 

Also, your plots say "2 parallel" and "at 160W".  I thought you were going to put 160 W on each, or am I just confused by the labelling here?

 

Anyway, good luck.  It'll be interesting to hear how this turns out.

 

Yeah no kidding on the PR efficiency. It is really just for distortion reduction and excursion control. The Vas of the drivers are too big in comparison to the enclosure size for the PR to have much, if any gains. An interesting note is how wide the bandwidth in which the PR provides excursion control. The excursion graph for the 2 L16 sealed in 5L is essentially a flat line at 6.2mm below 100Hz. Usually it is over a very narrow range near tuning, but in this case the PR is reducing excursion even an octave above tuning. I don't trust these drivers to stay clean at xmax in when playing ~40Hz stuff in such a small box. Looking at data-bass's measurements, most woofer seem to skyrocket in distortion well before hitting xmax when the frequency got very low, say <20Hz. Since this woofer is like 1/12th the size of a 18'', I expect the same phenomenon to happen at a much higher frequency well above 40Hz, and this is where the PR comes in to really reduce distortion and excursion.

 

Here's the simulation with 4 sealed with double the power with the same drivers compared to 2 woofers + PR. "2 parallel" at 160W means 160W between 2 woofers. You'll see in this graph it is 4 parallel with 320W, same 80W per driver, which is the max power rating on the driver. As you can see, there is no output advantage until 80Hz, since the doubling of drivers and power was negated by reducing box volume by half. There is no real advantage until ~110Hz, and I don't really "need" extra output in the 100-150Hz nearly as much as output below <100Hz. Would be nice, but the extra cost, electronics, and power requirement is too much. In reality, with this much power, I actually have to worry about thermals. There is next to no air inside the speaker to dissipate the heat off the voice coils, so realistically I can't put 320W in, more like ~200W to avoid cooking the woofer. Also, 4 sealed will require lesser drivers to avoid increasing box size, reducing the performance even more. 

post-1515-0-81773700-1470579326_thumb.png

post-1515-0-90631000-1470579328_thumb.png

post-1515-0-75237600-1470579330_thumb.png

Link to comment
Share on other sites

I agree this is a large discrepancy. I don't know how to measure the sensitivity of a speaker. I have no way of measuring the voltage from the amplifier.

 

Do you have a audio measurement system and know how to capture an anechoic or at least a gated response?  From there, you just need a digital multimeter with true RMS capability to measure the voltage you used for the sweep.  The best way is with a sine tone generator that's at the same level as the sweep.

 

It is unfortunately Hoffman's Iron Law is applying to tweeters as well... A CD + waveguide would be sweet, but to put it in some perspective of what I'm dealing with, the darn SEOS 12 waveguide itself is nearly as big than this entire speaker! When designing a speaker this small, I'm surprised I'm not facing more compromises! 

 

What about adding a second tweeter? I can low pass it at 5000Hz or so to avoid comb filtering, and I'll get 6dB more output in the lower treble region. Expensive way though. Need an extra amp and miniDSP. Originally I planned on not having a budget limitation for this project, but this has gotten incredibly expensive (and fun). I can build like a dozen SI HT18's with the same money if they were still available.  :o

 

DIYSG sells some horns that are a lot smaller, but I think you'd still have to upsize the speaker.

 

If you can get that second tweeter within 1/4 wavelength and can tolerate the cost and what not, that looks like a viable route.  But that adds a lot more complexity and expense.

 

Yeah no kidding on the PR efficiency. It is really just for distortion reduction and excursion control. The Vas of the drivers are too big in comparison to the enclosure size for the PR to have much, if any gains. An interesting note is how wide the bandwidth in which the PR provides excursion control. The excursion graph for the 2 L16 sealed in 5L is essentially a flat line at 6.2mm below 100Hz. Usually it is over a very narrow range near tuning, but in this case the PR is reducing excursion even an octave above tuning. I don't trust these drivers to stay clean at xmax in when playing ~40Hz stuff in such a small box. Looking at data-bass's measurements, most woofer seem to skyrocket in distortion well before hitting xmax when the frequency got very low, say <20Hz. Since this woofer is like 1/12th the size of a 18'', I expect the same phenomenon to happen at a much higher frequency well above 40Hz, and this is where the PR comes in to really reduce distortion and excursion.

 

Here's the simulation with 4 sealed with double the power with the same drivers compared to 2 woofers + PR. "2 parallel" at 160W means 160W between 2 woofers. You'll see in this graph it is 4 parallel with 320W, same 80W per driver, which is the max power rating on the driver. As you can see, there is no output advantage until 80Hz, since the doubling of drivers and power was negated by reducing box volume by half. There is no real advantage until ~110Hz, and I don't really "need" extra output in the 100-150Hz nearly as much as output below <100Hz. Would be nice, but the extra cost, electronics, and power requirement is too much. In reality, with this much power, I actually have to worry about thermals. There is next to no air inside the speaker to dissipate the heat off the voice coils, so realistically I can't put 320W in, more like ~200W to avoid cooking the woofer. Also, 4 sealed will require lesser drivers to avoid increasing box size, reducing the performance even more. 

 

I see.  The box is already so small that there's nothing to be gained in the low end by adding more drivers.  What does just the two sealed look like?  About the same in the low end?  If so, that might be the option I'd lean toward. 

 

The thing about the PR design is that it has a significant drawback, and that is the unloading.  Especially being that you want to run "flat with a house curve" down to 40 Hz you are going to have a serious problem with time domain ringing in your final response.  It's gonna go "boom boom boom" at 40 Hz or wherever your curve makes its transition.  I don't think the FIR filters can help you with correcting the excess phase down there because you don't have enough taps on the 2x4HD.  Indeed, I don't think "dead flat to 40 Hz" is realistic regardless unless you want the time domain to be trashed.  If you did a sealed system and did Linkwitz Transform (all IIRs) to reshape it to a Q=0.707 2nd order roll-off somewhere a bit lower than 40 Hz, you could maybe get close.

 

If you were willing to double the amps into the PR system, you could eek out some extra performance, but you would have to pay more attention to thermal issues.  The same doubling of amp could be done with the two woofer sealed system to eek out some more performance as well.  But again, you'd probably want some long term limiting to protect against thermal failure.  Then again, you might want some of that anyway.

 

I guess the other alternative would be to try to find a woofer with similar compliance and displacement but an even stronger motor.

Link to comment
Share on other sites

The thing about the PR design is that it has a significant drawback, and that is the unloading.  Especially being that you want to run "flat with a house curve" down to 40 Hz you are going to have a serious problem with time domain ringing in your final response.  It's gonna go "boom boom boom" at 40 Hz or wherever your curve makes its transition.  I don't think the FIR filters can help you with correcting the excess phase down there because you don't have enough taps on the 2x4HD.  Indeed, I don't think "dead flat to 40 Hz" is realistic regardless unless you want the time domain to be trashed.  If you did a sealed system and did Linkwitz Transform (all IIRs) to reshape it to a Q=0.707 2nd order roll-off somewhere a bit lower than 40 Hz, you could maybe get close.

 

I think you can still do the PR just fine, but SME's right, you have to account for time domain issues.

 

Group delay is related to the curvature of your final SPL. This means that it's treating say, 35Hz very differently from 40Hz. Intuitively, ask yourself "How is it going to tell the difference between these two frequencies, and how is it going to attenuate one but not the other?" This is the exact same concept that gives rise to the Heisenberg Uncertainty Principle. Knowing frequency with 100% precision requires infinite amount of time. Makes sense right? Frequency is cycles PER UNIT TIME.

 

Now your system sees an input curve. As voltage rises, it doesn't know if it's going to be a 40Hz sine wave, a gaussian function (bell curve), a 39Hz sine wave, or any other infinite number of shapes until your 40Hz sine has completed it's full cycle. That's 1/40th of a second, or 25 milliseconds. Let's say our sine wave stopped at 25.6ms. It was a 39Hz sine wave. A perfect brickwall 40Hz filter would eliminate it, but it wouldn't be able to tell the difference until that last 600 microseconds. Such a slope is so steep that 40Hz frequencies would be delayed infinitely until it could be certain there was no 39.99999...Hz content, then the 40Hz tone would pass through unadulterated and any <40Hz tones would be eliminated.

 

Real systems aren't perfect brickwalls, but attenuate by a certain amount. To attenuate power by 16x (12db) for every octave, we've created a second order rolloff. Sealed systems eventually progress to this steepness. A ported system will eventually have a 24db/octave slope. These slopes (and therefore delays) will stack with whatever active processing you add to it.

 

Option 1

Eq flat to 40hz with only a 4th order rolloff. This means no high pass. Loud notes below 35Hz will fart, but you'll still only have your 4th order rolloff and time domain behavior will be quite acceptable.

 

Option 2

Soften the corner of your final slope. Think Bessel filter, Zaph's quasi-EBS/QB3 alignment, Ricci's 25Hz ported BMS 18N862 system etc. i.e., a gradual increase in slope the further you drop in frequency. Response will be flat to say ~60Hz, slowly roll off to maybe -6dB at ~40Hz, then your high pass filter drops frequency response below tuning. Now all frequencies from 30-60Hz are delayed, but no frequency is delayed by more than one cycle, so time domain behavior will be acceptable. You can keep your high pass filter to eliminate below-tuning farts, you won't hear any ringing at the low corner, but 40Hz output is reduced.

 

Option 3

Eq flat to 40Hz and put in your 40Hz high pass filters. No farts, no reduced 40Hz output, but very high group delay. Maybe as much as 1.5 cycles, maybe more.

 

Now let's throw another wrench in the system. Room effects. Room effects will add all sorts of peaks and valleys everywhere in the low end. Even if Option 3 had barely acceptable group delay before you put the speaker in the room, it probably won't afterward. That's why a lot of speaker designers opt for Option 2 or Option 1. You'll have to decide where you want the tradeoffs to land.

