I was extremely unlucky. I bought two new units. One had a problem with the HF part, the other one with the MF part. O returned them under warranty and I am waiting to have them fixed.
I remembered someone on diyaudio also face the HF frequency response issue.
After the examination there's some glue left in the HF coil gap at that case.
Luckily mine are okay...I bought from abroad and it'll be painful if repair is needed for me...
Edit: this thread
https://www.diyaudio.com/community/threads/dcx464-busted-5db-difference-in-hf.368946/
After the examination there's some glue left in the HF coil gap at that case.
Luckily mine are okay...I bought from abroad and it'll be painful if repair is needed for me...
Edit: this thread
https://www.diyaudio.com/community/threads/dcx464-busted-5db-difference-in-hf.368946/
Yes @Stanislav : I agree.
After many experimentations and listening comparisons with 6dB/Oct. // and 12dB/Oct. // configurations, I'll even add some other advantages to the serial 6dB/Oct. crossover :
1 - the speakers are connected in same polarity. This means that on a transient wave, the cones will move in the same direction, without any other delay than the construction flaws of each speaker can provide, unlike on most higher order crossovers.
2 - the speakers are in Serie, and connected directly to the amplifier's terminals, without passing through a LC crossover component (wired in // to the speakers). This means that no possible phase gap or rotation exists than the construction flaws of each speaker can provide.
3 - the number of parts are minimalistic. This simplicity means that the insertion losses are reduced, and as they are not in serie with the speakers, they possibly have less impact on efficiency.
4 - each speaker "sees" the same LC assembly, conversely to the // 6dB/Oct. configuration, where each speaker "sees" a L or a C. A parallel 12dB/Oct. offers also a LC assembly, so it may be an explanation for the attenuation slopes of the Serie 6dB/Oct. closer to a 2nd order slope.
All my DIY speakers have now 6dB/Oct. Serie crossover, being 2-ways or 3-ways. I even went for this solution when I rebuilt the crossovers of my Magnepan SMGb. The results are amazing and at any rate, I wouldn't go back !
T
No, because I do not have the instruments to do this anymore (obsolete software, thank you Windows).
I think that the slope is between 6dB/Oct. and 12dB/Oct, if I rely on several (but few) studies I read on the subject (including those mentioned by @Stanislav in his post). I can mention others, but these are in French :
http://jimbee.over-blog.com/page-5371974.html
https://www.audiotechno.fr/php/filtre-serie.php
T
No: the -claimed- advantages are no longer there as soon is the slope is no longer 1st-order. Any decent simulator can sim parallel networks.
Interesting. I hope I don't get thread sidetracked, but what kind of rules is involved in this first-order series filter when thinking about the distances/ C to C between the drivers? Was it that all first order filters (parallel or series) require very short/shorter than usual distances to work properly? For example, If using Mabat´s Exar 400 horn and instead using steep filters at approx 700Hz would it be worth to try with first order series filter? C-to-C would be huge if using 15" ..
Last edited:
Same rules as for any other slope or order.
C to C ideally should by 1/4 wavelength of the XO frequency, in many cases not possible.
I have a pair of OB speakers with a 15" woof crossed at around 250 - 300 Hz, followed by an 8" FR (SB20). The C-C distance is roughly 35cm.
So 344/300=1.146m, this is the wavelength.
1.146/4=0.28m, just about right.
In my case a smaller FR will help.
You can't simulate these in any of the free crossover design packages yourself?
Attached. This is assuming a purely resistive 8ohm electrical load, flat FR, phase & coincident acoustic centres (so we're just looking at the electrical transfer functions: the second you apply real-world drivers to this, with their frequency-varying impedance loads, variable frequency/amplitude responses & offsets between the drivers in the 3-planes it all goes to pieces).
1st order parallel, 1st order series, 2nd order Butterworth & LR2 shown with a 1KHz nominal frequency. I included two different 2nd order filters since you didn't specify any particular type -in this case Butterworth for type-consistency & LR2 because it electrically speaking sums flat. The 2nd order filters both have the HF polarity flipped as they are 180 degrees out of phase in the same polarity. 1st order sums with 90 degrees phase difference, whether it's a series or parallel filter. No mysteries here. With a purely resistive electrical load & ignoring everything else, 1st order series & parallel are identical in terms of the summed / system impedance, frequency response & phase shift. It's when you get to real-world applications that you can start to see functional differences -but for that, we'd need to have the measured FR, phase & impedance data for a given speaker.
Attachments
Last edited:
Yes, that's it. I did my best to achieve that on my 375L Monitors (DIY) - and yes, this is not always possible... Here, FC are 5000Hz and 400Hz.
Below, at left, the 12dB/Oct. // version ; at right, the 6dB/Oct. Serie version - both listening-tested on the speakers above, in instant A/B test, with the same sources / amps / musical program and samples :
No picture : the 6dB/Oct. Serie configuration completely buries the 12dB/Oct. // configuration above, aurally speaking.
Yes. If we did not tested it - as objectively and seriously as possible - with Audio friends, we would never have believed it ! Like what, as often, la carte n'est pas le terrain ( the map is not the terrain)...
T
THANKS Scott. But I already knew the simplistic story of this four-way comp and it does not show/explain features such as slope and phase of 1st-order series vs the rest. Hence my request (even begging) for the difference to be sim'ed, if only for a "typical" or ideal set of drivers (impedance curves, phase) and test conditions (baffle layout, aligned acoustic centers, etc).
May I also propose a "null hypothesis" experiment: take two identical drivers, near-flat impedance area, crossover one to the other. Should be zero audible difference series vs parallel 1st-order, correct?
Last edited:
I'd also point out that I did show the individual filter phases in the four simple simulations above. See attached; phase indicated by the arrows. The filter & phase slopes are identical between 1st order series & parallel since this is an identical static / resistive load.
Attachments
Last edited: