Amen. 😀 that interview was one of the cornerstones in my starting to learn about DIY speakers.
On the last page i think, he talked about radiation angles of drivers vs their size.
Here's a chart I made some years ago (needs proofing) Green zone be good zone.
I'd love to have the Duntech Sovereign to experiment with. Given the design incorporates first order acoustic crossovers, it would make the perfect test bed for comparing higher-order linear phase crossovers. The hard work is already done for such a comparison !!
I try to compare low order vs high on my syn horns, but without the well behaved out of band extensions John D achieved, anything below 4th order becomes compromised. I don't think that's a problem really, due to the higher order syns sounding so damn good, but I always like to compare.
Which I bet John D would too. The full multi-channel active FIR DSP Magnus sounds like he was ready to take it to the next level. Multi-channel active has it's own inherent improvements over passive. And better time alignment, and ability to use more finely tuned EQ's are the obvious starter DSP improvements.
It would make an even more perfect test bed for higher order complementary linear phase xovers, imo. Unless first order was convincingly superior at normal listening levels, I'd still probably opt for higher order just to gain some safe SPL. First order acoustic has to be a hard thermal load for a driver.
Do you know if the Duntech Sovereign's had any problems with driver sections burning out?
I changed the crossover slope of the SUB/LOW from 24dB to 96dB because the phase trace difference between the left and right SUB/LOW at the 100Hz crossover point is smallest at 96dB—around 60 degrees. If I align the four phase traces of the left SUB, left LOW, right SUB, and right LOW (phase alignment), I need to increase the time delay of the right SUB and right LOW by 60cm. The result is a low-frequency response that is clean and powerful—unlike anything I’ve heard before!
To establish how you work with this project, we should start by knowing more about how you measure - IMO - so we can better understand from what data you adjust the DSP.
I definitely understand how you can be excited about active filtering, since I have been using DSP to filter everything I have been doing for many years now.
But I also know from experience - now, that is 😊- that it is so easy to make a mistake in a DSP, because of a mistake in a measurement. So this is why I kindly ask for reassurance, when it comes to any measurement and measurement technique.
So, first we need to know exactly how your setup looks like, when you measure. Distance from speaker to any reflective surface, mic-stand, distance between speaker and mic ( 5m sounds too much ), gating/smoothing and so on. This all plays a role to fully figure out whether your measured data is leading you to the best possible adjustment in the DSP.
Speakers only have several drivers because one can't do it all - and we actually want sound to come from one "spot" from each speaker, which again is the whole original idea of this speaker, namely to make several drivers sound like one speaker. By using 96dB slopes, you might get "tighter" bass - subjectively, but that is just one driver, you need the whole speaker to sum as one.
Before doing anything else - I would pull the speaker to the most free spot you possibly can, measure at maybe 1,5-2 meter distance - which is tricky - but you need to have distance that makes the drivers sum - not near field, because you might only get valid data down to 1kHz, but you can make it a little better with some damping at the first reflections.
Then measure at several angles without moving the mic, but only turning the speaker, so we get a rough picture of the radiation pattern of all drivers - since you changed them - before any idea of filter slopes and EQ can even remotely be settled.
I would confirm tweeter and midrange coherence and then focus on the woofers as you move the mic to the listening position. I still do not understand why you need to delay the woofers at all 🤔
You need good data to make a good speaker.
I definitely understand how you can be excited about active filtering, since I have been using DSP to filter everything I have been doing for many years now.
But I also know from experience - now, that is 😊- that it is so easy to make a mistake in a DSP, because of a mistake in a measurement. So this is why I kindly ask for reassurance, when it comes to any measurement and measurement technique.
So, first we need to know exactly how your setup looks like, when you measure. Distance from speaker to any reflective surface, mic-stand, distance between speaker and mic ( 5m sounds too much ), gating/smoothing and so on. This all plays a role to fully figure out whether your measured data is leading you to the best possible adjustment in the DSP.
Speakers only have several drivers because one can't do it all - and we actually want sound to come from one "spot" from each speaker, which again is the whole original idea of this speaker, namely to make several drivers sound like one speaker. By using 96dB slopes, you might get "tighter" bass - subjectively, but that is just one driver, you need the whole speaker to sum as one.
Before doing anything else - I would pull the speaker to the most free spot you possibly can, measure at maybe 1,5-2 meter distance - which is tricky - but you need to have distance that makes the drivers sum - not near field, because you might only get valid data down to 1kHz, but you can make it a little better with some damping at the first reflections.
Then measure at several angles without moving the mic, but only turning the speaker, so we get a rough picture of the radiation pattern of all drivers - since you changed them - before any idea of filter slopes and EQ can even remotely be settled.
I would confirm tweeter and midrange coherence and then focus on the woofers as you move the mic to the listening position. I still do not understand why you need to delay the woofers at all 🤔
You need good data to make a good speaker.
Sorry for the earlier mistake! The phase trace difference of 60 degrees between the left and right speakers occurred because I didn’t position the measurement microphone exactly at the midpoint between them. After correcting the microphone placement, there's no longer any phase trace difference between the L/R speakers.
When setting the SUB/LOW crossover frequency to 100 Hz with a -24 dB slope, the LOW section needs a delay of 30 cm to align the phase traces of the SUB .However, with a -96 dB slope at the same crossover frequency, no delay is needed for proper phase alignment between SUB and LOW.
I’m planning to test a -96 dB slope for the LOW/MID and MID/HIGH crossovers as well, to see if it results in better sound compared to the -24 dB setting.
I’ve listed the steps for SMAART V8 measurement as follows:
My interface is a MOTU896 Hybrid. For TF (Transfer Function) measurements, both the reference signal and the measured pink noise sent to the DF45 are output via SPDIF (using a Y-connector for dual output).
My interface is a MOTU896 Hybrid. For TF (Transfer Function) measurements, both the reference signal and the measured pink noise sent to the DF45 are output via SPDIF (using a Y-connector for dual output).
- Microphone height: Adjusted to match the vertical center of the DUNTECH's HIGH driver at 95 cm.
- Left-right microphone positioning: Set all DF45 SUB/LOW/MID/HIGH delays to 0. Adjust the microphone’s horizontal position so that the delay finder shows identical values for the LEFT and RIGHT HIGH (or MID) drivers (e.g., my value is around 18 ms).
- Measuring the phase trace between LEFT SUB and LOW:
- First, capture the phase trace of the SUB with delay finder set to 0 ms.
- Turn off the DF45 SUB output. Zoom in on the phase trace around 100 Hz.
- Turn on the DF45 LOW output. Adjust the DF45 LOW delay until the SUB and LOW phase traces overlap.
- Repeat the same procedure for the right side.
- Measuring the delay between LEFT MID and HIGH:
- First, use delay finder to measure the LOW delay value, and note this value (e.g., mine is 17.80 ms). This will be the system’s reference point (the time the sound reaches the listener).
- Then, measure the MID and HIGH delays using delay finder. Adjust the DF45 MID and HIGH delays until the measured delay equals the reference point.
- Repeat the same procedure for the right side.
Many people might ask: Why not just use the original Dynaudio drivers and passive crossovers?
The reason is that the Dynaudio drivers have been in use for over 30 years, and there are no new replacements available. The woofer surrounds are also in poor condition. In the future, they would need to be replaced every few years — a hassle that doesn’t necessarily restore the original performance.
I want to use woofers that don’t require surround replacement. And if new woofers with better performance become available (like Eton or AT), I’d want the flexibility to upgrade. Isn’t that the most exciting part of hi-fi and DIY? Otherwise, you might as well just buy Wilson Audio’s flagship speakers and do nothing.
The main reason, though, is that the stock DUNTECH Sovereign 2001 doesn’t meet my expectations for clean sound — especially in the bass; not tight enough. I’ve always felt the original 290Hz crossover point between SUB and LOW is too high, which makes female vocals sound unclear and nasal.
I want to lower the crossover point to 100Hz to improve vocal clarity and loose bass. The only way to do this is through active crossovers or by modifying the original low-pass section. I also crave steep-slope crossovers for the low end (tight bass), which again requires active crossover.
Combined with my strict demand for time and phase alignment, DSP-based digital crossovers become the only viable solution. Not even the analog 4-way electronic crossover in the Nautilus can achieve that.
The reason is that the Dynaudio drivers have been in use for over 30 years, and there are no new replacements available. The woofer surrounds are also in poor condition. In the future, they would need to be replaced every few years — a hassle that doesn’t necessarily restore the original performance.
I want to use woofers that don’t require surround replacement. And if new woofers with better performance become available (like Eton or AT), I’d want the flexibility to upgrade. Isn’t that the most exciting part of hi-fi and DIY? Otherwise, you might as well just buy Wilson Audio’s flagship speakers and do nothing.
The main reason, though, is that the stock DUNTECH Sovereign 2001 doesn’t meet my expectations for clean sound — especially in the bass; not tight enough. I’ve always felt the original 290Hz crossover point between SUB and LOW is too high, which makes female vocals sound unclear and nasal.
I want to lower the crossover point to 100Hz to improve vocal clarity and loose bass. The only way to do this is through active crossovers or by modifying the original low-pass section. I also crave steep-slope crossovers for the low end (tight bass), which again requires active crossover.
Combined with my strict demand for time and phase alignment, DSP-based digital crossovers become the only viable solution. Not even the analog 4-way electronic crossover in the Nautilus can achieve that.
Last edited:
I don't see the logic. Was there something wrong with the crossover?
