As a great sanity check, you can use CamillaDSP's file input/output options. Create some stereo pink noise and save it to a file. Create a CamillaDSP config.yml test file to read from and write to a file. Then run CamillaDSP with your XO FIR's, pink noise and test config.yml file. Then take the output file and analyze it in REW. It will show you exactly how CamillaDSP is applying your XO FIRs without testing it on your playback hardware. All files (not including the config.yml) will have to be raw PCM's in the same sample rates.
You can use SoX to convert wav files to raw PCM files and back again as needed.
i do that after a progression of FDW , start with big problems , and work my way to no FDWs and little to no smoothing
FWIW, I have not found the need to use FDW on RePhase's XO's unless you are trying to add speaker/room correction to the XO's. The pure XO's don't contain any measured signal components so they should be clean/pure synthetic constructs and ready to go without smoothing and/or FDWing.
I know you are, and I respect that, just saying, I found some AudioLense literature, but when did marketing, list weaknesses...
It if you go to the actual page it speaks of multiple pre-ring suppression methods.
Right, but you should be able to move the XO point and cover more bandwidth, at 0 transfer function, and to a lower point.
Attachments
Hey, my vote goes with the (very good) advice of dgerg, emailtim and Oaebeieo...
Sorry b-force, you are very assertive...but i think very incorrect. Steep complementary linear phase xovers work very well without pre-ringing.
And make tuning multi-ways considerably easier. And diminish lobing, and improve polar response.
Your pre-ringing comments are valid only at the system high-pass and low-pass ends, ime.
Sorry b-force, you are very assertive...but i think very incorrect. Steep complementary linear phase xovers work very well without pre-ringing.
And make tuning multi-ways considerably easier. And diminish lobing, and improve polar response.
Your pre-ringing comments are valid only at the system high-pass and low-pass ends, ime.
Last edited:
camplo there is a bit of difference between the type of filters being generated in the "room" correction software like Audiolense and Acourate and those which rephase can make. Acourate has quite a lot of pre-ringing suppression features that remove it while still allowing the wanted parts of the time correction. Filters created through windowing and inversion can be processed more easily in a way that leaves much of the pre-ringing out.
There is some information from the DRC docs here
http://drc-fir.sourceforge.net/doc/drc.html#sec33
It wasn't that long ago in the thread that pos and I showed some graphs from synthetic impulses that demonstrated the complete cancellation of all ringing pre and post when the crossover filters are complementary linear phase.
https://www.diyaudio.com/community/...eq-and-fir-filtering-tool.221434/post-6603929
While the speakers have consistent response between on and off axis the ringing will continue to be cancelled but at some point and in some designs that might happen much more quickly as you move off axis.
In case it wasn't obvious the 32 bit IEEE 754 .bin option in rephase writes the filter as raw 32 bit float, if you change the file type to .pcm after it is the same as a raw pcm file.
LIsten, gang
so , I Hope (okay I’m begging) for an answer on this
So I read the audio lense paper…. Intriguing,
ive been wondering for awhile….
why in the world do we use LRs or BWs in linear phase?
It makes no sense, the shape of the filter is designed for two drivers on the same plane to have a minimum phase reaction (like pos says and I like) a “phase tracking” attribute.
why is that even a thing? We should be liberated from shapes and alignments except the summing things we prefer based on the actual distance between the listener and drivers sensitivities and such, not from the minimum phase twisting that those alignments were designed for using drivers on the same plane.
the audio lense filters look a awful lot like Hk filters. Which I’ve used without an HK alignment and they work fantastic, in linear phase.
like , let’s say our drivers are not on the same plane. Shouldn’t there be a new alignment method made that incorporates the use of signal delays to remove GD from the LP with a minimum phase alignment?
back when Linkwitz-Riley made those alignment methods there was no signal delay or FIR. So why is it important to use LR shapes in FIR linear phase crossovers…
the freedom seems limitless….. what I have been dreaming about is a dual bandwidth FIR that allows a driver to play two short bandwidths. Imagine the benefits, especially when path lengths are unequal. You could have a dedicated driver fill in only the comb filter dips on a different axis and location close to the driver where it doesn’t have the same comb-filter attributes. To me it’s a dream, in a car audio system perhaps.
if I have a perfect system except the left side has a comb notch right in the midrange that is location dependent and can’t be delt with and requires a separate driver in another location to fill in..
often times it’s simply the axis in which it’s interacting with the room that causes that dip where a nearby driver on a slightly different axis can fill that one hole….
we’re good enough with this fir stuff to pull this off…
anyway my most recent thoughts…. Lately I’ve been ignoring “LR” alignments with any of my firs and just simply going -6db down to sum with radically different shapes on the knee that suit the room behavior…. I’ve been doing that by cascades of a few different standard alignments to get the shape I want down to -30db (where it matters)
so far so good….
would love your thoughts on this tho…
thanks in advance
so , I Hope (okay I’m begging) for an answer on this
So I read the audio lense paper…. Intriguing,
ive been wondering for awhile….
why in the world do we use LRs or BWs in linear phase?
