The quote below reflects Acourate approach to designing speaker correction filters with driver alignment and digital XO. (I've shortened the text)
My comments are below
The above procedure gives an impression that one can distinguish between speaker correction and room correction. And as the sequence above goes on, the speaker seems to be corrected first, whereafter the room is corrected at last. This all sounds good and sophisticated. But it doesn't work that way.
The final measurement, how it is treated and the "room correction" filter made from it will - for good and bad - correct everything that goes on before the sound reaches the microphone. And that includes the speakers and the newly generated nearfield correction filters.
When the "room correction" filter at the end of the day is convolved with the nearfield correction filter, it's the sonic signature from the room measurement that decide how the sound is corrected. We could probably throw a number of filters in there, and the end result would basically be the same.
My comments are below
ShinOBIWAN said:
The above procedure gives an impression that one can distinguish between speaker correction and room correction. And as the sequence above goes on, the speaker seems to be corrected first, whereafter the room is corrected at last. This all sounds good and sophisticated. But it doesn't work that way.
The final measurement, how it is treated and the "room correction" filter made from it will - for good and bad - correct everything that goes on before the sound reaches the microphone. And that includes the speakers and the newly generated nearfield correction filters.
When the "room correction" filter at the end of the day is convolved with the nearfield correction filter, it's the sonic signature from the room measurement that decide how the sound is corrected. We could probably throw a number of filters in there, and the end result would basically be the same.
Hi Bernt,
this is my last answer here on your message. I have tried myself to give some useful information to this thread (after I have been asked to do so) and not to hijack it or to advertise. As I still do not intend to do so this is my last response here to you. If some members like to experience a discussion what makes sense and what does not then please open a new thread. I have no problem to join.
Also ShinOBIWAN may argue for himself according to your last mail.
Anyway I'm wondering about your approach. Why don't you give an answer but proceed with some new statements? BTW there are clear arguments and good reasons to linearize drivers independantly of a global room corrections. Your point of view is WRONG. But please do not expect me to teach you how to write your software.
this is my last answer here on your message. I have tried myself to give some useful information to this thread (after I have been asked to do so) and not to hijack it or to advertise. As I still do not intend to do so this is my last response here to you. If some members like to experience a discussion what makes sense and what does not then please open a new thread. I have no problem to join.
Also ShinOBIWAN may argue for himself according to your last mail.
Anyway I'm wondering about your approach. Why don't you give an answer but proceed with some new statements? BTW there are clear arguments and good reasons to linearize drivers independantly of a global room corrections. Your point of view is WRONG. But please do not expect me to teach you how to write your software.
uli.brueggemann said:
Ths slight mismatch you point out is because the pulse wasn't totally at rest by the end.
The limitations at the extremes are deliberate. Silence cannot be inverted. But there's no difference between a smoothed and unsmoothed response there.
Edit: New inversion added.
This is a very interesting thread and I apology if my last two posts was too off topic.
Attachments
Wingfeather said:
Yea, that's kinda the same boat I'm in. I wrote some biquad code a few years back, but all it does is process sound files (not even doing 'live' audio coming into the PC).
It's the step of getting it from a nice plaything to a solution that might be usable 'for the masses' that's difficult.
There are a number of miniature embedded PC boards available, though the CPU performance isn't usually that high. I guess it would be possible to add a DSP board and some decent analogue audio hardware to such as system, but that might make the cost a little high.
I believe that FPGA development tools (and the raw hardware cost) is dropping significantly these days, but my limit is messing about with PICs.
What we need is a cheap (sub 200GBP) programmable box with enough power to run filters, and either a digital in/out or decent A/D - D/A stages. Anyone know of such a piece of kit?
Hi sploo,
Damn straight! That would be awesome. And in terms of raw hardware cost, if one were to build it himself, that should be achieveable. It's the development of it that's the kicker. A professionally-made 4-layer PCB, plus a suitable DSP, CPLD (for clock division and similar), and the necessary D/A hardware would probably add up to around there. If the development were done in the DIY fashion (i.e. for free) and people were willing to solder it up themselves (gulp), I think it would be possible.
I also did think about developing some sort of custom PCI interface that included a DSP and converters - but while it removes the need for tricky casework and a separate PSU you suddenly need to write drivers! Nightmare.
