A different kind of PC
As an example of the kind of hardware and internals that are also called PCs these days, check out the Hoojum Cubit P4. It even has the hunk of thick aluminium to block all them Evil Microphonics which all Real Audiophiles are so concerned about. Of course, for any digital xo work, you'll probably want to add your own sound card hardware... but it's a great starting point.
In fact, if the digital filtering is done in the user state outside the kernel, on Linux, like with BruteFIR, then you can even give a menu of sound cards for the customer to choose from.
Another product range to drool over is Hush Technologies' boxes. More than enough compute power to do a lot of digital filtering, and totally fanless and silent. And looks great on the audio rack too. One of the models, fully equipped with CPU, RAM, HDD, and CD-ROM drive, is available for EUR 666.00. And their boxes are also available in black, and you can show off all them heatsink fins and hint that you're using "Class A digital filtering." 😀
Tarun
As an example of the kind of hardware and internals that are also called PCs these days, check out the Hoojum Cubit P4. It even has the hunk of thick aluminium to block all them Evil Microphonics which all Real Audiophiles are so concerned about. Of course, for any digital xo work, you'll probably want to add your own sound card hardware... but it's a great starting point.
In fact, if the digital filtering is done in the user state outside the kernel, on Linux, like with BruteFIR, then you can even give a menu of sound cards for the customer to choose from.
Another product range to drool over is Hush Technologies' boxes. More than enough compute power to do a lot of digital filtering, and totally fanless and silent. And looks great on the audio rack too. One of the models, fully equipped with CPU, RAM, HDD, and CD-ROM drive, is available for EUR 666.00. And their boxes are also available in black, and you can show off all them heatsink fins and hint that you're using "Class A digital filtering." 😀
Tarun
In the Mini ITX world the Shuttle SS-51 has been getting very affordable. Outpost (Frie's Electronics), MEI Microcenter and Tiger Direct all carry it. Its worth checking out on the Mini-ITX Website. It uses fan less heat pipe technology to remove heat from the CPU.
Nice little aluminum box to.
-Dave
Nice little aluminum box to.
-Dave
How much power is enough power?
I've been looking into a similar project for a few months now. Wanting a Windows-based solution and not knowing how to program, I found a graphical programming language called Max/MSP, which I believe could be used to create a multi-channel convolution engine (to implement DRC, or Digital Room Correction, a freeware program available on freshmeat and discussed on the htpc section of www.avsforum.com and also at the hifi_dsp Yahoo! group (btw, there are people at both of these groups who are interested in a project similar to the one being discussed in this thread). Max/MSP would also allow for the implementation of user-defined high-order, linear phase xo and time delay per driver. I have no idea how much computing power this would take, though.
tcpip, you've written that the hush technologies box (a Via 1 ghz, if memory serves) is more than enough power.... More than enough power for what? I'm looking for enough power to process up to 16 paths (6 channels total - 2x4, 1x3 and 1x2) of high-order, linear phase filters.
Also, several people have mentioned BruteFIR and Linux. I'm a fan of both, but, there are some limitations. For instance, there are no Linux drivers for my Lynx Two-B soundcard. Also, for those who would use the PC as a front-end (music jukebox, DVD player, etc.) Windows programs and surround decoding algorithms are years ahead of what's available on Linux.
I've been looking into a similar project for a few months now. Wanting a Windows-based solution and not knowing how to program, I found a graphical programming language called Max/MSP, which I believe could be used to create a multi-channel convolution engine (to implement DRC, or Digital Room Correction, a freeware program available on freshmeat and discussed on the htpc section of www.avsforum.com and also at the hifi_dsp Yahoo! group (btw, there are people at both of these groups who are interested in a project similar to the one being discussed in this thread). Max/MSP would also allow for the implementation of user-defined high-order, linear phase xo and time delay per driver. I have no idea how much computing power this would take, though.
tcpip, you've written that the hush technologies box (a Via 1 ghz, if memory serves) is more than enough power.... More than enough power for what? I'm looking for enough power to process up to 16 paths (6 channels total - 2x4, 1x3 and 1x2) of high-order, linear phase filters.
