Inspired by the Rutcho Dm601 S3 mod, I decided to tackle the crossover on my pair of DM603 S3's using his design. It is my first time diving into this and I will admit I have no idea what I am doing. I imported the response graphs of the Kevlar woofer and tweeter into VituixCAD, and simulated the stock and Rutcho crossovers. I came up with a 1.5ohm R1, a 3.3uF C1, short of the resistor on the tweeter inductor. For the mid-woofer, I created a zobel network just like Rutcho, at 4.7uf cap, but a different resistor value of 4.7ohm.
Using a 2.5 Allen key, I removed the silver woofer. There are three screws - one at the middle and one at each bottom corner. The top corners are held by expansion pins. I used a pair of pliers to squeeze them and put my hand behind the board to apply pressure inward. For the zobel, I cut open and balded the blue and brown driver wires and mounted the capacitor and the resistor from the leads vertically. The mounted components are held by glue, and I had to use a knife to pry them off. I did this to both speakers but my beginner soldering skills had my zobel bridge compromised and open. This gave me a unique opportunity to A/B the before and after while running simultaneous modified tweeters.
The highly crossed Kevlar woofer was essentially distorting at the high frequencies. It sounded like outright white noise and heavily muddied the mid-range.
VituixCAD shows on the stock crossover a terrible phase shift after 7.5Khz and both the directivity graphs and phase patterns looked ugly. My post-mod speakers sound huge, delicate, spacious. it is definitely an improvement all around. However, I lost a lot of mid range and I suspect it is due to the 4.7ohm resistor on the zobel. I could play with 1.8ohms but the second problem is that the phase is nearly completely opposed from 2-4.7 kHz. The vocals sound delicate but they are also distant and lacking bark. I suspect I am not getting the full sound stage yet out of these drivers.
After some research, it occurred to me that the zobel increases the order of the filter. For parity, do I have to do the same to the tweeter? I ran this through VItuixCAD and came up with a 4.3uf C1 (I add a 1uf cap in parallel to the existing one), a 4th order filter created by a 5.6uf cap and a 200uh 15AWG inductor, and reducing the Zobel resistor to 3.3ohms.
Using a 2.5 Allen key, I removed the silver woofer. There are three screws - one at the middle and one at each bottom corner. The top corners are held by expansion pins. I used a pair of pliers to squeeze them and put my hand behind the board to apply pressure inward. For the zobel, I cut open and balded the blue and brown driver wires and mounted the capacitor and the resistor from the leads vertically. The mounted components are held by glue, and I had to use a knife to pry them off. I did this to both speakers but my beginner soldering skills had my zobel bridge compromised and open. This gave me a unique opportunity to A/B the before and after while running simultaneous modified tweeters.
The highly crossed Kevlar woofer was essentially distorting at the high frequencies. It sounded like outright white noise and heavily muddied the mid-range.
VituixCAD shows on the stock crossover a terrible phase shift after 7.5Khz and both the directivity graphs and phase patterns looked ugly. My post-mod speakers sound huge, delicate, spacious. it is definitely an improvement all around. However, I lost a lot of mid range and I suspect it is due to the 4.7ohm resistor on the zobel. I could play with 1.8ohms but the second problem is that the phase is nearly completely opposed from 2-4.7 kHz. The vocals sound delicate but they are also distant and lacking bark. I suspect I am not getting the full sound stage yet out of these drivers.
After some research, it occurred to me that the zobel increases the order of the filter. For parity, do I have to do the same to the tweeter? I ran this through VItuixCAD and came up with a 4.3uf C1 (I add a 1uf cap in parallel to the existing one), a 4th order filter created by a 5.6uf cap and a 200uh 15AWG inductor, and reducing the Zobel resistor to 3.3ohms.
Attachments
I see you could be reversing your woofer to mid polarity.
Have you used timed measurements? I see you applied the Z parameter with your tweeter which you'd only do if your measurements were not properly timed. Phase is being affected by the delay, at least by way of the tweeter Z setting and the mid Y, in connection with the vituixcad standard listening position.
