This project has finally evolved to where I can now change the volumes, crossover, effective curvature and EQ from my phone while listening--in real time. It's very cool, and fun to play with. The project was done primarily as a learning exercise, and if I had known how much I would have learned, I wouldn't have let this stretch out over 3 years.
It's posted here in the lounge because there isn't any other logical place on this forum for this type of project. It's a mix of line-level DSP, class D amps, multi-way, exotic drivers (FPS), line array design, host-side software and WiFi audio.
I've started an article about this on audiodevelopers.com, but I still have a long ways to go to finish it off. The article is at this link.
It's posted here in the lounge because there isn't any other logical place on this forum for this type of project. It's a mix of line-level DSP, class D amps, multi-way, exotic drivers (FPS), line array design, host-side software and WiFi audio.
I've started an article about this on audiodevelopers.com, but I still have a long ways to go to finish it off. The article is at this link.
There is a lot of information about the software on audiodevelopers.com. The line array software is an extension of the stereo 3-way code, modified to work with 3 DSP chips and support curvature as another parameter.
The fact that there is no logical place to discuss this project on diyAudio bothers me, probably more than it should. Take a look at the "market share" report chart below and you will see why. Audio is taking a sharp turn toward "smart speakers", with Wi-Fi, embedded DSP, built-in amps, and cell phone control. If you don't believe this chart, I'll post some others that have a similar message.
Audio is changing, but there are still many opportunities for DIY in this new order--that's what audiodevelopers.com is about. But diyAudio seems strangely out of step with this evolution, because there is no logical place to address these new technologies, evidenced by how poorly this line array project fits in the current forum structure. I've already made this point before and don't want to belabor it. But there are so many talented people looking at diyAudio that I want to do something to nudge it in the right direction. So, yes, this original post was something of a troll, but it had a good intention
.
The fact that there is no logical place to discuss this project on diyAudio bothers me, probably more than it should. Take a look at the "market share" report chart below and you will see why. Audio is taking a sharp turn toward "smart speakers", with Wi-Fi, embedded DSP, built-in amps, and cell phone control. If you don't believe this chart, I'll post some others that have a similar message.
Audio is changing, but there are still many opportunities for DIY in this new order--that's what audiodevelopers.com is about. But diyAudio seems strangely out of step with this evolution, because there is no logical place to address these new technologies, evidenced by how poorly this line array project fits in the current forum structure. I've already made this point before and don't want to belabor it. But there are so many talented people looking at diyAudio that I want to do something to nudge it in the right direction. So, yes, this original post was something of a troll, but it had a good intention
.
@bbutterfield: Our cabin has a "great room" with a 20-ft ceiling, log walls (not as reflective as sheetrock) and carpet. So reflections probably don't play as big a role as other venues. The speakers are on casters so I'll experiment with positioning when I start to take measurements. I've experimented a little bit with curvature, but it's hard to compare because the frequency response changes when you go from a concave "focused" array to a convex CBT. I can't draw any conclusions about line arrays right now other than that they sound nice and have a lot of potential.
@mark100: thanks. The control from a mobile device is a great convenience. But I wish someone would volunteer to help make some nice user interfaces using native Android or IOS tools. Audiodevelopers was set up to be a collaborative site, with multiple contributors, but in the last 12 years or so it has just been me *sigh*.
@mark100: thanks. The control from a mobile device is a great convenience. But I wish someone would volunteer to help make some nice user interfaces using native Android or IOS tools. Audiodevelopers was set up to be a collaborative site, with multiple contributors, but in the last 12 years or so it has just been me *sigh*.
Thought provoking. Makes me think about what could be done with 24 TC9s and a Motu 24Ao audio interface, times 2; but then where would I get 48 channels of amplification?
As to where to discuss it, how about a system design forum where the focus of the discussion isn't any particular kind of speaker but how to make its various parts work together as a whole and function as a system?
As to where to discuss it, how about a system design forum where the focus of the discussion isn't any particular kind of speaker but how to make its various parts work together as a whole and function as a system?
How about 48 of these; SSM3525 Datasheet and Product Info | Analog Devices
Digital input. Very few external components necessary and less than $5 each in small quantity. Practically speaking it's <20W but if you are using 24 per side that's probably more than necessary.
Digital input. Very few external components necessary and less than $5 each in small quantity. Practically speaking it's <20W but if you are using 24 per side that's probably more than necessary.
The amps I used are a combination of SSM3302 (woofers) and SSM2518 (tweeters). Each board has 22 channels of amplification. The Analog Devices amplifier chips use a spread-spectrum clocking technique, which makes power supply filtering and routing much easier than chips that use a common clock.
Each board has a total of 22 amplifiers, yet the board is fairly small. If I ever did this design again, I would use the SSM3582.
I'm 66, but I don't find soldering .5mm-spacing chips all that difficult. I've wrecked a couple of ADAU1701 chips using an SMD rework station, so I just use a 40-year-old Weller soldering iron with a fine tip. I put a "spare" SSM3302 chip on each board in case I messed up one of the amps, but all of the amps worked fine at initial power-on. I understand why most people don't want to even try those IC's with .5mm lead spacing, but it's really not that hard. I tend to think of it as DIY jewelry.
Each board has a total of 22 amplifiers, yet the board is fairly small. If I ever did this design again, I would use the SSM3582.
I'm 66, but I don't find soldering .5mm-spacing chips all that difficult. I've wrecked a couple of ADAU1701 chips using an SMD rework station, so I just use a 40-year-old Weller soldering iron with a fine tip. I put a "spare" SSM3302 chip on each board in case I messed up one of the amps, but all of the amps worked fine at initial power-on. I understand why most people don't want to even try those IC's with .5mm lead spacing, but it's really not that hard. I tend to think of it as DIY jewelry
.
Last edited:
I'm older and my eyes are worse than yours. I suppose I could see that small if I could focus in close enough. If one were going to do a lot of this it would probably pay to DIY a small reflow oven.
The SSM3582 and SSM3302 are really just 15W amps with a 12V supply and 4 ohm load. That's plenty of power for high volumes when you have 20 channels.
It's easiest to lay out the board and get the volumes matched if all of the amps are the same, so I would use the SSM3582 for every driver. There aren't any issues with running out of headroom with those 2W amps on the tweeters--it's just easier overall if all of the amps are identical, and it makes the design more flexible.
I had to use 3 ADAU1701 chips per board, each driven by the same I2S signal. Each ADAU1701 has 8 channels, so I had 24 channels of DSP to work with.
The software needs to keep track of which DSP processing cell in which chip goes with which channel, so there are lots of structures that look like this:
static single_biquad Sub_polarity[number_of_subs] = {{Gen_Filter5_4, ADAU1701_addr + 2}, {Gen_Filter6_4, ADAU1701_addr + 2}};
This goofy-looking software maps an array of subwoofer channels to specific biquad filters and DSP chip addresses.
On reflection, the software was MUCH more tedious than soldering the .5mm chips. I can solder chips while drinking beer. The software must be done while totally sober
It's difficult to compare curvature without also changing the frequency response, but right now I don't have code that changes both at the same time. For example, when the curvature is concave, there is a "sweet spot" that is much brighter than when the array is set to convex. So there are obvious differences in curvature settings, but what is most noticeable is the difference in frequency response. I can add that code, but I want to better understand what needs to be done first.
There was a nice thread on line arrays a while back with some mathematical models that I need to review, and I also need to better understand the Legendre shading that Keele describes in this article. I used Keele's equations to calculate the shading values, and provided 5 steps from full Legendre shading to no shading. But I haven't done a lot of experimentation with the shading parameter.
Right now the setup is a bit clumsy, where you need to connect to one speaker via Bluetooth, make changes, disconnect, and then connect to the other speaker. I need to make that process quicker and easier to facilitate A/B testing.
I just got past a huge hurdle: my wife had some friends over and they loved listening to music on these. So my wife is OK with keeping them in the living room!
There was a nice thread on line arrays a while back with some mathematical models that I need to review, and I also need to better understand the Legendre shading that Keele describes in this article. I used Keele's equations to calculate the shading values, and provided 5 steps from full Legendre shading to no shading. But I haven't done a lot of experimentation with the shading parameter.
Right now the setup is a bit clumsy, where you need to connect to one speaker via Bluetooth, make changes, disconnect, and then connect to the other speaker. I need to make that process quicker and easier to facilitate A/B testing.
I just got past a huge hurdle: my wife had some friends over and they loved listening to music on these. So my wife is OK with keeping them in the living room!
Actually soldering the ball grid arrays is a lot easier than you think. You must make a solder paste mask (I use plastic sheet) and screen print the solder on. Then simply place the parts on the pads. A hot plate or toaster oven can be jury rigged into a SMD solder station. I use a hot plate. The parts tend to self center on the pads and solder bridges are rare if you exercise care during screen printing (you can always wipe the solder off if you mess up screen printing).
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.