Hi Everyone,
First, thanks to everyone who has been willing to share their experiences on these topics with me before. I hope to kind of consolidate those discussions and share with you all my own history with the Linkwitz-Riley filters as a narrative backbone.
In the late 1980s I was becoming an audiophile at the same time I was working with a competitor to Dolby in terms of motion picture equipment. I worked for a company in the US state of Georgia which manufactured turnkey audio racks. From processors to amps. The amps were mostly sourced from Hafler with our own cases and power supplies. We literally kept them in business for several years.
This was after I'd audited Dr. Marshall Leach's class in Audio Engineering in the mid-1980s and first learned about filters and phase though I admit I did not then nor do I now really have the math I would need to fully appreciate his course. In any event I'm pretty sure he introduced me to the LR filters, which I'd be exposed to live and in person in a couple of ways. The Snell A/III and THX.
At the time, if you wanted to have a THX certified theater and met the criteria we used Dolby processors intead of ours and then THX would sell you their own active speaker crossovers which included electrical LR4 filters. Not sure how much thought they gave to the electro-acoustical sum, but that's what I remember.
Unknown to me, the speaker I'd be completely awed by, the Snell A/III would, according to Troels Gravesen, be a fantastic example of high filter order speakers. As you are all sick of hearing, I love those. Also related, I recently finished a center channel using LR4 and got great off-axis performance.
So here I was all happy with LR4 when it seems that in some cases the LR4 has gotten out of favor for the woofer-mid transition. Besides a couple of constructive comments here (thank you), I notice two other datapoints from Stereophile.
One, the Snell A/III impulse response:
So, clearly not a time/phase perfect design. Would we argue today that Snell could have sounded better by bringing in the woofer's group delay?
The other point, and forgive me for bringing up a $30,000 USD speaker was the Grimm with it's near perfect impulse response. Reasonably straightforward to achieve in a DSP world, but also appears to agree with the advice I was given, use lower order (maybe 2nd) filters for the woofers.
So I'd like to open up the floor to my more knowledgeable friends here. What do you think of all this? Could Snell have done better? Is the Grimm excessively focused on the impulse response?
Honestly as a beginning hobbyist I would be ecstatic if I produced anything remotely close to the A/III, with the understanding that a lot more went into the A/III performance than just the impulse response.
Thank you all for engaging. I look forward to learning from you what you think, and maybe re-think my own past as a result.
First, thanks to everyone who has been willing to share their experiences on these topics with me before. I hope to kind of consolidate those discussions and share with you all my own history with the Linkwitz-Riley filters as a narrative backbone.
In the late 1980s I was becoming an audiophile at the same time I was working with a competitor to Dolby in terms of motion picture equipment. I worked for a company in the US state of Georgia which manufactured turnkey audio racks. From processors to amps. The amps were mostly sourced from Hafler with our own cases and power supplies. We literally kept them in business for several years.
This was after I'd audited Dr. Marshall Leach's class in Audio Engineering in the mid-1980s and first learned about filters and phase though I admit I did not then nor do I now really have the math I would need to fully appreciate his course. In any event I'm pretty sure he introduced me to the LR filters, which I'd be exposed to live and in person in a couple of ways. The Snell A/III and THX.
At the time, if you wanted to have a THX certified theater and met the criteria we used Dolby processors intead of ours and then THX would sell you their own active speaker crossovers which included electrical LR4 filters. Not sure how much thought they gave to the electro-acoustical sum, but that's what I remember.
Unknown to me, the speaker I'd be completely awed by, the Snell A/III would, according to Troels Gravesen, be a fantastic example of high filter order speakers. As you are all sick of hearing, I love those. Also related, I recently finished a center channel using LR4 and got great off-axis performance.
So here I was all happy with LR4 when it seems that in some cases the LR4 has gotten out of favor for the woofer-mid transition. Besides a couple of constructive comments here (thank you), I notice two other datapoints from Stereophile.
One, the Snell A/III impulse response:
So, clearly not a time/phase perfect design. Would we argue today that Snell could have sounded better by bringing in the woofer's group delay?
The other point, and forgive me for bringing up a $30,000 USD speaker was the Grimm with it's near perfect impulse response. Reasonably straightforward to achieve in a DSP world, but also appears to agree with the advice I was given, use lower order (maybe 2nd) filters for the woofers.
