Nice speakers! I like the Elac look on the tweeter and midrange. What grills did you use on the woofers?
thanks , tweeter are scanspak , the woofers are volt rv3143 / mids are volt vm752.. Thats how the woofers come . made in UK..
Last edited:
The Bessel filter topology could be tricky to work with. It has quite quite a slow roll off rate and so the zone of interaction between LF and HF sections will be very broad.
The Hypex Filter Designer program user’s manual provides sets of biquad filters that can be used to create L–R filter functions. I’d suggest giving those a try. You’ll need two biquads per L–R to get the 4th-order response, in a similar manner to what was necessary for the 4th-order Butterworth filters.
Last edited:
Thanks Witwald, your advice is much appreciated, I am using the filters from Hypex woofer and tweeter that comes free with the download . Its a big pain in the *** but I guess has to be done.. All the very best from Amami Japan .. 👍
I'm not sure what you mean when you say the "filters from Hypex...that come free". From what I can see, the Hypex Filter Designer program enables you to create a wide range of filters that can suit a number of target applications.
From the Hypex Filter Designer manual, there is an example Linkwitz–Riley 4th-order crossover filter, LR4, designed with a cut-off frequency of 1kHz. Using that as the basis, the low-pass and high-pass LR4 filters with the desired cut-off frequencies are created using two cascaded low-pass or high-pass filter biquads (2nd-order general filters), as appropriate. These biquad filters all have Q factors of 0.71, and thus consist of cascaded 2nd-order Butterworth filters.
The approach presented in the manual can be customized to suit your drivers as follows.
For the woofer: the low-pass LR4 filter consists of two low-pass filter biquads both with cut-off frequencies of 380Hz and Q factors of 0.71.
For the midrange: a) the high-pass LR4 filter section consists of two high-pass filter biquads, both with cut-off frequencies of 380Hz and Q factors of 0.71; and b) the low-pass LR4 filter section consists of two low-pass filter biquads both with cut-off frequencies of 3800Hz and Q factors of 0.71.
For the tweeter: the high-pass LR4 filter consists of two high-pass filter biquads, both with cut-off frequencies of 3800Hz and Q factors of 0.71.
To help illustrate the above description to some degree, below is a snapshot of an LR4 filter design taken directly from the Hypex Filter Designer manual. Of course, the values of the cut-off frequencies would need to be changed accordingly.
From the Hypex Filter Designer manual, there is an example Linkwitz–Riley 4th-order crossover filter, LR4, designed with a cut-off frequency of 1kHz. Using that as the basis, the low-pass and high-pass LR4 filters with the desired cut-off frequencies are created using two cascaded low-pass or high-pass filter biquads (2nd-order general filters), as appropriate. These biquad filters all have Q factors of 0.71, and thus consist of cascaded 2nd-order Butterworth filters.
The approach presented in the manual can be customized to suit your drivers as follows.
For the woofer: the low-pass LR4 filter consists of two low-pass filter biquads both with cut-off frequencies of 380Hz and Q factors of 0.71.
For the midrange: a) the high-pass LR4 filter section consists of two high-pass filter biquads, both with cut-off frequencies of 380Hz and Q factors of 0.71; and b) the low-pass LR4 filter section consists of two low-pass filter biquads both with cut-off frequencies of 3800Hz and Q factors of 0.71.
For the tweeter: the high-pass LR4 filter consists of two high-pass filter biquads, both with cut-off frequencies of 3800Hz and Q factors of 0.71.
To help illustrate the above description to some degree, below is a snapshot of an LR4 filter design taken directly from the Hypex Filter Designer manual. Of course, the values of the cut-off frequencies would need to be changed accordingly.
Hay , yes indeed, Thanks for the great feed back much appreciated, I have been working on the filters all day /again / and think I have cracked the problems. Its still not quite right but much better. I will post the final outcome when I finish, Its always great to have help and good advice from you guys .. 👍
Glad you’re getting hang of wrangling filters 😉
That said, if these are filter responses then the combined plot which hypex shows is going to be wrong, and the filter responses won’t make sense as they’ll be the wrong slope and phase.
It’s not normal to have identical high and low pass.
