Thanks xrk!
Of course I know that the final acoustical responses what's matters. I flattened the driver responses so the applied xo filters causes (almost) textbook acoustical slopes.
I may try my question simpler: I set the low-pass BW4 by cascading a q=0.54 and a q=1.31 filters with cut-off both set at 2000Hz, then I need to set the high-pass q=0.58 with cut-off of 2000Hz? Or 2000/1.27=1575Hz to get 2nd order Bessel response with -3dB at 2000Hz? If I set the high-pass q=0.58 filter cut-off to 2000Hz then the -3dB point would be higher in frequency.
Of course I know that the final acoustical responses what's matters. I flattened the driver responses so the applied xo filters causes (almost) textbook acoustical slopes.
I may try my question simpler: I set the low-pass BW4 by cascading a q=0.54 and a q=1.31 filters with cut-off both set at 2000Hz, then I need to set the high-pass q=0.58 with cut-off of 2000Hz? Or 2000/1.27=1575Hz to get 2nd order Bessel response with -3dB at 2000Hz? If I set the high-pass q=0.58 filter cut-off to 2000Hz then the -3dB point would be higher in frequency.
Last edited:
I am not sure what you mean, as I thought setting a BW4 or BES2 sets the shape and steepness of the curve. At least that is my understanding - there is no independent Q adjustment. Or are you saying that you generated an approximate BW4 and BES2 via convolution of two parametric curves of set freq and two different Q's? I don't know the math off the top of my head but you can use a filter simulator to see if the combo you are using gives the same shape - that is how I would do it. I would not be too worried about achieving an exact textbook response - just close enough is good and avoids added phase distortions of the two combo filters. Does that answer your question?
An update for my Harsch xo attempt. It seems like there is no need to use the 1.27 factor formula for creating the BES2 response with a DSP like Hypex are using. Just use your target xo point with a Q=0.58 2nd order high(or low)-pass.
Below there is some of my measurements, showing the individual slopes with their target and the sum of a 3-way with 320Hz and 2000Hz xo target point and the step response. The woofer response is a victim of the gated measurement but I left it there.
The responses should be like that or I am missing something? Sounds really good but unfortunately the off-axis responses doesn't look as good as with the classic LR xo version.
Below there is some of my measurements, showing the individual slopes with their target and the sum of a 3-way with 320Hz and 2000Hz xo target point and the step response. The woofer response is a victim of the gated measurement but I left it there.
The responses should be like that or I am missing something? Sounds really good but unfortunately the off-axis responses doesn't look as good as with the classic LR xo version.
Attachments
You did a 3 way Harsch? Very cool!
The step response you have looks like what they look like when done in DSP. When I did the passive Harsch XO, it looks much closer to a right triangle - almost like what one looks like from a 1st order. I think you have it.
Plot the measured acoustic phase to make sure it doesn’t wrap. It should rise 55deg from flat baseline for Harsch. If the phase is done well, it should beat the LR’s I. Polars should look better with Harsch if done right, I think. Flat phase should give more consistent polars.
The step response you have looks like what they look like when done in DSP. When I did the passive Harsch XO, it looks much closer to a right triangle - almost like what one looks like from a 1st order. I think you have it.
Plot the measured acoustic phase to make sure it doesn’t wrap. It should rise 55deg from flat baseline for Harsch. If the phase is done well, it should beat the LR’s I. Polars should look better with Harsch if done right, I think. Flat phase should give more consistent polars.
Hey Jim! It's fresh for me too but good to hear that other members with Fusionamps wants to try this interesting crossover. If you decided to try, I would be interested on your results both subjectively and objectively. I may try to help you if you ever need.
YSDR - To achieve a 4th order BW low pass filter you cascaded two 2nd order filters with q=1.31 and q=0.54... Since q * q = 1.31 * 0.54 = 0.707, you achieve your goal. What is your thought process for selecting 1.31 and 0.54?
My first inclination would have been to set both of the 2nd order filters to q=0.84, which is the square root of 0.707... but I am probably missing something important... thoughts?
My first inclination would have been to set both of the 2nd order filters to q=0.84, which is the square root of 0.707... but I am probably missing something important... thoughts?
Haha, the process started at page 22 in the HFD manual, so it's not from the top of my head.
But why is these exact Q's unfortunately I don't know, I just followed the descriptions. Btw, some people are scared by the Hypex DSP usage, but the possibilities are much more expansive than with ordinary DSP imo.