I suppose an obvious solution would be to try a DSP development board - but these don't seem to have much in the way of useful I/O - either too few channels or just really really cheap A/D/A - and they're significantly more expensive to boot.
Bit of a bummer, really.
And Uli (btw, how do you attribute the quote boxes to people?!),
Oh, I know that! lol. But the clear blips that exist in the impulse response don't seem to be coming through into the step response. So the integrated result looks smoother than you might expect. Or at least smoother than I expect. Since I find it easier to visualise what's going in the time domain from looking at an impulse response rather than a step response, I wondered what was happening with that.
The impulse response plots you posted do seem to show that a decent amount (although not all) of the visible room reflections have been removed by your filter.
This is what my comments way back at the beginning were really about - about whether your automated method ends up correcting for the room (at the one measurement location) as well as the driver. Because if so then this seems like what the anti-room-correction crowd are warning against. What do you think?
Then again, I could just be reading the graph wrong!
Have you ever tried applying your correction and then moving the microphone and measuring the response there?
All I think was being said here is that it's mathematically possible to do this. Which of course it is. But due to the shocking lack of spatial robustness, I don't think you'll find anybody recommending it as a viable form of correction.
But how? Obviously, you start with the transfer function of the Butterworth/LR/etc filter. Which you mentioned. You can take that and apply the bilinear transform (for example - there are other methods too) to come up with a set of biquad coefficients for an IIR filter. And this gives you a minimum-phase digital version of the original minimum-phase filter.
Sure.
But how do you convert that transfer function to a linear-phase filter kernel for putting in an FIR filter? Pardon me if I'm missing something obvious
Damn straight! That would be awesome. And in terms of raw hardware cost, if one were to build it himself, that should be achieveable. It's the development of it that's the kicker. A professionally-made 4-layer PCB, plus a suitable DSP, CPLD (for clock division and similar), and the necessary D/A hardware would probably add up to around there. If the development were done in the DIY fashion (i.e. for free) and people were willing to solder it up themselves (gulp), I think it would be possible.
I also did think about developing some sort of custom PCI interface that included a DSP and converters - but while it removes the need for tricky casework and a separate PSU you suddenly need to write drivers! Nightmare.
I suppose an obvious solution would be to try a DSP development board - but these don't seem to have much in the way of useful I/O - either too few channels or just really really cheap A/D/A - and they're significantly more expensive to boot.
Bit of a bummer, really.
And Uli (btw, how do you attribute the quote boxes to people?!),
Oh, I know that! lol. But the clear blips that exist in the impulse response don't seem to be coming through into the step response. So the integrated result looks smoother than you might expect. Or at least smoother than I expect. Since I find it easier to visualise what's going in the time domain from looking at an impulse response rather than a step response, I wondered what was happening with that.
The impulse response plots you posted do seem to show that a decent amount (although not all) of the visible room reflections have been removed by your filter.
This is what my comments way back at the beginning were really about - about whether your automated method ends up correcting for the room (at the one measurement location) as well as the driver. Because if so then this seems like what the anti-room-correction crowd are warning against. What do you think?
Then again, I could just be reading the graph wrong!
Have you ever tried applying your correction and then moving the microphone and measuring the response there?
All I think was being said here is that it's mathematically possible to do this. Which of course it is. But due to the shocking lack of spatial robustness, I don't think you'll find anybody recommending it as a viable form of correction.
But how? Obviously, you start with the transfer function of the Butterworth/LR/etc filter. Which you mentioned. You can take that and apply the bilinear transform (for example - there are other methods too) to come up with a set of biquad coefficients for an IIR filter. And this gives you a minimum-phase digital version of the original minimum-phase filter.
Sure.
But how do you convert that transfer function to a linear-phase filter kernel for putting in an FIR filter? Pardon me if I'm missing something obvious
Hi Wingfeather,
two points I have not figured out yet how to do:
1. Quotes: actually I simply have marked some text or a quote, Ctrl-C and Ctrl-V in the query that comes up when hitting the quote button. Now your question has led me to check and I can see in some quotes something like "originally posted by". Is this done by manual editing?
2. Pictures
There are messafes where the picture is already in the message. How to do this? Up to now I used the Attach File function.
Integration:
Sometimes it is e.g. better to use a position chart, sometimes it is better to use a speed chard and sometimes an acceleration chart is more helpful. I live with all of them. The acceleration is most oscillating and the position is the smoothest curve.
So with speakers you can use the pulse response chart or the step response chart. In a pulse response chart it is difficult to see e.g. the bass driver behaviour whereas in the stepresponse it is more difficult to see reflections. So use both properly and all is ok.
Inversion of an unsmoothed response:
Ok, I know that! lol.
I can do that easily too with even better curves compared to the shown example. But have you really tried to create an inversion of an unsmoothed curve without any tricky treatment (like minphase and cutting at the upper and lower frequency range)? And please: the amplitude response is not so important, how does the pulse response look like then and is it useable?
Honestly speaking I wonder why many scientific papers speak about poles and zeroes and division by zero.
Creation of linearphase filters:
sometimes it is difficult to see a forest because of all the trees
www.dspguide.com may help.
two points I have not figured out yet how to do:
1. Quotes: actually I simply have marked some text or a quote, Ctrl-C and Ctrl-V in the query that comes up when hitting the quote button. Now your question has led me to check and I can see in some quotes something like "originally posted by". Is this done by manual editing?
2. Pictures
There are messafes where the picture is already in the message. How to do this? Up to now I used the Attach File function.
Integration:
Sometimes it is e.g. better to use a position chart, sometimes it is better to use a speed chard and sometimes an acceleration chart is more helpful. I live with all of them. The acceleration is most oscillating and the position is the smoothest curve.
So with speakers you can use the pulse response chart or the step response chart. In a pulse response chart it is difficult to see e.g. the bass driver behaviour whereas in the stepresponse it is more difficult to see reflections. So use both properly and all is ok.
Inversion of an unsmoothed response:
Ok, I know that! lol.
I can do that easily too with even better curves compared to the shown example. But have you really tried to create an inversion of an unsmoothed curve without any tricky treatment (like minphase and cutting at the upper and lower frequency range)? And please: the amplitude response is not so important, how does the pulse response look like then and is it useable?
Honestly speaking I wonder why many scientific papers speak about poles and zeroes and division by zero.
Creation of linearphase filters:
sometimes it is difficult to see a forest because of all the trees
www.dspguide.com may help.
Sanded and buffed the baffles and now just the tweeter enclosure to see to and then that's it but I've waited long enough to get my teeth stuck into these so that will have wait. Time to play!
I've noted something that could be an issue. Both of the mid and bass cabinets have their respective midrange and bass enclosures leaking into each other ie. push the bass cone and the mid moves out, let it go back to rest position and so does the mid. I know what it will be - to get the cables to the mid driver I had to drill a hole in the wall that partitions the bass and mid sections and this hole has had its seal broken with all the pulling of the wires.
I'm reluctant to fix it because that means removing most of the foam to get to the holes at the rear and it won't be easy job to get it all back in place afterwards.
How big an issue is/could this be? I'm guessing it will affect tuning frequency and the distortion of the mids in particular. It obviously needs fixing.
Here's the pics:
I've noted something that could be an issue. Both of the mid and bass cabinets have their respective midrange and bass enclosures leaking into each other ie. push the bass cone and the mid moves out, let it go back to rest position and so does the mid. I know what it will be - to get the cables to the mid driver I had to drill a hole in the wall that partitions the bass and mid sections and this hole has had its seal broken with all the pulling of the wires.
I'm reluctant to fix it because that means removing most of the foam to get to the holes at the rear and it won't be easy job to get it all back in place afterwards.
How big an issue is/could this be? I'm guessing it will affect tuning frequency and the distortion of the mids in particular. It obviously needs fixing.
Here's the pics:
An externally hosted image should be here but it was not working when we last tested it.
An externally hosted image should be here but it was not working when we last tested it.
An externally hosted image should be here but it was not working when we last tested it.
An externally hosted image should be here but it was not working when we last tested it.
An externally hosted image should be here but it was not working when we last tested it.
An externally hosted image should be here but it was not working when we last tested it.
Shin,
I'm stunned
by the looks of your speakers. Beautiful they are.
If you decide to open them up to tighten the air leak, I can heartly recommend that you treat the midrange cabinets with wool - in addition to, or instead of foam. Wool behind a midrange has a very positiv effect on midrange clarity. It is said to be due to superb diffraction ability, but I don't know. I only know that it works really well.
