Jump to content

The ultimate small speaker - final design peer review thread


lowerFE

Recommended Posts

Another update: I found a suitable tweeter to address the low sensitivity issue of the Scanspeak D3004/6020 tweeter I was originally planning on using. 

 

TW030WA03_04-250ps.jpg

 

I will be using the Wavecor TW030WA04. This is a 1.2'' tweeter flangeless 4 Ohm tweeter. This tweeter has a sensitivity of 94dB @ 2.83V, which is 7-8dB higher than the Scanspeak. Wavecor's specs are known to be very accurate, so I believe the 94dB sensitivity is real. The 57% larger Sd also means the tweeter will likely to play a lot lower with less distortion at high volumes. Since this is flangeless, I don't have to butcher an expensive tweeter's frame to reduce size, and it will make it easier for me to get very tight CTC spacing and a much better chance of getting the 1500Hz crossover for the 1/4'' wavelength crossover to work. While a few users have commented on the excellent sound quality of this tweeter, I don't know if it'll sound as good as the Scanspeak as they are the benchmark small format tweeter. I do expect it to sound very close, and the benefit of the vastly higher sensitivity for dynamics and higher peak SPL should outweigh the slight sound quality advantage of the Scanspeak. 

 

There is a non ferro fluid version of this tweeter, and the sensitivity is 1dB higher. Should I consider it?

Link to comment
Share on other sites

More updates:

 

Let's get the bad news out of the way first. It looks like I won't be able to fit this on a 14'' x 6'' baffle. I will need 14.5'' x 6'' if I want to rear mount all the drivers. I am considering reducing the side wall thickness of the enclosure and midrange chamber by half to 1/8'', so I'll save 4 x 1/8'' = 1/2'' to bring the dimension back down to 14'' x 6''. The other walls will remain 1/4''. While mechanical vibrations won't be a problem due to the dual opposed design, I'm not sure of the possibly consequences of reducing the side wall thickness to 1/8'' for cabinet resonances. 

 

Now for the good news!

 

The amplifier is getting a significant upgrade. I will be using the ICEpower 50ASX2 for the midrange and tweeter, which will provide 50W for each driver. The two woofers will be powered by the ICEpower 125ASX2, which will deliver 125W per channel to the woofers. The ICEpower amps should sound much better than most of the class D amplifiers on eBay and Alibaba. I never knew ICEpower made lower power units until now. They have integrated power supplies, the dimensions are incredibly small and they're not horribly expensive. That said, they are still more than double the price of the previous amp I was considering. However, there will be no doubts on the excellent sound quality, power output, and build quality of ICEpower amps. 

tw030wa13_14_front.jpg

 

I'm changing the tweeter to the Wavecor TW030WA13, which seems to be an upgraded version of the TW030WA04 with a nice chambered tweeter. The driver is very good at playing low with 60% more Sd and over 3x higher Xmax than the Scanspeak tweeter. Someone mentioned using this tweeter with a 1000Hz crossover and it had no problems even with volume in the high 90dB's. The only downside is the larger physical size. Even with the flange perfectly cut, it is still 7mm larger, requiring an even lower crossover frequency.

Link to comment
Share on other sites

I don't think 1/8" thickness will be satisfactory if you are using ply or MDF.  It will most likely be very lossy.  Even hardboard is quite floppy at 1/8" thick.  I'd be concerned about 1/4" being to be too thin, but it depends on the length and width of the unbraced spans.  IIRC, stiffness roughly increases by 8X every time you double thickness, so you give up a lot when you reduce thickness.

Link to comment
Share on other sites

I don't think 1/8" thickness will be satisfactory if you are using ply or MDF.  It will most likely be very lossy.  Even hardboard is quite floppy at 1/8" thick.  I'd be concerned about 1/4" being to be too thin, but it depends on the length and width of the unbraced spans.  IIRC, stiffness roughly increases by 8X every time you double thickness, so you give up a lot when you reduce thickness.

 

Agreed. Not to mention it will also become quite fragile to shocks or other unintended incidents. Getting enough material to seal all of the joints well could be tough also.

Link to comment
Share on other sites

I just got an interesting thought:

 

I based my enclosure design on BBC's research on cabinet thickness, and how they reached the conclusion that thinner walls are better for midrange than thicker walls because thinner walls pushes the panel resonance down to a lower frequency - about ~100Hz for a 1/4'' thickness wall. So that has me thinking: what thickness and what dimensions does the walls have to be to push the resonance down to 40-50Hz where it would be beneficial to have the box resonate and add bass output to a small speaker where every bit of bass output is needed. 

 

Are there simulations or formulas to predict box resonance?

Link to comment
Share on other sites

I just got an interesting thought:

 

I based my enclosure design on BBC's research on cabinet thickness, and how they reached the conclusion that thinner walls are better for midrange than thicker walls because thinner walls pushes the panel resonance down to a lower frequency - about ~100Hz for a 1/4'' thickness wall. So that has me thinking: what thickness and what dimensions does the walls have to be to push the resonance down to 40-50Hz where it would be beneficial to have the box resonate and add bass output to a small speaker where every bit of bass output is needed. 

 

Are there simulations or formulas to predict box resonance?

 

If you dig around, you might be able to find formulas for resonances of single panels in isolation, but they won't be convenient to use and will fail to fully describe of a completely cabinet consisting of multiple panels and whatever bracing scheme is used.  A more accurate solution for full cabinets is to use FEM solid mechanics modelling, but this will be even less convenient to use and more sensitive to input parameters that you most likely have no easy way to measure or determine.

 

I'd rather not go into the BBC study here and in fact, the subject of cabinet design is very complex and fraught with controversy over "which way is best".  I will say, however, that if the box walls are resonating enough to substantially alter the response, it will almost certainly do more harm than good.  You will have a hard time predicting how it will behave.  If you are that desperate for extra bass output, I think the time would be better spent developing custom rectangular drivers and/or passive radiators.  Serious.  If that sounds like a pain (and it does to me), then it's probably best to stick with what is known to work.

Link to comment
Share on other sites

If you dig around, you might be able to find formulas for resonances of single panels in isolation, but they won't be convenient to use and will fail to fully describe of a completely cabinet consisting of multiple panels and whatever bracing scheme is used.  A more accurate solution for full cabinets is to use FEM solid mechanics modelling, but this will be even less convenient to use and more sensitive to input parameters that you most likely have no easy way to measure or determine.

 

I'd rather not go into the BBC study here and in fact, the subject of cabinet design is very complex and fraught with controversy over "which way is best".  I will say, however, that if the box walls are resonating enough to substantially alter the response, it will almost certainly do more harm than good.  You will have a hard time predicting how it will behave.  If you are that desperate for extra bass output, I think the time would be better spent developing custom rectangular drivers and/or passive radiators.  Serious.  If that sounds like a pain (and it does to me), then it's probably best to stick with what is known to work.

 

That's probably true -- if box resonance is audibly contributing, that probably won't sound good. I was initially excited by the idea of squeezing even more bass out of a small speaker. So far the driver and passive radiators are contributing to the bass. It would be cool if the box also contributed to the bass output. 

 

I am curious what the box resonance frequency for the speaker I will be building is though. Is there any way to measure that? Any behaviour that could be seen though a CSD or something?

 

I actually researched about some pretty crazy things to do to get as much bass as possible in a small speaker. The most promising one was the use of gases less compliant than air to make the internal volume larger than it is. I remember increasing by a factor of 2 was quite achievable. In a speaker this small, I'm mostly limited by the power input as most small drivers cannot handle more than 100W of power. By reducing power requirements by a factor of 4, I could have gained 6dB more output either through the drivers able to use more of their excursion or doubling up on drivers. Unfortunately, gases leak over time, and even if the box can be completely sealed, the gas will leak through the driver cones and surround, and this idea was abandoned. 

 

So on things that I could realistically do, I will be making my own passive radiator by butchering the very high excursion Tang Band W5-1138SMF that will have at least twice the excursion than the off the shelf Peerless passive radiator. This is necessary to support the active woofers since each active driver only gets one passive radiator of the same Sd. 

Link to comment
Share on other sites

There will be multiple resonances.  And note that the cabinet resonances are entirely separate from the resonances of the air within the cabinet.  For an ideal single panel, the resonances are harmonically related and analogously similar to what you see with air space cabinet resonance and room modes.  For a completed cabinet, the resonances will not be harmonic and will be more difficult to predict.  The best way to measure cabinet resonances is with an accelerometer and maybe a CSD of that data.  Of course, you have to choose where to put the accelerometer for each measurement, and need multiple measurements to get a good idea of what's going on.  And of course, this still doesn't tell you how radiation from the cabinet impacts the acoustic response.

Link to comment
Share on other sites

Some more updates, starting with bass:

 

I measured my previous design to see what kind of group delay I was getting. The speaker is tuned to 47Hz, high passed at 40Hz, with ~12dB of bass boost that makes it perfectly flat to 45Hz. Unfortunately, the speaker can reach 2 - 2.5 cycles at and below tuning. This is clearly not good, at least by the textbook definition.

 

EDIT: One complication is that the SMSL amp I used to power the woofers seem to have a built in bass boost. The model shows I need up to 12dB of boost, but I only applied 9dB on the miniDSP to achieve it. This might have made the group delay a lot worse as the results are a lot worse than what WinISD predicted. One explanation could be that there were a total of 5 filters used (4 on the miniDSP, 1 in the amp) to shape the bass response. Would using 4 filters instead of 1 in the miniDSP worsen the group delay? I used 4 when I could have just used 2 (boost and high pass) because I was anal on getting the bass to be literally ruler flat. However, if more filters cause more delays, I will reduce the number of filters. 

 

That said, my subjective opinion of the bass quality is excellent. Maybe it's because I might prefer a boomier sounding bass, because I don't like the sound of many well regarded SQ subwoofers from the likes of JL, Martin Logan or Paradigm. Because of this finding, I will continue my previous plan to boost the bass flat despite the group delay problems. 

 

The second update is on the tweeter: Originally I switched out the Scanspeak tweeters because SME pointed out that their very low 85.9dB test bench sensitivity will hold back the performance. However, after some research, it turns out that it seems every tweeter that Medley's Musings test has 4-6dB lower sensitivity than the manufacturer's rating. Scanspeak is known to have accurate, if not conservative specs even with Klippel testing. Two other tests for two different tweeters in the same tweeter family showed that Scanspeak's sensitivity rating is actually 1dB conservative. Therefore I believe the D3004/602010 tweeter I previously used does have an honest 90dB sensitivity.

 

As for the Wavecor tweeter, I noticed the sensitivity rating is an average between 5-20KHz. However, the tweeter exhibits a rising top end, which means that the actual sensitivity is actually 1-2dB lower because of the lower 2-5KHz band. Therefore, the Wavecor tweeter is only at most 2dB more sensitive instead of 7dB that I previously thought. The Wavecor tweeter also has poorer off axis response partly due to the bigger Sd. The sound quality is also a gamble. The tweeter is also 10mm bigger, increasing CTC distance. Therefore, the Wavecor lost of its advantage in sensitivity and the disadvantages now overwhelm the advantages. The Scanspeak tweeter is back because it is smaller, better off axis dispersion, better sound quality, and having a protective grill. 

 

Because the Scan is back in the game, now the question is, should I do a true Ricci style build and upgrade to the beryllium version of the tweeter? :P What do you guys think? The price is $120 vs $280 for each tweeter. However, I just read a measurement of the Be tweeter and the measurements and especially the distortion data is simply outstanding. It would have no problem handling the 1500Hz crossover. The tester also mentioned that besides having less sensitivity, it measures and sounds essentially the same as the $450-$520 normal sized Scanspeak beryllium tweeter as they share the same diaphragm, just a weaker motor on the small one.

 

The devil's advocate side of me is saying no, at ~$2500, this silly speaker is already stupid expensive and way more expensive than expected. I also want to hear the difference between the old and the new and how audible the difference between the two are when they share the same mid and tweeter, but this one will have an improved crossover, diffraction control, DSP, amp, midrange damping and construction. 

 

Link to the Scanspeak small Be tweeter measurements

 

This is the distortion measured at 98dB at 1 meter, barely any distortion even at 1000Hz.

red-D3004-604010_II_dist_-12.2_96dB-1m.g

Link to comment
Share on other sites

 

The devil's advocate side of me is saying no, at ~$2500, this silly speaker is already stupid expensive and way more expensive than expected.

 

 

 

:o  :blink: That's one expensive micro speaker!

 

Since the budget is blown out already, was there any thought to using exotic materials like perhaps aluminum plate for the enclosure?

Link to comment
Share on other sites

Indeed.  With some kind of sheet metal, you can probably achieve sufficient enclosure stiffness with much thinner walls, but if you do this, you really need to be sure that the first resonant frequency is high enough that you won't hit it because sheet metal is very poorly damped.  Or you can try using some damping sheet with it, of which there are many products designed specifically for sheet metal.  This may require testing to ascertain what works and what's acceptable.

 

Edit: Note that the sheet metal still has to be sufficiently thick for adequate stiffness.  Most products are very thin and may not be stiff enough.  You're pretty much on your own here if you go this way though.

Edited by SME
Link to comment
Share on other sites

Metal would require lots of tooling. Folding, welding, maybe even casting, just way too much for probably no net benefit.

 

If wood won't cut it for you, then you need composites. 1/4" carbon fiber or fiberglass with epoxy is damn tough. Even 1/8" will go a long way. They can be laid up in the final shape, with virtually no tooling required. Adding bracing would be elementary. Some people even consider it easier than woodworking once you get the hang of it. Just carve out a foam mold, glass the whole thing, then remove your foam. Cost is quite reasonable also. For such a small enclosure, the materials cost will be miniscule compared to your current $2.5k electronics and transducers budget.

 

Useless in the workshop? Draw it in Solidworks or SketchUp and send that bitch to a 3D printer. You can brace it within an inch of its life for no extra work or money. Put gussets on top of gussets inside your gussets. You only pay per gram in 3D printing. You'll probably be using ABS or something. Fiber reinforced plastics can be stronger than many metals, but I don't know if they're printable.

  • Like 1
Link to comment
Share on other sites

:o  :blink: That's one expensive micro speaker!

 

Since the budget is blown out already, was there any thought to using exotic materials like perhaps aluminum plate for the enclosure?

 

Yeah it is nuts! Every component is top notch. It is essentially taking the drivers of a very high end 3 way speaker, choose the smaller diameter driver in the same family, and cramming them into this tiny box. I wasn't kidding when I said the value factor is a -1. And you guys aren't helping me with fiscal responsibility! :P That said, if things turn out well, it should be able to compete toe to toe against some very high end "Hi-Fi" style loudspeakers. Now that would be nuts!

 

The main reason is because I couldn't reuse any of my previous components. Originally I thought cost wise this is not gonna be so bad because I can reuse the miniDSP, amp, and the mid and tweeters. But then the HD version of the miniDSP came out and it is priced far too high (should have replaced the 6 year old 2x4 instead of charging double the price as a higher end model), but I gotta have it for the digital input, FIR filters and impulse correction. Then I decided to use 4 of the pricey ICEpower amps to clean up the clutter and a properly designed amp, so I couldn't use the 4 amps I already bought. SME mentioned I should get the 4 ohm midrange to get more out of the amp, so I couldn't reuse the 8 ohm midrange in the previous speaker. So I thought, screw it, might as well keep the old speaker for comparison and sentimental reasons and put it in another room. The miniDSP's and the 4 ICEpower amps represent nearly half the cost of the build. 

 

Fuck it, I'm gonna get the beryllium tweeters. It's *THE* ultimate small speaker, right? :lol: I better not screw up when I cut the face plate off of these!

 

I could make an enclosure out of aluminum, but I cannot put a nice fancy finish on it, and I want this speaker to have a very nice finish to it. 

Link to comment
Share on other sites

Useless in the workshop? Draw it in Solidworks or SketchUp and send that bitch to a 3D printer. You can brace it within an inch of its life for no extra work or money. Put gussets on top of gussets inside your gussets. You only pay per gram in 3D printing. You'll probably be using ABS or something. Fiber reinforced plastics can be stronger than many metals, but I don't know if they're printable.

 

You could 3d print them but the cheaper FDM machines, which produce the cheap parts don't finish that well on parts that big. The materials aren't all that strong without some thickness either. any more than 30% fill can cause huge issues with parts warping / pulling on bigger pieces. SLS or SLA would be a LOT better but the money goes way up too.

Link to comment
Share on other sites

I'm not an engineer.  :(  I wish I had CAD skills to draw this up and send it to a 3D printer or something. Certainly would be much easier than working with wood!

 

Just ordered the woofers and tweeters. Will be ordering parts slowly throughout the next week or so to soften the financial sting  :lol:

 

Next up, designing a grill for these speakers to hide the "magic" design. I look forward to seeing some shocked faces when demoing this :D .

Link to comment
Share on other sites

Houston, we have a problem. 

 

In fact, this is sort of a big problem. The Tang Band W5-1138SMF cannot be converted into a useful passive radiator. I was under the impression that the amount of mass to tune a PR for a given Sd, frequency and enclosure size is a constant when this isn't the case. The amount of mass needed is dramatically changed by the Fs of the PR. The lower the Fs, the less mass it needs. Therefore, Cms can dramatically affect the mass needed. The TB driver has a very stiff suspension, so the mass requirement for tuning is much much greater, about 4x higher than the original Mms, and there's no way the suspension can handle this much weight. 

 

The backup option is to use the Peerless 5.25'' PR. However, it only has a 6mm Xmax when I need around 11mm. Any ideas? Or am I forced to just set the limiter higher so it doesn't blow up the Peerless PR near/below tuning?

Link to comment
Share on other sites

I use https://shapeways.coma fair bit for prints. The quality is very high, but it's still FDM, so things usually have a slightly rough finish. They do a variety of materials including fiber reinforced (extra cost and requires a human to verify your model) and a bunch of metals (speakers still too inexpensive? Try printing the enclosures out of platinum!)

 

It gets pricey fast however, and the maximum dimensions are typically small (e. g. < 30cm), but that's easy enough to check. All of their material pricing is on the site as well, without having to upload a model.

 

All that said, I don't know why you'd want to print over using wood. Maybe to make complex curves/shapes?

Link to comment
Share on other sites

So I did some playing around, and I could not find a reasonable cost driver suitable to be converted to a passive radiator. There are ~$100 drivers that could, but they're a bit too expensive to experiment with. 

 

So the option that I'm considering is just to use the Peerless 5.25'' PR for the following reasons:

 

1. The Peerless PR has extremely suitable PR parameters: High Cms, Qms, Mms, low Fs, and decent Vas. The Tang Band has low Cms, low Vas, decent Qms and high-ish Fs. 

 

2. Even though it only has 6mm of Xmax while modeling shows I need up to 11mm of excursion on the PR at the deepest usable frequency, the modeling software does not account for PR losses, which I think is non-trivial. Looking at my previous speakers where I used 2 of the 4'' version of the Peerless PR, the actual PR excursion was only a little more than half of what modelling suggests (partially because I used stuffing).  Note this is non scientific, just comparing with my eye and comparing it with the ruler I had in hand. I saw a video of someone that used a Tang Band 3.5'' RBM woofer (7mm Xmax!) with just one of the 4'' Peerless PR (jump to 11:20). Modelling shows one would need *4* of those PR's to keep the PR excursion below the 5mm Xmax. Yet, in the video he clearly drove the RBM woofer to the max and the lonely 4'' PR did not go flying into the room (although it was doing way more excursion than what I can drive my PR's and looks like it was doing a lot more than 5mm of excursion).

 

3. The 11mm excursion is only achieved with 150 watts of input power. With that much power, I expect at least 3dB of compression, which would further reduce PR excursion. That said, I do still have to watch out for transients. I'm hoping the 6mm rating is actually Xmax, and the Xmech is higher (hopefully a lot higher)

 

Has anyone blown a passive radiator before? I'm thinking of sacrificing 2 PR's by turning it louder and louder until the PR's are physically damaged from over-excursion in order to find the limits so I can dial the compressor properly. Good idea? (Probably not  :lol:). Main concern is hoping the broken parts don't damage anything inside the speaker. 

 

I use https://shapeways.coma fair bit for prints. The quality is very high, but it's still FDM, so things usually have a slightly rough finish. They do a variety of materials including fiber reinforced (extra cost and requires a human to verify your model) and a bunch of metals (speakers still too inexpensive? Try printing the enclosures out of platinum!)

It gets pricey fast however, and the maximum dimensions are typically small (e. g. < 30cm), but that's easy enough to check. All of their material pricing is on the site as well, without having to upload a model.

All that said, I don't know why you'd want to print over using wood. Maybe to make complex curves/shapes?

 

 

The only reason I mentioned printing is because it'll be a lot easier than going to the woodshop. But realistically, I'm going to the woodshop for this. There's no fancy curves on the enclosure. I just want a good looking finish on the wood. 

Link to comment
Share on other sites

Passive radiators may be a bit more forgiving than drivers as there are fewer parts to potentially ram each other.  However, the suspension of any driver can still be damaged if it's over-driven repeatedly.  For your application, I think Xmech may be more important.  You have to make many compromises to achieve big sound from a small box, and I think a bit of THD in the bass from high excursion at loud playback levels will probably not be a big deal.

Link to comment
Share on other sites

Passive radiators may be a bit more forgiving than drivers as there are fewer parts to potentially ram each other.  However, the suspension of any driver can still be damaged if it's over-driven repeatedly.  For your application, I think Xmech may be more important.  You have to make many compromises to achieve big sound from a small box, and I think a bit of THD in the bass from high excursion at loud playback levels will probably not be a big deal.

 

 

With a better modelling tool, the Peerless is a noticeably more "efficient" passive radiator (1-2dB) than the Tang Band. Looks like I will just use the Peerless PR's. I tried very hard to have as small of compromises as possible with this design, but it looks like this will have to be something I have to compromise in this ultra small speaker. Good thing the limitations won't show until it gets close to the limits of the speaker. 

 

I've ordered the 4 ICEpower amps and the miniDSP's. It's coming along, slowly but surely.

Link to comment
Share on other sites

  • 1 month later...

As I am working on tuning this speaker on the DSP and making it linear phase, I was reminded of a problem that I will have horrible group delay in the bass region because of the combination of a vented enclosure and massive shelf boost to flatten the bass response. 

 

I remember for my previous speaker, group delay exceeded 2 cycles below a certain frequency, let's say 60Hz. I don't have anywhere near enough taps to brute force correct the phase shift. But what if, say I separated the <60Hz band, and I delay the >60Hz band by the equivalent delay in ms of 2 cycles at 60Hz, which is 33ms. Would doing this improve sound quality more than it hurt it?

Link to comment
Share on other sites

As I am working on tuning this speaker on the DSP and making it linear phase, I was reminded of a problem that I will have horrible group delay in the bass region because of the combination of a vented enclosure and massive shelf boost to flatten the bass response. 

 

I remember for my previous speaker, group delay exceeded 2 cycles below a certain frequency, let's say 60Hz. I don't have anywhere near enough taps to brute force correct the phase shift. But what if, say I separated the <60Hz band, and I delay the >60Hz band by the equivalent delay in ms of 2 cycles at 60Hz, which is 33ms. Would doing this improve sound quality more than it hurt it?

 

Probably not, or at least, not enough to make it worth the effort.  Part of the trouble is that the filters you need to separate the high and low contribute group delay themselves, and the summed response has some non-minimum phase group delay character as well.

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...