No, not without a reason.
Have you used timed measurements? I see you applied the Z parameter with your tweeter which you'd only do if your measurements were not properly timed. Phase is being affected by the delay, at least by way of the tweeter Z setting and the mid Y, in connection with the vituixcad standard listening position.
Really interesting work! I can tell you’ve put serious thought into this crossover upgrade, especially referencing Rutcho’s work and simulating with VituixCAD. Looking forward to seeing how your changes perform in real-world testing.
Just for inspiration, here is @PKAudio's crossover modification for a 802d2:
https://pkaudio.webnode.cz/bw802d2/
https://pkaudio.webnode.cz/bw802d2/
Not to rag on B&W, but I want to rag on B&W. One of the most consistent aspects of their speaker designs, regardless of price, is their use of peaky tweeters. These then force their hands with the rest of the design.
If you are forced to use a tweeter like this you are left with two options. One, add a lot of EQ to flatten the response and also lowering the sensitivity or cross the mid relatively high to avoid those issues. The latter of course causes off-axis response issues, which some may argue helps with perceived imaging at the cost of narrow sweet spots.
Again, the argument can be made that you could take advantage of a tweeter like this by building in a loudness curve.
Most of all I think their curious choices kind of leave us observers a fun set of puzzles, like design archeology to try to work our way backwards into the though patterns of the designers.
Almost all the ways in which I'd want to alter a B&W crossover end up with a lot of parts, which kind of begs them to be rethought with active crossovers instead of attempting significant crossover mods.
If you are forced to use a tweeter like this you are left with two options. One, add a lot of EQ to flatten the response and also lowering the sensitivity or cross the mid relatively high to avoid those issues. The latter of course causes off-axis response issues, which some may argue helps with perceived imaging at the cost of narrow sweet spots.
Again, the argument can be made that you could take advantage of a tweeter like this by building in a loudness curve.
Most of all I think their curious choices kind of leave us observers a fun set of puzzles, like design archeology to try to work our way backwards into the though patterns of the designers.
Almost all the ways in which I'd want to alter a B&W crossover end up with a lot of parts, which kind of begs them to be rethought with active crossovers instead of attempting significant crossover mods.
Both the tweeter circuit and models on the DM601S3 and DM603S3 are identical. Rutcho measured a 0.16mh inductor but B&W documentation presents the same 0.07mh as my speaker. Rutcho measured the responses of the tweeter and woofer drivers independently and these are the responses I utilized.
The only difference between the kevlar woofer circuits is the DM601 has a 1.1mh inductor value vs 0.7mh on mine. Im convinced the kevlars are the same drivers in different baskets. It would make the most sense from a mass production standpoint.
I saw in another thread that the tweeter is slightly recessed into a waveguide at the baffle measured at 26mm. A key point of the Rutcho mod is the tweeter polarity inversion for better phase alignment. When I zero out the Z and Y values on a simulation of the original Rutcho replicated circuit, I get a deep -25db notch at 5khz. Setting the tweeter to a standard positive polarity on VituixCAD shows a flat response.
I can test this in the upcoming days. My current mod simulation shows a theoretical flat response when zeroing out the axis values and resetting the tweeter to its original positive polarity. It demonstrates a 1 for 1 perfect phase alignment and also boosts the midrange area I am missing (if you ignore the 5khz notch). If it happens in real-world then it fully supports your theory of zeroing the axis values.
Did you measure this? Knowing would help to make this easier to follow. I'm reluctant to accept a null check as a sign of correctness.
Do you see how your lower cross in the sim has a polarity problem?
Do you see how your lower cross in the sim has a polarity problem?
I was mostly using spectroid measurements as measuring the drivers while mounted to the cabinet was too much of an endeavor. I also don't own a measurement microphone, just an SM57 which is known to be trebly.
I am pausing this as I have royally screwed up and will post another thread about that particular snafu.
I am pausing this as I have royally screwed up and will post another thread about that particular snafu.