Can ScopeFIR take a measurement, a target curve, and generate the FIR file that corrects the measurement to match the target curve.
Where the target is defined by both mag and phase ? That's what has value for my purposes, and I'm always searching/studying these....
When do you find Kaiser-Bessel most helpful?
I've played with it some, adjusting alpha and comparing against the more commons...tapered cosine, Hanning, Blackman- Harris, Mod Adrienne, etc....
but nearly always end up just sticking with Cosine tapered..
Where the target is defined by both mag and phase ? That's what has value for my purposes, and I'm always searching/studying these....
When do you find Kaiser-Bessel most helpful?
I've played with it some, adjusting alpha and comparing against the more commons...tapered cosine, Hanning, Blackman- Harris, Mod Adrienne, etc....
but nearly always end up just sticking with Cosine tapered..
ScopeFIR doesn't do that -- but you can take the coefficients produced by another program, convolve with a crossover FIR, and get a brand new filter with both characteristics. The new filter length will be the sum of the two initial filter lengths, however the program has a truncation option either by length or by quantizing the coefficients and throwing away the zero values. Truncating to the same precision as the input coefficients most often yields a filter no longer than the bigger one of the pair.
K-B is useful where I want to tweak rolloff vs stopband attenuation: if you're not too concerned about getting -120 dB stopband (60 dB is more than enough for woofers or midranges) you can get a quicker rolloff at the crossover point for a given length of filter.
Note the coefficients for highpass and lowpass are not exactly complementary: they differ by a couple of percent, inaudible in the grand scheme of things. You can make a complementary filter by exporting the LPF coefficients into a text file, negating everything, then adding 1.0 to the central value. Voilà, instant filter pair summing to a one-sample impulse response.
K-B is useful where I want to tweak rolloff vs stopband attenuation: if you're not too concerned about getting -120 dB stopband (60 dB is more than enough for woofers or midranges) you can get a quicker rolloff at the crossover point for a given length of filter.
Note the coefficients for highpass and lowpass are not exactly complementary: they differ by a couple of percent, inaudible in the grand scheme of things. You can make a complementary filter by exporting the LPF coefficients into a text file, negating everything, then adding 1.0 to the central value. Voilà, instant filter pair summing to a one-sample impulse response.
Try hard actually cool
OK so let me understand you correctly hear me out on this
If I have a Mini Dsp 2 x 4HD
And I do driver FIR on 4 FIRs
You’re telling me I can measure the sums of left and right separately and merge the sums FIR with the two other FIRs on each respective ch?
That is actually kinda cool
And keep the same amount of taps?
Can taps be assigned and does it do minidsp filters ?
OK so let me understand you correctly hear me out on this
If I have a Mini Dsp 2 x 4HD
And I do driver FIR on 4 FIRs
You’re telling me I can measure the sums of left and right separately and merge the sums FIR with the two other FIRs on each respective ch?
That is actually kinda cool
And keep the same amount of taps?
Can taps be assigned and does it do minidsp filters ?
From the help page:
You can try ScopeFIR free for thirty days to see if it works for you, after which it wants a license key.
Disclosure: I make no money or get any other recompense, just a satisfied user.
So ..
Let’s say I have an fir for a driver in a two way but have 4 fir ch…..
To do driver and global correction over top of eachother in 2 iterations.
I can have it add the left channel let’s say correction to the two left ch speakers that already have an fir made for driver level tuning ?
Let’s say I have an fir for a driver in a two way but have 4 fir ch…..
To do driver and global correction over top of eachother in 2 iterations.
I can have it add the left channel let’s say correction to the two left ch speakers that already have an fir made for driver level tuning ?
If I understand your question correctly, you have a FIR for correcting a woofer to flat then you want to add a LPF FIR for crossing over said woofer. Then yes, you can convolve both FIRs so you have one FIR which includes both driver correction and crossover filter. The same applies to the tweeter, so two FIRs in total, one for woofer and one for tweeter, which include both crossover filter and frequency response EQ. It's actually a fine way to run since funnies outside driver passbands are sorted before they can affect the summed response.
HOWEVER: equalizing a woofer flat to 20 kHz is not recommended, nor is EQing a tweeter flat to 20 Hz. Generate correction filters only until the crossover filter stopbands, say one octave above crossover for the woofer and one octave below crossover for the tweeter. That way you avoid amplifying math roundoff errors outside the drivers' range of interest and filter stopbands are kept sane instead of rising some ridiculous number of dB at the extrema.
HOWEVER: equalizing a woofer flat to 20 kHz is not recommended, nor is EQing a tweeter flat to 20 Hz. Generate correction filters only until the crossover filter stopbands, say one octave above crossover for the woofer and one octave below crossover for the tweeter. That way you avoid amplifying math roundoff errors outside the drivers' range of interest and filter stopbands are kept sane instead of rising some ridiculous number of dB at the extrema.
Yeah you sorta got it right
Let me try one more time
Let’s say you have 4 for ch , each ch is 1024taps
You have a 2 way system
You use the 4 ch and make a driver correction (eq and phase) and make a crossover for two woofers and two tweeters to make a L and R full range system.
you get the system aligned and it sounds good, altho you want to do a gloabal correction on left and right
So the gloabal left and right would mean 2 additional FIRs that you don’t have.
Could you add that correction over the top of the 4 firs that you already have?
Hope that makes more sense…..
Let me try one more time
Let’s say you have 4 for ch , each ch is 1024taps
You have a 2 way system
You use the 4 ch and make a driver correction (eq and phase) and make a crossover for two woofers and two tweeters to make a L and R full range system.
you get the system aligned and it sounds good, altho you want to do a gloabal correction on left and right
So the gloabal left and right would mean 2 additional FIRs that you don’t have.
Could you add that correction over the top of the 4 firs that you already have?
Hope that makes more sense…..
You can convolve two filters together with trace arithmatic in REW, if you window the output you can maintain the same number of taps.
Gotcha, thx.
I don't really need the ability to convolve files into one. The few times I've toyed with it, just for curiosity's sake, I used REW.
My files sizes are always pretty small...16 k taps maximum (at 48kHz) . My guess is you use considerably larger?
Yep. In my book 60 dB is more than enough for anything....
One sample impulse results from summing are so easy with the simple complementary xover types from rePhase or other FIR generators, I can't see need for further complication ????
So, I’ve done a few systems now in cars with the OpenDrc 2ch upstream of a standard DSP….
Compared to my car with fully FIR crossover and a gloabal fir
Guess what…..
So as long as the phase is good on the standard dsp (crossovers tracking decently and any natural GDs that look like an APF are sorted out by adding an APF to all the other drivers in the modal and lows that don’t have the issue)
Then, the result is the same damn thing
So , 2ch FIR is all anyone needs for 2ch audio
Now the sub adds a layer of complexity because you can’t really add the sub into both ch….. if you want to phase correct below 100hz , the sub will need its own fir OR do the left and right phase correction with sun off , then measure the sub and make it match whichever ch (left or right) and inset it on that ch (whichever matches the best)
Then it might be necessary to add an APF to both Left and Right (no sub) to match the sub after the correction
But , why not a 2ch correction……. I’ve discovered it works great. Just get good phase first .
Compared to my car with fully FIR crossover and a gloabal fir
Guess what…..
So as long as the phase is good on the standard dsp (crossovers tracking decently and any natural GDs that look like an APF are sorted out by adding an APF to all the other drivers in the modal and lows that don’t have the issue)
Then, the result is the same damn thing
So , 2ch FIR is all anyone needs for 2ch audio
Now the sub adds a layer of complexity because you can’t really add the sub into both ch….. if you want to phase correct below 100hz , the sub will need its own fir OR do the left and right phase correction with sun off , then measure the sub and make it match whichever ch (left or right) and inset it on that ch (whichever matches the best)
Then it might be necessary to add an APF to both Left and Right (no sub) to match the sub after the correction
But , why not a 2ch correction……. I’ve discovered it works great. Just get good phase first .
Last edited:
hi there
long time reader, first time poster here. i'm very interested in the paper by greg birchin about the perfect reconstruction crossover, as i'm in the middle of building a wallmounted nested array (symmetrical pairs) speaker. greg, if your reading this, could you please send me the paper via pm (i'm not allowed to pm anybody as i have to less posts)?
i appreciate this community and am happy to contribute, as soon as i have something relevant to ask or share 🙂
best regards, joachim
long time reader, first time poster here. i'm very interested in the paper by greg birchin about the perfect reconstruction crossover, as i'm in the middle of building a wallmounted nested array (symmetrical pairs) speaker. greg, if your reading this, could you please send me the paper via pm (i'm not allowed to pm anybody as i have to less posts)?
i appreciate this community and am happy to contribute, as soon as i have something relevant to ask or share 🙂
best regards, joachim
Hey again ! 🙂
How many ways / xovers were in the comparison you made, global vs doing individual drivers?
With boxes I've tried that comparison on, that were two ways, it didn't matter IF the tuning started out decent. Global equaled individual.
But starting out decent, meant truly decent !!!!!!
Like a Meyer two-way or something....or any box with an honest smoothly sloping phase curve....... either active analog, or passive xover in place, didn't matter much.
When I've tried comparison with 3-ways or above....global FIR has always been pretty yuck, compared to individual sections with separate FIR filter embedded with linear phase xovers.
And i do mean always.
So very curious about your comparisons. Thx 🙂
@mark100
Yeah wow that syn looks like a beast! What a cool project.🥳🥳🥳🥳
So yeah with multi-way unwrapping, the phase needs to track per channel….
What I do is view the phase with the crossovers off, and compare all the ways, and then do what’s needs to be done to get them to track the same with no filters and the magnitude on each driver flattened with no crossovers, the entire natural response.
If I see a wrap (frequencie dependent delay of sort that shows as a phase wrap) I’ll add APFs if needed to get them to track the same.
Then I’ll build the minimum phase crossover, and it’s not so important that left and right match because we can fix all of that phase in the gloabal, it’s more important that same side drivers match in phase…
Then a gloabal unwrap works fine , at least it has worked fine up to now
Yeah wow that syn looks like a beast! What a cool project.🥳🥳🥳🥳
So yeah with multi-way unwrapping, the phase needs to track per channel….
What I do is view the phase with the crossovers off, and compare all the ways, and then do what’s needs to be done to get them to track the same with no filters and the magnitude on each driver flattened with no crossovers, the entire natural response.
If I see a wrap (frequencie dependent delay of sort that shows as a phase wrap) I’ll add APFs if needed to get them to track the same.
Then I’ll build the minimum phase crossover, and it’s not so important that left and right match because we can fix all of that phase in the gloabal, it’s more important that same side drivers match in phase…
Then a gloabal unwrap works fine , at least it has worked fine up to now
Last edited:
I'm creating a document that contains pretty much everything I know about Matched Delay Subtractive crossovers, including aspects discussed in this thread. It is complete, but still in draft form. I have attached the table of contents. If anybody is interested in this, PM me and I will send the whole document -- it's BIG, currently 116 pages and 5 MB, but that's mostly graphics. I'd appreciate feedback about errors, omissions, ambiguities, etc. Thanks.
When I find the time to do so I will definitely try the two-stage subtracive delay with sigma studio.
Regards
Charles
Regards
Charles