Building a 2-way small active speaker with software crossover

[DanielT]

Impedance plots of the drivers within the cabinet

View attachment 314310

View attachment 314311

Measurements on the speaker elements in one cabinet, but have you measured the speaker elements in both cabinets? And thus examined if there is any difference between them?

You may so far only built one speaker box?

 

[pma]

You may so far only built one speaker box?

Correct, just one speaker box. As I have stated in the post #1 and several times later, this project is just an "exercise" for me, to try and to learn to work with SW crossover and also my curiosity to play with a horn tweeter. I will not build a second box, this will never be my main system.

However, I liked this lesson and learned new things. First, it was confirmed that the SW crossover method is much more effective to optimize final response. The resulting sound is neutral and uncoloured. It, of course, suffers from a mini woofer and not enough low frequency extension. It is a small speaker. It will stay near my PC as a small monitor speaker, as it is much more neutral compared to my JBL Control 1 Pro. The horn tweeter is also an interesting experience - it sounds open and very different from 1" fabric dome tweeters. Overall, the sound of this small box is very clean and uncoloured.

 

[dualazmak]

Correct, just one speaker box. As I have stated in the post #1 and several times later, this project is just an "exercise" for me, to try and to learn to work with SW crossover and also my curiosity to play with a horn tweeter. I will not build a second box, this will never be my main system.

However, I liked this lesson and learned new things. First, it was confirmed that the SW crossover method is much more effective to optimize final response. The resulting sound is neutral and uncoloured. It, of course, suffers from a mini woofer and not enough low frequency extension. It is a small speaker. It will stay near my PC as a small monitor speaker, as it is much more neutral compared to my JBL Control 1 Pro. The horn tweeter is also an interesting experience - it sounds open and very different from 1" fabric dome tweeters. Overall, the sound of this small box is very clean and uncoloured.

I really look forward to hearing your possible progress in implementing PC(Windows)_DSP-EKIO based multichannel setup in your main stereo audio system at your listening room/acoustics.

 

[pma]

I really look forward to hearing your possible progress in implementing PC(Windows)_DSP-EKIO based multichannel setup in your main stereo audio system at your listening room/acoustics.

Well, this is unlikely to happen in the near future . But, you are right. If I was to design and build a new 3-way speaker, it would definitely be with a software crossover and a multichannel DAC .

 

[dualazmak]

I am sorry to be a little bit out of the scope of this thread, but please allow me having this post.

Just a small remark, the white noise is definitely not a good method to test the speaker.

After having your above message, I have carefully thought/considered again (only my brain storming) "that issue" of pros and cons of two Fq response measurement methods, i.e. "Fq sine sweep method" and my "cumulative (recorded) white noise averaging method (ref. here and here)".

I assume the pros and cons of the two methods may vary depending on what kind of Fq response information we would like to have.

I essentially agree with your point that "to test the speaker" or "to test anechoic behavior/characteristics of a SP driver or a multi-way SP unit", sine sweep at as near as possible to "the speaker" would be preferable.

On the other hand, in case we we would like to have "total" Fq response data/information on our whole audio system in our room acoustic environment especially at our listening position, I mean including the effects of all the "room modes", white noise averaging may have several pros, I assume.

As I wrote here, I frequently compared the two methods, i.e. "rapid sine sweep" versus "cumulative (recorded) white noise averaging", at my rather large listening/living room (the multichannel full audio system, the latest setup here) and at my rather small office upstairs (desktop DSP stereo audio system with one subwoofer); I found that the resulting Fq response curves obtained by the two methods are/were not always similar/identical with each other. And, the difference are/were always more significant at the small office upstairs which has very little room acoustic treatments.

I assume/guess that "cumulative (recorded) white noise averaging" would sometimes better "simulate" our usual music listening situation since all the SP drivers receive/sing "white noise", not the sharp/narrow-Fq sine sweep, and the sound actually activate our room air as such.

