I'm not sure if I raise relevant question or nor, and whether to post here or in the room correction forum. My wondering seems just in between...
I stumbled on the Kii Three and D&D 8c pitch that they can achieve almost perfect operation in "normal" rooms. This seems to rely on the consistent off axis response of the speakers.
Then going through different threads this leaded me to FIR correction as performed with Acourate or Audiolense products (and DRC-FIR for the open source more steep learning curve). This was looking really interesting to improve the system in the room, at listening position. I have bought and read "Accurate sound reproduction" from Mitchco.
Then to Grimm LS1... and the associated whitepaper https://www.grimmaudio.com/wordpress/wp-content/uploads/speakers.pdf . That whitepaper was highlighting the drawbacks/restictions of the FIR approach, especially on the off axis response... Arghh so some promote FIR to solve key issues, but this may impact significantly the off axis response, which is important for imaging and timbre and the secret of LXminis/Kii/D&D.
I started to get bugged about the interest or not to invest time (and money) in FIR correction, either for complete speaker filtering (each driver ground up FIR filtered for listening position) or for final lifting to adjust speakers for listening position.
Is the Grimm whitepaper a bit pushy comparing very different designs related to off axis response (controlled directivity LS1 to MTM arrangement) ?
Can the FIR final corrections be adequate with Cadioid like speakers like my LX-Mini, and would not degrade too much the Off-Axis response?
Is the jury still out, or my problem hill defined ?
(my use case, out of the curiosity/want to learn part, is to have as good performance as possible with my LX-Mini at listening position, in a difficult room (not symetrical, low ceiling, little acoustic treatment possible...).
JMF
I stumbled on the Kii Three and D&D 8c pitch that they can achieve almost perfect operation in "normal" rooms. This seems to rely on the consistent off axis response of the speakers.
Then going through different threads this leaded me to FIR correction as performed with Acourate or Audiolense products (and DRC-FIR for the open source more steep learning curve). This was looking really interesting to improve the system in the room, at listening position. I have bought and read "Accurate sound reproduction" from Mitchco.
Then to Grimm LS1... and the associated whitepaper https://www.grimmaudio.com/wordpress/wp-content/uploads/speakers.pdf . That whitepaper was highlighting the drawbacks/restictions of the FIR approach, especially on the off axis response... Arghh so some promote FIR to solve key issues, but this may impact significantly the off axis response, which is important for imaging and timbre and the secret of LXminis/Kii/D&D.
I started to get bugged about the interest or not to invest time (and money) in FIR correction, either for complete speaker filtering (each driver ground up FIR filtered for listening position) or for final lifting to adjust speakers for listening position.
Is the Grimm whitepaper a bit pushy comparing very different designs related to off axis response (controlled directivity LS1 to MTM arrangement) ?
Can the FIR final corrections be adequate with Cadioid like speakers like my LX-Mini, and would not degrade too much the Off-Axis response?
Is the jury still out, or my problem hill defined ?
(my use case, out of the curiosity/want to learn part, is to have as good performance as possible with my LX-Mini at listening position, in a difficult room (not symetrical, low ceiling, little acoustic treatment possible...).
JMF
The reason FIR isn't used is because of the pre-ringing when sharp filters are used. If you use them for each driver then the preringing artefacts won't cancel out perfectly off axis.
This problem only occurs when you use sharp FIR for 2 different sources (i.e. the crossover) hence explained in the Grimm paper.
That being said, Grimm isn't against the use of FIR. They use it themselves after all. However it's the application that matters. We don't care about the phase of the individual drivers, so why would we use FIR for that if it only creates the previously stated problems? We do care about the sum of the phase of the drivers (that is what we hear after all). We can correct for that using fir!
There is a particular property of filters that says that if the impulse response is symmetrical, the phase will be linear. We can then take the impulse response of the system and flip it's impulse response in time and then convolute it with our system. You might wonder a few things:
Q1. Won't this make the system non-causal?
Q2. Isn't the impulse response of minimum phase systems infinite?
A1: Yes, hence delay is added until the system is causal again. This does require a fair bit of memory and introduces a lot of delay (=not suitable for live applications). For this reason Grimm amde an option to turn this off (they call it low latency mode).
A2: Yes, but eventually the amplitude is so small that it's not relevant anymore. You can just cut it off then. A reasonable estimate is made later in this thread: Understanding Grimm Audio XO
The system won't be TRUELY linear phase, however the Grimm's aren't truely linear phase either. They have a ~45 degree phase shift at ~200Hz. They are linear phase from about 2kHz and up though.
Fun fact: The LS1 was actually intended to also be cardioid! Hence the very wide baffle!
This problem only occurs when you use sharp FIR for 2 different sources (i.e. the crossover) hence explained in the Grimm paper.
That being said, Grimm isn't against the use of FIR. They use it themselves after all. However it's the application that matters. We don't care about the phase of the individual drivers, so why would we use FIR for that if it only creates the previously stated problems? We do care about the sum of the phase of the drivers (that is what we hear after all). We can correct for that using fir!
There is a particular property of filters that says that if the impulse response is symmetrical, the phase will be linear. We can then take the impulse response of the system and flip it's impulse response in time and then convolute it with our system. You might wonder a few things:
Q1. Won't this make the system non-causal?
Q2. Isn't the impulse response of minimum phase systems infinite?
A1: Yes, hence delay is added until the system is causal again. This does require a fair bit of memory and introduces a lot of delay (=not suitable for live applications). For this reason Grimm amde an option to turn this off (they call it low latency mode).
A2: Yes, but eventually the amplitude is so small that it's not relevant anymore. You can just cut it off then. A reasonable estimate is made later in this thread: Understanding Grimm Audio XO
The system won't be TRUELY linear phase, however the Grimm's aren't truely linear phase either. They have a ~45 degree phase shift at ~200Hz. They are linear phase from about 2kHz and up though.
Fun fact: The LS1 was actually intended to also be cardioid! Hence the very wide baffle!
With multiple drivers their relative phase can matter. It isn't important if you change both by the same amount.
Be careful trying to correct for your listening position.
Be careful trying to correct for your listening position.
Only the relative phase near the crossover is really important (as the grimm paper suggests). However it doesnt matter if there is a phase shift, as long as they both shift the same amount (= delay)
Electrical filters, iir and fir, are one dimentional.
It's impossible for them to adjust for an acoustical filter, Iow the speaker casing, that is three dimentional. Exept for one point in space.
The design of the speaker box itself is therefore crucial and can't be adjusted for with other filters.
It's impossible for them to adjust for an acoustical filter, Iow the speaker casing, that is three dimentional. Exept for one point in space.
The design of the speaker box itself is therefore crucial and can't be adjusted for with other filters.
The electrical filters (regardless of form and domain) are one dimensional. However they still matter off axis when you use multiple drivers that play the same frequency range. Multiple drivers don't have the same dispersion and (unless you use coaxials) are not centered in the same space.
Speakers like the Kii abuse this effect to guide the dispersion with multiple drivers! D&D 8C does it with gaps in the enclosure that acousticly filter the response. Nasty to design but very effective.
Ouch, not sure to understand all the answers. Especially the idea of "one dimensional" filters.
And maybe I have to refine my question... Which could be: under which conditions, FIR filters (using Acourate or Audio lense) could help improve frequency and phase response at a "wide" (about 1m area) listening position, improving the perceived response (on axis + off axis reflections ; have the benefits of the FIR adjustment on axis without degrading too much the off axis response of a well designed speaker) ?
JMF
And maybe I have to refine my question... Which could be: under which conditions, FIR filters (using Acourate or Audio lense) could help improve frequency and phase response at a "wide" (about 1m area) listening position, improving the perceived response (on axis + off axis reflections ; have the benefits of the FIR adjustment on axis without degrading too much the off axis response of a well designed speaker) ?
JMF
I am not too keen on the term 1-dimensional filters either. What he means is that if you place a filter in front of the driver, it will effect the response the same way no matter if you listen on axis or off axis. This is of course true, you can't control on & off axis of a single driver using electronic filters. However you can do so by using multiple drivers that play the same frequency range.
Personally I like the Grimm audio approach. I am not sure what Acourate or Audio lense uses as I am not familiar with their tools.
What kind of improvements of phase response are you expecting? The changes in phase shift is hardly audible (if audible at all).
JMF11, ordinarily you cannot fix an acoustic problem with any kind of signal processing. Put simply you cannot change off-axis response with DSP.
Its the math that needs to be used.
To calculate iir or fir filters, you only need one dimention.
To calculate a polar response, you need three dimentions.
OK understood the one dimesion idea and the fact that the FIR filter can't tailor differently the on axis and off axis response of the speaker. So same correction is applied to all directions.
My concern is to identify if there are conditions where the corrections calculated based on the on axis measurements have still acceptable impact on the off axis. Cases where the concerns raised in Grimms white paper about using blind/brute FIR correction are not so relevant as the issues that it would raise are still much compensated by the benefits for normal listening at listening position.
My concern is to identify if there are conditions where the corrections calculated based on the on axis measurements have still acceptable impact on the off axis. Cases where the concerns raised in Grimms white paper about using blind/brute FIR correction are not so relevant as the issues that it would raise are still much compensated by the benefits for normal listening at listening position.
Yes, with some acoustic issues there is a partial EQ that can help, even though it will not sound perfect at any setting. Adjusting exactly as you see at the listening position can be too much.
That depends on how you do it.In a crossover there are always two sources playing at once. Changing effects of one source likely always effects the off axis response. AFAIK the only way to solve it is by changing both sources an equal amount.
I am not sure I fully understand what you mean here. Could you elaborate?
What is acceptable is a personal choice of course. There are scenario's where phase can be corrected without having the issues of FIR. I explained them earlier and they are also described by Grimm's publication. Perhaps there are other ways, but I don't know (I'm by no means a DSP expert).
Check out this post by Mitch, (you have his e-book already):
https://www.diyaudio.com/forums/multi-way/349162-steep-crossovers-3-steps-forward-2-steps-7.html#post6075818
In your use case, yes, you could benefit from using programs like Acourate, Audiolense or the free DRC-FIR. (meaning using it with the LX-Mini)
A lot depends on how the speaker drivers are positioned and where they hand over (at what frequency).
Kind regards, a happy DSP user.
https://www.diyaudio.com/forums/multi-way/349162-steep-crossovers-3-steps-forward-2-steps-7.html#post6075818
In your use case, yes, you could benefit from using programs like Acourate, Audiolense or the free DRC-FIR. (meaning using it with the LX-Mini)
A lot depends on how the speaker drivers are positioned and where they hand over (at what frequency).
Kind regards, a happy DSP user.
yes this is the hard part of it. I've had success by doing a polar response outside with a good amount of windowing time. its possible to see how sound is radiating off axis then, and issues which are common to all axis are very obvious. then you can correct the "average" problems. the result is much better this way
Thanks all for your feedbacks. So in my understanding, it is worth a try, but there is no 100% guarantee for the result. On a case by case basis (different room and speakers), result has to be checked. And corrections have to be as far as possible focused on "average" problems appearing on and off axis to keep the polar response consistent.
I will try to have a look at DRC-FIR to see if the learning curve is not too steep, as Audiolense and Acourate are a significant investment compared to my whole system.
JMF
I will try to have a look at DRC-FIR to see if the learning curve is not too steep, as Audiolense and Acourate are a significant investment compared to my whole system.
JMF
The polar response will never change. The corrections have to be focussed on average so you don't 'correct' something that can't be corrected.
Correct 🙂
@JMF11 - All types of EQ - no matter if it's IIR, FIR or processed in any type DSP imaginable. There is just no way to get around the fact, that you can't change the sound i more than ONE exact point of space - exactly where you measure - without changing it somewhere else too!
My point is. If you do not measure exactly what the speakers do, without the room and understand exactly what can be corrected and what can't. Then it makes little sense to correct it, since you can't be sure that you did not just make something worse..... even though you might get a smoothed average curve.
Average is to find a trend. But if you use FIR, then you need to be very sure to hit the spot with the right amount and type of correction. Or else, you will simply ruin something that was maybe better before.
Maybe FIR can sound better, in a few occasions. But you have to be absolutely certain, that there is no screw-up in any of the following first:
Cabinet construction
Resonances
Driver placement, size, combination and anything that causes any type of non linear distortion.
Filters
Room
Damping
Diffusion
Frequency response
Listening position
.... and the list goes on.
All speakers are each of their own, an individual compromise. So you have to list the things you can't live with, and then find the best of what can be made with the leftovers 😉
Also. Do you have a good smooth extended bass response throughout the room? More than 30% of our judgement, when it comes to good sound reproduction - comes from great bass.
I believe you would never hear a small phase error in the higher octaves, if there is any error in the bass response. Simply too big of a chance, that some masking will kill the tiny advantage that FIR might bring.
Please prove me wrong 🙂 It's an interesting subject and new techniques are sometimes being developed to get the best of a given technology - with a minimum of compromises.
JMF11, here's a thread dedicated to DRC-FIR with some scripts etc. to make it easier:
A convolution based alternative to electrical loudspeaker correction networks
If you want to correct per driver, you'll have to be a bit more creative. But it's doable.
A tool like DRC-FIR uses frequency dependent windows(*) that are user adjustable to filter the measurement(s). While the user still needs to make judgement calls, it does help to filter out the room in the top octaves.
One could use REW to average multiple measurements, though after having done that I still prefer the results of single point measurements in a 'controlled' environment. (meaning I've absorbed early reflections and that cleaned up the little variations between various measurement points, enough for the results to be valid along a wider listening area. Checked by taking multiple measurements after correction)
(*)= just like Acourate and Audiolense
A convolution based alternative to electrical loudspeaker correction networks
If you want to correct per driver, you'll have to be a bit more creative. But it's doable.
A tool like DRC-FIR uses frequency dependent windows(*) that are user adjustable to filter the measurement(s). While the user still needs to make judgement calls, it does help to filter out the room in the top octaves.
One could use REW to average multiple measurements, though after having done that I still prefer the results of single point measurements in a 'controlled' environment. (meaning I've absorbed early reflections and that cleaned up the little variations between various measurement points, enough for the results to be valid along a wider listening area. Checked by taking multiple measurements after correction)
(*)= just like Acourate and Audiolense