Link to comment
Share on other sites

Do you have a audio measurement system and know how to capture an anechoic or at least a gated response?  From there, you just need a digital multimeter with true RMS capability to measure the voltage you used for the sweep.  The best way is with a sine tone generator that's at the same level as the sweep.

 

I do have audio measurement capabilities and know how to do a gated measurement. I don't have access to an anechoic chamber anymore. I don't have a multimeter, I may get one. However wouldn't the only purpose of that to satisfy the curiosity of what the sensitivity of the tweeter would be?

 

I see.  The box is already so small that there's nothing to be gained in the low end by adding more drivers.  What does just the two sealed look like?  About the same in the low end?  If so, that might be the option I'd lean toward. 

 

The thing about the PR design is that it has a significant drawback, and that is the unloading.  Especially being that you want to run "flat with a house curve" down to 40 Hz you are going to have a serious problem with time domain ringing in your final response.  It's gonna go "boom boom boom" at 40 Hz or wherever your curve makes its transition.  I don't think the FIR filters can help you with correcting the excess phase down there because you don't have enough taps on the 2x4HD.  Indeed, I don't think "dead flat to 40 Hz" is realistic regardless unless you want the time domain to be trashed.  If you did a sealed system and did Linkwitz Transform (all IIRs) to reshape it to a Q=0.707 2nd order roll-off somewhere a bit lower than 40 Hz, you could maybe get close.

 

If you were willing to double the amps into the PR system, you could eek out some extra performance, but you would have to pay more attention to thermal issues.  The same doubling of amp could be done with the two woofer sealed system to eek out some more performance as well.  But again, you'd probably want some long term limiting to protect against thermal failure.  Then again, you might want some of that anyway.

 

I guess the other alternative would be to try to find a woofer with similar compliance and displacement but an even stronger motor.

 

Yeah the box is ridiculously small. It is around 10x smaller than the recommended enclosure volume for these woofers. Because of that the graph for 2 L16 sealed looks exactly the same as the 2 L16 with PR except the roll off below tuning. The excursion graph is essentially a flat line at 6mm from 90Hz and below, which is the xmax of the driver. I worry about the distortion from running the driver at xmax. 

 

I cannot double the amps into the PR system. The amp I'm using is already the most powerful small amplifier I can find. Even if I can, the driver cannot safely handle more than 80W. 

 

This woofer is about as good as it gets for bass. There are woofers more capable for bass, like the Tang Band W5-1138SMF, but the bass and lower midrange sound quality would be far inferior. There are woofers with stronger motors, such as the Scanspeak 15WU, but at $317 it is over 3x the price and the improvement is likely not big.

 

I'm not sure if I want to add a thermal protection system. I'm not aware of any commercial product with that capability, and I'm not designing this for a commercial product. But for curiosity sakes, what were you thinking of using to add a thermal protection system?

 

 

I think you can still do the PR just fine, but SME's right, you have to account for time domain issues.

 

Group delay is related to the curvature of your final SPL. This means that it's treating say, 35Hz very differently from 40Hz. Intuitively, ask yourself "How is it going to tell the difference between these two frequencies, and how is it going to attenuate one but not the other?" This is the exact same concept that gives rise to the Heisenberg Uncertainty Principle. Knowing frequency with 100% precision requires infinite amount of time. Makes sense right? Frequency is cycles PER UNIT TIME.

 

Now your system sees an input curve. As voltage rises, it doesn't know if it's going to be a 40Hz sine wave, a gaussian function (bell curve), a 39Hz sine wave, or any other infinite number of shapes until your 40Hz sine has completed it's full cycle. That's 1/40th of a second, or 25 milliseconds. Let's say our sine wave stopped at 25.6ms. It was a 39Hz sine wave. A perfect brickwall 40Hz filter would eliminate it, but it wouldn't be able to tell the difference until that last 600 microseconds. Such a slope is so steep that 40Hz frequencies would be delayed infinitely until it could be certain there was no 39.99999...Hz content, then the 40Hz tone would pass through unadulterated and any <40Hz tones would be eliminated.

 

Real systems aren't perfect brickwalls, but attenuate by a certain amount. To attenuate power by 16x (12db) for every octave, we've created a second order rolloff. Sealed systems eventually progress to this steepness. A ported system will eventually have a 24db/octave slope. These slopes (and therefore delays) will stack with whatever active processing you add to it.

 

Option 1

Eq flat to 40hz with only a 4th order rolloff. This means no high pass. Loud notes below 35Hz will fart, but you'll still only have your 4th order rolloff and time domain behavior will be quite acceptable.

 

Option 2

Soften the corner of your final slope. Think Bessel filter, Zaph's quasi-EBS/QB3 alignment, Ricci's 25Hz ported BMS 18N862 system etc. i.e., a gradual increase in slope the further you drop in frequency. Response will be flat to say ~60Hz, slowly roll off to maybe -6dB at ~40Hz, then your high pass filter drops frequency response below tuning. Now all frequencies from 30-60Hz are delayed, but no frequency is delayed by more than one cycle, so time domain behavior will be acceptable. You can keep your high pass filter to eliminate below-tuning farts, you won't hear any ringing at the low corner, but 40Hz output is reduced.