Also, several people have mentioned BruteFIR and Linux. I'm a fan of both, but, there are some limitations. For instance, there are no Linux drivers for my Lynx Two-B soundcard. Also, for those who would use the PC as a front-end (music jukebox, DVD player, etc.) Windows programs and surround decoding algorithms are years ahead of what's available on Linux.
Re: How much power is enough power?
Years... i wouldnt be too sure of that...
But we shouldnt be looking at taking the HTPC path here...
This project should be confined to a nice active xover, room eq... thats it.
I say forget ADC/DACs as as said before, youll never please poeple this way... everyone has thier opinions on what DAC sounds better, etc.
So, stick with:
SPDIF In -> EQ -> XOvers -> 6/8 SPDIF Out.
This will keep things nice and simple, and also help keep the cost down for us poor guys 😉
Years... i wouldnt be too sure of that...
But we shouldnt be looking at taking the HTPC path here...
This project should be confined to a nice active xover, room eq... thats it.
I say forget ADC/DACs as as said before, youll never please poeple this way... everyone has thier opinions on what DAC sounds better, etc.
So, stick with:
SPDIF In -> EQ -> XOvers -> 6/8 SPDIF Out.
This will keep things nice and simple, and also help keep the cost down for us poor guys 😉
I guess you can look at my statement two ways... I didn't necessarily mean that it will take years for Linux to catch up to the current state of Windows, but, I did mean that Linux is now where Windows was years ago. More development is being aimed at Windows. If what you're looking for is simple a Linux box with spdif in and out, I think the solution already exists - BruteFIR on a Hush box with the appropriate soundcard (maybe RME makes something? There are Linux drivers for most, if not all, of their cards). Creating a Windows based solution would be much more of a challenge, but would allow use of the Lynx Two-B and AES3 ins and outs (Lynx makes a 16-channel AES3 card - the Lynx AES16). Or, use the DACs on the Lynx, which have repeatedly been positively compared to the best standalone DACs available.
Re: Re: How much power is enough power?
Tarun
I asked questions about computer power in a thread I started and the answers surprised me. I found those responses quite eye-opening.dc said:
Personally, I feel that we may not suffer in spite of lack of support for some of the high-end cards for this application. Already, others seem to feel that we need only digital in/out, in which case probably lower-end cards can do. Or else go with some of the supported high-end cards, either supported by ALSA or OSS. Anything special I'm missing about the Lynx Two?
DVD playback yes, because of the conflict of open source and CSS. But for just audio playback, I don't see any constraints at all. And for this box, we are not talking any front-end. This is supposed to be a "standard black box in the audio rack", which is programmed over USB (or Ethernet?) from another PC. That PC can run any OS it wants. So for this box, what's the constraint?
For whatever it's worth, I agree. 🙂MWP said:
Tarun
TAS3103
I have used this device to implement a 2 and 3 way digital crossover and had great results. TI's software is a little clumsy to use but the device has preformed as well as the A/D and D/A converters and analog ground scheme allow it to. Just make sure to run it at 96 KHz.
The advantage is that TI's software supplied will allow you to implement all the functions, with no programming even though it is clumsy and there is a bit of a learning curve.
Also, their software, on the CD I got with the evan board does not support burning EAPROMS. However, I nagged them and they sent me a build that does support EAROMS to run the thing is master mode.
If anyone writes any additional software for operating this DSP, contact me at dmf@renkus-heinz.com. We might be willing to buy it.
I have used this device to implement a 2 and 3 way digital crossover and had great results. TI's software is a little clumsy to use but the device has preformed as well as the A/D and D/A converters and analog ground scheme allow it to. Just make sure to run it at 96 KHz.
The advantage is that TI's software supplied will allow you to implement all the functions, with no programming even though it is clumsy and there is a bit of a learning curve.
