|
Post by patate91 on Jul 20, 2018 5:35:45 GMT
I'm waiting for my MiniDSP to arrive and I was reading about IIR and FIR. I plan to use the MiniDSP as an active crossover as well as an EQ. I would like to add that I'm in the process to get the best results I can with room treatment first. So the crossover part is more important for me. Anyway, I'm having a hard time to understand FIR and if there is benefits in a well treated room over IIR, which appears easier to implement with REW. I'm wondering if there's an Ethan's article I missed (again )
|
|
|
Post by Michael Lawrence on Jul 20, 2018 14:20:31 GMT
FIR filtering is useful when you need to generate a filter that has a phase response OTHER THAN the minimum phase response of the amplitude response. It's often used in beam-steering applications, for example, where we want to change the phase response without changing the amplitude response. Room issues can be non-minimum phase but I'm not sure it makes sense to go through that process in this context. There is an excellent series about FIR filtering on ProSoundWeb written by SynAudCon guru Pat Brown. Part 1
Part 2
Part 3I may be missing a part in there, but that will tell you more about FIR than I could in a forum post. I also recommend this. FIR is a buzzword right now but it's not a cure-all. When used in NON-MP, for any decent LF resolution, the filter needs a LARGE number of taps, which means latency becomes a serious issue very quickly. This is part of why its use is often precluded in live sound situations. To be honest with you, it's not often that I would encounter a small room acoustics problem for which I would recommended FIR filtering as a solution.
|
|
|
Post by patate91 on Jul 20, 2018 16:03:22 GMT
FIR filtering is useful when you need to generate a filter that has a phase response OTHER THAN the minimum phase response of the amplitude response. It's often used in beam-steering applications, for example, where we want to change the phase response without changing the amplitude response. Room issues can be non-minimum phase but I'm not sure it makes sense to go through that process in this context. There is an excellent series about FIR filtering on ProSoundWeb written by SynAudCon guru Pat Brown. Part 1
Part 2
Part 3I may be missing a part in there, but that will tell you more about FIR that I could in a forum post. I also recommend this. FIR is a buzzword right now but it's not a cure-all. When used in NON-MP, for any decent LF resolution, the filter needs a LARGE number of taps, which means latency becomes a serious issue very quickly. This is part of why its use is often precluded in live sound situations. To be honest with you, it's not often that I would encounter a small room acoustic problem for which I would recommended FIR filtering as a solution. Part 3 and "this" links are not working. My speakers have woofers and horn loaded compression drivers, for what I understand it could be a good thing or help to have fir filters as a crossover.
|
|
|
Post by Michael Lawrence on Jul 20, 2018 16:28:10 GMT
Links fixed, I always mess those up. In Part 2, Pat addresses the applicability of using FIR for crossover filtering. If you want my opinion, I don't make a habit of deviating from manufacturer's recommended crossover parameters unless I've determined that they absolutely cannot work. Many hours and dollars go into loudspeaker R+D, with tools we don't have access to in the field, and in general it's unlikely that a homegrown solution will end up superior to what they've developed. Also worth noting that the filter response in isolation means nothing. The drivers have their own amplitude rolloff and non-linear phase responses which will be combined with that of the filters. Only by viewing everything in context together - with a dual-channel analyzer - can enough information be gathered to make an informed decision. Unless there are serious issues with the crossover, I spend my often-limited optimization time on other things. Just my experiences, take it or leave it.
|
|
|
Post by patate91 on Jul 20, 2018 17:19:41 GMT
|
|
|
Post by patate91 on Jul 20, 2018 17:21:55 GMT
|
|
|
Post by Michael Lawrence on Jul 20, 2018 19:28:30 GMT
In general it is more dangerous to run the HF driver too low than to run the LF driver too high. Overexcursion is very likely and will quickly damage a driver. That extra half an octave means about 6dB more excursion at the previous Fc. I'd have to do the math but I think that's right off the top of my head. In this case it appears that the HF driver in question is rated for the application but use caution because lowering the Fc makes its job harder. Also be aware that it will require a new phase alignment with the woofer as a result. I assume you're using a biamp setup?
|
|
|
Post by patate91 on Jul 20, 2018 20:37:39 GMT
In general it is more dangerous to run the HF driver too low than to run the LF driver too high. Overexcursion is very likely and will quickly damage a driver. That extra half an octave means about 6dB more excursion at the previous Fc. I'd have to do the math but I think that's right off the top of my head. In this case it appears that the HF driver in question is rated for the application but use caution because lowering the Fc makes its job harder. Also be aware that it will require a new phase alignment with the woofer as a result. I assume you're using a biamp setup? As you can see Pioneer/TAD already crossover at 650hz using the same slopes as my TN-4 crossover Yes I plan to biamp (a Pioneer and luxman).
|
|
|
Post by patate91 on Jul 20, 2018 20:39:09 GMT
|
|
|
Post by patate91 on Jul 20, 2018 20:45:52 GMT
|
|
|
Post by patate91 on Jul 20, 2018 22:06:44 GMT
Also found this, post 29 Its not a big issue.
