Building a 2-way small active speaker with software crossover

[dualazmak]

If you have any residual noise with that sensitive tweeter, you can replace your 1st-order DSP filter with a capacitor (5kHz in your 1st iteration example). It would knock down any remaining noise, and provide some measure of protection against issues and accidents.

Yes, I essentially agree with you regarding the needs of low-pass UHF-cut filter(s) in DSP EKIO and also physical protection capacitor(s) for tweeter (and super-tweeter).

Just for your possible reference,,,

As for high-cut (low-pass) filter cutting-off possible UHF (ultra-high Fq) noises, in my almost completed/established DSP(EKIO)-based multichannel setup (ref. here), I set -48 dB/Oct low-pass (high-cut) LR filters at 25 kHz for my midrange-Beryllium-squawker, Beryllium tweeter and also for metal horn super-tweeter (ref. here for summary of "my" rationales.)

As for physical protection capacitors, I use 68 microF (400 VDC) film cap for Beryllium midrange (covering 500 Hz - 6 kHz), 10 microF (400 VDC) film cap for Beryllium tweeter (covering 6 kHz - ca. 15 kHz), and also 10 microF (400 VDC) filme cap for metal horn super-tweeter (covering ca. 8.8 kHz to 25 kHz) (ref. here). Of course, I carefully measured the Fq response before and after the protection capacitors (ref. here and here).

 

[pma]

So - that's it, for the moment . Final 14kHz lowpass was added based on measurements and listening, to cut ugly behaviour of the horn tweeter above 15kHz and also the artifacts

.

 

[dualazmak]

May I ask reasons and pros/cons for your using LR and BW filters?

 

[pma]

HP and LP LR filters when summed at same crossover frequency do not create +3dB peaking, as BW filters do. Here, with not equal cut-off frequencies it may be questionable which filter to use. LR filter is a cascade of 2 BW filters.

 

[dualazmak]

Just let me ask you one more simple question.

As for the -14 dB gain for CH2, we can set it also in EKIO's output panel, and 0 dB (no gain set) in CH2 EQ panel like this;

Which do you prefer setting -14 dB, in output panel or in EQ setting, and why?

 

[voodooless]

Have you done any time alignment? That may be advantageous. Check out the impulse response of the whole.

 

[pma]

Check out the impulse response of the whole.

Do you think I do not?

 

[voodooless]

Do you think I do not?

Do you think I think you do not?

 

[pma]

Do you think I think you do not?

Changing delay in both directions, between W and T branches, makes this, in measurements:

Orange is "no delay", Green is "optimum delay" suggested by VCad optimizer. So, any delay, + or - , only creates bigger or smaller dip near crossover frequency. Horn tweeter driver is relatively deep in the Z axis, which probably creates natural necessary delay.

 

[voodooless]

Changing delay in both directions, between W and T branches, makes this, in measurements:

View attachment 313880

Orange is "no delay", Green is "optimum delay" suggested by VCad optimizer. So, any delay, + or - , only creates bigger or smaller dip near crossover frequency. Horn tweeter driver is relatively deep in the Z axis, which probably creates natural necessary delay.

Thanks, it’s good to show this I think. Since you’re publicly showing your experiment, it nice to cover these aspects as well. Others can learn from it

It indeed makes sense that no-delay should be close to optimal, given the physical dimensions.

 

[dualazmak]

Changing delay in both directions, between W and T branches, makes this, in measurements:

View attachment 313880

Orange is "no delay", Green is "optimum delay" suggested by VCad optimizer. So, any delay, + or - , only creates bigger or smaller dip near crossover frequency. Horn tweeter driver is relatively deep in the Z axis, which probably creates natural necessary delay.

Yes, very nice work!
This is another clear-cut example case for "advanced simulation software does not always suggest proper parameter(s) for real world optimisation", and therefore measurement of real air sound using calibrated microphone is always indispensable.
I too have experienced many similar situations/cases in my DSP(EKIO)-based multichannel project (the latest optimisation including time alignments can be found here).

