...Created Equal
If you'll allow me to answer: the Processor manufacturers all have development tools and code libraries to get you going, the best and most useable of these are, at some cost, available from the larger companies with the most well known names. Lesser lights abound, search for Math Libraries and Graphical User Interface tools and you're off to a start.
But; from experience I can tell you reliable good results come only from discovering through investigation which aspects of the current (off the shelf) state of the art are amenable to audibly beneficial improvement and then knowing how through the appliance of mathematics and science which means and methods are needed to realise them.
Unfortunately because that discovery is not readily obvious nor those realisations easily attainable. I suggest that today downloadable freeware suites like DRC are the best way to cut your teeth and get a flavour of what's possible. There's even freeware available on this site if you look for it! At least then you will have an aiming point.
But; even trying to achieve just that much from scratch is a massively respectable undertaking. Getting from there to matching the best of what's available from the real specialists is probably beyond all but a very few on minds the planet. It's really that tricky so:
Earlier in this thread I tried to set the scene...
With a follow up here...
If you're still reading; then I guess that's you and the rest of us all together in one big happy huddle then. Or is it?
Firstly, it's all about software; the OS, then it's all about more software; the player and then its all about even more software again; the DSP package. In fact its all; all about the software. Or is it?
There are thousands of folks who started out with the self same idea and slowly came to the conclusion that once you get performance of the signal conversion I/O gear into the upper reaches of the pro-audio quality level. Aside from finding some elegant way to manage gain structure (not so easy or cheap to do well) and achieve the right measures of galvanic isolation (even nastier than it sounds) having overcome both these big issues and found solutions to the other tortuous matters like system scalability and timing stability then everything will be fine. So now it seems like it's all about the hardware. Or is it?
Let's see; I've had quite a few proprietary active systems here and these have presented me with the most frustration, the greatest disappointment and some of my largest financial losses in this area to date. So when it comes to that approach i.e. trusting a vendor's reputation and feeling comfortable in making what's really a quite substantial investment with them; in dragon's den speak: 'I'm out'.
I should say at this point that DSP based active systems per se are at best a half way (dolls) house and although noticeable improvements are achievable. It's only when you successfully begin to remove from the listening equation large chunks of the room effects that the jaw dropping/happy face time really starts happening. Not so trivial to get there but once you do there's no going back; ever!
I've been travelling down this road of active systems since the 1980's and because of my work background I've been in and out of academic, governmental and industrial places that most folks don't even know exist i.e. places where they have and keep up to date 15 year product and 25 technology roadmaps with hundreds of people involved in underwriting and validating them.
There are more of these places than you may expect. Everything in this arena is happening in accordance with these plans; albeit subject to delivery schedules affected by prevailing influences from sociological, ecological, economic and even political trends.
In our time of relative industrial maturity keeping one eye always open for when some point of technology interception will enable a new wave of marketplace disruption is slightly more challenging than train spotting without either a timetable or a map of the station locations i.e. not all that difficult but you really have to want to do it and be prepared to live on an aeroplane for 20 years just to get up to speed.
Oh and then when you get there they have to let you in through the door so you can give what's wanted; get out with what's needed and be back on the plane again before it leaves. You'd think e-mail and video conferencing would be better way's to get this done and while they do help that's just not how it works. Yes papers get published and post doc's get promoted. Patents are filed, CEO's and VP's come and go. But the plans, what feeds them and keeps them alive, always remain.
Key to the productisation of all this planning are standards and by this I mean internationally recognised properly accredited standards coming from competent and accepted bodies. Not that set of exploitative gangsterised poorly motivated muscle-bound market mafia advert driven hyped up free for all nonsense which abounds in their absence and never does any of US any real good. Remember Try (to your entire satisfaction) Well Before You Buy!
Interoperability is always an absolute must have; I lobbied for years to get the producer of a well known software Music Player to include DSP capabilities in their offering and I believe it's been a very successful move for them and for many of us too. So it can be done and everybody benefits.
But better even than interoperability alone is a De Facto standard. These are at least better than none and can be started by a trade association or group of manufacturers actively seeking to benefit the customer by differentiating their products on cost/performance within a common framework of deliverables. But they've got to want to; and who I wonder, could possibly make them wish to do that?
So to the current state of audio DSP and active systems. We are now well past the point of supply required to support the emergence of (near) perfect products for our purposes. But who's ever going to agree on what they are? The same as was true for cars early in the previous century; this answer is of course to be found in the future(ologist). Convergence of product format, function and specification is an inevitable consequence of commonality in requirements arising from similarity in purpose of application.
Simple really, you'd almost be forgiven for thinking manufacturers weren't doing this because they hadn't sat down and thought about it. But they did, while rubbing their hands together and silently thanking us for not having done the same. It's the 21st century folks; the old ways of being a consumer and silently suffering a vendor's opportunistic predation are just so yesterday.
In this still immature but presently emerging niche product sector I would rather buy from someone like the thread starter or indeed anyone who at least looks like they've given some thought to what was best for me rather than just themselves. So old fashioned; I know!
I've worked in the conception, production, ratification and acceptance of standards like Ethernet and PCI Express to name but two. It's frustratingly obvious to me that:
We i.e. Us. All need those benefit's arising from our would be suppliers having gotten their act together and done the decent thing. Will they ever; well, I guess that's up to you?
So this audio DSP active speaker thing; is it all about the software, is it all about the hardware or in the end is it all about something as seemingly irrelevant as silly old standardisation?
You choose, more to follow...
If you'll allow me to answer: the Processor manufacturers all have development tools and code libraries to get you going, the best and most useable of these are, at some cost, available from the larger companies with the most well known names. Lesser lights abound, search for Math Libraries and Graphical User Interface tools and you're off to a start.
But; from experience I can tell you reliable good results come only from discovering through investigation which aspects of the current (off the shelf) state of the art are amenable to audibly beneficial improvement and then knowing how through the appliance of mathematics and science which means and methods are needed to realise them.
Unfortunately because that discovery is not readily obvious nor those realisations easily attainable. I suggest that today downloadable freeware suites like DRC are the best way to cut your teeth and get a flavour of what's possible. There's even freeware available on this site if you look for it! At least then you will have an aiming point.
But; even trying to achieve just that much from scratch is a massively respectable undertaking. Getting from there to matching the best of what's available from the real specialists is probably beyond all but a very few on minds the planet. It's really that tricky so:
Earlier in this thread I tried to set the scene...
With a follow up here...
If you're still reading; then I guess that's you and the rest of us all together in one big happy huddle then. Or is it?
Firstly, it's all about software; the OS, then it's all about more software; the player and then its all about even more software again; the DSP package. In fact its all; all about the software. Or is it?
There are thousands of folks who started out with the self same idea and slowly came to the conclusion that once you get performance of the signal conversion I/O gear into the upper reaches of the pro-audio quality level. Aside from finding some elegant way to manage gain structure (not so easy or cheap to do well) and achieve the right measures of galvanic isolation (even nastier than it sounds) having overcome both these big issues and found solutions to the other tortuous matters like system scalability and timing stability then everything will be fine. So now it seems like it's all about the hardware. Or is it?
Let's see; I've had quite a few proprietary active systems here and these have presented me with the most frustration, the greatest disappointment and some of my largest financial losses in this area to date. So when it comes to that approach i.e. trusting a vendor's reputation and feeling comfortable in making what's really a quite substantial investment with them; in dragon's den speak: 'I'm out'.
I should say at this point that DSP based active systems per se are at best a half way (dolls) house and although noticeable improvements are achievable. It's only when you successfully begin to remove from the listening equation large chunks of the room effects that the jaw dropping/happy face time really starts happening. Not so trivial to get there but once you do there's no going back; ever!
I've been travelling down this road of active systems since the 1980's and because of my work background I've been in and out of academic, governmental and industrial places that most folks don't even know exist i.e. places where they have and keep up to date 15 year product and 25 technology roadmaps with hundreds of people involved in underwriting and validating them.
There are more of these places than you may expect. Everything in this arena is happening in accordance with these plans; albeit subject to delivery schedules affected by prevailing influences from sociological, ecological, economic and even political trends.
