As some of you know, we have had a project to design an open-source speaker based on Purifi driver. It has been led by our member, @Rick Sykora who has named it "Directiva." You can read about it here: https://www.audiosciencereview.com/...r-open-source-platform-speaker-project.20407/
The idea came about from a hypothesis: if a speaker design had access to Klippel Near-field scanner, would he be able to design a super speaker? I own the speaker but I am not a speaker designer so it made sense to have others help create one. The results of that work is now here in the form of a prototype built by Rick with much help on design side from member @ctrl.
Other than the crinkled woofer from Purifi, there is not much to distinguish this speaker from any other DIY prototype:
I let Rick chime in with the choice of tweeter. On the back side, we have an SB Acoustics passive radiator:
My understanding is that a traditional port design is difficult for the Purifi driver and hence the use of the radiator.
Connectivity is provided through a SpeakOn connector which provides dual set of inputs, one for tweeter and one for woofer.
This is an active design so I build a compact system to drive it comprising of the Minidsp 2x4 HD and SMSL VMV A2 desktop amplifier. The combination sat on top of the speaker with a pad under it. It was actually a bit smaller than it which helped with not creating diffraction error or changing baffle compensation.
Two scans were performed with the second one employing minor refinements.
ASR Directiva Measurements
Here is our spin graph:
I must say, the flatness of the on-axis response put a big smile on my face. I am only used to seeing such an accurate response when measuring DSP based active monitors. Not only that, the early window directivity mirrors the on-axis smoothly and only has a minor dip between 2 and 3 kHz.
Drilling into the early window reflections, we see that it is the vertical axis that is responsible for that dip:
Hard to do much better without a coaxial driver. Fortunately the floor is most responsible for that so as usual, put a thick rug there and you should be able to get better response than that. Assuming you don't do that, you still get a very nice predicted in-room frequency response:
Driver and radiator response is clean:
Maybe having a little filter to pull that bump above 10 kHz down a couple of dBs but likely won't be audible.
Our beamwidth measurement is excellent:
There is slight narrowing but we need that to keep the speaker from sounding too bright as it lights up all surfaces otherwise.
Horizontal directivity as a result looks pretty:
We have the usual non-perfect vertical directivity but it is better than many 2-way speakers:
This is a speaker that is room friendly.
CSD/Waterfall shows some resonances:
Some of this could be because of the DSP and and amp being on top of the amp as the enclosure feels solid as a rock.
We see more evidence of the resonances in distortion measurements which otherwise are excellent:
Directiva Listening Tests
These days it takes me so little time to tell if overall tonality is right. And such was the case here. I felt no need to reach for EQ although at times I thought upper bass was a bit weak. I think speakers with distortion generate harmonics that land in that region. Take that away and the perception is a cleaner but slightly leaner upper bass. I am emphasizing upper bass as Directiva was able to deliver good bit of sub-bass. I didn't want to mess with EQ as you really need to measure the room+speaker response and correct for that.
The sound was very clean, making me want to really crank up the volume. When I pushed that hard, I could sometimes hear a bit of ticking sound. I am not sure if the the minidsp was saturating its input or the driver was the issue. By then the volume was pretty loud and with two speakers you would be fine.
I tested for tweeter hiss and unfortunately it was bad in this configuration. I could easily hear it up to about 2 feet away. Minidsp supports Toslink and ideally you would be using that input and avoid its ADC and my pre-amp. That would have required me ripping out my listening room even more so I did not go there. Fortunately unlike powered monitors, with a kit like this, you can play with such options (and amplification) to get the best noise and power capability you need.
DIY Build Cost
Rick tells me that the cost for the components to build a pair is about US $1,300 to $1,400. A commercial implementation with fancy finish and such would then land at $4K to $5K rang which would be pretty expensive given the lack of DSP and amplification cost on top of that. For DIY though, it makes sense.
Conclusions
Well, the hypothesis is proven: get a few smart people together and give them a Klippel NFS measurement system and they too can produce world-class speaker performance. I am very pleased with how the project has come out, showing the power of the community to come together to follow the best in science, engineering and simulation technology. Thanks to all of you who contributed. I am proud of you all!