This is also questionable. What benefit do you expect it to bring?
Thanks allenB! ChatGPT have good answers for me.
- When all frequency bands arrive at the ears simultaneously (in phase), the imaging becomes more precise and the soundstage more stable.
- Vocals and instruments sound clearer, more focused, and detailed.
✅
- Phase alignment prevents interference and coloration around the crossover regions.
- It significantly improves issues such as boomy bass, nasal mids, and harsh highs.
✅
- When drivers are out of phase, overlapping frequency ranges can cause constructive and destructive interference, resulting in frequency dips or peaks.
- Time and phase alignment reduces this effect, creating a smoother and more accurate frequency response.
✅
- For percussive or transient-rich sounds (like drums or piano), aligned drivers produce sharper, faster, and more impactful sound.
✅
- In a 4-way active DSP crossover system (like the Accuphase DF-45), precise delay and phase control is available. However, without proper alignment, these features can actually degrade sound.
- Alignment ensures the DSP’s potential is fully realized, allowing the system to perform at its best.
In short, time and phase alignment is fundamental to achieving high-fidelity reproduction in a multi-way speaker system. This is why it’s a standard in professional active monitors and high-end audio systems.
I assume that the chat has simply repeated what others have said. I wouldn't put too much confidence in it.
I assume you posted this thread just for conversation, and enjoyment of the hobby. Some of us nerds like to see measurements, and talk about x-overs. AS big as those speakers are, how do you get them up away from boundaries to take measurements? They must weigh a ton.
I assume you posted this thread just for conversation, and enjoyment of the hobby. Some of us nerds like to see measurements, and talk about x-overs. AS big as those speakers are, how do you get them up away from boundaries to take measurements? They must weigh a ton.
If anyone is interested in seeing a full measurement video using SMAART V8, I'm not very familiar with video editing, but I'm willing to give it a try. That way, everyone can better understand the process. As for the big speakers—they’re not going anywhere; it's just me running back and forth!
Time and again these misunderstandings and incorrect information goes around.
If time/phase misalignment were that easy to hear, then most commercially available speakers wouldn't sell at all! Even for myself, the perceived improvement from time/phase alignment might be purely psychological rather than auditory. However, lowering the crossover frequency from 290Hz to 100Hz and increasing the slope from 24dB to 96dB made a very noticeable improvement in sound quality!
What are you using to adjust SPL levels of each driver? Are you also changing polarity when needed between tweeter / mids / woofers? Are you measuring the individual driver responses?
It's strange. It seems obvious that a flat phase response gives less linear distortion. Which is important when working with sound in recording studios.
In order for all frequencies to come to the ears at the same time, you must obtain a similar phase-frequency response in the listening area.
Unfortunately, IIR crossovers such as LR24 and similar cannot provide such a phase-frequency response. A flat phase-frequency response can only be obtained with two types of crossover filters, these are FIR and subtractive.
I would not claim that subtraction is needed. What you need to get a linear phase-response (when not using FIR, or subtractive-delay) is a so-called constant-voltage crossover. One means of building them is by subtraction but there other possibilities as well. And don't forget to take the bare driver's response into consideration also.
Regards
Charles
Regards
Charles
True.. everything is a compromise 😉 Still, many still debate whether a linear phase really is an absolute requirement for high audio fidelity.
But the OP should definitely consider what the actual goal is with this design, especially when most people who try to advise him, would have done it quite differently.
I don't see any reason for debate. If the condition is to get high audio fidelity, then the linear phase as well as the linear frequency response are the conditions of this very high audio fidelity. Or by high audio fidelity everyone understands something different.
Fidelity in my language means to get exactly the same as the original. Phase distortions, like frequency response distortions, are a type of linear distortion, so to get an accurate reproduction of the original signal we must eliminate linear distortions, i.e. must have a linear phase and a linear frequency response.
I also meant, that many still debate whether it is actually audible to have a linear phase. I've heard it several times, but still I have yet to experience this bliss of high fidelity, just because the phase is linear.
I would like it to be so, in a way that was night and day, so I knew whether to chase it or not. But to this day, I haven't heard a single proper blind test or demonstration between a linear or non-linear phase, which clearly defined the audible difference - not the measured - but the audible.
Smooth and even power response + room correction below Schröder - yes - that I can hear, and that is something I believe is high fidelity enough for me - for the time being 🙂
You can read this study. It is slightly biased towards the audibility of pre-ringing in FIR filters, but in general it also conducts studies on the audibility of phase shift in IIR filters of the LR24 type. In general, as I recall, the audible limits of nonlinear phase have long been determined. Roughly speaking, if the group delay is below 3.5 ms, you don't have to worry too much. But if we are talking specifically about high audio fidelity, then the need to have a linear phase comes from the very definition of high audio fidelity.