In our time of relative industrial maturity keeping one eye always open for when some point of technology interception will enable a new wave of marketplace disruption is slightly more challenging than train spotting without either a timetable or a map of the station locations i.e. not all that difficult but you really have to want to do it and be prepared to live on an aeroplane for 20 years just to get up to speed.
Oh and then when you get there they have to let you in through the door so you can give what's wanted; get out with what's needed and be back on the plane again before it leaves. You'd think e-mail and video conferencing would be better way's to get this done and while they do help that's just not how it works. Yes papers get published and post doc's get promoted. Patents are filed, CEO's and VP's come and go. But the plans, what feeds them and keeps them alive, always remain.
Key to the productisation of all this planning are standards and by this I mean internationally recognised properly accredited standards coming from competent and accepted bodies. Not that set of exploitative gangsterised poorly motivated muscle-bound market mafia advert driven hyped up free for all nonsense which abounds in their absence and never does any of US any real good. Remember Try (to your entire satisfaction) Well Before You Buy!
Interoperability is always an absolute must have; I lobbied for years to get the producer of a well known software Music Player to include DSP capabilities in their offering and I believe it's been a very successful move for them and for many of us too. So it can be done and everybody benefits.
But better even than interoperability alone is a De Facto standard. These are at least better than none and can be started by a trade association or group of manufacturers actively seeking to benefit the customer by differentiating their products on cost/performance within a common framework of deliverables. But they've got to want to; and who I wonder, could possibly make them wish to do that?
So to the current state of audio DSP and active systems. We are now well past the point of supply required to support the emergence of (near) perfect products for our purposes. But who's ever going to agree on what they are? The same as was true for cars early in the previous century; this answer is of course to be found in the future(ologist). Convergence of product format, function and specification is an inevitable consequence of commonality in requirements arising from similarity in purpose of application.
Simple really, you'd almost be forgiven for thinking manufacturers weren't doing this because they hadn't sat down and thought about it. But they did, while rubbing their hands together and silently thanking us for not having done the same. It's the 21st century folks; the old ways of being a consumer and silently suffering a vendor's opportunistic predation are just so yesterday.
In this still immature but presently emerging niche product sector I would rather buy from someone like the thread starter or indeed anyone who at least looks like they've given some thought to what was best for me rather than just themselves. So old fashioned; I know!
I've worked in the conception, production, ratification and acceptance of standards like Ethernet and PCI Express to name but two. It's frustratingly obvious to me that:
We i.e. Us. All need those benefit's arising from our would be suppliers having gotten their act together and done the decent thing. Will they ever; well, I guess that's up to you?
So this audio DSP active speaker thing; is it all about the software, is it all about the hardware or in the end is it all about something as seemingly irrelevant as silly old standardisation?
You choose, more to follow...
Last edited:
...not created equal
Steven, my first post on this subject is #261 within this thread.
This is my second in this series responding to this comment:
Originally Posted by Tranquility Bass -
The future is looking bright for the Analog Devices SHARC DSP family.
Next generation SHARC+ technology with embedded ARM-CORTEX-A5 core !!
ADSP-SC58x and ADSP-2158x Series | Analog Devices
Nasty IIR filters are right out of the question for me and as it's just another source of error I'm not a great fan of decimation either. So until these chips get much more powerful Double Precision Floating Point Units, properly usable SIMD Pipelined Vector Processing Instructions and an easily workable software API. In my view it's like choosing a toy when you really need a tool i.e. vaguely similar but one of these just won't do the (64K/128K+ taps x N channels) job.
Maybe a whole bunch of them could do it but then on overall system cost bang per buck will probably still favour the big boys with their own production fabs and well established code base. It's their goal to stay on top of this pile and it looks like they're going to be there for quite some time to come. Not for ever maybe but I'd give it at least another two or three process nodes or say 7 or 8 years before that even looks like changing. Nothing to stop anyone trying the other way though.
In this game without any recognised international standards it's just historical prejudice, hyped up fashion and some vendors wishes to leverage proprietary packaging based on the 'lock-in' profit motivation along with poorly thought through system partitioning that's holding us all back by perpetuating even more confusion. Still; the market(ing?)'s always right donchano: it's the way of the world. But as the customer gets older growing an ever more investment minded perspective the web helps them to get wiser. Spending big bucks only to end up with expensive built in obsolescence looks just so last century.
Right now my money's with what the grownups in the pro-space go for. After all that I/O and signal conversion stuff is what they used to make the recordings with in the first place. But this could all change if someone found the right mix of hardware and software for the future of audio DSP and Build Your Own Active Systems. Not a trivial undertaking though.
I think you have to have designed and implemented a number of active systems from a variety of technologies before you can say what's missing today and what's really needed in the future. Even if we had exactly that; decently priced and properly proportioned it still might not sell though because not many can be bothered to wade through the appalling complexities, gather the relevant experience and work out precisely what to do about it all. Customers included.
Even if someone did produce the 'missing link' there's no guarantee that any one vendor will get all their ducks in a row so it works well for your next build or the one after that and in a world without any proper standards that's where open systems interoperability steps in to save the day. Only customer's best interests and system architectural realities can be allowed to dictate which parameters are used in the multiple constraints satisfaction of optimal functional partitioning. Everything else is just more of the same piled higher and deeper.
For the foreseeable future is something (worth having) for (next to) nothing really on the menu? Ask me again in ten years time and by today's measures the answer could perhaps be yes.
But; by then of course the game will have moved on and all of that costly closed off expensive proprietary kit along with the buyers cash will have gone to live in the big skip in the sky.
However, a significant proportion of a purchaser's open systems components may well still be in use.
Steven, my first post on this subject is #261 within this thread.
This is my second in this series responding to this comment:
Originally Posted by Tranquility Bass -
The future is looking bright for the Analog Devices SHARC DSP family.
Next generation SHARC+ technology with embedded ARM-CORTEX-A5 core !!
ADSP-SC58x and ADSP-2158x Series | Analog Devices
Nasty IIR filters are right out of the question for me and as it's just another source of error I'm not a great fan of decimation either. So until these chips get much more powerful Double Precision Floating Point Units, properly usable SIMD Pipelined Vector Processing Instructions and an easily workable software API. In my view it's like choosing a toy when you really need a tool i.e. vaguely similar but one of these just won't do the (64K/128K+ taps x N channels) job.
Maybe a whole bunch of them could do it but then on overall system cost bang per buck will probably still favour the big boys with their own production fabs and well established code base. It's their goal to stay on top of this pile and it looks like they're going to be there for quite some time to come. Not for ever maybe but I'd give it at least another two or three process nodes or say 7 or 8 years before that even looks like changing. Nothing to stop anyone trying the other way though.
In this game without any recognised international standards it's just historical prejudice, hyped up fashion and some vendors wishes to leverage proprietary packaging based on the 'lock-in' profit motivation along with poorly thought through system partitioning that's holding us all back by perpetuating even more confusion. Still; the market(ing?)'s always right donchano: it's the way of the world. But as the customer gets older growing an ever more investment minded perspective the web helps them to get wiser. Spending big bucks only to end up with expensive built in obsolescence looks just so last century.
Right now my money's with what the grownups in the pro-space go for. After all that I/O and signal conversion stuff is what they used to make the recordings with in the first place. But this could all change if someone found the right mix of hardware and software for the future of audio DSP and Build Your Own Active Systems. Not a trivial undertaking though.
I think you have to have designed and implemented a number of active systems from a variety of technologies before you can say what's missing today and what's really needed in the future. Even if we had exactly that; decently priced and properly proportioned it still might not sell though because not many can be bothered to wade through the appalling complexities, gather the relevant experience and work out precisely what to do about it all. Customers included.
Even if someone did produce the 'missing link' there's no guarantee that any one vendor will get all their ducks in a row so it works well for your next build or the one after that and in a world without any proper standards that's where open systems interoperability steps in to save the day. Only customer's best interests and system architectural realities can be allowed to dictate which parameters are used in the multiple constraints satisfaction of optimal functional partitioning. Everything else is just more of the same piled higher and deeper.