Just pad back the horn 3 db using 10 ohms as a base line to calculate the attenuation.
Failing that buy a couple of 8 ohm L pads from Parts Express.
Somewhere on an old server I modelled the Tad crossover. The 36 db crossover low pass slope does two jobs. Its steep slope removes the response hash above 650 herts that the Tad cone has and also creates a time domain delay on the woofers to align the woofers and the horn.
Way ahead of JBL crossover network design at the time
www.audioheritage.org/vbulletin/showthread.php?33526-some-TAD-help-needed-if-possible&p=338942&viewfull=1#post338942
|
|
|
Post by patate91 on Jul 21, 2018 23:44:57 GMT
Of look at this picture from Arai's web site it seems possible to work on group delay with horn placement. Since my horns are not fixe on the speaker cabinet can I use REW to measure it? There's a tab for group delay but I don't know what I'll have to look for. For time alignement it will be easy to do it with the minidsp
|
|
|
Post by Michael Lawrence on Jul 22, 2018 13:39:41 GMT
"Spatial" time alignment yields to phase alignment once the drivers get this close. Using a delay to compensate for the physical distance offset does not guarantee that you have phase alignment through the crossover range, because the two drivers are different and thus have different phase responses. I made this incorrect assumption a few times when I was learning how to do this stuff. With a full-range speaker stacked on a sub, you can compensate with delay to the millimeter based on a physical measurement and still be quite a ways off of the correct phase alignment.
If the group delay is flat (constant line slope on phase plot when viewed with on linear scale) the problem can be corrected with either physical displacement or electronic delay. If it's not, you have to focus your efforts on the crossover region, centered on the frequency where both drivers are playing the same frequency at the same level. This is where the are most interactive, so this is where it's most important to align them.
Real-world loudspeaker drivers are not phase-linear, so unfortunately measuring physical displacement doesn't give us all the information. Remember also that a crossover filter will add phase shift of its own, and not all filter pairs will add symmetrical phase shift.
This is just the tip of the iceberg of a complex topic, which quickly gets into nasty math. In my opinion, REW or other single-channel measurement software is not the best choice for this task. You really want a full dual-channel FFT for this stuff.
|
|
|
Post by patate91 on Jul 22, 2018 14:50:43 GMT
This bring me back to iir vs fir works with phase so theoricaly a fir crossover will make it easier to get everything time and phase aligned. After reading this paper another question raise : is it audible? At a certain point with in a music listening situation is the phase alignement issues are audible? www.rane.com/note160.htmlAnother question the paper brings me is about crossover slope. As stated the sharper are the slopes lesser there's issue with phase (the window becomes really small). It looks like FIR gives the opportunity to have very sharp slopes, 96db I think).
|
|
|
Post by Michael Lawrence on Jul 22, 2018 15:49:40 GMT
I would say that's an oversimplification. Both FIR and IIR can be used to create a minimum phase response - a response in which the phase response is the derivative / conjugate of the amplitude response. In addition FIR can be used in a linear phase configuration to create a phase response other than MP. The natural response of a loudspeaker rolloff is as far as I know always an MP phenomenon so those problems could be dealt with using IIR filters. The DSP I use regularly can generate 48 dB / octave Linkwitz-Riley crossover slopes without need for FIR and I can't imagine a need for something steeper than that for this purpose. A LP FIR filter of that order would impart a considerable latency, as described in Pat Brown's articles.
In general, steeper slopes mean less overlap range between drivers but the extra phase shift means it can be harder to align. It's a tradeoff. When setting crossover slopes manually (that is, other than what the manufacturer recommends) the best choice is determined by situational variables. Bob McCarthy's book "Sound Systems: Design and Optimization" covers this in detail.
Many studies have been conducted concerning the audibility of phase shift, including some by Ethan. In all cases it has been found that phase shift itself is inaudible under normal listening conditions.
|
|