 

Option 3

Eq flat to 40Hz and put in your 40Hz high pass filters. No farts, no reduced 40Hz output, but very high group delay. Maybe as much as 1.5 cycles, maybe more.

 

Now let's throw another wrench in the system. Room effects. Room effects will add all sorts of peaks and valleys everywhere in the low end. Even if Option 3 had barely acceptable group delay before you put the speaker in the room, it probably won't afterward. That's why a lot of speaker designers opt for Option 2 or Option 1. You'll have to decide where you want the tradeoffs to land.

 

Urghhh group delay, I never thought that would be an issue. Although I wonder what horrible group delay sounds like. 

 
I definitely want to boost the bass to at least flat to 40Hz and more. Could my Option 4 work?
 
What if I EQ'd flat to 40Hz and set the high pass at 35Hz? Would that alleviate the group delay problem a bit? I did this on my current speaker where I boosted it flat to 45Hz and a high pass was set at 40Hz.
Link to comment
Share on other sites

 

I do have audio measurement capabilities and know how to do a gated measurement. I don't have access to an anechoic chamber anymore. I don't have a multimeter, I may get one. However wouldn't the only purpose of that to satisfy the curiosity of what the sensitivity of the tweeter would be?

 

 

Yeah the box is ridiculously small. It is around 10x smaller than the recommended enclosure volume for these woofers. Because of that the graph for 2 L16 sealed looks exactly the same as the 2 L16 with PR except the roll off below tuning. The excursion graph is essentially a flat line at 6mm from 90Hz and below, which is the xmax of the driver. I worry about the distortion from running the driver at xmax. 

 

I cannot double the amps into the PR system. The amp I'm using is already the most powerful small amplifier I can find. Even if I can, the driver cannot safely handle more than 80W. 

 

This woofer is about as good as it gets for bass. There are woofers more capable for bass, like the Tang Band W5-1138SMF, but the bass and lower midrange sound quality would be far inferior. There are woofers with stronger motors, such as the Scanspeak 15WU, but at $317 it is over 3x the price and the improvement is likely not big.

 

I'm not sure if I want to add a thermal protection system. I'm not aware of any commercial product with that capability, and I'm not designing this for a commercial product. But for curiosity sakes, what were you thinking of using to add a thermal protection system?

 

 

 

Urghhh group delay, I never thought that would be an issue. Although I wonder what horrible group delay sounds like. 

 
I definitely want to boost the bass to at least flat to 40Hz and more. Could my Option 4 work?
 
What if I EQ'd flat to 40Hz and set the high pass at 35Hz? Would that alleviate the group delay problem a bit? I did this on my current speaker where I boosted it flat to 45Hz and a high pass was set at 40Hz.

 

 

"It is around 10x smaller than the recommended enclosure volume for these woofers"

 

You might want to find a driver with more BL product and a lower Q, otherwise its going to be tricky to EQ the thing. The other thing you could do is drop a size down and go with a 4" or 5" woofer. Your PR is already smaller than your driver which is unconventional to start with.

Link to comment
Share on other sites

I do have audio measurement capabilities and know how to do a gated measurement. I don't have access to an anechoic chamber anymore. I don't have a multimeter, I may get one. However wouldn't the only purpose of that to satisfy the curiosity of what the sensitivity of the tweeter would be?

 

Yeah.  I think the tweeter sensitivity is a pretty fundamental consideration given that it's similar to the sensitivity of the other drivers.  It could seriously hold back the design.  If you run with a lot of house curve, then maybe it won't.  I don't know.  You'll have to decide based on your preferences and whatever data you can get on its sensitivity.

 

Yeah the box is ridiculously small. It is around 10x smaller than the recommended enclosure volume for these woofers. Because of that the graph for 2 L16 sealed looks exactly the same as the 2 L16 with PR except the roll off below tuning. The excursion graph is essentially a flat line at 6mm from 90Hz and below, which is the xmax of the driver. I worry about the distortion from running the driver at xmax. 

 

I cannot double the amps into the PR system. The amp I'm using is already the most powerful small amplifier I can find. Even if I can, the driver cannot safely handle more than 80W. 

 

This woofer is about as good as it gets for bass. There are woofers more capable for bass, like the Tang Band W5-1138SMF, but the bass and lower midrange sound quality would be far inferior. There are woofers with stronger motors, such as the Scanspeak 15WU, but at $317 it is over 3x the price and the improvement is likely not big.

 

I'm not sure if I want to add a thermal protection system. I'm not aware of any commercial product with that capability, and I'm not designing this for a commercial product. But for curiosity sakes, what were you thinking of using to add a thermal protection system?