For the foreseeable future is something (worth having) for (next to) nothing really on the menu? Ask me again in ten years time and by today's measures the answer could perhaps be yes.
But; by then of course the game will have moved on and all of that costly closed off expensive proprietary kit along with the buyers cash will have gone to live in the big skip in the sky.
However, a significant proportion of a purchaser's open systems components may well still be in use.
Logician,
I think you will get many opinions on the use of IIR or FIR type filters. I guess it really comes down to what you are trying to do with the filter. I have seen where IIR filters were used in an FIR application. As you say the FIR is going to take much more processing capability but with that comes the latency problems in some applications. I don't think it is correct to discount the IIR filters out of hand but I will leave that conversation for others with much more knowledge than me to discuss. For network applications for xo's I think the actual speakers themselves and the configuration has a large part of this decision. A speaker with very good phase properties and some physical alignment of two or more drivers goes a long way to making the situation much simpler to solve. Now adding all the other functions we have been talking about earlier may well demand the move to the FIR solutions.
ps. I hadn't realized you were pointing back at an earlier post in this thread, I just got a fault when trying to click on your active link in your post.
I think you will get many opinions on the use of IIR or FIR type filters. I guess it really comes down to what you are trying to do with the filter. I have seen where IIR filters were used in an FIR application. As you say the FIR is going to take much more processing capability but with that comes the latency problems in some applications. I don't think it is correct to discount the IIR filters out of hand but I will leave that conversation for others with much more knowledge than me to discuss. For network applications for xo's I think the actual speakers themselves and the configuration has a large part of this decision. A speaker with very good phase properties and some physical alignment of two or more drivers goes a long way to making the situation much simpler to solve. Now adding all the other functions we have been talking about earlier may well demand the move to the FIR solutions.
ps. I hadn't realized you were pointing back at an earlier post in this thread, I just got a fault when trying to click on your active link in your post.
Logician,
You say you've bought several DSP speaker products before and been disappointed. Can you tell us which ones & perhaps why?
Are you involved with a commercial product? It would be good to know your association with any of the groups actively involved in DSP for speakers and perhaps what your day job is.
_________________
My suggestion of a dedicated micro ATX PC board running Linux for a 'digital speaker' was precisely to address the 'lock in' issues you mention.
It also addresses connectivity by having ethernet. The de facto standard for 21st century sound is looking to be 'web' based.
The 'digital speaker' will simply look at a URL on the LAN and play it continuously.
__________________________
When I started playing with DSP in the early 90's, a 200 pt FIR (or 180 pt IIR) with TI 320c20 was SOTA. In the early 90's with the Essex lot, this was still the case but the DACs had improved cos we could modify the S/PDIF stream.
When I returned to civilisation (??) this Millenium after more than a decade in the bush, I find that even the cheapest PC had more than sufficient power to do stuff I only dreamt about in da old days.
It's no longer necessary to use DSP processors to do fancy real time EQ for speakers. You could programme a standalone programme or VST to do it all using C to run on a cheapo PC.
It's this type of computing power that gets more powerful and cheaper fastest.
So your recommendation to try DRC or Accourate is spot on. I'm not sure anything more powerful is necessary .. certainly for stereo.
I do have a caveat about room EQ but only perhaps cos I've no hands on experience with DSP for this and I have been studying room response for a long time.
______________________
For me, a surround sound system is a speaker ... a device you put an 'electrical' signal in at one end and get a noise out of the other.
DSP for a single speaker ignores the fact that a speaker ISN'T a 1D in/out chain. It's 1D in & 3D out.
Using many speakers in the room gives you a much better chance of doing something sensible with the room.
_____________________
Would you like to expand on your dislike for IIRs?
You say you've bought several DSP speaker products before and been disappointed. Can you tell us which ones & perhaps why?
Are you involved with a commercial product? It would be good to know your association with any of the groups actively involved in DSP for speakers and perhaps what your day job is.
_________________
My suggestion of a dedicated micro ATX PC board running Linux for a 'digital speaker' was precisely to address the 'lock in' issues you mention.
It also addresses connectivity by having ethernet. The de facto standard for 21st century sound is looking to be 'web' based.
The 'digital speaker' will simply look at a URL on the LAN and play it continuously.
__________________________
When I started playing with DSP in the early 90's, a 200 pt FIR (or 180 pt IIR) with TI 320c20 was SOTA. In the early 90's with the Essex lot, this was still the case but the DACs had improved cos we could modify the S/PDIF stream.
When I returned to civilisation (??) this Millenium after more than a decade in the bush, I find that even the cheapest PC had more than sufficient power to do stuff I only dreamt about in da old days.
It's no longer necessary to use DSP processors to do fancy real time EQ for speakers. You could programme a standalone programme or VST to do it all using C to run on a cheapo PC.
It's this type of computing power that gets more powerful and cheaper fastest.
So your recommendation to try DRC or Accourate is spot on. I'm not sure anything more powerful is necessary .. certainly for stereo.
I do have a caveat about room EQ but only perhaps cos I've no hands on experience with DSP for this and I have been studying room response for a long time.
______________________
For me, a surround sound system is a speaker ... a device you put an 'electrical' signal in at one end and get a noise out of the other.
DSP for a single speaker ignores the fact that a speaker ISN'T a 1D in/out chain. It's 1D in & 3D out.
Using many speakers in the room gives you a much better chance of doing something sensible with the room.
_____________________
Would you like to expand on your dislike for IIRs?
Last edited:
kgrlee,
I understand that most music will be supplied by streaming media at some point. I also understand many functions of the audio system being controlled by a computer. But how or where do you propose to do a crossover? I can see perhaps using a passive or active analog crossover in the speaker and doing any correction in the computer but I don't see how you could have the computer do the crossover unless you are going to use a multiple band output from the computer then? That would mean bi wiring or even tri wiring if each speaker was to have a discrete output from the computer to each component. How do you propose to distribute all the functions, all on the computer or in multiple places along the chain.
ps. not going to be up much longer, it is almost midnight and I have to be somewhere early in the morning.
I understand that most music will be supplied by streaming media at some point. I also understand many functions of the audio system being controlled by a computer. But how or where do you propose to do a crossover? I can see perhaps using a passive or active analog crossover in the speaker and doing any correction in the computer but I don't see how you could have the computer do the crossover unless you are going to use a multiple band output from the computer then? That would mean bi wiring or even tri wiring if each speaker was to have a discrete output from the computer to each component. How do you propose to distribute all the functions, all on the computer or in multiple places along the chain.
ps. not going to be up much longer, it is almost midnight and I have to be somewhere early in the morning.
The mini-ATX computer is on the speaker. It connects to your music source via your LAN.
It receives the digital stream via the LAN, processes it and gives from 2 to 8 analogue o/ps from the built in stereo to 7.1 DACs on board.
A mini-ATX board with stereo out can do a 2 way xover and apply whatever EQ you want. This is the 3-way system in my post #270
I would still do the usual 3-5kHz xover passive though it can be very simple as it only has to get the transition 'smooth' .. leaving 'flat' to the digital EQ.
kgrlee, that might be another thing that has changed while you were in the bush. Tweeters can play lower now than for example the T27. At the same time, much more has become known about the importance of off axis behaviour. In that context, 3KHz is already on the high end, 5KHz is a no-no imo. Also, when you cross over that high, pathway differences between the two drivers can become very pronounced. In order to get the main lobe in the right place, usually the tweeter will have to be delayed. This is not trivial to do with a passive filter.
multi way from computer
Hello Kindhornman. Those questions are valuable and sometime ego I was asking myself. My solution is an exterior multichannel DAC. The best values are those from the pro world. Then you have a simple link between the computer (crossovers and DRC) and the DAC (USB or chose your flavor) and then analogue connections to each amplifier/speaker, just as usually done except for the number of channels.
I have a 5 way system with closed bass box and horns from 250 Hz up.