P.S. Just in case @MZKM is sleep right now, he was kind enough to give me his preference scores:
Preference Rating
SCORE: 6.5
SCORE w/ sub: 8.0
Frequency response: +/- 1.9dB 40Hz-20kHz
LFX: 34Hz
The idea came about from a hypothesis: if a speaker design had access to Klippel Near-field scanner, would he be able to design a super speaker? I own the speaker but I am not a speaker designer so it made sense to have others help create one. The results of that work is now here in the form of a prototype built by Rick with much help on design side from member @ctrl.
Other than the crinkled woofer from Purifi, there is not much to distinguish this speaker from any other DIY prototype:
I let Rick chime in with the choice of tweeter. On the back side, we have an SB Acoustics passive radiator:
My understanding is that a traditional port design is difficult for the Purifi driver and hence the use of the radiator.
Connectivity is provided through a SpeakOn connector which provides dual set of inputs, one for tweeter and one for woofer.
This is an active design so I build a compact system to drive it comprising of the Minidsp 2x4 HD and SMSL VMV A2 desktop amplifier. The combination sat on top of the speaker with a pad under it. It was actually a bit smaller than it which helped with not creating diffraction error or changing baffle compensation.
Two scans were performed with the second one employing minor refinements.
ASR Directiva Measurements
Here is our spin graph:
I must say, the flatness of the on-axis response put a big smile on my face. I am only used to seeing such an accurate response when measuring DSP based active monitors. Not only that, the early window directivity mirrors the on-axis smoothly and only has a minor dip between 2 and 3 kHz.
Drilling into the early window reflections, we see that it is the vertical axis that is responsible for that dip:
Hard to do much better without a coaxial driver. Fortunately the floor is most responsible for that so as usual, put a thick rug there and you should be able to get better response than that. Assuming you don't do that, you still get a very nice predicted in-room frequency response:
Driver and radiator response is clean:
Maybe having a little filter to pull that bump above 10 kHz down a couple of dBs but likely won't be audible.
Our beamwidth measurement is excellent:
There is slight narrowing but we need that to keep the speaker from sounding too bright as it lights up all surfaces otherwise.
Horizontal directivity as a result looks pretty:
We have the usual non-perfect vertical directivity but it is better than many 2-way speakers:
This is a speaker that is room friendly.
CSD/Waterfall shows some resonances:
Some of this could be because of the DSP and and amp being on top of the amp as the enclosure feels solid as a rock.
We see more evidence of the resonances in distortion measurements which otherwise are excellent:
Directiva Listening Tests
These days it takes me so little time to tell if overall tonality is right. And such was the case here. I felt no need to reach for EQ although at times I thought upper bass was a bit weak. I think speakers with distortion generate harmonics that land in that region. Take that away and the perception is a cleaner but slightly leaner upper bass. I am emphasizing upper bass as Directiva was able to deliver good bit of sub-bass. I didn't want to mess with EQ as you really need to measure the room+speaker response and correct for that.
The sound was very clean, making me want to really crank up the volume. When I pushed that hard, I could sometimes hear a bit of ticking sound. I am not sure if the the minidsp was saturating its input or the driver was the issue. By then the volume was pretty loud and with two speakers you would be fine.
I tested for tweeter hiss and unfortunately it was bad in this configuration. I could easily hear it up to about 2 feet away. Minidsp supports Toslink and ideally you would be using that input and avoid its ADC and my pre-amp. That would have required me ripping out my listening room even more so I did not go there. Fortunately unlike powered monitors, with a kit like this, you can play with such options (and amplification) to get the best noise and power capability you need.
DIY Build Cost
Rick tells me that the cost for the components to build a pair is about US $1,300 to $1,400. A commercial implementation with fancy finish and such would then land at $4K to $5K rang which would be pretty expensive given the lack of DSP and amplification cost on top of that. For DIY though, it makes sense.
Conclusions
Well, the hypothesis is proven: get a few smart people together and give them a Klippel NFS measurement system and they too can produce world-class speaker performance. I am very pleased with how the project has come out, showing the power of the community to come together to follow the best in science, engineering and simulation technology. Thanks to all of you who contributed. I am proud of you all!
P.S. Just in case @MZKM is sleep right now, he was kind enough to give me his preference scores:
Preference Rating
SCORE: 6.5
SCORE w/ sub: 8.0
Frequency response: +/- 1.9dB 40Hz-20kHz
LFX: 34Hz