It makes no sense, the shape of the filter is designed for two drivers on the same plane to have a minimum phase reaction (like pos says and I like) a “phase tracking” attribute.
why is that even a thing? We should be liberated from shapes and alignments except the summing things we prefer based on the actual distance between the listener and drivers sensitivities and such, not from the minimum phase twisting that those alignments were designed for using drivers on the same plane.
the audio lense filters look a awful lot like Hk filters. Which I’ve used without an HK alignment and they work fantastic, in linear phase.
like , let’s say our drivers are not on the same plane. Shouldn’t there be a new alignment method made that incorporates the use of signal delays to remove GD from the LP with a minimum phase alignment?
back when Linkwitz-Riley made those alignment methods there was no signal delay or FIR. So why is it important to use LR shapes in FIR linear phase crossovers…
the freedom seems limitless….. what I have been dreaming about is a dual bandwidth FIR that allows a driver to play two short bandwidths. Imagine the benefits, especially when path lengths are unequal. You could have a dedicated driver fill in only the comb filter dips on a different axis and location close to the driver where it doesn’t have the same comb-filter attributes. To me it’s a dream, in a car audio system perhaps.
if I have a perfect system except the left side has a comb notch right in the midrange that is location dependent and can’t be delt with and requires a separate driver in another location to fill in..
often times it’s simply the axis in which it’s interacting with the room that causes that dip where a nearby driver on a slightly different axis can fill that one hole….
we’re good enough with this fir stuff to pull this off…
anyway my most recent thoughts…. Lately I’ve been ignoring “LR” alignments with any of my firs and just simply going -6db down to sum with radically different shapes on the knee that suit the room behavior…. I’ve been doing that by cascades of a few different standard alignments to get the shape I want down to -30db (where it matters)
so far so good….
would love your thoughts on this tho…
thanks in advance
Last edited:
Could be, although in practice one would almost never use fully complementary filters. (they really only exist in theory)
My whole point was that using higher order systems is just not the way to go and making practical things more difficult.
It's always a little strange that one tiny little aspect of it is being blown up, where you guys all jump onto and forgetting all the other aspects. (in general, not towards you specifically)
As @HammerSandwich correctly stated, most higher order filters can give even an higher excursion (I forgot about that for a second).
The benefits are very small and crossing lower than the Fs is a bad idea to begin with.
One can get very good lobing and polar repsonse with just some easy 4th order filters, no problem.
It can even be easily done with 2nd and 3rd order filters.
If not, than your acoustic design is flawed to begin with.
I use fully complementary xovers pretty much all the time. On 2,3,4 or 5-ways.
Works great provided prerequisite in-band and out-of-band magnitude flattening has been accomplished on a driver by driver basis.
The sum of one side of a fully complementary xover with the flattening EQs effecting it, is of course not complementary electrically.
But the sum of the xover, the EQ's, and the raw driver response, does make for fully complementary acoustic xovers.
Or at least fully acoustically complementary on the reference tuning axis.
It should be noted, there will be no pre-ringing from using fully linear phase xovers, provided there is not serious lobing that destroys complementary acoustic summation.
One great way to reduce the potential for lobing is through the use of steep xovers. If there is any non-theoretical audible pre-ring from off axis summation problems, reducing the width of summation frequencies through steep xovers has to drive the issue towards moot.
My experience is that it makes xovers, combining drivers...so much easier ....(again, provided steep are fully complementary linear phase.)
Even drivers that have barely overlapping response are easy, because you don't need much overlapping response.
And if drivers do have a wide range of overlap, you can choose any frequency for the xover within that range.....IOW, choose the frequency that gives best polar response.
Sure, good design works.
But you can get even better polar response with the right kind of steep
First post in this thread. I am thinking of modifying my system by adding three subs below 80Hz to 1) help me with a more uniform bass response and 2) to dedicate my mid-bass driver to frequencies above 80hz. The three subs would be operated mono (one signal for all) and they would have three different positions in the room.