 

The PR can give you lower distortion as long as things work out as you hope.  But I see it as being too risky vs. the benefits.  OTOH, you won't know unless you try.

 

By thermal protection system, I just mean some kind of circuit or DSP logic that limits long term output to keep things from overheating.  Most amps have some kind of thermal protection system, even if it just switches off when it overheats.  Nicer amps will simply throttle power back unless the temperature continues to rise for some reason (like poor ventilation).  Amp thermal protection systems are also more likely to act on direct temperature measurements of the components.  With speakers, the design resembles a compression and limiter circuit with a long time constant.  I don't know how feasible it would be to implement such logic with the MiniDSP HD 2x4, and I guess it would require even more hardware to deal with the fact that you'll be using that MiniDSP compressor for controlling bass peaks as well.

 

I have no idea which commercial products use thermal protection systems, but I see no reason why such a system is not present in at least some designs.  It's not likely to get advertised as it is likely to come with negative connotations.  (I.e., "built in thermal throttling for driver protection" doesn't really seem to flatter the product, now does it?)  Nevertheless, if for some reason such a system uses amps that are strong enough to overheat any drivers, it will almost certainly have thermal protection because customers expect these things to be bullet proof.

 

Urghhh group delay, I never thought that would be an issue. Although I wonder what horrible group delay sounds like.

 

The group delay causes those frequencies to arrive late in the attack and linger for longer in the decay.  It will also tend to make transients emphasize those frequencies because they play for longer (loudness correlates to duration as well as SPL) and because they dominate the decay spectrum.  This can make kicks and things like that rather one-note-ish.  These characteristics are at least partly responsible for the negative reputation that ported systems tend to have for sound quality vs. sealed.  On the other hand, some people like "fuller, heavier sound" that results.  A sharper knee with a shorter transition between flat and N-th order roll-off (where N is at least 6 for your system) will have more of this sound characteristic.

 

Of course, room acoustic problems like modal resonances can do the same thing to the sound and are often worse.  So it may not make as much difference if room problems dominate.  And in this case, your in-room response will be a long way from flat anyway.  Supposing that your room had excellent bass decay in the lowest frequencies (rare) or that you added room correction EQ, the high group delay is more likely to stand out.  The same will be true when you are outdoors.

 

I definitely want to boost the bass to at least flat to 40Hz and more. Could my Option 4 work?

 
What if I EQ'd flat to 40Hz and set the high pass at 35Hz? Would that alleviate the group delay problem a bit? I did this on my current speaker where I boosted it flat to 45Hz and a high pass was set at 40Hz.

 

 

Well, if you boost bass to be flat to 40 Hz and *then* apply a 35 Hz high pass, it won't be flat at 40 Hz any more.  If you go back and boost the 40 Hz to flat with the 35 Hz high pass filter, then you are basically sharpening the knee a bit more and increasing group delay.  How you get there isn't important.  It's the shape of the final response that matters.

Link to comment
Share on other sites

"It is around 10x smaller than the recommended enclosure volume for these woofers"

 

You might want to find a driver with more BL product and a lower Q, otherwise its going to be tricky to EQ the thing. The other thing you could do is drop a size down and go with a 4" or 5" woofer. Your PR is already smaller than your driver which is unconventional to start with.

 

This woofer has the lowest Q (0.33) and highest motor strength that I can find in drivers with non insane prices. I believe the lowest Q woofer is Scanspeak 15WU at 0.23 and around 50% more BL^2/Re, but it is $634 for 2 of them.

 

You're right, it will be tricky to EQ. However, If I drop a size down I lose out on Sd and power handling, both will reduce maximum clean bass output. 

 

The good thing is I'm probably gonna go with a DIY PR with twice the excursion as the active woofers so the PR's don't become a limiting factor. 

 

Yeah.  I think the tweeter sensitivity is a pretty fundamental consideration given that it's similar to the sensitivity of the other drivers.  It could seriously hold back the design.  If you run with a lot of house curve, then maybe it won't.  I don't know.  You'll have to decide based on your preferences and whatever data you can get on its sensitivity.

 

I think I'm going to leave the tweeter selection the way it is. I understand where you're coming from for HT use and extreme sensitivity compression drivers. However I just don't see this being a problem because if I use a 3000Hz crossover, I don't see myself playing the speaker loud enough for the tweeter to struggle. Given the average energy spectrum of sound in music, the mids will be playing ~5-10dB louder than the tweeter, so the mid would likely to be the first to struggle. At that point we're talking about ~95dB playback just the mids and highs. If the song has proper bass, the average SPL is probably closer to 110dB, which is really loud, much louder I'd ever listen to normally, especially on a small speaker. 

 

By thermal protection system, I just mean some kind of circuit or DSP logic that limits long term output to keep things from overheating.  Most amps have some kind of thermal protection system, even if it just switches off when it overheats.  Nicer amps will simply throttle power back unless the temperature continues to rise for some reason (like poor ventilation).  Amp thermal protection systems are also more likely to act on direct temperature measurements of the components.  With speakers, the design resembles a compression and limiter circuit with a long time constant.  I don't know how feasible it would be to implement such logic with the MiniDSP HD 2x4, and I guess it would require even more hardware to deal with the fact that you'll be using that MiniDSP compressor for controlling bass peaks as well.