Chris
This tread is about that: http://www.diyaudio.com/forums/multi-way/266524-you-active-multi-way.html
Hello Kindhornman. Those questions are valuable and sometime ego I was asking myself. My solution is an exterior multichannel DAC. The best values are those from the pro world. Then you have a simple link between the computer (crossovers and DRC) and the DAC (USB or chose your flavor) and then analogue connections to each amplifier/speaker, just as usually done except for the number of channels.
I have a 5 way system with closed bass box and horns from 250 Hz up.
Chris
This tread is about that: http://www.diyaudio.com/forums/multi-way/266524-you-active-multi-way.html
Thanks everyone for your responses. I was just conceptually thinking of the computer being remote from the speakers and having multiple outputs for the speaker at a distance. I wasn't thinking for sure about two atx type of computers one at each speaker and a lan to connect this all to something like a NAS. Even with a mini atx or even something like a pair of Arduino boards this seems to be overly complex and costly in the end by the time you string all of this together.
That seems to be the advantage of current dsp's and dac chips being placed strategically along the chain. I surely want to stay away from passive network components in the speaker enclosure beyond say some simple zobel or LRC tank circuits to do some impedance corrections right at the devices. Beyond that I would think just because of the efficiency difference at least a Saleen-Key type of active network would be more appropriate today.
But if we can do these functions with a small dsp and some dacs why even bother going the Saleen-Key route? When we are talking about an entire unitized system things can be done with a computer based software solution and external dsp speaker based solution but what if you are only supplying a speaker with built in amplification and xo, how you divide the different functions without knowing what the end customer is going to have becomes a big question.
I know with my cone driver going as low as if does and how high it can go before things start looking somewhat ragged I have chosen about 2.5khz for the xo point. Using a dome tweeter without a horn or waveguide I don't think I would want o go any lower than this crossover point, any lower than that and I think you are just asking to much of the tweeter. As also noted the match between the two in radiation pattern is another concern.
I think that 2.5Khz is high enough to keep the critical bandwidth within one device is also one of the goals, in larger systems it always takes some thinking of at what frequencies you want to do the divisions, phase alignment and radiation angle becomes more of the story the larger the system.
If I was doing a horn loaded system with more than one horn than my thinking shifts to other considerations.
I don't know enough to consider what happens when you start using more than one dsp in different locations doing different functions. I have to think about a consumer who is not going to understand any of this they aren't going to be doing any convolution or phase correction, they just want to use whatever source component they have in the easiest manner.
Whether that is by a simple wired connection or something like usb or Bluetooth and now even the WiSA wireless. A lan connection to a nas server would be probably about the highest level I would expect from an informed consumer that has some knowledge of building a higher end sound system and then I would expect that this would also include some video functions for movies and such. At that point using FIR type filters becomes another issue if you need to sync with a video stream.
That seems to be the advantage of current dsp's and dac chips being placed strategically along the chain. I surely want to stay away from passive network components in the speaker enclosure beyond say some simple zobel or LRC tank circuits to do some impedance corrections right at the devices. Beyond that I would think just because of the efficiency difference at least a Saleen-Key type of active network would be more appropriate today.
But if we can do these functions with a small dsp and some dacs why even bother going the Saleen-Key route? When we are talking about an entire unitized system things can be done with a computer based software solution and external dsp speaker based solution but what if you are only supplying a speaker with built in amplification and xo, how you divide the different functions without knowing what the end customer is going to have becomes a big question.
I know with my cone driver going as low as if does and how high it can go before things start looking somewhat ragged I have chosen about 2.5khz for the xo point. Using a dome tweeter without a horn or waveguide I don't think I would want o go any lower than this crossover point, any lower than that and I think you are just asking to much of the tweeter. As also noted the match between the two in radiation pattern is another concern.
I think that 2.5Khz is high enough to keep the critical bandwidth within one device is also one of the goals, in larger systems it always takes some thinking of at what frequencies you want to do the divisions, phase alignment and radiation angle becomes more of the story the larger the system.
If I was doing a horn loaded system with more than one horn than my thinking shifts to other considerations.
I don't know enough to consider what happens when you start using more than one dsp in different locations doing different functions. I have to think about a consumer who is not going to understand any of this they aren't going to be doing any convolution or phase correction, they just want to use whatever source component they have in the easiest manner.
Whether that is by a simple wired connection or something like usb or Bluetooth and now even the WiSA wireless. A lan connection to a nas server would be probably about the highest level I would expect from an informed consumer that has some knowledge of building a higher end sound system and then I would expect that this would also include some video functions for movies and such. At that point using FIR type filters becomes another issue if you need to sync with a video stream.
A couple thoughts:
1. ATX, even "micro ATX" systems are fairly large to incorporate into a small speaker box, HOWEVER, there are indeed a number of functionally similar single-board computers that have much smaller form factors.
2. I would not include Windows on any embedded system. First, in small quantities the costs will kill you, and second, it exposes you to customer "fiddling" and support calls. Use Linux. Use one of the free distributions, and use all free software. LADSPA plugins are available to do pretty much anything you need, and work in Ecasound, also free. Develop on your separate computer, upload and play.
1. ATX, even "micro ATX" systems are fairly large to incorporate into a small speaker box, HOWEVER, there are indeed a number of functionally similar single-board computers that have much smaller form factors.
2. I would not include Windows on any embedded system. First, in small quantities the costs will kill you, and second, it exposes you to customer "fiddling" and support calls. Use Linux. Use one of the free distributions, and use all free software. LADSPA plugins are available to do pretty much anything you need, and work in Ecasound, also free. Develop on your separate computer, upload and play.
...piled higher and deeper
kgrlee,
I've put my responses between your questions:
You say you've bought several DSP speaker products before and been disappointed. Can you tell us which ones & perhaps why?
I’ve had a couple of Meridian actives, later one from Tact and then another from Lyngdorf. In our brief meeting I didn’t really get on with Peter but I enjoyed dealing with his folks and each speaker was definitely better than the last. All were impressively well-made and excellent as objet d’art, though ultimately each was unsatisfying in some way or another. Initially I thought it was just the drive units as like CD surely this digital stuff was all ‘Perfect Sound Forever’; naive I know but I was younger and a lot more trusting then. Eventually over time through activating other manufacturer’s speakers I slowly came to the conclusion that it was the DSP filtering letting them all down. I had a brief stint with a DEQX but there I learned that the FIR filters were too short, the intricacy of the processing perhaps too limited to be properly effective and their signal conversion technology not to my taste. I said most of this to Kim last time we spoke so no big surprises here.
Are you involved with a commercial product? It would be good to know your association with any of the groups actively involved in DSP for speakers
After later spending years going through just about everything audio DSP both H/W & S/W based; because the range of filter specification parameters is sufficiently large to allow me to get a good result from almost any drive unit I include I now use audiolense in my own system. I feel I have a good relationship with the author but there's no commercial activity at present. I am hoping a way forward will eventually emerge. I would like to do something focussed around the software but presently available hardware platforms are so very partial at best and at worst; they are not really fit for this purpose. Of course audiolense is merely my personal favourite and as far as anyone else is concerned it should be entirely interchangeable with any of its many, more or less compatible alternatives. Interoperability is the watchword and without application specific De Facto product standards it’s all we’ve got.
and perhaps what your day job is.
For the past five years or so I would have to say it was mainly surviving the big C.
But as the surgeons recently kicked me out minus some parts and said don’t come back. I’m looking forward to getting on with the rest of my life. If wishing can come true I’ll be spending it doing something I have a real aptitude for and abiding passion with. In interludes between hospital visits and for some while before that I’ve spent much time and money finding out what matters and why; evaluating various pieces of the puzzle and measuring and testing them. Initially more as a learning exercise than with any real commercial intent. But lately; I’ve been thinking and my recent conclusions (as posted) are not at all what I may have predicted in the beginning.
Previously I founded a few companies; one of these is now owned by probably the second largest of the most well-known semiconductor manufacturers. I was later the Chief Scientist for another NASDAQ listed company which was bought by the biggest of all the storage manufacturers.
