Accidentally posted before I was done typing.
I think most of us would do things different than each other. That's part of the art of speaker design.
I think most of us would do things different than each other. That's part of the art of speaker design.
This thread reminds me of what not to not post here (DSP dial-in data)...
However, based on John Dunlavy's last interview at Stereophile and his admission that he was intending to build a fully active (DSP) version of his "Magnus" design, so it does hold some interest:
![Magnus2b[1].jpg](https://www.diyaudio.com/community/data/attachments/1366/1366373-0a6ebcac1ac67ba4dd14ff80f503cbb6.jpg?hash=mQQbfFqpti "Magnus2b[1].jpg")
Chris
However, based on John Dunlavy's last interview at Stereophile and his admission that he was intending to build a fully active (DSP) version of his "Magnus" design, so it does hold some interest:
Chris
Last edited:
Interesting paper - but only ONCE - is music mentioned in the test used - and I quote:
"The conclusions of various tests have been that large enough group delay errors may produce audible errors, but when listened to music or some other real sound material, the differences are often inaudible"
I'm not saying that we should not make further effort to improve everything possible - but we should also consider the challenges every time we bring a new tool to the workshop. An active system brings many good things to the table - many advantages. But again, these are only advantages when used cleverly.
In most cases, people still battle with some of the most basic in loudspeaker design like resonances, break-up modes, filter slopes, power response, EQ, baffle design, room acoustics, over-all setup etc.
I could agree, that linear phase could be the icing on the cake, when everything else is absolutely done almost perfectly..... but until then... back to the drawing board
😉I do think that he may need some help on measurement technique. ( That's just an assumption based on what's been posted. ) What you suggested is good. I'm still asking questions trying to see what he's actually done so far. He may not want my advice either. I assume that he has made on axis measurements, but he has not actually said so. There are some fine points in making a good x-over that the OP may not be aware of. Or, since he has been at this a while, he may just not have mentioned some of the details, or maybe he has figured out other ways to get the job done. I'm trying to help if needed, but not force my design ideas on him. I don't think you are either.
In general I agree, the problems you listed have a much greater impact on the final result than the linear phase. The linear phase cannot eliminate the influence of the room on the sound, etc., etc. Therefore, yes, the linear phase is the icing on the cake, not the cake itself.
It is not always clear that some advice comes from practical experience. When you focus too closely on one thing, other things can suffer... (and not for want of trying, sometimes the practical compromises force it to be that way.)
To summarize the discussion, seems to me like the OP
- uses DF75 preset crossovers, so only cutoff freq and slope adjust, no level and eq https://www.accuphase.com/model/df-75.html
- uses no equalization for each pair of drivers/way to adjust baffle diffraction and cone resonances etc
- uses smaart to adjust delay for each way and perhaps spl response at listening spot (5 meters away!)
- we have not seen full system spl and phase response measurement...
- measurements at listening spot are corrupted with room reflections and modes
- to handle baffle diffraction and driver nonlinearities in acoustic domain (=equalization, requires quasi-anechoic measurements) like the original passive speaker did
- to check/adjust delays so that the work with the xo slopes (IIR)
- - measuring individual ways and matching delays does not take into account crossover-induced phase shift as part of a system
- the OP is measuring the speaker with set xos and trying to force step response to single peak by adusting delays radically.
- the OP is ignoring spl response and directivity
Last edited:
Just use 4-5 measurements in the listening position, push a button and let DIRAC figure it out - voilà - you now have a perfect speaker. I'm kidding, of course 😉
Indeed. For laughs, here's what I would do (or have done just not so BIG). Three matched integrated amps (e.g. my TPA3221 bpl) plus a subwoofer plate amp with continuous phase knob for the bottom. With each driver-pair in LX position wired opposite polarity determine its combined null i.e. acoustic center offset. Mount the drivers using shims or custom adapters to ensure time alignment (as Dunlavy intended). Cover baffle surfaces with textured wool felt; add rounded moulding to cabinet hard edges. LCR notch filter to rid all drivers of resonance peaks within two octaves of intended range. Balance them with individual volume controls. And finally --
Set upper active crossovers all to 1st-order. By trial-and-error tweak each XO frequency up/down until phase is matched. This can be determined by listening to test music, and found/verified playing test tone at said XO frequency -- shifting driver offset would decrease combined volume, or undo null if reverse polarity. Done.
Other than present-day finer aspects of in-room response/directivity control, any technical reasons this wouldn't work in principle, everyone? Thanks.