 

I have no idea which commercial products use thermal protection systems, but I see no reason why such a system is not present in at least some designs.  It's not likely to get advertised as it is likely to come with negative connotations.  (I.e., "built in thermal throttling for driver protection" doesn't really seem to flatter the product, now does it?)  Nevertheless, if for some reason such a system uses amps that are strong enough to overheat any drivers, it will almost certainly have thermal protection because customers expect these things to be bullet proof.

 

Oh I didn't mean commercial speakers not having thermal protection. I'm sure many speakers have them so they're bulletproof. 

 

What I meant is I don't know of a commercial product that'll help me implement this. MiniDSP does not do this. 

 

But, I did think of something interesting. I recently realized just how small the PCBs for class D amplifier boards and power supplies are when they're out of the casing. This makes me think I can just put them inside the speaker and reduce a ton of clutter and cable mess. The amps have thermal protection, so when it senses the air inside the speaker is getting too warm, it probably means the woofer's voice coil is crying for help too, so the thermal protection will act to protect the amp and the drivers themselves.

 

The group delay causes those frequencies to arrive late in the attack and linger for longer in the decay.  It will also tend to make transients emphasize those frequencies because they play for longer (loudness correlates to duration as well as SPL) and because they dominate the decay spectrum.  This can make kicks and things like that rather one-note-ish.  These characteristics are at least partly responsible for the negative reputation that ported systems tend to have for sound quality vs. sealed.  On the other hand, some people like "fuller, heavier sound" that results.  A sharper knee with a shorter transition between flat and N-th order roll-off (where N is at least 6 for your system) will have more of this sound characteristic.

 

Of course, room acoustic problems like modal resonances can do the same thing to the sound and are often worse.  So it may not make as much difference if room problems dominate.  And in this case, your in-room response will be a long way from flat anyway.  Supposing that your room had excellent bass decay in the lowest frequencies (rare) or that you added room correction EQ, the high group delay is more likely to stand out.  The same will be true when you are outdoors.

 

 

Well, if you boost bass to be flat to 40 Hz and *then* apply a 35 Hz high pass, it won't be flat at 40 Hz any more.  If you go back and boost the 40 Hz to flat with the 35 Hz high pass filter, then you are basically sharpening the knee a bit more and increasing group delay.  How you get there isn't important.  It's the shape of the final response that matters.

 

I'm gonna have to research more about this as I'm still not very clear of the relationship between final response and group delay. I wanna apply a ton of bass boost, and flat to 40Hz is just the bare minimum. I want at least 5-6dB hot from flat.

Link to comment
Share on other sites

I'm gonna have to research more about this as I'm still not very clear of the relationship between final response and group delay. I wanna apply a ton of bass boost, and flat to 40Hz is just the bare minimum. I want at least 5-6dB hot from flat.

 

As far as group delay, it's not the amount of bass boost that affects things, it's the shape of the final response.  More specifically, it's a question of how much the slope of the response is forced to change and how narrow of a frequency region does this transition occur over.  The more abrupt the transition, the higher the group delay.  (This is a rough explanation, but I think it should still be intuitively useful.)

 

Your problem is that you want a response that's flat or even boosted to some frequency very close to the PR tune.  However, you *need* a response that's rolled off by a certain amount by a certain frequency to avoid over excursion and distortion.  Between those points, your system response slope has to transition from flat to whatever the roll-off is.  The key thing is that with a sealed system, that roll-off is only 2nd order, but with the PR system it will be at least 6th order.  The PR system *requires* this steeper roll-off to avoid unwanted distortion and protect the drivers. This means that the response will have a more abrupt transition and thus higher group delay.  The only way to avoid this problem with the PR system is to find a way to tune even lower so that you can shape the response for a smoother transition.  Or else, don't go as low or as flat.

 

If your sim allows you to apply arbitrary EQ filters, you can use those to experiment with signal shaping to see what it does to the group delay and /or phase.  The higher the group delay, the more phase shift, and the more phase shift, the more temporal distortion induced by the roll-off.

Link to comment
Share on other sites

Hey LowerFe...Have you looked at any of the Aurum Cantus drivers? They have a couple which may work better than the SEAS for this type of project.

 

Any chance you would increase the enclosure size to 6.5x6.5 or maybe 7x7? That opens up a lot of more driver choices with a bit more SD. Also can you really fit the 3.5" mid and tweet in between the two drivers on a 14x6 baffle?

Link to comment
Share on other sites

Hey LowerFe...Have you looked at any of the Aurum Cantus drivers? They have a couple which may work better than the SEAS for this type of project.

 

Any chance you would increase the enclosure size to 6.5x6.5 or maybe 7x7? That opens up a lot of more driver choices with a bit more SD. Also can you really fit the 3.5" mid and tweet in between the two drivers on a 14x6 baffle?