Last time I counted over twenty of my patents are still active worldwide. I’ve had articles published in IEEE Journals and several other papers published not only by the IEEE but also in many other conference proceedings from around the world. I did lots of standards work too at various stages with Intel and the PCI SIG, the Global Grid Forum and other well-known industry bodies.
I have provided consultancy to very large scale projects run by the Technical Directorate of the European Union Commission. My last job before having to stop for health reasons in 2011 was in fulfilment of a brief to further industrial relevance in science and academia.
_________________
My suggestion of a dedicated micro ATX PC board running Linux for a 'digital speaker' was precisely to address the 'lock in' issues you mention.
This is my conclusion too but I’m interested in scalability as well. At least galvanic isolation is there in the PHY layer on the network side but sadly not on the audio side. The audible consequences of this can be quite upsetting when you first hear just how bad the on board signal conversion sounds without this.
It also addresses connectivity by having ethernet. The de facto standard for 21st century sound is looking to be 'web' based.
Ethernet or IEEE 802 is not a De Facto standard it’s an Internationally Recognised Properly Accredited Formalised one with lots of specialised sub-classes. Ethernet can only be used to deliver audio to one single endpoint on a network because there's no reliable consistency in the timing of packet delivery .i.e. although audio traffic is generally ergodic over time it is for a variety of reasons entirely stochastic at the link level.
After much professional involvement with 802.3 the one I was personally most interested in is 802.1 AVB and what lies beyond. This arose to serve Audio Video Bridging requirements but to be honest it’s not quite up to what we need for scalability in this application because the best they can do today; even with expensive specialised hardware switches only guarantees 4us ‘packet jitter’ between endpoints and it says nothing about cross domain clock drift. So on top of all this some extra but unspecified synchronisation is still needed to preserve proper audio timing.
Anyway nowadays these guys are more interested in factory automation, so don’t hold your breath. But as they always say, in the long run. Never bet against Ethernet! You don’t really have to try too hard to guess who it is that says that though!
The 'digital speaker' will simply look at a URL on the LAN and play it continuously.
Which is all fine until you need an extra output channel or two; then you’re really stuffed. This is where ‘going pro’ becomes necessary and as I wrote; that opens up a whole new can of worms.
__________________________
When I started playing with DSP in the early 90's, a 200 pt FIR (or 180 pt IIR) with TI 320c20 was SOTA. In the early 90's with the Essex lot, this was still the case but the DACs had improved cos we could modify the S/PDIF stream.
I began work in the Mainframe Systems Development Division of what’s today called Fujitsu, where we had just started making the first 20 layer PCB’s in Europe. To do this we used a massive hand wired discrete transistor processor with a huge 24K Word ferrite bead core memory that got its software off fixed head multi track drum stores the size of a car these were loaded from vast banks of mag tape decks. The programs, written in machine code using a teletype feeding onto paper tape were manually edited by cutting and splicing were transferred on to punched cards and fed in through a reader the size of a small tank. The line printer it was as big as a train and noisier than gunfire.
The whole system was bigger than two houses and less powerful than the cheapest modern mobile phone. We later had a similar system for engineering use that grew so much its weight eventually caused the building it was in to lean over and collapse. This was replaced by a much more powerful minicomputer only the size of a van. So yes I know just how fast this technology moves!
When I returned to civilisation (??) this Millenium after more than a decade in the bush, I find that even the cheapest PC had more than sufficient power to do stuff I only dreamt about in da old days.
Exactly, it’s not the computer that’s important any longer it’s just the software and the audio peripherals that matter now.
I welcome you back from the wilds and offer you my personal apology for not having made more progress in your absence but ,I was only working on the processor and interconnect side of things for much of that time. I would be pleased if you felt that those countless meetings with all the senior folks from the Technical Director’s Office at Intel have eventually paid off. But I do hope you won’t hold the sad and sorry state of audio I/O peripherals for active speaker systems against me as of late, I’ve not been really been well enough to get it done; yet.
It's no longer necessary to use DSP processors to do fancy real time EQ for speakers. You could programme a standalone programme or VST to do it all using C to run on a cheapo PC.
This is my point precisely and as a starter for 10 it’s great, but; what if the winning score has to be 180? What do we need to do then?
I started poking around with the TMS32010 in the early 80’s and still can’t really see that I have any use for it’s modern brethren even now. Always best to keep a watching brief though.
It's this type of computing power that gets more powerful and cheaper fastest.
As I have been saying for over a decade and keep on saying to anyone who will listen!
It’s apposite that you use the word power; because if they’d been left alone; Intel’s thermal footprint would probably have stayed up around 150 Watts or more for a lot longer. We, amongst others, spent years persuading them that our big data centre and large supercomputer customers were not yet ready to get into the private power station business and explaining how they were all becoming quite concerned about the risks associated with such high thermal densities etc. etc.
The first CORE 2 DUO processors emerged out of a series of what were, long before any public announcements, super top secret processor code names like Bear Tooth; for lower power with high performance and Larabee; for the big boys. Being involved in arguing specifications of these processors and lobbying for some of the feature sets in them was great fun. But; it’s all history now these having mutated into the remarkably powerful modern ~35 Watt devices that serve us all so well. As for Larabee it became the Xeon multi core coprocessor the so the supercomputing guy’s eventually got there too. Today you could get what you want for IIR from a ~5 Watt Atom or maybe even less?
So your recommendation to try DRC or Accourate is spot on. I'm not sure anything more powerful is necessary .. certainly for stereo.
Best results come from the direct connection of an amplifier to each drive unit so channel count in the signal processing S/W and I/O conversion H/W is very variable. Then the all that other stuff I’ve already written about at length really does start to matter.
I do have a caveat about room EQ but only perhaps cos I've no hands on experience with DSP for this and I have been studying room response for a long time.
DRC is the best place to start with this, but; real perseverance in mastering that beast is a big must. Threads on here can help you with that.
In a well-designed loudspeaker system active or not anechoic measurements may be almost ideal perhaps to within a dB or so of the desired response. But once put in a listening room. If the acoustic discontinuities can be limited to only 15db or so then this would be a typical to good result.
Near anechoic behaviours can today be achieved in ordinary listening rooms through the use of corrective software. In my experience the differences made by this far and away outstrip the benefits of costly cable swapping or even those smaller but still expensive incremental system upgrades we all wish for. Everything else being in order, non-corrective active systems are all well and good. But even with the very best of them; room influences will still bear heavily on listener satisfaction.
______________________
For me, a surround sound system is a speaker ... a device you put an 'electrical' signal in at one end and get a noise out of the other.
Agreed; the future will fully address wavefield synthesis but will we still be around to see it done properly?
DSP for a single speaker ignores the fact that a speaker ISN'T a 1D in/out chain. It's 1D in & 3D out.
Accepted; with the reservation that stereo (multi sub) done as I said earlier is enough for me and OK for today. There are just too many complex variables for the current/affordable state of the art to manage the resultant wavefield well enough at higher frequencies to let me just relax and listen without feeling consciously critical of these easily notable shortcomings.
Using many speakers in the room gives you a much better chance of doing something sensible with the room.
Acknowledged; and all the more so at lower frequencies, but; higher up the spectrum though it’s all a bit of a crap shoot.
The usage case I'm most interested in is stereo domestic listening through active loudspeakers with individually amplified drive units. But with more processing power and appropriate trade-offs in the FIR filter coefficient configurations latency can be made negligible so I don't see that as a real issue for other applications.
_____________________
Would you like to expand on your dislike for IIRs?
Perhaps I should start out by saying it’s not just IIR it’s also some forms of FIR too.
The big breakthrough for me came after spending huge amounts of time level matching proper side by side comparisons with different forms of FIR and IIR filter processing. While I accept there may be some controversy with this. My opinions are qualified by extensive experience with IIR & FIR and with various ways of using both types together. My certain impression is that an IIR will always behave as an IIR and in some aspect always sound like an IIR. FIR systems are as wide ranging in their technical specifications as they are in their audio performance. To me some FIR filtering sounding not so dissimilar in the annoying degree of their ‘unwanted artefact dependant qualities’ to IIR processing; but each acoustic signature sounds very different.