I've settled for ordinary wool, but these speakers probably deserves kashmir
I'm stunned
by the looks of your speakers. Beautiful they are.If you decide to open them up to tighten the air leak, I can heartly recommend that you treat the midrange cabinets with wool - in addition to, or instead of foam. Wool behind a midrange has a very positiv effect on midrange clarity. It is said to be due to superb diffraction ability, but I don't know. I only know that it works really well.
I've settled for ordinary wool, but these speakers probably deserves kashmir
Wingfeather said:
For getting cheap PCBs made, some friends have had success with this place: http://www.pcbcart.com/
I do have a particular bit of (ready made) hardware in mind, but I can't talk about it due to commercial issues - I am working on it though, and will detail it if I can get approval.
Wingfeather said:
Like this... [ QUOTE ][ i ]Originally posted by TheUser'sName[ /i ][ B ]Remove the spaces within the brackets to make the tags work.[ /B ][ /QUOTE ]
ShinOBIWAN said:
Superb Ant. I think we've become 'de-sensitized' to seeing the LGTs, and forget just how special the design and execution really is. There really are very few DIY efforts that come close.
Shin
that last but one photo is almost identical to some of the initial CAD proposals, amazing.
However I'm disgusted that you've left the tweeter cabinet unfinished! how very unlike you
Honestly though, superb work, if i ever achieve 1/10th of that quality I'd be amazed. My Kit assemblys arent anywhere near that...
Nick.
that last but one photo is almost identical to some of the initial CAD proposals, amazing.
However I'm disgusted that you've left the tweeter cabinet unfinished! how very unlike you
Honestly though, superb work, if i ever achieve 1/10th of that quality I'd be amazed. My Kit assemblys arent anywhere near that...
Nick.
sploo said:
Cheers Gordon,
There's an easy fix for that. Pop round and see them in the flesh
Even better, wait a couple of years, I'll have the other one finished and you can listen to them 
They're a stunning testiment to my little nephew and a great present to myself.
I'm sure he'd approve if he knew what a speaker was!
Heh, it really is shocking how much they resemble your drawing. What kind of kit are you going to hook them up to in order to give them a shot? BTW, if it were me, I think I'd have to fix the chamber leak before giving them a go. The whole time I was listening to them I'd just be wondering what they'd sound like if only I'd fixed the leak. Also, how is your hardware coming (amps, preamp, etc.)?
What, an air leak ?
I'm truely shocked that you would have neglected such a detail.
If the cable entry between the cabinets is difficult to access, would it be possible to pull the bass driver and 'run' some epoxy down the cable ?
As in pour the epoxy straight onto the speaker cable so it follows it into the hole and seals it.
I'm truely shocked that you would have neglected such a detail.
If the cable entry between the cabinets is difficult to access, would it be possible to pull the bass driver and 'run' some epoxy down the cable ?
As in pour the epoxy straight onto the speaker cable so it follows it into the hole and seals it.
Grumpy_Git said:
m0tion said:
I had to look twice myself to make sure you hadn't pulled a swifty on us ha ha. Uncanny, it shows two things, how accurate the cad program is and how well you can follow instructions!
I can't follow instructions at all so in my case they would have looked very different indeed.
I would be more worried that the bass driver pressure could actually damage the mid driver, though it seems the leak is to small to allow that to happen.
How many man hours do you estimate you have spent on these??
You will probably groan when you think of that, but all of us who haven't spent that time or effort on them reckon your effort was worth it, if that helps you feel better!
Magnificent.
Outstanding!!!
I've been following this thread since beginning, and waiting to see these beasts finished. They look amazing and as mOtion said, they look identical to your project drawing at the beginning of the adventure.
Hope that you will fix that leak and finish amplifiers. This must be one of the best projects i ever saw, and even one of the best speakers made in general.
Good luck with finishing these beauties
I've been following this thread since beginning, and waiting to see these beasts finished. They look amazing and as mOtion said, they look identical to your project drawing at the beginning of the adventure.
Hope that you will fix that leak and finish amplifiers. This must be one of the best projects i ever saw, and even one of the best speakers made in general.
Good luck with finishing these beauties
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.