 

Been very busy lately, so I haven't been able to spend time to digest SME's suggestions and give a proper response, and won't be able to for a few more days. I also have some substantial improvements in mind, but don't have time to flesh out the details yet. But I can respond to yours quickly.

 

Yes, I actually did look at them, specifically the AC120, and the specs are unreal. However, when I asked opinions of them, multiple people chimed in to say the woofers measure *extremely* off spec, like Qts of 0.55 instead of 0.18. 

 

14x6x6 is already bigger than I want and at the edge of what I'd consider to be a small speaker. Increasing it further will no longer make it a small speaker and becomes dangerously close to the size of a bookshelf speaker. I'd say 80% of the fun is the challenge of keeping it small and the satisfaction of being able to claim "mine is smaller and sounds way better". Unfortunately at 14x6x6 it is already bigger than most sound docks. I absolutely want it noticeably smaller than the Sonos Play 5, by far the best sounding lifestyle speaker on the market, and be able to claim "mine is smaller and sounds way better". You built the M.A.U.L. to be the coolest kid on the block, I'm sure you'd understand where I'm coming from  :D The value factor on this project is a solid 0, but the fun factor is 12/10.

 

And yes, I can fit the 2 woofers, mid and tweeter on the 14x6 baffle if I machine off the frame for every driver. I'm trying extremely hard to keep the box size small! Without doing this, the dimensions would have to increase by 2'' in width and 1'' in height. 

Link to comment
Share on other sites

Yes, I actually did look at them, specifically the AC120, and the specs are unreal. However, when I asked opinions of them, multiple people chimed in to say the woofers measure *extremely* off spec, like Qts of 0.55 instead of 0.18. 

 

  :D The value factor on this project is a solid 0, but the fun factor is 12/10.

 

That's the exact driver I was thinking of and the first I've heard of that. I guess those specs are too good to be true. :(

Do you have any links to measurements or discussion? I was thinking about using the AC165 in a compact bookshelf design but not if its a fairy tale driver.

Link to comment
Share on other sites

That's the exact driver I was thinking of and the first I've heard of that. I guess those specs are too good to be true. :(

Do you have any links to measurements or discussion? I was thinking about using the AC165 in a compact bookshelf design but not if its a fairy tale driver.

 

 

Here's the thread

 

http://techtalk.parts-express.com/forum/tech-talk-forum/1294948-anyone-used-the-aurum-cantus-ac120

 

I wanna announce several improvements I plan on incorporating. First, the woofer, since this is data-BASS :P .

 

I found a much better woofer for my application - the little known Wavecor WF152BD05. This is the driver I was hoping for, sacrifice Fs to tighten the suspension and drop the Vas. It has a nice low Qts of 0.29 and a lot of motor strength. The Q in a sealed 5L enclosure for a pair of them is 0.69! In a PR enclosure, it delivers around 2dB more output than the SEAS L16RNX from 40-55Hz with the same input power because the PR's actually boost output. This is very close to an ideal specially designed small subwoofer driver like the Tangband W5-1138SMF that brings around 3dB more. I've attached simulations for a comparison between the Wavecor and the SEAS L16RNX. One person mentioned that the sound quality is extremely good and comparable to a Scanspeak Revelator 15W (!), and behaves very well when pushed very hard. Looks like this is about as good of a woofer for my speaker. Bass, sound quality, non crazy price, check check check. 

 

Next, I'm going to try an internally amplified design. I'm thinking of trying my hands on some electronics and put a 4 channel amplifier and power supply right into the speaker. Both components are small enough to just fit in the speaker, and would massively declutter the need for 2 external amps and 2 power bricks for *each* speaker.

 

http://www.parts-express.com/sure-electronics-aa-ab33182-4x100w-at-4-ohm-class-d-digital-audio-amplifier-board-sta508-(t--320-335

http://hifimediy.com/power-supplies/diy-power-supplies/SMPS300R-28V-110V

 

Unfortunately the amp won't be as powerful, 64W vs 125W per channel (1% THD), but this is close enough to the 70W power rating on the woofers. However, even if the woofers can take the 125W, after power compression the difference is probably around 1dB. So I'll sacrifice 1dB for the massive convenience from 4 less boxes and 4 pairs less wires. One other advantage is that this amp could help to thermally protect the speaker. It has a temperature sensor that'll throttle power if it gets too hot, I believe at >80C. If it gets this hot inside the speaker, the woofer's voice coil is probably straining as well, and it would be good to limit power. While this is not the ideal amp, this is more than enough when I have no experience with DIY electronics. An ideal amp would be a really nice sounding mid powered amp like the TPA3116 for the tweeter and midrange, and a powerful amp like the TDA7498E for the woofers, but that would be too much of a challenge for a first time electronic DIY. Maybe next time.