In other words a very clever FIR system can be made to get out of its own way but an IIR will always be trying to fall over its own feet. It’s a little like all that debate the pixel peepers have over on the digital darkroom photography sites about; soft focus and excessive colour saturation (IIR) vs washed out overdone edge sharpening (FIR) i.e. neither are correct but both are obvious. In the end I wanted to find something better than either of these.
Put another way various artefact elimination strategies using forms of apodisation and other compensatory techniques can be used to make FIR time domain behaviours near perfect. But in an open loop system nothing can be done for an IIR and for a set of IIR filters with overlapping responses; time domain behaviours become even less desirable. Even in a closed loop scenario IIR responses tend to dominate and whenever they do some life or liveliness disappears from the music.
Our (my) hearing or perhaps more precisely, feeling of ease in the listening experience, is very sensitive to these time domain disturbances. They are not always instantly apparent and it may sometimes take weeks or even months before being able to characterise, qualify and then quantify what has caused this. This is where my initial dissatisfaction with the more expensive active systems first originated. Although each new speaker was definitely better than its predecessor listening could be likened to looking at a famous piece of art in the wrong wavelengths of lighting. I could still ‘see’ it but was certain that it should not ‘look’ that way.
But invariably at some juncture past the point of no returns each time following the arrival of a new speaker it occurred to me that something was missing and so after securing the cash it was on to the next one. Eventually after the last attempt I just gave up in disgust, got rid, and set about finding out in particular scientific detail why this could be. It took about seven years and involved contracts in several countries and collaborations on two continents along with ever more expense. Eventually I was and am still satisfied with my limited but improved understanding and although continuously upgrading it, also with my system.
The best of the FIR implementations are so much better than IIR or poorly implemented FIR it simply has to be heard to be believed. In essence it’s probably not so much the Finite Impulse Response aspect itself that matters. Given adequate tap length and properly configured it's more the richness of specification provided by the higher degrees of freedom from the far greater number of FIR filter coefficients themselves over and above an IIR that I think makes the biggest and most striking differences possible. If I were to try to sum up what’s missing with IIR and poorly implemented FIR in one word, as an amateur musician who’s spent far too long standing next to various kinds of drummer that word would be: MICRODYNAMICS!
My simplest explanation of why the above is true lies not in considering IIR & FIR frequency or phase response which at least in one aspect may be identical but in thinking about their very different native time domain behaviours. IIR systems can only ever degrade timing and poorly implemented FIR systems will corrupt it. Temporal aberrations exist for both IIR & FIR but the nature of the time domain artefacts from each type is almost opposite.
In my view there are two main formats of mixed FIR & IIR systems; contiguous and separate. In the first format the two types of DSP are combined in a closed loop correction system which really just gives the FIR filter a harder job to do and is used by some vendors to compensate for less than capable FIR filter coefficient specification. This format has sub optimal IIR and FIR behaviours and gives a result to my ear of having perhaps the worst of both types of DSP. In the second format open loop operation of the IIR sections may overlay on either closed or open FIR filtering systems. Either way; it seems to me that the acoustic signature of the IIR tends to dominate and this, I feel, always detracts from the integrity of the sound.
kgrlee,
I've put my responses between your questions:
You say you've bought several DSP speaker products before and been disappointed. Can you tell us which ones & perhaps why?
I’ve had a couple of Meridian actives, later one from Tact and then another from Lyngdorf. In our brief meeting I didn’t really get on with Peter but I enjoyed dealing with his folks and each speaker was definitely better than the last. All were impressively well-made and excellent as objet d’art, though ultimately each was unsatisfying in some way or another. Initially I thought it was just the drive units as like CD surely this digital stuff was all ‘Perfect Sound Forever’; naive I know but I was younger and a lot more trusting then. Eventually over time through activating other manufacturer’s speakers I slowly came to the conclusion that it was the DSP filtering letting them all down. I had a brief stint with a DEQX but there I learned that the FIR filters were too short, the intricacy of the processing perhaps too limited to be properly effective and their signal conversion technology not to my taste. I said most of this to Kim last time we spoke so no big surprises here.
Are you involved with a commercial product? It would be good to know your association with any of the groups actively involved in DSP for speakers
After later spending years going through just about everything audio DSP both H/W & S/W based; because the range of filter specification parameters is sufficiently large to allow me to get a good result from almost any drive unit I include I now use audiolense in my own system. I feel I have a good relationship with the author but there's no commercial activity at present. I am hoping a way forward will eventually emerge. I would like to do something focussed around the software but presently available hardware platforms are so very partial at best and at worst; they are not really fit for this purpose. Of course audiolense is merely my personal favourite and as far as anyone else is concerned it should be entirely interchangeable with any of its many, more or less compatible alternatives. Interoperability is the watchword and without application specific De Facto product standards it’s all we’ve got.
and perhaps what your day job is.
For the past five years or so I would have to say it was mainly surviving the big C.
But as the surgeons recently kicked me out minus some parts and said don’t come back. I’m looking forward to getting on with the rest of my life. If wishing can come true I’ll be spending it doing something I have a real aptitude for and abiding passion with. In interludes between hospital visits and for some while before that I’ve spent much time and money finding out what matters and why; evaluating various pieces of the puzzle and measuring and testing them. Initially more as a learning exercise than with any real commercial intent. But lately; I’ve been thinking and my recent conclusions (as posted) are not at all what I may have predicted in the beginning.
Previously I founded a few companies; one of these is now owned by probably the second largest of the most well-known semiconductor manufacturers. I was later the Chief Scientist for another NASDAQ listed company which was bought by the biggest of all the storage manufacturers.
Last time I counted over twenty of my patents are still active worldwide. I’ve had articles published in IEEE Journals and several other papers published not only by the IEEE but also in many other conference proceedings from around the world. I did lots of standards work too at various stages with Intel and the PCI SIG, the Global Grid Forum and other well-known industry bodies.
I have provided consultancy to very large scale projects run by the Technical Directorate of the European Union Commission. My last job before having to stop for health reasons in 2011 was in fulfilment of a brief to further industrial relevance in science and academia.
_________________
My suggestion of a dedicated micro ATX PC board running Linux for a 'digital speaker' was precisely to address the 'lock in' issues you mention.
This is my conclusion too but I’m interested in scalability as well. At least galvanic isolation is there in the PHY layer on the network side but sadly not on the audio side. The audible consequences of this can be quite upsetting when you first hear just how bad the on board signal conversion sounds without this.
It also addresses connectivity by having ethernet. The de facto standard for 21st century sound is looking to be 'web' based.
Ethernet or IEEE 802 is not a De Facto standard it’s an Internationally Recognised Properly Accredited Formalised one with lots of specialised sub-classes. Ethernet can only be used to deliver audio to one single endpoint on a network because there's no reliable consistency in the timing of packet delivery .i.e. although audio traffic is generally ergodic over time it is for a variety of reasons entirely stochastic at the link level.
After much professional involvement with 802.3 the one I was personally most interested in is 802.1 AVB and what lies beyond. This arose to serve Audio Video Bridging requirements but to be honest it’s not quite up to what we need for scalability in this application because the best they can do today; even with expensive specialised hardware switches only guarantees 4us ‘packet jitter’ between endpoints and it says nothing about cross domain clock drift. So on top of all this some extra but unspecified synchronisation is still needed to preserve proper audio timing.
Anyway nowadays these guys are more interested in factory automation, so don’t hold your breath. But as they always say, in the long run. Never bet against Ethernet! You don’t really have to try too hard to guess who it is that says that though!
The 'digital speaker' will simply look at a URL on the LAN and play it continuously.
Which is all fine until you need an extra output channel or two; then you’re really stuffed. This is where ‘going pro’ becomes necessary and as I wrote; that opens up a whole new can of worms.
__________________________
When I started playing with DSP in the early 90's, a 200 pt FIR (or 180 pt IIR) with TI 320c20 was SOTA. In the early 90's with the Essex lot, this was still the case but the DACs had improved cos we could modify the S/PDIF stream.
I began work in the Mainframe Systems Development Division of what’s today called Fujitsu, where we had just started making the first 20 layer PCB’s in Europe. To do this we used a massive hand wired discrete transistor processor with a huge 24K Word ferrite bead core memory that got its software off fixed head multi track drum stores the size of a car these were loaded from vast banks of mag tape decks. The programs, written in machine code using a teletype feeding onto paper tape were manually edited by cutting and splicing were transferred on to punched cards and fed in through a reader the size of a small tank. The line printer it was as big as a train and noisier than gunfire.
The whole system was bigger than two houses and less powerful than the cheapest modern mobile phone. We later had a similar system for engineering use that grew so much its weight eventually caused the building it was in to lean over and collapse. This was replaced by a much more powerful minicomputer only the size of a van. So yes I know just how fast this technology moves!
When I returned to civilisation (??) this Millenium after more than a decade in the bush, I find that even the cheapest PC had more than sufficient power to do stuff I only dreamt about in da old days.
Exactly, it’s not the computer that’s important any longer it’s just the software and the audio peripherals that matter now.
I welcome you back from the wilds and offer you my personal apology for not having made more progress in your absence but ,I was only working on the processor and interconnect side of things for much of that time. I would be pleased if you felt that those countless meetings with all the senior folks from the Technical Director’s Office at Intel have eventually paid off. But I do hope you won’t hold the sad and sorry state of audio I/O peripherals for active speaker systems against me as of late, I’ve not been really been well enough to get it done; yet.
It's no longer necessary to use DSP processors to do fancy real time EQ for speakers. You could programme a standalone programme or VST to do it all using C to run on a cheapo PC.
This is my point precisely and as a starter for 10 it’s great, but; what if the winning score has to be 180? What do we need to do then?
I started poking around with the TMS32010 in the early 80’s and still can’t really see that I have any use for it’s modern brethren even now. Always best to keep a watching brief though.
It's this type of computing power that gets more powerful and cheaper fastest.
As I have been saying for over a decade and keep on saying to anyone who will listen!
It’s apposite that you use the word power; because if they’d been left alone; Intel’s thermal footprint would probably have stayed up around 150 Watts or more for a lot longer. We, amongst others, spent years persuading them that our big data centre and large supercomputer customers were not yet ready to get into the private power station business and explaining how they were all becoming quite concerned about the risks associated with such high thermal densities etc. etc.
The first CORE 2 DUO processors emerged out of a series of what were, long before any public announcements, super top secret processor code names like Bear Tooth; for lower power with high performance and Larabee; for the big boys. Being involved in arguing specifications of these processors and lobbying for some of the feature sets in them was great fun. But; it’s all history now these having mutated into the remarkably powerful modern ~35 Watt devices that serve us all so well. As for Larabee it became the Xeon multi core coprocessor the so the supercomputing guy’s eventually got there too. Today you could get what you want for IIR from a ~5 Watt Atom or maybe even less?
So your recommendation to try DRC or Accourate is spot on. I'm not sure anything more powerful is necessary .. certainly for stereo.
Best results come from the direct connection of an amplifier to each drive unit so channel count in the signal processing S/W and I/O conversion H/W is very variable. Then the all that other stuff I’ve already written about at length really does start to matter.
I do have a caveat about room EQ but only perhaps cos I've no hands on experience with DSP for this and I have been studying room response for a long time.
DRC is the best place to start with this, but; real perseverance in mastering that beast is a big must. Threads on here can help you with that.
In a well-designed loudspeaker system active or not anechoic measurements may be almost ideal perhaps to within a dB or so of the desired response. But once put in a listening room. If the acoustic discontinuities can be limited to only 15db or so then this would be a typical to good result.
Near anechoic behaviours can today be achieved in ordinary listening rooms through the use of corrective software. In my experience the differences made by this far and away outstrip the benefits of costly cable swapping or even those smaller but still expensive incremental system upgrades we all wish for. Everything else being in order, non-corrective active systems are all well and good. But even with the very best of them; room influences will still bear heavily on listener satisfaction.
______________________
For me, a surround sound system is a speaker ... a device you put an 'electrical' signal in at one end and get a noise out of the other.
Agreed; the future will fully address wavefield synthesis but will we still be around to see it done properly?
DSP for a single speaker ignores the fact that a speaker ISN'T a 1D in/out chain. It's 1D in & 3D out.
Accepted; with the reservation that stereo (multi sub) done as I said earlier is enough for me and OK for today. There are just too many complex variables for the current/affordable state of the art to manage the resultant wavefield well enough at higher frequencies to let me just relax and listen without feeling consciously critical of these easily notable shortcomings.
Using many speakers in the room gives you a much better chance of doing something sensible with the room.
Acknowledged; and all the more so at lower frequencies, but; higher up the spectrum though it’s all a bit of a crap shoot.
The usage case I'm most interested in is stereo domestic listening through active loudspeakers with individually amplified drive units. But with more processing power and appropriate trade-offs in the FIR filter coefficient configurations latency can be made negligible so I don't see that as a real issue for other applications.
_____________________
Would you like to expand on your dislike for IIRs?
Perhaps I should start out by saying it’s not just IIR it’s also some forms of FIR too.
The big breakthrough for me came after spending huge amounts of time level matching proper side by side comparisons with different forms of FIR and IIR filter processing. While I accept there may be some controversy with this. My opinions are qualified by extensive experience with IIR & FIR and with various ways of using both types together. My certain impression is that an IIR will always behave as an IIR and in some aspect always sound like an IIR. FIR systems are as wide ranging in their technical specifications as they are in their audio performance. To me some FIR filtering sounding not so dissimilar in the annoying degree of their ‘unwanted artefact dependant qualities’ to IIR processing; but each acoustic signature sounds very different.
In other words a very clever FIR system can be made to get out of its own way but an IIR will always be trying to fall over its own feet. It’s a little like all that debate the pixel peepers have over on the digital darkroom photography sites about; soft focus and excessive colour saturation (IIR) vs washed out overdone edge sharpening (FIR) i.e. neither are correct but both are obvious. In the end I wanted to find something better than either of these.
Put another way various artefact elimination strategies using forms of apodisation and other compensatory techniques can be used to make FIR time domain behaviours near perfect. But in an open loop system nothing can be done for an IIR and for a set of IIR filters with overlapping responses; time domain behaviours become even less desirable. Even in a closed loop scenario IIR responses tend to dominate and whenever they do some life or liveliness disappears from the music.
Our (my) hearing or perhaps more precisely, feeling of ease in the listening experience, is very sensitive to these time domain disturbances. They are not always instantly apparent and it may sometimes take weeks or even months before being able to characterise, qualify and then quantify what has caused this. This is where my initial dissatisfaction with the more expensive active systems first originated. Although each new speaker was definitely better than its predecessor listening could be likened to looking at a famous piece of art in the wrong wavelengths of lighting. I could still ‘see’ it but was certain that it should not ‘look’ that way.
But invariably at some juncture past the point of no returns each time following the arrival of a new speaker it occurred to me that something was missing and so after securing the cash it was on to the next one. Eventually after the last attempt I just gave up in disgust, got rid, and set about finding out in particular scientific detail why this could be. It took about seven years and involved contracts in several countries and collaborations on two continents along with ever more expense. Eventually I was and am still satisfied with my limited but improved understanding and although continuously upgrading it, also with my system.
The best of the FIR implementations are so much better than IIR or poorly implemented FIR it simply has to be heard to be believed. In essence it’s probably not so much the Finite Impulse Response aspect itself that matters. Given adequate tap length and properly configured it's more the richness of specification provided by the higher degrees of freedom from the far greater number of FIR filter coefficients themselves over and above an IIR that I think makes the biggest and most striking differences possible. If I were to try to sum up what’s missing with IIR and poorly implemented FIR in one word, as an amateur musician who’s spent far too long standing next to various kinds of drummer that word would be: MICRODYNAMICS!
My simplest explanation of why the above is true lies not in considering IIR & FIR frequency or phase response which at least in one aspect may be identical but in thinking about their very different native time domain behaviours. IIR systems can only ever degrade timing and poorly implemented FIR systems will corrupt it. Temporal aberrations exist for both IIR & FIR but the nature of the time domain artefacts from each type is almost opposite.