Just for example, when we play full orchestra symphony finale "tutti" as well as huge pipe organ performance "tutti" with using so many organ pipes, the sound has very wide distribution of high gain Fq (16 Hz to 22 kHz) peaks including so many harmonics tones "at once/simultaneously" which in-all-together activate our room acoustics including the various room modes (resonances, reflections, reberbrations, etc.) And I myself am rather focusing-on/interested-in Fq response measurement of "such total sound I listen to" at my listening position in my room acoustics with good reproducibility and less statistical FFT fluctuations in the obtained data.

I also have intensively applied the "cumulative (recorded) white noise averaging" for Fq response measurements at various "stages" in my DSP-based audio system, I mean in digital domain and various stages of analog line level, and of course finally in actual room air sound. The summary of such measurements shown in identical X-Y scale is like in this diagram (ref. here);

As you may kindly agree, I could have various information and insights throughout the above systematic Fq response measurements, and of course the "consistency" of the measurement method should be indispensable for any comprehensive discussion(s) on any combination of these Fq response data.

Just for your info, I once asked very briefly to Dr. Floyd Toole regarding my above "thoughts", and he kindly responded (ref. here);
"If properly done both swept tone and noise analysis should give identical answers. It is a choice. The principal difference is in the heating of the drivers in sustained tests at high sound levels - power compression. Low frequencies require longer averaging times."
At that time, I dared not to further bother Dr. Toole by asking "how about the inclusion of room modes?", since such question would be too naive and primitive for his attention.

I assume, however, at least his point of "Low frequencies require longer averaging times." would support/justify my "cumulative (recorded) white noise averaging" for Fq response measurements, since the FFT calculation is applied on much "rich" raw data (I usually record the white noise sound for one minute or longer) so that the statistical/mathematical fluctuation(s) in low Fq can be minimized to give very good reproducibility.

I can apply FFT size of even "65536" in averaging Fq analysis on the "selected whole 1 min white noise track (very rich raw data)" using Blackman-Airris window-set of Adobe Audition which gives consistent and reproducible Fq curve covering 15 Hz to 22 kHz. As you know, usually I do not like applying too much smoothing, such as REW's psychoacoustic smoothing*, on Fq response curve; I fully agree with Dr. Toole who kindly pointed at the end of his post here;
Don't worry about little ripples. When I see exceptionally smooth high-resolution room curves I strongly suspect that something wrong has been done. The measurement microphone is no substitute for two ears and a human brain. Happy landings!

In any way, in my future really critical and important room air sound Fq response measurement, I would like to carefully compare the results given by both methods, "rapid sine sweep" and "cumulative (recorded) white noise averaging".


*In REW user manual: "Psychoacoustic smoothing uses 1/3 octave below 100Hz, 1/6 octave above 1 kHz and varies from 1/3 octave to 1/6 octave between 100 Hz and 1 kHz. It also applies more weighting to peaks by using a cubic mean (cube root of the average of the cubed values) to produce a plot that more closely corresponds to the perceived frequency response."

 

[pma]

We need to measure the speaker as properly as possible, in the design phase. That means, without the influence of room modes and reflections, as much as possible. That's why we use near field/far field merging, windowing of the impulse response etc. Room modes and reflections are a separate issue and during the speaker design, they are only confusing the issue. With the cumulative white noise method, we are unable to separate direct sound from the reflected sound, but our ear is able to do so. That's why human voice or music recorded in the room always sounds different, during replay, compared to the sound listened directly in the room. Microphone measures sum of all sounds, it does not have brains capability to distinguish between direct and reflected sound etc. Cumulative white noise method in the listening room is unusable during speaker design phase.

 

[dualazmak]

We need to measure the speaker as properly as possible, in the design phase. That means, without the influence of room modes and reflections, as much as possible. That's why we use near field/far field merging, windowing of the impulse response etc. Room modes and reflections are a separate issue and during the speaker design, they are only confusing the issue. With the cumulative white noise method, we are unable to separate direct sound from the reflected sound, but our ear is able to do so. That's why human voice or music recorded in the room always sounds different, during replay, compared to the sound listened directly in the room. Microphone measures sum of all sounds, it does not have brains capability to distinguish between direct and reflected sound etc. Cumulative white noise method in the listening room is unusable during speaker design phase.

Yes, I well understood your point especially "Cumulative white noise method in the listening room is unusable during speaker design phase."

BTW, today I could pre-register to Tokyo International Audio Show 2023 (TIAS 2023) to be held at Tokyo International Forum, November 3 through 5.

During last year's TIAS 2022, I could talk to several professional SP designers and engineers (from Japan including YAMAHA and TAD people) and from abroad, of course they use SOTA anechoic chambers/laboratories for design development and tuning of their past-present-future SP drivers and SP units. I believe I will be able to talk to them again during coming TIAS 2023 in November.

I would like to unofficially ask them, especially to YAMAHA and TAD engineers, about whether they would measure their SPs in anechoic chamber using not only rapid sine sweep but also cumulative white noise averaging, or not. I am just curious if they would be interested in ”intermodulation distortion measurement/assessment" within single SP driver by using "white noise averaging" and/or similar "mixed multiple Fq tone signals"...

After TIAS 2023, I will get back to you here on this thread and/or on my project thread, if I could get some (official or unofficial) information in this regard.

 

[pma]

Passive crossover

I have made an attempt to make a passive crossover for this speaker box. The crossover is too big to be placed inside the box, and the tuning with a solder iron in hand is tiring.

Whatever I do, I am unable to avoid a dip near crossover frequency, with the passive xover. These 2 drivers definitely need sharp filters to be usable in a 2-way box. Below the comparison of on-axis response with the passive xover (green), and active SW xover (orange). Like a chalk and cheese. Please note that the HF roll-off in the orange response is intentional, to cut ugly behaviour of the tweeter above 15kHz.

 

[dualazmak]

Thank you again and again for your sharing your hard work which is very much interesting and worthwhile for me (and for all of us, I believe).

Your work in building passive LCR network outside of the SP box reminds me my similar struggle in building (all soldering!) renovated LCR network outside of my YAMAHA NS-1000 (ref. here), and that was the start of my multichannel exploration. I still keep that L & R passive outer-box LCR network for easy and quick rollback of my entire audio rig back to my single-amp reference sound system as it was in early 2019 also using the SP-cabling board (ref. here).

 

[Thomas_A]

Passive crossover

I have made an attempt to make a passive crossover for this speaker box. The crossover is too big to be placed inside the box, and the tuning with a solder iron in hand is tiring.

View attachment 314545

Whatever I do, I am unable to avoid a dip near crossover frequency, with the passive xover. These 2 drivers definitely need sharp filters to be usable in a 2-way box. Below the comparison of on-axis response with the passive xover (green), and active SW xover (orange). Like a chalk and cheese. Please note that the HF roll-off in the orange response is intentional, to cut ugly behaviour of the tweeter above 15kHz.

View attachment 314549

Som what target did the passive filter have?

 

[dualazmak]

May I ask a kind of stupid question? Of course, you may just ignore this inquiry.

Can you try SW(EKIO) digital XO/EQ ustream in your PC and dare to feed the signal into DAC and then to feed into your "passive LCR network" experimental setup so that the air sound could be flatter around the XO fq??

I recently suggested this exceptional approach to someone who would like to just test DSP(XO/EQ/Delay) in his single amplifier passive network 3-way SP system as shown in this diagram;

 

[Ken Tajalli]

Whatever I do, I am unable to avoid a dip near crossover frequency, with the passive xover. These 2 drivers definitely need sharp filters to be usable in a 2-way box. Below the comparison of on-axis response with the passive xover (green), and active SW xover (orange). Like a chalk and cheese. Please note that the HF roll-off in the orange response is intentional, to cut ugly behaviour of the tweeter above 15kHz.

Reverse the polarity of the tweeter wiring.

 

[pma]

Reverse the polarity of the tweeter wiring.

Thank you for the valuable advice, I have of course done it as a 1st attempt to cure (I am really not an idiot in electroacoustics). It is then even worse, though at another frequency. Please note that the dip is quite wide. Maybe a time delay would help, though not for sure, but who would be trying it to make the passive xover even more complex and big. The responses of the drivers are so messy near 3 kHz that they have to be cut sharply, and the active LR 48dB/oct is an ideal tool to fix it.