 

[pma]

Finally, in-room measurements, calibrated at 86 dBA SPL (pink noise calibration with a sound level meter)

1. Frequency response to a sine sweep, no smoothing, no windowing (500ms)

2. Same but with 1/6 octave smoothing

3. Distortion measured at 86 dBA SPL

(as you can see, I prefer logarithmic scale on Y-axis)

I finished with EKIO driving iConnectAudio4+ at 96kHz Fs. 48kHz was too low, as there were SW filter artifacts just above 20kHz.

--------------

.... and, shots of my workplace

 

[dualazmak]

Thank you for sharing your hard work!
May I understand that the dips at around 160 Hz and 205 Hz would be caused by room modes?

 

[pma]

Thank you for sharing your hard work!
May I understand that the dips at around 160 Hz and 205 Hz would be caused by room modes?

Thank you for your interest in this project. Yes, you are right, the dips are exactly the result of room modes. Frankly, the result shown is very good, regarding room modes influence. Usually it is much worse. But of course, if I was measuring from a longer distance, I would get much worse response below 400-500Hz. May I recommend a white paper by Jeff Bagby

http://audio.claub.net/software/FRD_Blender/White%20Paper%20-%20Accurate%20In-Room%20Frequency%20Response%20to%2010Hz.pdf

He states the conditions for microphone placement to be in the woofer farfield and to avoid baffle step influence. For a small box like this, with a mini woofer with 85mm cone radiating diameter, proper results are achieved in the distance of 40cm and more from the front baffle. It makes no sense to measure it from 1m or 2m, because the response is then polluted mostly by room modes. Of course with a bigger speaker box you need to measure from larger distance. I strongly recommend the paper on the link above.

 

[dualazmak]

Yes, you are right, the dips are exactly the result of room modes. Frankly, the result shown is very good, regarding room modes influence. Usually it is much worse. But of course, if I was measuring from a longer distance, I would get much worse response below 400-500Hz.

Thank you, fully agree and understood. And I will carefully read the suggested paper.

I too did intensive Fq response measurements at nearfield and listening position, and in my case I have been rather focusing on the data at listening position while all the L&R SP drivers are singing together (of course reflecting many room modes) which simulating my real music listening sansation with my ears and brain at listening position.

And, I fully agree with you that microphone calibration is important in our rather intensive Fq response measurements.

Just for your possible reference and interest, the best tuned Fq response data as of August 03 2023 I shared here is showing such Fq response at my listening position in my room acoustics. The fine up-down structure is highly reproducible reflecting room modes.

BTW, my ECM8000 measurement microphone (Made in Germany 2008) looks still in really good shape, I mean keeping almost within +/-1.3 dB deviation from full-flat throughout 20 Hz to 20 kHz. Yesterday, we did very preliminary cross-calibration by sending the white noise recording data to my friend who has precisely calibrated Earthwork M50 pairs. I hope I will bring my ECM8000 to his home in early October for intensive calibration using his nice audio setup (of course capable of 16 Hz to 30 kHz), and the result(s) will be shared on my project thread. The precise relative difference between ECM8000 and M60 will be reflected to the calibration curve for M50 to get my ECM8000 calibration curve.

 

[DanielT]

Finally, in-room measurements, calibrated at 86 dBA SPL (pink noise calibration with a sound level meter)

1. Frequency response to a sine sweep, no smoothing, no windowing (500ms)

View attachment 314036

2. Same but with 1/6 octave smoothing

View attachment 314037

3. Distortion measured at 86 dBA SPL

View attachment 314038
(as you can see, I prefer logarithmic scale on Y-axis)

I finished with EKIO driving iConnectAudio4+ at 96kHz Fs. 48kHz was too low, as there were SW filter artifacts just above 20kHz.

--------------

.... and, shots of my workplace

View attachment 314040 View attachment 314041

Good job PMA.
One thing I wonder about, why that distortion peak around 3kHz?

That peak is a bit reminiscent of the Parts Express DIY C-Note Speaker that Amir tested:

Parts Express DIY C-Note Speaker Review

This is a review and detailed measurements of the Parts Express C-Note DIY Kit. It was kindly purchased, put together and shipped to me by member @Winkleswizard. A pair of these with everything you need including the cabinet "flat pack" costs just US $100. And that includes free shipping if...

audiosciencereview.com

Incidentally, a nice tweeter in the C-Note Speaker:

Dayton Audio - ND25FW-4 1" Soft Dome Neodymium Tweeter with Waveguide 4 Ohm

ND25FW-4 1" Soft Dome Neodymium Tweeter with Waveguide 4 Ohm

www.daytonaudio.com

Costs around $17.
A really nice low-budget tweeter, as long as XO is not too low and of lower order XO then.

Order Dayton Audio ND25FW-4 Dome Tweeter With Waveguide

Buy the ND25FW-4 online at SoundImports ✓ Same day shipping ✓ Low shipping rates

www.soundimports.eu

Edit:
Otherwise low distortion then you can use this tweeter that many DIYers now use (and also Revel for example). The performance for it is top notch and thus the price of $50 is incredibly low, considering what you get.
XO maybe even down to 2-2.3 KHz while maintaining low distortion. Provided steep XO.

SB Acoustics SB26ADC-C000-4 | HiFiCompass

The THD measurements at distance of 315 mm and voltage levels 2.83 - 8 Volt were carried out with High Pass Filter "on" (2nd order Buttreworth type, 500 Hz cutoff, HPF2-500). The THD measurements at distance of 315 mm and voltage level 11.2 Volt were carried out with High Pass Filter "on" (2nd...

hificompass.com

 

[pma]

The horn tweeter HT-88 has resonance at 2.5kHz and is suggested to be used above 4kHz. My crossover frequency is near 3kHz and what you see is the tweeter distortion, unfortunately. Seen in individual driver measurement as well. It goes down above 4kHz, or at lower level. It is fine up to some 80dB SPL.

 

[pma]

Thank you, fully agree and understood. And I will carefully read the suggested paper.

I too did intensive Fq response measurements at nearfield and listening position, and in my case I have been rather focusing on the data at listening position while all the L&R SP drivers are singing together (of course reflecting many room modes) which simulating my real music listening sansation with my ears and brain at listening position.

And, I fully agree with you that microphone calibration is important in our rather intensive Fq response measurements.

Just for your possible reference and interest, the best tuned Fq response data as of August 03 2023 I shared here is showing such Fq response at my listening position in my room acoustics. The fine up-down structure is highly reproducible reflecting room
BTW, my ECM8000 measurement microphone (Made in Germany 2008) looks still in really good shape, I mean keeping almost within +/-1.3 dB deviation from full-flat throughout 20 Hz to 20 kHz. Yesterday, we did very preliminary cross-calibration by sending the white noise recording data to my friend who has precisely calibrated Earthwork M50 pairs. I hope I will bring my ECM8000 to his home in early October for intensive calibration using his nice audio setup (of course capable of 16 Hz to 30 kHz), and the result(s) will be shared on my project thread. The precise relative difference between ECM8000 and M60 will be reflected to the calibration curve for M50 to get my ECM8000 calibration curve.

I would kindly recommend you to use 50dB span on Y-axis, it is being a standard in acoustical measurements. In case you use much larger span, as you do (your span is 90dB) the response curve looks smoother, but only at first view. 50dB is good to use for a quick visual comparison.
Just a small remark, the white noise is definitely not a good method to test the speaker.

 

[DanielT]

The horn tweeter HT-88 has resonance at 2.5kHz and is suggested to be used above 4kHz. My crossover frequency is near 3kHz and what you see is the tweeter distortion, unfortunately. Seen in individual driver measurement as well. It goes down above 4kHz, or at lower level. It is fine up to some 80dB SPL.

That seems reasonable. It is not possible with any type of crossover, analog, digital, filter and so on to conjure up the inherent limitations that drivers have. They are what they are.
Of course you know this, but having said that, what I find most interesting about this thread, is how much you can get out of the drivers you use.

Edit:
Low distortion in all its glory but it doesn't beat a good FR.

 

[pma]

Impedance plots of the drivers within the cabinet