In my view there are two main formats of mixed FIR & IIR systems; contiguous and separate. In the first format the two types of DSP are combined in a closed loop correction system which really just gives the FIR filter a harder job to do and is used by some vendors to compensate for less than capable FIR filter coefficient specification. This format has sub optimal IIR and FIR behaviours and gives a result to my ear of having perhaps the worst of both types of DSP. In the second format open loop operation of the IIR sections may overlay on either closed or open FIR filtering systems. Either way; it seems to me that the acoustic signature of the IIR tends to dominate and this, I feel, always detracts from the integrity of the sound.
It is precisely for reasons of off-axis behaviour that I would probably choose a xover nearer 5kHz than 2.5kHz.
Of course this depends on the precise drive units chosen. My study of this subject goes back to before
Is Linear Phase Worthwhile?
There is no problem in getting the main lobe in the right place .. even for a cuboid box speaker. Just choose the right passive xover. But you need to know a bit more about speakers & xovers than Linkwitz to do this though.
PS There's at least one treble unit which I might use down to 2.5kHz with an appropriate bass unit but its not my favourite unit.
Last edited:
The ATX solution means you can try it out easily with the computer remote etc beforehand too.
If the computer is also your main multi-media machine, then all you need is a soundcard with the correct number of clean audio o/ps.
We can work towards that but it means we'll probably have to run EVIL Windoze. Let call it the Single Computer Solution (SCS)
This is only if the speaker is a purely digital speaker eg fed via a LAN. If it is, some ATX boards have up to 8 channels of good DACs built in.
Only the crossover between treble & mid/bass needs to be passive. DON'T LEAVE THIS OUT. YOU HAVE BEEN WARNED.
We'll let logician answer this question about synch.
My feeling is if you are not using EVIL Windoze, you just need to start each DSP playing the track at the same time.
For the SCS feeding your multi-amped speakers, you need ASIO if using EVIL Windoze or you'll lose synch (if you ever get synch in the first place).
The 'correct' way to do this is when you sell your speakers, you sell them the soundcard which installs a driver that replaces the EVIL Windoze Mixer. The driver has all the DSP. This is definitely possible as the Creative drivers manage it. But Redmond go out of their way to make it difficult & secret.
Linux is much simpler.
But if you adopt my strategy of passive xover and single big amp with Digital EQ, your PC/laptop probably has good enough audio on its stereo headphones. (I withdraw that. I'm seen some really foul h/p outlets
Just check that the H/P outlet has textbook 16b performance and if not, you need a good soundcard.)Then Accourate or DRC. Run the filters in JRiver Media software, or Windows MP with Sourceforge Convolver for Windows as Barley recommends.
What are the speaker & amps you use with audiolense?
I'm sorry to hear that. Hope everything gets better or at least as you plan.
The purely digital speaker (which looks like more that what Steven is after) would be part of the package built and hence controlled by us. It would be integrated with probably Hypex PSs and amps .. so grounding etc which is the main determinant of such matters is our fault.
If the ATX board was tested & chosen to have textbook 16b 48kHz performance, there's no reason why this shouldn't provide SoTA performance for the system.
4us is 0.05". Good enough for units in a speaker and ample for speaker to speaker.
Are you saying there's no easy way to synch over an existing ethernet protocol, say between musical excerpts to deal with clock drift?
OK. You got me beat
I can only boast punched cards on ICL1900 and unix on PDP11 which was the fastest computer on site at the time.Well I believe a score of 10 today (Accourate & DRC) is good enough .. even for da deaf Golden Pinnae. I'm proving this in a similar field and would love to have another play at it with speakers again ... especially the room stuff for which I go back a loo.oong time.
______________________
This is "wavefield synthesis" and we discuss the limits where this applies and where a psychoacoustic approach needs to be taken.
Is My Decoder Ambisonic?
It's only the 3rd of a series of papers by BLaH.
Would you like to expand on your dislike for IIRs?
.. loadsa stuff
Thanks for this. My own prejudices with some supporting evidence is in the AES paper I quoted earlier on IIRs.
It is precisely for reasons of off-axis behaviour that I would probably choose a xover nearer 5kHz than 2.5kHz.
Is Linear Phase Worthwhile?
There is no problem in getting the main lobe in the right place .. even for a cuboid box speaker. Just choose the right passive xover. But you need to know a bit more about speakers & xovers than Linkwitz to do this though.
PS There's at least one treble unit which I might use down to 2.5kHz with an appropriate bass unit but its not my favourite unit.
Can you please explain why you think a xover nearer 5kHz has better off axis behaviour than one at 2.5kHz? This would only be true for either an extremely small mid driver, or an extremely narrow wave guide. The reason is that you want as much as possible the same radiation characteristics for the two drivers around xover. Otherwise, off axis behaviour will always be compromised.
This has nothing to do with linear phase, so I don't understand your link to a 1981 paper on that topic.
It never works to claim expertise without arguments, but rather by denigrating a well known expert who´s work has been validated by thousands.
There are plenty fine tweeters out there that work great at even <2.5kHz. Get current.
kgrlee,
could you put up a block diagram of your proposed layout?
I see you proposing a mix of passive xo and active correction, at least that is what I think you are proposing. Why stay with the passive networks, why not at least move to active xo, even if it stays analog.
I was thinking why not move the passive network if you are still going down that road to between the preamp and power amp and be able to use much smaller components, no more huge inductors and large packs of film caps that have to handle high voltages. You could use small passive components and even put an active buffer between the network and the power amp for impedance matching. I guess you would still want to leave a series cap in between the power amp and a dome tweeter but otherwise get away from the passive network as it has been done forever.
I'm still watching and trying to see the argument that Logician has just posted about his dislike of the IIR filters and desire for the FIR filtering. There seems to be two camps taking different sides of this argument. The small amount of filtering resources needed to do the iir vs fir's much larger computing power needed.
I really think the consumer would not want to have to install specific software or a new soundcard to make this happen, we aren't going to change the large manufacturers ideas of how to store music on a device or how to serve it up. To many already have their music on their phones or on a hard drive and they just want to push play and have music appear in their ears. Everyone is going to have their own music player, whatever came with the computer or the software installed, whether it is Apple or Microsoft or even Linux based players. I assume that every desktop computer has a soundcard installed today, laptops will have built in audio with not much more than a headphone output, and the same for a cell phone, headphone or today's Bluetooth output for sound. Wisa will happen over time if the users can hear the difference through better speakers, otherwise Bluetooth wireless is a dead end for better sound quality.
could you put up a block diagram of your proposed layout?
I see you proposing a mix of passive xo and active correction, at least that is what I think you are proposing. Why stay with the passive networks, why not at least move to active xo, even if it stays analog.
I was thinking why not move the passive network if you are still going down that road to between the preamp and power amp and be able to use much smaller components, no more huge inductors and large packs of film caps that have to handle high voltages. You could use small passive components and even put an active buffer between the network and the power amp for impedance matching. I guess you would still want to leave a series cap in between the power amp and a dome tweeter but otherwise get away from the passive network as it has been done forever.
I'm still watching and trying to see the argument that Logician has just posted about his dislike of the IIR filters and desire for the FIR filtering. There seems to be two camps taking different sides of this argument. The small amount of filtering resources needed to do the iir vs fir's much larger computing power needed.
I really think the consumer would not want to have to install specific software or a new soundcard to make this happen, we aren't going to change the large manufacturers ideas of how to store music on a device or how to serve it up. To many already have their music on their phones or on a hard drive and they just want to push play and have music appear in their ears. Everyone is going to have their own music player, whatever came with the computer or the software installed, whether it is Apple or Microsoft or even Linux based players. I assume that every desktop computer has a soundcard installed today, laptops will have built in audio with not much more than a headphone output, and the same for a cell phone, headphone or today's Bluetooth output for sound. Wisa will happen over time if the users can hear the difference through better speakers, otherwise Bluetooth wireless is a dead end for better sound quality.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.