That said, if these are filter responses then the combined plot which hypex shows is going to be wrong, and the filter responses won’t make sense as they’ll be the wrong slope and phase.
It’s not normal to have identical high and low pass.
After having a long conversation with Hypex tech support, they inform me that adjusting the gain on the crossover stage , will not affect the crossover points, as this was one of the big problems i had becouse the mids were much too loud, having to drop the gain to almost 14db is very strange.. You guys are cool thanks for the help im sure i will need more.. 👍
The high-pass and low-pass slopes of the filters need to take into account the natural response roll-offs of the drivers. Hence, as mentioned, the filter slopes may be different in order to achieve a relatively flat summation of the individual response functions.
However, a 4th-order Linkwitz–Riley, LR4, filter has a powerful ability to imprint its own magnitude and phase response on any driver to which it might typically be applied. That's a result of it being a filter of quite high order, which can be easily achieved in an active design, and not so easily in a passive design.
To help illustrate the behavior of LR4 filters, I've set up a VituixCAD model that includes some basic driver acoustic responses similar to what are possessed by the Scanspeak+Volt+Volt drivers in the 3-way system that is being worked on. With a slight shift in the frequency of one of the filters, plus some amplification here and there, a quite flat summed response can be achieved. This comes courtesy of the properties of high-order Linkwitz–Riley filters.
Below is the simulated response of the 3-way system, where the low-pass and high-pass filtered responses have roughly the same slope.
The VituixCAD model used to produce the above response is shown below. The filter sections are on the left. The amplifiers connecting to each driver are in the middle. The second-order high-pass and low-pass filters are used to simulate the natural responses of each of the drivers, through a suitable choice of cut-off frequencies and Q factors.
Last edited:
Thank you all again for your ongoing help and support it is very much appreciated. The the crossover in relation to my original plan has stayed the same 380Hz - 3.8KHz. The main issue is moving the Hypex filter points around to get the over all sound I would like , once you have the baseline starting point which is no problem and you guys have explained 👍My friend who is a Jazz singer and myself are just moving the Q / boost cut / allPass / shelf filters around the get the sound we think is best. This involves playing a hand full of music that we know well, and adjusting, it’s painstakingly slow . and we have had to reduce the midrange gain down almost by 14db ‘ . It’s all sounding quite good. Looks a bit crazy buts sounds ok 👍 Hypex support tell me that this is how it’s done and will not effect the crossover. I’ll post a photo later. Thanks everyone.
That's all as expected. Once the gain is set appropriately, the crossover points should come out correct. In actuality, there will be an effect from the natural responses of the drivers that will come into play. This will usually widen the crossover region a little, which will need one or both of the filters to have their cut-off frequency adjusted. Low-order filters will generally be more affected by this than high-order filters. And some filter types, such as Butterworth, are quite sensitive to errors in phase and magnitude response in their crossover regions, while others, like Linwkitz-Riley filters, aren't as sensitive.
That 14dB gain adjustment is very large, and it has me puzzled as to what made it necessary. Maybe you just like a more laid-back midrange sound quality for vocals? 🙂 Thinking about it some more, could the input sensitivity of the midrange channel somehow be set much higher than that of the other channels?
In any case, I'd suggest using a microphone and some measurement software to try and get the filter settings under control. Adjusting these types of things by ear can be notoriously difficult, and it's a less-than-ideal approach. It might work out okay for the final adjustment "voicing" in your room's acoustic space, but for the initial setting of parameters I think it's an approach that poses many difficulties.
@Amami Life Based on your measurements, the response that you have decided upon is very unusual. With a joke in mind, it would appear that you have chosen to apply the philosophy of the "BBC Dip" to an extreme degree. 🙂
On a positive note, the response of the midrange driver over the decade between 400Hz and 4000Hz is relatively flat, with a ±2dB range. For the tweeter, however, I am quite concerned about the large peak that is evident at 5000Hz, which is about 6dB greater than the tweeter's response between 8kHz and 12kHz. That is really going to put a lot of added "brightness" into whatever you are listening to, and may explain why the level of the midrange has been pushed down so far.