Last edited:
After replacing the Accuphase DF-45 with the DF-75, I discovered that the S/PDIF digital signal from the Zidoo UHD8000 is incompatible with the DF-75. All the LEDs on the DF-75’s front panel kept flashing. At first, I thought the brand-new DF-75 was defective. After checking the user manual, I found out that the DF-75 is only compatible with S/PDIF signals that include volume information (IEC60958/AES3), whereas the DF-45 is compatible with S/PDIF signals without volume information (JEITA CP-1201/AES3).
The S/PDIF signal from the Zidoo UHD8000 is of the type without volume information, so it must first pass through the Accuphase DC-330, which converts it into an S/PDIF signal with volume information in order to be compatible with the DF-75.
When I first set it up, the sound was muffled, lacking resolution and spatial depth. So I replaced the Zidoo with an AURALiC Aries G1, connecting its AES/EBU output to the AES/EBU input of the DC-330, then DC-330 connect to DF-75 with HS-Link cable. The sound immediately returned to normal.
Using SMAART V8, I re-measured and adjusted all the DF-75’s settings (initially copied from the DF-45). The final crossover settings are:
I would rate the sound quality of my current active crossover system (DA + Scan drivers) at 90 points, compared to the original passive crossover system (all DA drivers) at 70 points. I have no regrets about the changes I’ve made so far.
The S/PDIF signal from the Zidoo UHD8000 is of the type without volume information, so it must first pass through the Accuphase DC-330, which converts it into an S/PDIF signal with volume information in order to be compatible with the DF-75.
When I first set it up, the sound was muffled, lacking resolution and spatial depth. So I replaced the Zidoo with an AURALiC Aries G1, connecting its AES/EBU output to the AES/EBU input of the DC-330, then DC-330 connect to DF-75 with HS-Link cable. The sound immediately returned to normal.
Using SMAART V8, I re-measured and adjusted all the DF-75’s settings (initially copied from the DF-45). The final crossover settings are:
- Crossover points: 100 / 2000 / 7000 Hz
- Slopes: all at -96 dB/oct
- Delays: SUB = 0 cm, LOW = 0 cm, MID = 19 cm, HIGH = 35 cm
I would rate the sound quality of my current active crossover system (DA + Scan drivers) at 90 points, compared to the original passive crossover system (all DA drivers) at 70 points. I have no regrets about the changes I’ve made so far.
Last edited:
At first glance this evidence suggested to me the critical coherent frequency range is 2-7khz covered by the original single driver; the XOs perhaps not so critical for coherence phase-alignment-wise, so long as time-delays synchronized all the drivers to arrive concurrently. This interpretation seemed problematic. So were they time-aligned or not i.e. DSP-equidistant from each acoustic center to listener's ear?
Last edited:
Green trace is original DUNTECH Sovereign 2001(passive Xover 290/2000/6000 Hz+all Dynaudio units)
Red trace is DF75 active 4-ways Xover 100/2000/7000 Hz+SUB/LOW/HIGH ScanSpeak/MID Dynaudio units.
Measured in axis and front of the Left speakers about 2.5M, with UMIK2+REW+1/12 smooth Graph.
Attachments
In the mid-to-high frequency range, time-alignment—having each driver’s sound arrive at the listener’s ear simultaneously—is more important than phase-alignment. However, in the low-frequency range, phase-alignment—aligning the phase of the sounds arriving at the listener’s ear—is more important than time-alignment.
Therefore, when measuring with SMAART V8, I adjusted the time delay of the LOW channel on the DF75 so that the phase trace of the SUB overlaps. Then, I set the time delays for the MID and HIGH channels of the DF75 to match the LOW channel’s time delay.
Thirty years ago, before measurement software like SMAART became available, I also used a method similar to yours for system tuning. The difference was that, at the time, I couldn’t find a subwoofer amplifier with a continuous phase knob—only ones with a simple 0/180-degree phase switch were available.
Looks pretty good. What does it look like when gated?
Also, as a x-over nerd, I like to see individual drivers, and see how they cross, and sum.
Also, as a x-over nerd, I like to see individual drivers, and see how they cross, and sum.
DF75 has level adj. in every channel, but has no EQ. You have to put ACCUPHASE DG68 (GEQ) between DC330 preamp and DF75 or gear with DIRAC (ex. minidsp SHD studio), but I prefer to keep it simple, nature, and clean. I've tried all of those before — they made the sound unnatural and less effortless.