Denes
I tried the same thing (Q = 0.84) and found out it didn't work. Then I found this link: Cascading filters that explains the math behind why Q = 1.31 and Q = 0.54 are the correct values.
Neat stuff guys! Btw, Sigma DSP Studio (free from Analog Devices) allows some very powerful graphical programming via striungs and nodes of just about any DSP function if used with their ADAU series DSP chips. I got their eval board for the ADAU1452 and it has an 8ch output with FIR, Toslink in/out, analog in/out, I2S, and SPI. Very flexible.
You are most welcome, HiFiJim. Hope you get to use it - it sounds very very good.
Btw, Bwaslo (author of Xsim) reminded me that by delaying the "amp" (delay between entire speaker and mic), you can get the phase to align itself and unwrap to see if you get the 55deg bump. In my Xsim simulation, I simply added -1.12in delay to the "amp" and I got this for the predicted phase plot - note the 55deg bump is now visible:
So YDSR, try changing overall delay between mic and speaker to see if you can get your phase wraps to go away.
You are most welcome, HiFiJim. Hope you get to use it - it sounds very very good.
Btw, Bwaslo (author of Xsim) reminded me that by delaying the "amp" (delay between entire speaker and mic), you can get the phase to align itself and unwrap to see if you get the 55deg bump. In my Xsim simulation, I simply added -1.12in delay to the "amp" and I got this for the predicted phase plot - note the 55deg bump is now visible:
So YDSR, try changing overall delay between mic and speaker to see if you can get your phase wraps to go away.
Attachments
Last edited:
Indeed, no wrap.
Should you need it later on, in XSim, double click on the amp pictogram and you'll see the pop up window with the changes that are available, just like when you double click on the speakers or parts pictograms.
I just created a Harsch filter in my Hypex Fusion amp. Crossover frequencies are 200 Hz and 2000 Hz.
No measurements yet, but I gave it a listen. The differences between it and the baseline LR4 filters is subtle, but within a few minutes I can notice it. I think I am getting better blending between the woofer and the mid driver. a more seamless transition.
More listening, and more measurements to come.
YSDR - What do you think of the Harsch crossover in your system? What is your subjective opinion on music?
Jim
No measurements yet, but I gave it a listen. The differences between it and the baseline LR4 filters is subtle, but within a few minutes I can notice it. I think I am getting better blending between the woofer and the mid driver. a more seamless transition.
More listening, and more measurements to come.
YSDR - What do you think of the Harsch crossover in your system? What is your subjective opinion on music?
Jim
Hi Jim!
I found my 3-way Harsch xo maybe a slightly more "immediate" (than the LR one), which I liked at first, but after some listening in my case, the tonal balance is shifted towards higher frequencies and gained some nervousness and sibilance to my ears. The sound is thinned out, but the low-mid-high range levels was not altered electronically compared to LR.
Then I measured the off-axis response and found some abrupt change at different degrees around the xo points whereas the LR xo was more smooth.
Moreover with the Harsch xo, I found that a slight change in the timing alters the frequency response too much compared to LR xo.
After all, I returned to the LR xo and I say I am happy with that.
I am using a LR2 between bass-mid and LR4 between mid-high. The Harsch xo I used was the BW4 low-passes and BES2 high-passes version. What is yours?
You said the bass-mid integration was improved in your case, maybe it's worth a try a LR2 instead of LR4 at that range. I use shallower slopes whenever is possible. Just a thought.
I am also curious about your further listening and measurements too.
Cheers!
I found my 3-way Harsch xo maybe a slightly more "immediate" (than the LR one), which I liked at first, but after some listening in my case, the tonal balance is shifted towards higher frequencies and gained some nervousness and sibilance to my ears. The sound is thinned out, but the low-mid-high range levels was not altered electronically compared to LR.
Then I measured the off-axis response and found some abrupt change at different degrees around the xo points whereas the LR xo was more smooth.
Moreover with the Harsch xo, I found that a slight change in the timing alters the frequency response too much compared to LR xo.
After all, I returned to the LR xo and I say I am happy with that.
I am using a LR2 between bass-mid and LR4 between mid-high. The Harsch xo I used was the BW4 low-passes and BES2 high-passes version. What is yours?
You said the bass-mid integration was improved in your case, maybe it's worth a try a LR2 instead of LR4 at that range. I use shallower slopes whenever is possible. Just a thought.
I am also curious about your further listening and measurements too.
Cheers!