So I'd like to open up the floor to my more knowledgeable friends here. What do you think of all this? Could Snell have done better? Is the Grimm excessively focused on the impulse response?
Honestly as a beginning hobbyist I would be ecstatic if I produced anything remotely close to the A/III, with the understanding that a lot more went into the A/III performance than just the impulse response.
Thank you all for engaging. I look forward to learning from you what you think, and maybe re-think my own past as a result.
I don't usually try for a particular alignment. Most of my recent speakers are 2-ways, and the boxes are on the small side usually. It gets crowded trying to shoehorn in a x-over while keeping the coils away from each other, and aimed properly. I think adjusting BSC is easier with a simple 2nd order filter, than with a 4th order. Most of the time, I use a sloped baffle to make alignment easier. A recent speaker uses natural roll-offs, and simple filters, and notches. It sounds good to me, and is probably the best I've built. The roll-offs are shallow with a lot of overlap. I might be giving up some power handling, but I don't think it's noticeable. I'm not sure what I would call the x-over. The impulse looks like a 2nd order to me. The slopes are far from it.
I totally appreciate that. I'm thinking about this a lot because my next project is a DSP powered 3 way. I've discussed this before here, but the Snell A had a lot of space and the crossovers were huge and ugly looking. Sorry, Peter, but they were. The reason I'm so fixated on LR4 right now is that the off axis response I got in my center was just outstanding. I would never have attempted this with an internal passive design and I doubt the acoustic offsets would have let me either.
One of my recent centers was plus or minus 2dB at 45 degrees, for most of the frequency range. I think I used 3rd order filters. That design was scrapped for use as a center. TM design. The design above is also very good off axis horizontally. Actually, they were the same speaker, but with different x-overs, and voicing. The center had very little BSC, the x-over for use as mains is close to full BSC. I built another center, as this one was going to be used for my mains..
Last edited:
Using DSP FIR filters, a "near perfect impulse response" (flat phase and amplitude response over the audio bandwidth) far better than the Snell A/III can be achieved with any crossover slope from ~6dB to 100dB (or more) per octave if you have the processing power to accomplish it, which wasn't available back in the 1980s.
Implementing FIR processing is probably not something a "beginning hobbyist" would find easy, but is certainly is easier than trying to get anything resembling flat phase response over the audio bandwidth from IIR filters.
Art
That the crossover points were invisible in the horizontal plane and still barely visible in the vertical.
Interesting point, and a lot of this is determined by the geometry. I suspect it might be why we see horizontal polars much more often than verticals even though verticals can be where attention is needed.
@AllenB Having looked at several Stereophile reviews, the vertical response, especially above the tweeter, is usually the worst. I mean, there are so many expensive speakers that do horribly horizontally too, but even those that do well horizontally do poorly vertically. Here's an example from my own center, where I really wanted excellent off-axis response as I move around a lot while watching TV. Please note the black line is 45 degrees above axis, a punishing measurement for most speakers. Stereophile only measures up to +15 I think.
Crossover points are at 300 and 3kHz.
The green line is 45 degrees horizontally. I do think of this as a win in terms of accomplishing off-axis response, and that's why the LR4 filters did well for me.
For comparison here is the Grimm's horizontal and vertical response plots, which are also excellent. The Snell review only has horizontal and it's equally good. Based on all this data and practical experience, it seems the LR4 can, with the right drivers end up with excellent off axis responses:
Crossover points are at 300 and 3kHz.
The green line is 45 degrees horizontally. I do think of this as a win in terms of accomplishing off-axis response, and that's why the LR4 filters did well for me.
For comparison here is the Grimm's horizontal and vertical response plots, which are also excellent. The Snell review only has horizontal and it's equally good. Based on all this data and practical experience, it seems the LR4 can, with the right drivers end up with excellent off axis responses:
Last edited:
Yes, this is it. It's a common problem for sure, but here in DIY land not only do we get to fit a room and speaker together, we can also determine what the speaker does to help it fit.
Like for example, the main lobe of the tweeter to mid x-over in the center channel above, could be aimed upwards, which could move the null at 45 degrees to about 65 degrees. That would not be good for watching while laying on the floor though, as the lower null would be a problem.