Another point to mention, distortion is much higher with the passive xover, especially that of the horn tweeter. The drivers seem to benefit from direct connection to amp output in the active version.

 

[Ken Tajalli]

Thank you for the valuable advice, I have of course done it as a 1st attempt to cure (I am really not an idiot in electroacoustics). It is then even worse, though at another frequency. Please note that the dip is quite wide. Maybe a time delay would help, though not for sure, but who would be trying it to make the passive xover even more complex and big. The responses of the drivers are so messy near 3 kHz that they have to be cut sharply, and the active LR 48dB/oct is an ideal tool to fix it.

Another point to mention, distortion is much higher with the passive xover, especially that of the horn tweeter. The drivers seem to benefit from direct connection to amp output in the active version.

No offence intended, I, myself, sometimes miss something rudimentary.
asalam o aleykom! (peace be upon you, I think!)

 

[DanielT]

Another point to mention, distortion is much higher with the passive xover, especially that of the horn tweeter.

So much higher than the 3% you already had around 3kHz. Is it therefore audible?

It is interesting for us to read about how you try to get the best out of the drivers you have. The problems or challenges you face and how you solve them.

 

[dualazmak]

The drivers seem to benefit from direct connection to amp output in the active version.

Yes, I believe this would be one of the important and excellent pros of fully active, direct connection to amps, in active setup.

I too was very much impressed by audible and measurable less distortion and improved impulse response (transient behavior) of not only my midrange but also my 30 cm woofer (now directly driven by YAMAHA A-S3000) when I first made the whole system fully active with complete elimination of passive LCR network and attenuators.

I (we) know well that passive LCR network (and attenuators) greatly waste/consume the input power into just heat even how cleverly we designed them.

In LCR network for woofer, I found the rather large inductor(s) (coils) were significantly deteriorating the very nice transient characteristic (spec) of the driver. On the other hand, in active setup with direct connection to powerful HiFi amplifier with excellent damping factor, we can fully utilize/extract woofer's original own maximum capabilities, I believe (ref. here).

Furthermore, we cannot deal with the "time alignment" in passive system; only the active DSP(EKIO) setup can precisely manage the time-domain tuning including time alignment for all the SP drivers. (I already know this tuning would be not needed for OP's present 2-way experimental setting though.)

 

[thewas]

Another point to mention, distortion is much higher with the passive xover, especially that of the horn tweeter. The drivers seem to benefit from direct connection to amp output in the active version.

I doubt that is a result of the crossover being active but just of the different transfer function.
Also to tell the other side of the story, a passive notch filter can in some cases even reduce the distortion compared to an active one:
https://purifi-audio.com/wp-content/uploads/2022/03/220211_R05-Notchfilter.pdf

I too was very much impressed by audible and measurable less distortion and improved impulse response (transient behavior) of not only my midrange but also my 30 cm woofer (now directly driven by YAMAHA A-S3000) when I first made the whole system fully active with complete elimination of passive LCR network and attenuators.

Do you have distortion measurements of both?

I (we) know well that passive LCR network (and attenuators) greatly waste/consume the input power into just heat even how cleverly we designed them.

Which is usually not an issue from sound quality point of view though.

In LCR network for woofer, I found the rather large inductor(s) (coils) were significantly deteriorating the very nice transient characteristic (spec) of the driver. On the other hand, in active setup with direct connection to powerful HiFi amplifier with excellent damping factor, we can fully utilize/extract woofer's original own maximum capabilities, I believe (ref. here).

It seems to me you had not a lowpass function enabled on the active crossover, which again would be an pears with apple comparison as different transfer functions, please correct me if I am wrong.

Furthermore, we cannot deal with the "time alignment" in passive system;

It can be partially done with a passive all pass filter, I agree though it is more expediently done in an active crossover, but both are second choice as constant time delay doesn't get a linear phase (summation) like an adjusted FIR crossover.