 

After much thought, it may not be worth chasing the 1/4'' wavelength crossover with the midrange and tweeter. So I want to try something else - an cylindrical midrange chamber. In my last speaker, the midrange chamber is essentially a cube, the least ideal enclosure shape. This time, I will use a stuffed cylindrical midrange chamber to better deal with standing waves and resonances. Because of the cylindrical shape, the midrange and tweeter cannot be in the same chamber as it will take up too much internal volume because there's no way of reclaiming the internal volume behind the tweeter like I can do with wood. This results in an increase in the distance between the mid and tweeter to a point where the crossover frequency will be infeasibly low for the tweeter.  However, I do expect a noticeably cleaner midrange with this change. 

 

Man, I keep thinking this is about as good as it gets, but I keep proving myself wrong. I like it!

post-1515-0-79855800-1471036981_thumb.png

post-1515-0-26584100-1471036984_thumb.png

Link to comment
Share on other sites

...

 

After much thought, it may not be worth chasing the 1/4'' wavelength crossover with the midrange and tweeter. So I want to try something else - an cylindrical midrange chamber. In my last speaker, the midrange chamber is essentially a cube, the least ideal enclosure shape. This time, I will use a stuffed cylindrical midrange chamber to better deal with standing waves and resonances. Because of the cylindrical shape, the midrange and tweeter cannot be in the same chamber as it will take up too much internal volume because there's no way of reclaiming the internal volume behind the tweeter like I can do with wood. This results in an increase in the distance between the mid and tweeter to a point where the crossover frequency will be infeasibly low for the tweeter.  However, I do expect a noticeably cleaner midrange with this change. 

 

...

 

The standing wave resonances of the inside of a cylinder are only slightly less bad than in a cube with all sides equal but still worse than if you just used a box with dissimilar dimensions.  The curved sides don't have any kind of magic standing-wave suppressing properties.  Any shaped interior with have standing waves.  Shapes with symmetry are worse because there you see multiple resonances with different geometry but the same frequency.  Besides, good absorber inside the box will suppress almost all the interior standing wave resonances.

 

Of course, the wood panels and box itself also have standing wave resonances, but these are different and depend more on the plywood thickness, the bracing scheme, and any damping that's present, than the shape.

Link to comment
Share on other sites

The standing wave resonances of the inside of a cylinder are only slightly less bad than in a cube with all sides equal but still worse than if you just used a box with dissimilar dimensions.  The curved sides don't have any kind of magic standing-wave suppressing properties.  Any shaped interior with have standing waves.  Shapes with symmetry are worse because there you see multiple resonances with different geometry but the same frequency.  Besides, good absorber inside the box will suppress almost all the interior standing wave resonances.

 

Of course, the wood panels and box itself also have standing wave resonances, but these are different and depend more on the plywood thickness, the bracing scheme, and any damping that's present, than the shape.

 

 

You're right. I mistakenly remembered that cylindrical midrange enclosures are significantly better than rectangular enclosures. I remember this from the Linkwitz LXmini, but the use of a cylindrical enclosure was for optimal diffraction instead of internal standing waves. I will stick with rectangular midrange enclosures as they can act as a brace and allow a closer midrange and tweeter placement. I might try one with non parallel walls depending on difficulty. 

 

After some thought, it doesn't seem there is a nice way to deal with the group delay problem without serious sacrifices of real world performance. The bass will be significantly boosted, there's no doubt about it for a satisfying listening experience. The use of a steep high pass slightly below tuning is necessary to prevent PR over excursion. On top of that, the compressor is going to increase the group delay as well. So either way the group delay around tuning is going to be bad. Honestly, I probably will enjoy the fat sounding ringing bass like how I enjoy my Logitech Z-5500 (it actually measures extremely well, much better than it's reputation). I don't actually like the sound of most high end subwoofers using heavy duty drivers from the likes of JL, Martin Logan, Paradigm, or DIY like the Dayton UM woofers that people say as "tight" (I unfortunately own two of these kind of subs). I much prefer the thicker sounding bass from subs like JTR or the PA style subs. But this is another (long) story. 

 

On the other hand, I found a much more suitable amplifier. 4 channels, much higher power (100W @ 0.05% THD), integrated power supply, small form factor, low price. Unfortunately, I have to buy a minimum of 10 units when I only need 2. I'm trying to figure out if there is a way around the MOQ. 

 

http://www.aliexpress.com/store/product/4CHANNEL-CLASS-D-SMPS-400W-PRO-AUDIO-AMPLIFIER-MODULE-4CH-OR2CH-OR-2-1CH-SET-EASILY/601158_460608255.html

 

I've also updated the first post with the upgrades.

Link to comment
Share on other sites

Join the conversation

You can post now and register later. If you have an account, sign in now to post with your account.

Guest
Reply to this topic...

×   Pasted as rich text.   Paste as plain text instead

  Only 75 emoji are allowed.

×   Your link has been automatically embedded.   Display as a link instead

×   Your previous content has been restored.   Clear editor

×   You cannot paste images directly. Upload or insert images from URL.

Loading...
×
×
  • Create New...