Returning to the midrange, the 14dB of cut that you have is way too much. The output level of the midrange really needs to be adjusted back up so that it is within a few dB of the woofer and tweeter levels, if a midrange dip is preferred, or at the same level if more accurate on-axis response is desired. If you don't do that, it's going to be a very unusual, non-linear sounding speaker, with a significant deemphasis of music notes in the midrange's frequency range.
You could try using pink noise as an audio test signal as well. It will usually help to show up aberrations in the frequency response, without being music-track specific. If you have one or more pairs of headphones, then you could acclimatise yourself to the sound quality of pink noise, and that would hopefully help to guide your frequency response adjustments.
Two monitor loudspeakers from the 1970s, such as the B&W DM6 and IMF TLS80 Mk II, both had a relatively flat on-axis response, as shown by the measurements below. Your system should be able to produce something approaching that type of flat response characteristic in order to have a reasonable sound quality on a wide range of music program material.
On a positive note, the response of the midrange driver over the decade between 400Hz and 4000Hz is relatively flat, with a ±2dB range. For the tweeter, however, I am quite concerned about the large peak that is evident at 5000Hz, which is about 6dB greater than the tweeter's response between 8kHz and 12kHz. That is really going to put a lot of added "brightness" into whatever you are listening to, and may explain why the level of the midrange has been pushed down so far.
Returning to the midrange, the 14dB of cut that you have is way too much. The output level of the midrange really needs to be adjusted back up so that it is within a few dB of the woofer and tweeter levels, if a midrange dip is preferred, or at the same level if more accurate on-axis response is desired. If you don't do that, it's going to be a very unusual, non-linear sounding speaker, with a significant deemphasis of music notes in the midrange's frequency range.
You could try using pink noise as an audio test signal as well. It will usually help to show up aberrations in the frequency response, without being music-track specific. If you have one or more pairs of headphones, then you could acclimatise yourself to the sound quality of pink noise, and that would hopefully help to guide your frequency response adjustments.
Two monitor loudspeakers from the 1970s, such as the B&W DM6 and IMF TLS80 Mk II, both had a relatively flat on-axis response, as shown by the measurements below. Your system should be able to produce something approaching that type of flat response characteristic in order to have a reasonable sound quality on a wide range of music program material.
What didn’t get mentioned is that the hypex filter implementation is not a crossover simulator and isn’t meant to be used to design crossovers. It’s meant as a last step to implement them after working on your simulator.
That said, you’re using the old school no measurement style design trying by ear. You may get good results this way, there are limitations to what you can tell by ear, especially when you use higher order filters.
A difference of 90 degrees here, 180 degrees there, remains as a fixed difference regardless of the slope. It doesn’t get diluted in the fast changing response.
Here’s an example of how 4th order can be simple, done passive.
https://www.diyaudio.com/community/...overs-without-measurement.189847/post-2834287
So yes initially I started off with a flat responce over the range, with the mid inverted . adding some filters this gave me a flat responce ' hope thats the right terminology?' Now, I loaded the Hypex filter woofer and tweeter data whitch gave me the next outcome. The midranged leval is very very noticeably loud. Once I got the levals under control it was a matter of shall we say 'tuning the over all sound. ' I did this by first using sonarworks reference 4 studio edition, with that infomation I got the over all outcome. So not only was the mids very loud , sonarworks actually pulled the mids down and bass up in the adjustment. Saying all that I dont particularly like sonarworks and i have it turned off , and only use for inital measurements. I have tryed from scratch buiding the filter design but the outcome is much the same without using the Hypex data .. It is puzzaling that tech support at Hypex said that moving the filters will not affect the crossover points, although I did explain the large gain adjustments, they actually pointed me in the direction of you guys in an email they sent me. basically what they were saying is that I should know what to do before I purchase their product , TRUE, but a dick ish mentality .. Thank you all once again .. Bring on the feed back..
Attachments
Yes we are using the old school way, we are both experienced with studio sound, my old Genelec 1033A are still in my head , saying all that I am a total dummy when it comes to all this side if speaker management, and new tech. I origenly thought it would not be a problem BUT now I understand its a pain in the *** especially when you dont have a clue what your doing .. thats why i am here ,,,