Also, their software, on the CD I got with the evan board does not support burning EAPROMS. However, I nagged them and they sent me a build that does support EAROMS to run the thing is master mode.
If anyone writes any additional software for operating this DSP, contact me at dmf@renkus-heinz.com. We might be willing to buy it.
Re: TAS3103
I've been playing around with a "paper design" using the TAS3103 for a short while and it appears to be perfect for my application (a 3-way powered speaker with provision for room EQ and other features). I don't have an eval board (yet) but I'm wondering just how dismal the software/development support might be for this device (?)...
Compared to a pcb stuffed full of op-amps and many, many configuration options (e.g. phase alignment, notch filters, etc.) this device might make the ideal "black box". It would take care of the immediate xover, slope, phase, and EQ-fixits required for a given enclosure/driver design and (even better) would allow for later room-EQ and other tweaks in real installations. And all this without the need for a multitude of high-tolerance matched components made of "unobtainium" and the endless fiddling that real world "tuning" involves.
The real challenge, if it's to be useful in the DIY world, would be in creating a decent software package (and GUI) which would allow the user to easily take advantage of all the device has to offer and to then download and test the results. Not easy! But worth it, if it can be done.
Count me in if anyone wants to pursue a widget based on this approach - otherwise I'm off to bash on the thing on my own ('cause I think the potential is too great to ignore)....
Regards,
Bill
I've been playing around with a "paper design" using the TAS3103 for a short while and it appears to be perfect for my application (a 3-way powered speaker with provision for room EQ and other features). I don't have an eval board (yet) but I'm wondering just how dismal the software/development support might be for this device (?)...
Compared to a pcb stuffed full of op-amps and many, many configuration options (e.g. phase alignment, notch filters, etc.) this device might make the ideal "black box". It would take care of the immediate xover, slope, phase, and EQ-fixits required for a given enclosure/driver design and (even better) would allow for later room-EQ and other tweaks in real installations. And all this without the need for a multitude of high-tolerance matched components made of "unobtainium" and the endless fiddling that real world "tuning" involves.
The real challenge, if it's to be useful in the DIY world, would be in creating a decent software package (and GUI) which would allow the user to easily take advantage of all the device has to offer and to then download and test the results. Not easy! But worth it, if it can be done.
Count me in if anyone wants to pursue a widget based on this approach - otherwise I'm off to bash on the thing on my own ('cause I think the potential is too great to ignore)....
Regards,
Bill
Hi; Great Idea. Although I'm not a programer, I'd love to have something like this in my system. Analog based xovers just doesn't cut it, personally, I think any opamps in the signal path doesn't cut it. I was on Ti's site yesterday and I noticed this device and I was thinking the same thing, will it work. I'm not sure, but does it have what it takes to be in the audiophile class though.
Oh we can do all that in software.....
"Analog based xovers just doesn't cut it, personally, I think any opamps in the signal path doesn't cut it."
"SPDIF In -> EQ -> XOvers -> 6/8 SPDIF Out.
This will keep things nice and simple, and also help keep the cost down for us poor guys."
You don't have to use op amps. A really good sounding SPDIF interface is much more difficult to design than an analog stage. A simple Spice model is much easier to write than software. You can model the driver rolloff and crossover slopes pretty easily as a starting point for crossover adjustment. There are also programs for hobbiest to model speaker response curves and passive networks to get the desired crossover slopes. Copying the transfer fuction with an active nework after that is not as difficult as you would think.
Which do you think has the easier learning curve and is the path of least resistance? For a decent design the question of which will sound better is a given...... analog, even with some of the better sounding op amps if designing discrete transistor circuits is not your thing.
"Analog based xovers just doesn't cut it, personally, I think any opamps in the signal path doesn't cut it."
"SPDIF In -> EQ -> XOvers -> 6/8 SPDIF Out.
This will keep things nice and simple, and also help keep the cost down for us poor guys."
You don't have to use op amps. A really good sounding SPDIF interface is much more difficult to design than an analog stage. A simple Spice model is much easier to write than software. You can model the driver rolloff and crossover slopes pretty easily as a starting point for crossover adjustment. There are also programs for hobbiest to model speaker response curves and passive networks to get the desired crossover slopes. Copying the transfer fuction with an active nework after that is not as difficult as you would think.
Which do you think has the easier learning curve and is the path of least resistance? For a decent design the question of which will sound better is a given...... analog, even with some of the better sounding op amps if designing discrete transistor circuits is not your thing.
I like op-amps...
I'm actually a big fan of op-amps and planned to use lots of them in an active crossover...🙂 The devices are so good these days that I'm sure I personally can't hear anything they might contribute - but others may disagree. However, I'm thinking that something like the TAS3103 might be worth trying as an alternate (because A/D and D/As are also pretty good these days 😉 ).
If you take a look at the excellent work done by Linkwitz (specifically his active design for the Orion, et al) he has obviously done an amazing amount of testing, tweeking and optimising. He's also explained it all very well and if you take it a section at a time it's perfectly logical and clear.
With a device like the TAS3103 you can implement the same functional blocks (ala Linkwitz) with the added advantage that you can very quickly make changes (e.g. throw in a notch filter at the last minute if needed) based on your test results. What's more, you can also add delay, amplitude compensation curves, etc., etc. - all without touching a soldering iron. Instead, you merely update the parameter table to be loaded into the TAS3103 to reflect your changes and you're done (at least until you change your mind again 🙄 )...
So, the question for me at least is what it sounds like and whether or not it's worth the trouble. I'm thinking it may well be worth it although Dan has pointed out above that the tools are fairly "clumsy". The prospect of a single A/D --> TAS3103 --> 3 D/A solution (all fairly cheap!) looks attractive if it plays out as well in reality as it does on paper.
Just my 0.02..........................
Bill
"So, is the universe really analog - or digital-with-infinite-resolution ?"
I'm actually a big fan of op-amps and planned to use lots of them in an active crossover...🙂 The devices are so good these days that I'm sure I personally can't hear anything they might contribute - but others may disagree. However, I'm thinking that something like the TAS3103 might be worth trying as an alternate (because A/D and D/As are also pretty good these days 😉 ).
If you take a look at the excellent work done by Linkwitz (specifically his active design for the Orion, et al) he has obviously done an amazing amount of testing, tweeking and optimising. He's also explained it all very well and if you take it a section at a time it's perfectly logical and clear.
With a device like the TAS3103 you can implement the same functional blocks (ala Linkwitz) with the added advantage that you can very quickly make changes (e.g. throw in a notch filter at the last minute if needed) based on your test results. What's more, you can also add delay, amplitude compensation curves, etc., etc. - all without touching a soldering iron. Instead, you merely update the parameter table to be loaded into the TAS3103 to reflect your changes and you're done (at least until you change your mind again 🙄 )...
So, the question for me at least is what it sounds like and whether or not it's worth the trouble. I'm thinking it may well be worth it although Dan has pointed out above that the tools are fairly "clumsy". The prospect of a single A/D --> TAS3103 --> 3 D/A solution (all fairly cheap!) looks attractive if it plays out as well in reality as it does on paper.
Just my 0.02..........................
Bill
"So, is the universe really analog - or digital-with-infinite-resolution ?"
TAS3103
Our listening tests does put this device in the audiophile class. The total THD+N, input to output on the DSP I made with the TAS3101 is in the area of 0.012%. Running at 96KHz and 24 bits. Used good AKM A/D and D/A converters with 110db+ dynamic range. For opamps on the output filters, I used OP275s. The input and output on my DSP is fully balanced as it is for pro use.
If anyone is trying this device and wants some help, using my experience, write me at dmfraser@sbcglobal.net.
If you write a better GUI for it, contact me too.
Our listening tests does put this device in the audiophile class. The total THD+N, input to output on the DSP I made with the TAS3101 is in the area of 0.012%. Running at 96KHz and 24 bits. Used good AKM A/D and D/A converters with 110db+ dynamic range. For opamps on the output filters, I used OP275s. The input and output on my DSP is fully balanced as it is for pro use.
If anyone is trying this device and wants some help, using my experience, write me at dmfraser@sbcglobal.net.
If you write a better GUI for it, contact me too.
I personally don't like opamps, from what I've heard of them in the past. Granted, I have not heard any of the lastest super opamps. I just think the less active circuitry in the path the better. I once built a 3 way active xover(i don't recall off hand what I was using for opamps). I thought it sounded great until I decided to try your basic 6db xovers at the front end of each amp. A world of difference, it just opened up. So what was at fault, the design as a whole, the opamps used, I don't know. But I know of other people who went the same route and they won't go back. But thats just my 2 cents worth.
Now as far as doing it in the digital domain, I'm sure that comes with its own pros and cons. I think I'd rather do it the digital domain. My system now is strictly digital right up to the pwramps(biamp) so it would be kinda hard to implement an active analog based xover.
dmfraser,I'm glad to hear that the TAS3101 rates right up there. There seems to be alot of talk about behringer units, do you think this one could surpase it or equal it.
Now as far as doing it in the digital domain, I'm sure that comes with its own pros and cons. I think I'd rather do it the digital domain. My system now is strictly digital right up to the pwramps(biamp) so it would be kinda hard to implement an active analog based xover.
dmfraser,I'm glad to hear that the TAS3101 rates right up there. There seems to be alot of talk about behringer units, do you think this one could surpase it or equal it.
Behringer
Behringer stuff is decent but not super. They make compromises to get the price so low. More bar band equipment than real [ro.
Behringer stuff is decent but not super. They make compromises to get the price so low. More bar band equipment than real [ro.
AD1953/1954
The Analog Devices widgets look very interesting but according to their data sheet "max 48 kHz sample rate".........🙁
Also, they have alot of extraneous features and options which seem geared more towards car audio or lower end 2.1 applications. Still, they're on the right track...sort of......🙂
Regards,
Bill
The Analog Devices widgets look very interesting but according to their data sheet "max 48 kHz sample rate".........🙁
Also, they have alot of extraneous features and options which seem geared more towards car audio or lower end 2.1 applications. Still, they're on the right track...sort of......🙂
Regards,
Bill
48KHz sample rate
I found that with a 48KHz sample rate the THD+N is about 3 times higher. On the same test unit I get 0.048% THD+N at 48KHz and 0.016% at 96 KHz.
This translates into a somewhat grittier sound at 48KHz though the extra THD does make to sound seem a little brighter. Adding some dither helps with the distortion at a cost of more noise. I have to design for 100db of more S/N so I have to do 96 KHz. However, for consumer use, where the sound source is 16 bit consumer grade CDs sampling at 44.1 KHz, playing through low powered (under 200W) low efficiency speakers (<95 db 1W/1M) then 48KHz is likely OK. Its just not audiophile.
I'm using, from AKM, the AKM5385A A/D and the AKM4395 D/A. I also use the Cirrus CS4362 in some products. All these are balanced input/output devices as I cannot get over 90db S/N or dynamic range out of single ended devices when I have 1000W Class-D power amplifiers as part of the system.
By going fully balanced and running at 96 KHz, I can get up to 107db dynamic range from the system.
Dan
I found that with a 48KHz sample rate the THD+N is about 3 times higher. On the same test unit I get 0.048% THD+N at 48KHz and 0.016% at 96 KHz.
This translates into a somewhat grittier sound at 48KHz though the extra THD does make to sound seem a little brighter. Adding some dither helps with the distortion at a cost of more noise. I have to design for 100db of more S/N so I have to do 96 KHz. However, for consumer use, where the sound source is 16 bit consumer grade CDs sampling at 44.1 KHz, playing through low powered (under 200W) low efficiency speakers (<95 db 1W/1M) then 48KHz is likely OK. Its just not audiophile.
I'm using, from AKM, the AKM5385A A/D and the AKM4395 D/A. I also use the Cirrus CS4362 in some products. All these are balanced input/output devices as I cannot get over 90db S/N or dynamic range out of single ended devices when I have 1000W Class-D power amplifiers as part of the system.
By going fully balanced and running at 96 KHz, I can get up to 107db dynamic range from the system.
Dan
Well, since it's been a month now since the last post - we must all have had a chance to think about it, right?..........🙂 🙂
I'd still be interested in the TAS3103 or AD1953/4 approach but can't bring myself to hand over $700 or so for an eval board just to play with either device. The alternative would be to glean as much as possible from the published app notes and then lash together a preliminary design to hack at but, unfortunately, the quantity of material available (e.g. not even a published reference design ??!) seems pretty sparse.
Dan, if you have an eval board - did they provide much in the way of support materials with it, or do you pretty much have to wing it from scratch?
Bill
I'd still be interested in the TAS3103 or AD1953/4 approach but can't bring myself to hand over $700 or so for an eval board just to play with either device. The alternative would be to glean as much as possible from the published app notes and then lash together a preliminary design to hack at but, unfortunately, the quantity of material available (e.g. not even a published reference design ??!) seems pretty sparse.
Dan, if you have an eval board - did they provide much in the way of support materials with it, or do you pretty much have to wing it from scratch?
Bill
TI Eval Board
TI included a GUI with the TAS3004 and TAS3103 eval boards I bought. They also included a piece of software called the Automatic Loudspeaker Equalizer where you import MLSSA data and it will automatically come up with a set of EQs for you to make your speaker match the curve you give it. They are asking $500.00 for that by itself though I don't know if anyone will pay that.
As well, when I found a feature on the GUI was not operative, I complained to TI and they made the feature operative and sent me a new version of the software. However, it is likely what they are shipping to others is sill unable to write to an EAROM for storing setups.
The DSP eval board for the TAS3103 was fixed at a 48 KHz sample rate and I had to change the crystal to make it run at 96K. As well, the A/D and D/A on the eval boards were only consumer grade and I had to use the I2C inputs and outputs on the EVM to go through decent converters.
However, I was able to get a functioning DSP built without having to write any code for it. And I got over 100db S/N out of it.
I just got my prototype PCBs today for a 1 In/3 or 4 out DSP using the TAS3103 and I hope this will work as good as the 1 in/2 out did for me.
Dan
TI included a GUI with the TAS3004 and TAS3103 eval boards I bought. They also included a piece of software called the Automatic Loudspeaker Equalizer where you import MLSSA data and it will automatically come up with a set of EQs for you to make your speaker match the curve you give it. They are asking $500.00 for that by itself though I don't know if anyone will pay that.
As well, when I found a feature on the GUI was not operative, I complained to TI and they made the feature operative and sent me a new version of the software. However, it is likely what they are shipping to others is sill unable to write to an EAROM for storing setups.
The DSP eval board for the TAS3103 was fixed at a 48 KHz sample rate and I had to change the crystal to make it run at 96K. As well, the A/D and D/A on the eval boards were only consumer grade and I had to use the I2C inputs and outputs on the EVM to go through decent converters.
However, I was able to get a functioning DSP built without having to write any code for it. And I got over 100db S/N out of it.
I just got my prototype PCBs today for a 1 In/3 or 4 out DSP using the TAS3103 and I hope this will work as good as the 1 in/2 out did for me.
Dan
It sounds like the support stuff TI provides is at least somewhat useful but I suspect that if I try the "lone hacker" approach (i.e. without eval board, etc.) I probably won't get too far. And without a "real" (that is - "commercial") application and an equivalent "real" customer I can't realistically expect much help from TI...🙂 Can't fault them for that really - they need to filter out the "tirekickers" (like ME 🙂 ) from live customers after all. Unfortunately, since I don't have my old day-job to use as a cover story..🙄 .. I can't just get in touch with TI's product manager and ask for any favors.
Anybody want to rent out an eval board?............
Anybody want to rent out an eval board?............
- Status
- Not open for further replies.