 

[pma]

Yes, I believe this would be one of the important and excellent pros of fully active, direct connection to amps, in active setup.

I too was very much impressed by audible and measurable less distortion and improved impulse response (transient behavior) of not only my midrange but also my 30 cm woofer (now directly driven by YAMAHA A-S3000) when I first made the whole system fully active with complete elimination of passive LCR network and attenuators.

I (we) know well that passive LCR network (and attenuators) greatly waste/consume the input power into just heat even how cleverly we designed them.

In LCR network for woofer, I found the rather large inductor(s) (coils) were significantly deteriorating the very nice transient characteristic (spec) of the driver. On the other hand, in active setup with direct connection to powerful HiFi amplifier with excellent damping factor, we can fully utilize/extract woofer's original own maximum capabilities, I believe (ref. here).

Furthermore, we cannot deal with the "time alignment" in passive system; only the active DSP(EKIO) setup can precisely manage the time-domain tuning including time alignment for all the SP drivers. (I already know this tuning would be not needed for OP's present 2-way experimental setting though.)

We know that the electro-dynamic speaker electro-mechanical model assumes that the speaker is driven from an ideal voltage source. Let me mention universally accepted speaker models like

or

Of course, any change in generator output impedance, i.e. the speaker is not driven from voltage source in case of the passive crossover, must change the speaker transfer function (and distortion) and we have to deal with a new electro-mechanical device.

 

[dualazmak]

Do you have distortion measurements of both?

Sorry, currently I have no objective distortion data before and after. But in 2019, I have confirmed it (improvements by elimination of passive network) by contacting with the retired YAMAHA SP engineer who actually designed and developed NS-1000, NS-100M, NS-1000x, NS-2000.

Which is usually not an issue from sound quality point of view though.

I believe it does have inferior effect on sound quality by "change with age"; especially if electrolyte capacitor(s) are used, and also "the heat" gradually damage the attenuator(s) with year-time.

In my case, even though the attenuator(s) for midrange and tweeter are/were really pro-grade heavy duty ones, I actually found "age dependent" deterioration when I intensively did DIY overhaul on them. I did not take photos during my overhaul, but this reference page exactly shows what I have done on the attenuators. (The perfectly cleaned attenuators are fully bypassed now, though.)

It seems to me you had not a lowpass function enabled on the active crossover, which again would be an pears with apple comparison as different transfer functions, please correct me if I am wrong.

Details of my DSP EKIO XO/deay configurations (ref. here);

It can be partially done with a passive all pass filter, I agree though it is more expediently done in an active crossover, but both are second choice as constant time delay doesn't get a linear phase (summation) like an adjusted FIR crossover.

Yes, I essentially agree with you.

For the completed 3-way (4-way, or more) rather large SP system, time alignment for all the SP drivers should be ideally/hopefully established by the "company" through physical design and tuning, and many of the SOTA HiFi SP producing "companies" take much efforts in the regard, I believe.

Even with my rather "vintage" YAMAHA NS-1000 (now in fully active configuration in my project), my best tuning of time alignment is only 0.3 msec between woofer and midrange+tweeter+super-tweeter (no delay setting needed between midrange, tweeter, super-tweeter) (ref. here).

As for time alignment with L&R sub-woofers (in my case large heavy YAMAHA YST-SW1000), however, I found that I need to have 16.0 msec group delay for woofer+midrange+tweeter+superteeter to time-align with sub-woofers (ref. here and here) which cannot be achieved by usual/standard passive setup. If I would like to achieve this by physically moving the sub-woofers ahead towards my listening position, 16.0 msec corresponds to "5.5 m push ahead" which would be really unrealistic since 5.5 m is far beyond my distance (about 3.3 m) from the SP system. (ref. here for detailed discussion in this regard).

 

[restorer-john]

I have just confirmed it (improvements by elimination of passive network) by contacting with the retired YAMAHA SP engineer who actually designed and developed NS-1000, NS-100M, NS-1000x, NS-2000.

Now that's very cool.

The engineer deserves serious respect.