If vertical was important to me for standing while viewing, I'd try 8th order filters on the tweeter / mid. The null would be much narrower. Other filter types could help as well, if you can accept a slight peak on axis at the x-over point.
If vertical was important to me for standing while viewing, I'd try 8th order filters on the tweeter / mid. The null would be much narrower. Other filter types could help as well, if you can accept a slight peak on axis at the x-over point.
Last edited:
@temp25 - The lobe is perfect for my listening locations. The +45 degree measurement was just for illustration. My listening area is about 8' away.
If I needed to raise the lobe I could also increase the tweeter delay, virtually pushing the tweeter back, but you are correct of course, the steeper the crossover the less destructive interaction.
If I needed to raise the lobe I could also increase the tweeter delay, virtually pushing the tweeter back, but you are correct of course, the steeper the crossover the less destructive interaction.
Changing the delay is the way I'd do it as well. What I was getting at is that a designer has options, and can make what's most suitable.
Delay would vary the amount of tilt at different frequencies, but adjusting phase instead would work well.
Well...where we are now...
First, i think it is a mistake to design a speaker for a particular room...unless you know the speaker will never be used any where else, and the room won't change either. Good luck with that !
Time honored anechoic tuning, or as best we DIYer's can accomplish that, is always the best approach imo.
Which means, once a speaker's physical acoustic design is built...the box and drivers... that there is not a whole lot of decision latitude for achieving its best tuning.
Crossover points are relatively fixed for best combination of on and off-axis response.
And since drivers need to have complementary acoustic response through those xovers, phase between them becomes fixed too.
Delays, when a speaker does have complementary acoustic response thru xovers, end up being simply offsetting the time-of-flight distance differences to acoustic centers, so they too have no latitude.
About the only thing up for decision, is the acoustic order of the crossovers. And the latitude for that is defined by the lowest order that the speaker stays behaved....we can always go up in order from there, but we can't go lower.
So generally with IIR, ....given that impulse, step, group delay, reduced phase rotation,.... all go together with the lowest IIR xover orders with which the speaker stays behaved...even their orders becomes comparatively fixed in the sense that the lowest you can get away with is optimal.
Bottom line...a whole lot less room to juggle than often imagined, in terms of xovers, phase, delays, and filters.....
Most the allowable juggling comes down to the technique used to achieve the chosen acoustic xover order. Whatever technique used, if done well, will end up with the same net filter as any other technique...(so no latitude again!)
Where we are today....like Art said, any order acoustic xover with flat phase can be achieved with FIR based DSPs. A linear-phase LR24 isn't considered high-order anymore. The old juggle this vs that world just doesn't make sense anymore...imso.
First, i think it is a mistake to design a speaker for a particular room...unless you know the speaker will never be used any where else, and the room won't change either. Good luck with that !
Time honored anechoic tuning, or as best we DIYer's can accomplish that, is always the best approach imo.
Which means, once a speaker's physical acoustic design is built...the box and drivers... that there is not a whole lot of decision latitude for achieving its best tuning.
Crossover points are relatively fixed for best combination of on and off-axis response.
And since drivers need to have complementary acoustic response through those xovers, phase between them becomes fixed too.
Delays, when a speaker does have complementary acoustic response thru xovers, end up being simply offsetting the time-of-flight distance differences to acoustic centers, so they too have no latitude.
About the only thing up for decision, is the acoustic order of the crossovers. And the latitude for that is defined by the lowest order that the speaker stays behaved....we can always go up in order from there, but we can't go lower.
So generally with IIR, ....given that impulse, step, group delay, reduced phase rotation,.... all go together with the lowest IIR xover orders with which the speaker stays behaved...even their orders becomes comparatively fixed in the sense that the lowest you can get away with is optimal.
Bottom line...a whole lot less room to juggle than often imagined, in terms of xovers, phase, delays, and filters.....
Most the allowable juggling comes down to the technique used to achieve the chosen acoustic xover order. Whatever technique used, if done well, will end up with the same net filter as any other technique...(so no latitude again!)
Where we are today....like Art said, any order acoustic xover with flat phase can be achieved with FIR based DSPs. A linear-phase LR24 isn't considered high-order anymore. The old juggle this vs that world just doesn't make sense anymore...imso.