The “problem” I am facing is that I want to keep on using my stereo tube preamplifier, which I am very fond of and has a stereo AVC.
For the subs, I’d take the output of the analogue preamp into a 2x4 mini DSP or similar and then to an independent amplifier for each sub. If possible, I’d use MSO for the multiple sub optimization. For the mids, I’d have a very simple first order electrical HP. The mid-highs would go from the preamp directly to an amplifier (i.e. no independent DSP on the mid-highs)
So, my question here is if I could perform a linearization of the system as whole with rephase. I would first define a LP filter on the miniDSP for subs, run the MSO (optimizing the subs only) and once the sub response is optimized, I would turn on to rephase to optimize the time/phase relation between mids and subs (and do perhaps other general EQ/linearization corrections that could be of benefit).
Can this be done with rephase? Is the procedure I have outlined above the way to do it (define a LP for the subs with miniDSP + a HP for the mids at the amp input or integrated to the amp circuit, then run MSO for subs only, then linearize with rephase)?
The “problem” I am facing is that I want to keep on using my stereo tube preamplifier, which I am very fond of and has a stereo AVC.
For the subs, I’d take the output of the analogue preamp into a 2x4 mini DSP or similar and then to an independent amplifier for each sub. If possible, I’d use MSO for the multiple sub optimization. For the mids, I’d have a very simple first order electrical HP. The mid-highs would go from the preamp directly to an amplifier (i.e. no independent DSP on the mid-highs)
So, my question here is if I could perform a linearization of the system as whole with rephase. I would first define a LP filter on the miniDSP for subs, run the MSO (optimizing the subs only) and once the sub response is optimized, I would turn on to rephase to optimize the time/phase relation between mids and subs (and do perhaps other general EQ/linearization corrections that could be of benefit).
Can this be done with rephase? Is the procedure I have outlined above the way to do it (define a LP for the subs with miniDSP + a HP for the mids at the amp input or integrated to the amp circuit, then run MSO for subs only, then linearize with rephase)?
Thanks fluid! I am also not that enthusiastic about the misniDSP, but I thought it can make things simpler. A RPi4 should be more capable, but then I'd have to do ADC and then get multichannel out to DACs. I know there is USB to i2s, multichannel, (xmos) from minidsp and and others. Can you comment on a good solution for multichannel out from a computer or Rpi to i2s or on the capabilities of the Rpi4 for the processing of the subs?
Multichannel output from an Rpi needs to be USB or possibly HDMI. A pro soundcard is a simple solution to the analogue input too. DIYINHK has a USB to i2s board and there are probably others but very few 8 channel DAC's or kits.
This thread here is pretty helpful with guidance and recommendations
https://www.audiosciencereview.com/forum/index.php?threads/rpi4-camilladsp-tutorial.29656/
This thread here is pretty helpful with guidance and recommendations
https://www.audiosciencereview.com/forum/index.php?threads/rpi4-camilladsp-tutorial.29656/
No, unless the specific convolution engine is capable of converting the filters on the fly to match the processing sample rate.
Matched filter and playback:
12345 12345 12345
12345 12345 12345
Mismatched filter and playback rates:
12345 12345 12345
123456 123456 123456
This will move where your corrections are applied and gets worse the more the sample rates differ.
- 48/44.1 is 1.09X off
- 48/96 is 2X off
- 48/192 is 4X off
- 48/384 is 8X off
- etc.
Last edited:
MiniDSP products use a sample rate convertor chip to convert anything not at the native rate.
You need to use the bin format to load an FIR filter into the MiniDSP from rephase.
Hey gang , I been lookin at this dsp from audison
it has its own fir calc built in and likely sets delays for filters internally
can someone take a peek Into to this and help me figure what they mean by a 9 pole fir and 10 pole fir, I’m not exactly understanding because I learned fir filters have no poles and is why there stable
Please help
https://www.audison.eu/products/bit-one-hd-virtuoso/
https://www.audison.eu/blog-post/virtuoso-sw/
it has its own fir calc built in and likely sets delays for filters internally
can someone take a peek Into to this and help me figure what they mean by a 9 pole fir and 10 pole fir, I’m not exactly understanding because I learned fir filters have no poles and is why there stable
Please help
https://www.audison.eu/products/bit-one-hd-virtuoso/
https://www.audison.eu/blog-post/virtuoso-sw/
Last edited: