Nucleus: my attempt at building the best sounding small speaker I can

Status
Not open for further replies.
I'd like to share with you guys something I've been working on and off for a while now. I call it, the Nucleus, and it is my attempt at making the best possible sounding small speaker I can possibly make without too much regards to budget. I thought it would be much more interesting to try to build something that far outperforms anything commercially available, and something that hasn't really been tried before: making a small speaker sound truly good.

Goal: To make an extremely high fidelity small sounding speaker with strong bass down to 45Hz in an enclosure no bigger than 0.1 cu ft.

An externally hosted image should be here but it was not working when we last tested it.


I wanted something that's visually obtrusive, so after much modelling, I ended up settling for a speaker that is 13'' wide, 5'' high, and 5'' deep. This is built on 1/4'' plywood of the highest grade to maximum internal volume. The midrange chamber inside acts as bracing for the speaker to provide additional strength. A nice rosewood stain was used to finish the enclosure.

The drivers consist of the Scanspeak Illuminator D3004/6020 for the tweeter crossing to the the Scanspeak Discovery 10F/8424G midrange at 3000Hz and crossing to the Tangband W4-1720 for the woofer at 300Hz. 2 custom Peerless 4'' passive radiators were used to tune the speaker to 46Hz. A miniDSP 2x4 is used as an active crossover as well as 2 SMSL SA-50 amps powering the drivers individually. The miniDSP was absolutely critical in bringing out the magic from these amazing drivers. Not only this is used to do 4th order crossovers, but it was necessary to correct box resonance from the high Q, diffraction effects of the baffle, and most importantly, the bass boost for woofer to bring out the bass, as well as the high pass filter and the limiter to keep the woofer in check.

If I'm going this far for this project, I might as well pull out all the stops. Extensive testing was done to tune the speakers. And the results? A speaker.that measures ±1dB from 100Hz to 16KHz with a -3dB roll off at 20KHz and a rise in bass response starting from 100Hz to +3dB at 50Hz. Here's a measurement done in anechoic chamber. There is no smoothing applied to the graph. Any smoothing to the graph will result in essentially a straight line. All the little bumps are actual response of the speaker as the microphone is very very high resolution. The chamber was only good down to about 100Hz, so that's where it stops. The second picture is an outdoor ground plane measurement for the bass. Again, that's with no smoothing.

An externally hosted image should be here but it was not working when we last tested it.


An externally hosted image should be here but it was not working when we last tested it.


So how does it sound? Just utterly fantastic. It is surreal hearing such high fidelity coming out of something so small. The Scanspeak drivers are truly excellent that just oozes detail and resolution. The clarity and detail is just outstanding that rivals some of the best speakers I've ever heard. The bass coming out of this speaker is truly astounding. The Tangband woofer is outstanding. It just pumps smooth and strong deep bass that is rather unbelievable from a speaker so small. At normal to above average listening volume, it absolutely sounds like a subwoofer is playing as the distortion is very low even at very high excursion levels. I've bought this to the office where I was pounding the speakers with rap and it just blew me away how much deep bass this speaker can produce. One caveat though, some music that really uses the stereo effect sound a little weird a mono, but that's just because there's only one speaker. But still, I am beyond happy how this turns out. It was an absolute blast designing and building this speaker, and it is truly satisfying when things come together beautifully. Small speakers can't sound good? You can't get bass from a small speaker? Oh yes, you absolutely can.
 
Last edited:
There's that small hill visible that starts at about 950Hz up to 1600Hz. It is in the very sensitive part of listening spectre. You could treat it with low Q mild notch. Other than that it looks great.
 
Last edited:
Superb choice of drivers and excellent execution. Great work!

My only suggestion that could improve this to the next level (and you have it at your fingertips) is to make it transient perfect, or quasi transient perfect. Percussion, pianos, plucked bass rimshots, will sound entirely convincingly real.

To do that, you have to get away from LR4 symmetric XO's as they are phase wrapped and you won't be able to reproduce a square wave. I know, music ain't square waves - but, if you can do them it means your transients will be snappy and time accurate.

Try the Harsch XO for the woofer to mid, and maybe 1st order symmetric from mid to tweeter.

So woofer LPF is 4th order BW at say 700Hz, mid HPF is 2nd Bessel at 700Hz (with delay added to mid and tweet equal to half period cycle at 700Hz), and mid 1st order LPF at 5kHz, and tweeter 1st order HPF 5kHz. All drivers +ve polarity and tweeter may require a little delay adjustment. I assume you have FRD's for all already so, just pull up Xsim, PCD, or Boxsim, whatver you use to simulate. Put the textbook targets from above and start tweaking the XO curves and EQ. EQ flat first then apply XO.

Once you hear transient perfect - you won't go back. 😀

http://www.diyaudio.com/forums/multi-way/277691-s-harsch-xo.html
 
Or switch to a MiniDSP 2x4HD and use its FIR EQ and Rephase software to alter the phase to transient perfect. It could only be so horizontally at one height, though, since the non-coincident driver arrangement would prevent that vertically. (Not saying that it necessarily matters....)
 
Very nice! I may be a heretic, but I would love to see your in-room measurements. 🙂 What is it like at your chair? Those mini-Scanspeak tweeters are the bomb.

Best,


Erik

Here it is. This is from ~12 feet away, with a ~8dB bass boost and 1/24 smoothing. The room acoustics is not great...

An externally hosted image should be here but it was not working when we last tested it.


Superb choice of drivers and excellent execution. Great work!

My only suggestion that could improve this to the next level (and you have it at your fingertips) is to make it transient perfect, or quasi transient perfect. Percussion, pianos, plucked bass rimshots, will sound entirely convincingly real.

To do that, you have to get away from LR4 symmetric XO's as they are phase wrapped and you won't be able to reproduce a square wave. I know, music ain't square waves - but, if you can do them it means your transients will be snappy and time accurate.

Try the Harsch XO for the woofer to mid, and maybe 1st order symmetric from mid to tweeter.

So woofer LPF is 4th order BW at say 700Hz, mid HPF is 2nd Bessel at 700Hz (with delay added to mid and tweet equal to half period cycle at 700Hz), and mid 1st order LPF at 5kHz, and tweeter 1st order HPF 5kHz. All drivers +ve polarity and tweeter may require a little delay adjustment. I assume you have FRD's for all already so, just pull up Xsim, PCD, or Boxsim, whatver you use to simulate. Put the textbook targets from above and start tweaking the XO curves and EQ. EQ flat first then apply XO.

Once you hear transient perfect - you won't go back. 😀

http://www.diyaudio.com/forums/multi-way/277691-s-harsch-xo.html

The next speaker will be a much better design than this one that'll correct all that. It'll be a 3 way speaker that sounds and measures like a point source. It will be near phase perfect with FIR filters on the upgraded miniDSP 2x4 HD and a crossover that'll result in no horizontal or vertical lobing error.

I learned a lot from designing and building this speaker. I wish I can change the title, because it is not the best sounding small speaker ever. The next one though will be very, very close.
 
Last edited:
Hi bcodemz and all, as bwaslo meant, midrange and tweeter are not acoustically coupled; thou could have at least slipped the rim of the latter under the one of the former. And as X meant, impulse respone needs further work (the advantage of Harsch X is that it is digital time only, say a CMOS bucket chain, but not digital time and amplitude, say a clocked binary computer).
 
Of course it is not.



When for so many years you cannot improve on the design, then it will be the best sounding ever 😉

While I have no problem with critiques, in fact I welcome it. I really appreciate everybody who took the time to post suggestions here to help me. However, I don't appreciate you devoting a post not for a critique, but just to attack, especially when it was a response to a self criticism of a poor choice in wording.

You show us frequency response data with REW so it's there already. Just click on the distortion plot tab.

The above is from REW right?

Here it is. I didn't post this because the noise floor was not great for a 85dB measurement when I was measuring this outdoors with all the wind and distant cars driving by. It was enough for FR measurements, but I'm not sure so much for distortion. Ideally a nearfield measurement would be taken to get a more accurate distortion measurement.

An externally hosted image should be here but it was not working when we last tested it.
 
However, I don't appreciate you devoting a post not for a critique, but just to attack, especially when it was a response to a self criticism of a poor choice in wording.

It wasn't an attack. It was a "clue" to inform you of a certain phenomenon.

You can see many (speaker) manufacturers where their first product is the best. No matter how hard they tried the next products are only more expensive, not better...

You can also design an amplifier, and you will need years of trying to improve on one successful design...

Best efforts are IMO always like that.
 
An externally hosted image should be here but it was not working when we last tested it.


Thanks for that. It's very respectable for a small speaker. Rises linearly in 2nd harmonic distortion to -10dB at 40Hz which is very deep for how small the box is. 3H is lower and not broad. I think this would actually sound very good at moderate volumes. That bass digs deep for how small it is and is still under 10% HD at the limit.
 
The picture and the box dimensions you give, suggests that the speakers are being used horizontally. Would this be because of constraints as to where they are being used, or an aspect of the design?

No, I just use it like that for aesthetics to mimic a sound dock, and keeping the tweeter at ear level when sitting down. I use it mostly on axis, so the problems of lobing and stuff aren't much of a problem. It is definitely not optimal being laid sideways, and that's a compromise I made. It is definitely used vertically when I listen off axis, or in stereo.
 
Last edited:
My only suggestion that could improve this to the next level (and you have it at your fingertips) is to make it transient perfect, or quasi transient perfect. Percussion, pianos, plucked bass rimshots, will sound entirely convincingly real. To do that, you have to get away from LR4 symmetric XO's as they are phase wrapped and you won't be able to reproduce a square wave. I know, music ain't square waves - but, if you can do them it means your transients will be snappy and time accurate.
Hello, please can you explain to me better the meaning of "phase wrapped" about LR4 symmetric Xover (I'm a beginner..)? What's the best topology of passive Xover from this point of view?
Thanks.
 
Status
Not open for further replies.