Yes, often the electrical filter is not Bessel or BW4 and plus requires some PEQ to effect the textbook BW4 on woofer and Bes2 on tweeter. Best way is to use simulator program and tweak raw measurements until it fits. PCD or Xsim are great for this.
Yes the dcx2496 has both bw4 and bes2, the problem I encountered is that xsim only has bes2 lowpass, so I'll have to model it by custom q filters or BYRTT's workaround. Or try this PCD software
Project summary
I followed your advice and gone back to Xsim with raw driver response curves, flattened them (only a few low-Q bands were needed) out and then applied filters/delays, thanks BYRTT for the LR2+PEQ tip!
Everything seemed alright so I proceeded to implement all that in DCX2496. And oh my gosh - I finally have a working Harsch alignment 😎 Please ignore the room effects below 200hz...
For reference I also measured LR2 at 300Hz which I also liked:
Phase of both setups:
Two things I like most about these speakers:
1. Relatively deep ant tight bass, with some mild EQ I can easily hit 40Hz
2. Very clean sound overall - I'm having trouble measuring distortion even at moderately loud levels...
I also looked at waterfall but it's so smooth I'm beginning to doubt if I'm doing it correctly:
For future reference, drivers used:
SB Acoustics SB23NRXS45-8, 8 ohm version
Tymphany TG9FD10-04, 4 ohm version
All measurements performed indoors with a ECM8000 microphone calibrated at Cross Spectrum Labs.
Sealed cabinets were made from a single sheet of 15mm Baltic birch plywood, front baffle doubled for rigidity, one cross section. External dimensions: 280x350x450mm. All joints glued using PVA glue (D3 grade). Fullrange enclosure is conical polycarbonate flower pot, internal volume about 3.3l, half stuffed with felt and rockwool as per xrk's instructions.
Front baffle chamfered with 1/2" roundover bit, all drivers flush mounted. Final finish - glossy polyurethane varnish. They won't win any awards for innovative design but WAF requirements are met 😀 Here's the final look:
Thank you guys for all the help and advice while stumbling towards this result! 🙂
I followed your advice and gone back to Xsim with raw driver response curves, flattened them (only a few low-Q bands were needed) out and then applied filters/delays, thanks BYRTT for the LR2+PEQ tip!
Everything seemed alright so I proceeded to implement all that in DCX2496. And oh my gosh - I finally have a working Harsch alignment 😎 Please ignore the room effects below 200hz...
For reference I also measured LR2 at 300Hz which I also liked:
Phase of both setups:
Two things I like most about these speakers:
1. Relatively deep ant tight bass, with some mild EQ I can easily hit 40Hz
2. Very clean sound overall - I'm having trouble measuring distortion even at moderately loud levels...
I also looked at waterfall but it's so smooth I'm beginning to doubt if I'm doing it correctly:
For future reference, drivers used:
SB Acoustics SB23NRXS45-8, 8 ohm version
Tymphany TG9FD10-04, 4 ohm version
All measurements performed indoors with a ECM8000 microphone calibrated at Cross Spectrum Labs.
Sealed cabinets were made from a single sheet of 15mm Baltic birch plywood, front baffle doubled for rigidity, one cross section. External dimensions: 280x350x450mm. All joints glued using PVA glue (D3 grade). Fullrange enclosure is conical polycarbonate flower pot, internal volume about 3.3l, half stuffed with felt and rockwool as per xrk's instructions.
Front baffle chamfered with 1/2" roundover bit, all drivers flush mounted. Final finish - glossy polyurethane varnish. They won't win any awards for innovative design but WAF requirements are met 😀 Here's the final look:
Thank you guys for all the help and advice while stumbling towards this result! 🙂
Attachments
I think you have the drivers' polarity inverted or have the delay between them set wrong, as the woofer and mid-tweet don't seem to sum up in this graph.
P.S. Harsch crossover does indeed sound nice. I was very sceptical along the way and tried to avoid placebo effect at all costs. However DCX2496 is capable of storing complete setups, including delays, filters and polarities, effectively enabling real-time switching between different crossover setups. The crispiness of percussive sounds on Harsch is indeed very obvious. I'll have to do some more listening to find out if I like it universally but yes, there's at least some truth to the Harsch hype.
I think you have the drivers' polarity inverted or have the delay between them set wrong, as the woofer and mid-tweet don't seem to sum up in this graph.
Hi Perceval,
I see why one might think that, but the phase plot would not have a rather gradual monotonic variation if the polarity was inverted:
Plus, that dip is on the midtweet all by itself (blue curve), so it cannot be result of XO-induced cancellation:
So in short, I think he has got the Harsch XO implemented correctly. The two giveaways are the slight reduction in the woofer to the left of the XO, and the step response showing the positive midtweeter peak on top of the woofer "breadloaf" 🙂
Congratulations Wxn, you got it dialed in correctly. Enjoy your new speakers. Now, the problem is, when you go to listen to commercial speakers costing many times more in a listening boutique, you will most likely come away unimpressed as your home speakers sound better. That is a very nice distortion plot!' The DCX2496 seems to be doing a formidable job here, very nice to know that it works so well.
And that waterfall is gorgeous looking and sounding! 😀
p.s., I have several sets of 3/4in BB ply baffles all precut, rebated flush and rear-chamfered for for the SB23's plus Visaton B80's or another set for 10F's. I have the SB23's sitting in a box still so maybe time to try making another set of speakers? Will fuddle my way through a passive XO if possible but looks like the SB23 might need some PEQ to work properly.
Last edited:
Beautiful looking speaker build and research skills there wxn 🙂 will imagine that DCX2496 unit is very handy dial in and research the least small bits of improvements for tonality and etc should there be any.
Hi Byrtt,
Happy belated Xmas! IIRC, did you not once perform surgery on your DCX2496 in order to eek out the most performance with changing the output driver opamp?
Cheers,
X
Happy belated Xmas! IIRC, did you not once perform surgery on your DCX2496 in order to eek out the most performance with changing the output driver opamp?
Cheers,
X
Hi X,
Happy Xmas, myself unfortunately have no DCX2496 unit but 4 times of DEQ2496 units which in past i used to do XO duty because they have the power and features to also do that task although its not directly advertised, about ab fabric output stages in those Behringer units i'm not so sure they so bad in reality compared to other mass fabricated stuff, will say at least my diy lab gear can not tell of any bad numbers inside normal audio band, that said i did modify output stages for two reasons, one is these units belong to pro gear sector and therefor output voltage is so huge one can imagine its close to also swing real tubes circuits 😀 the other reason was i'm a sucker for sound signature of DC coupled circuits.
Happy Xmas, myself unfortunately have no DCX2496 unit but 4 times of DEQ2496 units which in past i used to do XO duty because they have the power and features to also do that task although its not directly advertised, about ab fabric output stages in those Behringer units i'm not so sure they so bad in reality compared to other mass fabricated stuff, will say at least my diy lab gear can not tell of any bad numbers inside normal audio band, that said i did modify output stages for two reasons, one is these units belong to pro gear sector and therefor output voltage is so huge one can imagine its close to also swing real tubes circuits 😀 the other reason was i'm a sucker for sound signature of DC coupled circuits.
Last edited:
perceval's observation got me thinking so I redid some measurements. First the same graph as before, just repositioned speaker and the microphone. This is with 1.31ms delay btw:
Anyway, I did series of measurements with different midtweeter delays from 1.16ms (blue) to 1.75ms(brown) and there doesn't seem to be an obvious answer which is best. You either get better spl summation or less phase shift. I guess the takeaway here is that Harsch provides plenty of opportunities for finetuning 🙂
My DCX2496 is also modded. I added USB>i2s>src input path, replaced all the output opamps and capacitors in the signal chain and lowered analog stage gains to ~2Vrms output levels. Oh and the psu died a couple of years back, so I replaced the 5 legged smps driver with a more powerful version and changed all caps for good measure. Hasn't failed since.
Anyway, I did series of measurements with different midtweeter delays from 1.16ms (blue) to 1.75ms(brown) and there doesn't seem to be an obvious answer which is best. You either get better spl summation or less phase shift. I guess the takeaway here is that Harsch provides plenty of opportunities for finetuning 🙂
My DCX2496 is also modded. I added USB>i2s>src input path, replaced all the output opamps and capacitors in the signal chain and lowered analog stage gains to ~2Vrms output levels. Oh and the psu died a couple of years back, so I replaced the 5 legged smps driver with a more powerful version and changed all caps for good measure. Hasn't failed since.
Attachments
Last edited:
I've been reading the topic but cant find cabinet measurements with the internal specs needed. Is there a pdf for this?
Thanks and have a great 2019
Congrats for the design
Thanks and have a great 2019
Congrats for the design
Im not experienced with asymmetrical xos, but for LR typical delay for mid is only 0,2 -0,5ms when crossing at hundreds of Hz. My experiments with S.Harsch were not successful, I was never happy with measurements or sound.
I find step better to look at than impulse. Reversing polarity helps to see time alignment. When timing gets right, the sound gets "natural". Spl measurements are difficult because of reflections, ground plane at 1m with the speaker sideways on the floor are easiest for xo trimming.
I find step better to look at than impulse. Reversing polarity helps to see time alignment. When timing gets right, the sound gets "natural". Spl measurements are difficult because of reflections, ground plane at 1m with the speaker sideways on the floor are easiest for xo trimming.
It’s in post #1 and has be re-asked many times. Maybe I need to highlight it in color font?
4th sentence in very first paragraph of post 1.
Last edited:
Thank you and sorry. i Saw the internal dimensions but my ignorance is blocking me from seeing how to do a dagger for this. Never saw one or understand the concept of it. Some guidance, please.
Have been reading with interest.
Complete noobie question:
I have a pair of Dayton RS225-8 and a Pair of HIVI B3Ns - how would they pair up together?
Complete noobie question:
I have a pair of Dayton RS225-8 and a Pair of HIVI B3Ns - how would they pair up together?
Further update on my built.
Getting ready for paint Job. This is the first layer. I am using sanding sealer to seal the open pores of MDF.
Sadik
An externally hosted image should be here but it was not working when we last tested it.
Getting ready for paint Job. This is the first layer. I am using sanding sealer to seal the open pores of MDF.
An externally hosted image should be here but it was not working when we last tested it.
An externally hosted image should be here but it was not working when we last tested it.
Sadik
It might work, however, this driver has barely enough sensitivity at 80dB to 82dB may match RS225-8 after baffle step losses. I think You would be better off using a TC9FD for the same price and know that it will work. Try it without committing to the baffle rebate flush cut so you can still mount other drivers if needed.
Drivers that I have tried and can work with only adjusting the amount of attenuation of tweeter via resistor:
-10F/8424 (I would imagine 8414 would work too)
-TG9FD-8
-TC9FD-8
-3FE22-8
-3FE25-8
-B80
-PA95-8
-Peerless 3in alum cone full range (several variants) 8ohm
Last edited: