DIY active crossover in development - looking for inputs

[DannerD3H]

Hey ASR!

I am currently working on a rather ambitious project, but so far progress is steady and very promissing. I am trying to build an active crossover with 8 analog outputs to power a set of 4 way active speakers. For now, this is what it looks like(still a work in progress):

The goal is to replace the fusion amps in my current main speakers with my own DSP system and amps. This means I need to meet or exceed the performance of these modules... ambitious yes xD
I was hoping that some of you vetreans in here could provide som advice and general inputs along the way.

 

[DannerD3H]

Background
I like speaker building. However, I do not like to build analog crossovers. I also belive that a seperate amplifier for each driver has the potential to yield better results than traditional setups. It also allows to use high power Class D amps for the woofers and use AB or tube amps for the mids and highs if one so desires.

I have, as mentioned, build a pair of 3-way speakers based on fusion amps. They sound good. But I hate the Hypex software.. And I find the noise floor on the tweeter amp disapointing.
Otherwhise I love the concept and signal detect features and what have you.

I then went and build some rather special pc speakers only possible with hefty DSP use. Here I wnated to try and do it cheaply. So I used a Wondom JAB5.. Noise floor was horendous for nearfield PC monitors... And after a couple of months one of the amplifier chips blew up.. So no more Wondom for me. But maaan, I fell in love with Sigma Studdio..

I looked and looked but could only find wondom/dayton based products that allowed the use of Sigma Studio.. So I did some digging. I realized that the fusion amps used the same DSP chips and sigma studio under the hood. Sadly they block the acces to sigma studio for multiple reasons I wont get into. So I started trying to design a DSP core.. One thing took the other and now theres no turning back xD

 

[DannerD3H]

Current state
Im a mechanical engineer by trade with a big passion for electronics and audio. This is a hobby project so progress isn't at lightning speed. I spent a year to get to where I am now. I had to learn all the basics of PCB design and general mixed signal designs(never designed a PCB in my life). I would never have gotten this far without people like Phil's lab, Kaamostech, The FreeDSP team, XRK on DIY audio and the helpful guys at AD EngineeZone and many more. I also found som good books by Douglas Self and a ton of other resources I have devoured

So far this is what I have done:

1. First i managed to build the DSP core. Worked in the first try! :O This really sparked my interest. Had I failed this PCB I probably wouldn't have continued. It was the first PCB I ever made.
   
   
2. Then I had some noise issues with my cheap china modules and PSU's
3. Then I made a USB-C power delivery module along with some interface PCB's for the off the shelf DAC's and ADC's I currently use.
   
   
4. This helped but the ADC is simply just too noisy
5. Then I made a SPDIF interface board
   
   
6. This eleminated the noise from the ADC completely as well as removing potential for ground loops with my PC and allowed me to use its optical output at 24bit 96kHz instead of the properly horrible jack output
7. Finally the latest PCB is an amplifier input buffer heavily inspired by XRK's panel mount BTSB to properly test some different amplifiers.
   
   
   

My current prototype looks like this:

It just use cheap PCM5102 DAC's from aliexpress and a Cheap ADC from Audiophonics/aliexpress(too noisy to really use for anything serious). But I must say, I am already impressed.. And if these DAC's can sound this good, then I am very optimistic I can actually reach my goal and beat/meet the Fusion Amps!
Along with the DSP I have build a couple of amplifiers to test different modules. So far I tested the ICE200AS2 and the 3E TPA3255 based PFFB stereo board. They look like this:

Currently I use the DSP as a preamp for the 3E amp with a pair of bookshelf analog speakers I build long ago. And I must say.. I never enjoyed music listenig this much. I prefer it over my main setup.. It may just be that it is near field and no real room interactions, but it sure is satisfying and gives me motivation to continue!

 

[DannerD3H]

Some details about the current design and plans.

• Modular design
  • This means that PSU, DSP core, DAC's, ADC's and so on all will be seperate PCB's connected with wires.
  • This allows for faster and MUUCH CHEAPER experimentation and respins of PCB's. Tetsing new DAC's, ADC's, PSU and so on.
  • All digital signals, for instance the I2S lines from the DSP core to the DAC's will be carried over U.Fl cables.
  • The connectors on the back are also seperate PCB's like this:
    
    
    
    More on this later!
• DSP core
  • At the moment I am using the ADAU1466 dsp chip. This could potentially be replaced with anything, sharc DSP, pi based or whatever. However, I really wanted to use SigmaStudio and thats why I went with this chip. So far I am also far from hitting its limitations.
• DAC's
  • Modules with one 2-channal DAC chip on each PCB.
  • OPA1612, OPA1656, based output circuitry.
  • Input via U.Fl
  • Balanced 4Vrms Output via Molex connectors to match the connector PCB's shown above.
  • I plan to use AKM DAC's like AK4493SEQ, but nothing set in stone.
• Basic functions
  • SPDIF in and out
    • I only use digital sources so I need this input and its a built in feature of the DSP
    • This would theoretically allow to daisy chain modules as hypex does. This will come in handy for people with multible subs
  • 8 analog outputs
  • 2 analog inputs
    • I will mostly use these for dooing measurments with my scarlet 2i2. However, it can also be used for preouts from an AVR or other analog sources.
  • Programmable over USB-C - Via Sigma Studio.
    • Possibly also accept audio over USB, but we will see about that. Software aint my strongest so I will rely on modules like Amanero for this.
  • 4 I2S inputs.
    • The ADC uses one. USB would use another. This leaves two available slots for mor ADC modules, Bluetooth or just a I2S over HDMI interface if desired
  • Volume knob
    • For a start I will just use digital volume control, but will later investigate relay based latter volume control, but thats waaay out for now.
  • At least two presets
    • Selectable with a push button on the front
  • Input selection
    • For a start I will also use a button on the front to cycle inputs.
    • May later be some auto signal detection stuff if I get to that.

Future plans
When all the above is working I have a few things that I would like to try.

1. The first thing adresses the main issue I have with something like this that the fusion amps so neatly avoids. And that is the ratsnest of wires this will create.. First 8 XLR cables from the DSP to power amps. Then 8 sets of speaker cables from power amps to speakers..
   To solve this I Want to:
   • Get rid of all the XLR connectors and replace them with two Ethercon connectors. I will then make 2 4-channel amplifiers with each their ethercon input.
   • Then the speaker terminals ans cables will be replaced with 4 pole Speakon connectors and cables.
   • This will not only reduce the cables to what would look like a normal analog system, but also eleminate any risk of messing up when moving the system and hooking it up.
2. Add a simple micro controller like an arduino to handle a remote for volume, input, presets and so on

Inputs?
Is there any obvious features that I have left out?
Anything you guys would like from a system like this?
Is there any obvious mistakes/flaws in the concept that I haven't thought of?

 

[DannerD3H]

This is how I envision my final setup to look like, maybe stacked if thermals allow. And with ethercon and speakon if it works, which I expect

 

[DannerD3H]

Right now I am working on building the chassis shown for the DSP as well as all the interface PCB's.

Imediate plan from here on:

1. Next up is a PSU heavily inspired by the one found in Topping DX9.
2. Following that will probably be a PCM5102 based DAC module with a proper opamp based balanced output stage to test the concept and the new PSU.
3. Probably a ADC or updated DSP board
   

 

[jamescarter1982]

very impressive you basically accomplished everything I've been daydreaming about for years in a few months it seems like . Well done . like the speakon and ethercon (that's a new one on me ) idea . that's a big problem with multiway active speakers solved . I've always like the idea of incorporating the multichannel power amps into the stand for a pair of large stand mount speakers .I've seen pics of someone do this with a hypex amp and a large pair of sonus faber spekaers . cutting down on the clutter again

 

[Keith_W]

Very nice! I am super impressed that you taught yourself how to do all this, and make sensible design choices along the way!

If you are planning to commercialize the product, bear in mind that there is an 800lb gorilla in the room - MiniDSP. In fact the market for "IIR DSP in a box" has a lot of competition, and you need to ask yourself how you are going to differentiate yourself in a crowded market. I think MiniDSP has the market sewn up for the entry level. It doesn't let you do much, but it lets you do what you want easily. They offer great product support at a competitive price, and they have been around for a long time so there is market trust and acceptance. It will be difficult to compete against them unless you are prepared to get some serious investment behind you with a large marketing budget.

The problem with MiniDSP is that they don't cater for advanced users. For e.g. no MiniDSP product supports more than 8 channels, and I don't think you can gang two or more units together and get them to work as one. Also, there is no way (I am aware of) to bypass the MiniDSP control panel and directly program the thing yourself and change the block diagram (for e.g. the PEQ is always applied before delay). So it's not very flexible, and it does lock you in to a certain way of doing things.

You may also have a problem with software. Since biquads have poor portability, software has to be designed specifically for your DSP unit. SigmaStudio works for you, but ask yourself whether you think customers should be using it. It might be fine if you are aiming your product at engineering nerds or DSP nerds and don't expect to sell many, but there is already a product for those nerds (Danville DSP Nexus).

I can think of one market with no competition. There is NO commercially available "DSP in a box" that can do FIR filtering with >64k taps per channel for 8 channels, that costs reasonable money. There is DEQX, but they have priced themselves out of the market. The Analog Devices ADAU1466 chip you are using won't have enough grunt for that kind of processing, so you would need a Pi with an ARM as a minimum.

It will need a convolver. There are a couple of free options - CamillaDSP or BruteFIR. Both run on Linux. Unlike IIR, FIR is very portable so any FIR capable software can make filters for it.

 

[DannerD3H]

very impressive you basically accomplished everything I've been daydreaming about for years in a few months it seems like . Well done . like the speakon and ethercon (that's a new one on me ) idea . that's a big problem with multiway active speakers solved . I've always like the idea of incorporating the multichannel power amps into the stand for a pair of large stand mount speakers .I've seen pics of someone do this with a hypex amp and a large pair of sonus faber spekaers . cutting down on the clutter again

Thanks!!

The speakon and ethercon are pretty common in pro audio and studio gear, so I do not see why it shouldn't work for home audio as well?

Good idea with the amp in the base! Didn't think of that. But yeah kinda like the dutcdutch c8. However, another reason for this was that now ha ING build the speakers with fusion amps, I kinda just want my speakers to be a box with preferably one connector for all the drivers. I kinda like the electronics to be separated from the speaker it self. Not really sure why, maybe just for experimentation and future projects.

 

[DannerD3H]

Very nice! I am super impressed that you taught yourself how to do all this, and make sensible design choices along the way!

If you are planning to commercialize the product, bear in mind that there is an 800lb gorilla in the room - MiniDSP.

Thanks for the kind words and also all your inputs!

But regarding commercializong it.. I never planned to. I do this only for my own sake. I wanted this product but couldn't find it. That's why I'm trying to make it. I have no interest in fighting with miniDSP or in general provide customer support.. it would kill all the fun in the project...
That being said, if anyone on here, DIYAudio or my friends would be interested in one when it's done and if it works I wouldn't mind selling a few while rejecting all responsibility xD

But no.. the audio market is not one I wanna be a part of.. To much brand loyalty, snake oil and stubborness

A big problem with commercializing as you mention is Sigma Studio. First of all Analog devices does not really want you to make it part of a commercial product. Guess that's why only Chinese companies like wondom which generally don't care about anything makes products that use sigma studio.
Add to that if users really mess up in sigma studio they may actually damage the hardware, and or speakers.. no company wants that. That's probably why hyoex and minidsp not only makes their own gui but also blocks the dsp acces for users.
But I really like sigma studio, especially in combination with vituix cad. Those two programs together make it SO easy to test stuff..

Wasn't aware of the Danville DSP Nexus. That's basically exactly what I'm trying to build as far as I can see. It even have seperate DAC modules xD
At a serious price point though :O
Pretty sure this DIY project will be much cheaper than that.. At least I hope.. xD more fun at least

Again thanks for the inputs! Certainly valuable inputs! And now I have one more inspiration source in the Nexus, thanks!

 

[DWPress]

I haven't played with Sigma Studio but last time I checked it out it seemed to have all the tools to make decent IIR filters. As Keith mentioned, there is a lot of extra benefit to FIR filters beyond the portability. Given your obvious joy in the hobby and technical skills I look forward to seeing what you come up with in your build. Starting with a Minidsp 4x10 a decade ago myself (that I built into a full preamp with phono stage) I'm always amazed at the evolution of the science and technology we now have access to.

Congratulations on building this thing out. I'm sure it'll end up sounding great but it's even better when you've some personal involvement in making it and learning along the way!

 

[Keith_W]

Thanks!!

The speakon and ethercon are pretty common in pro audio and studio gear, so I do not see why it shouldn't work for home audio as well?

It will work for home audio. The problem is market resistance. Speakon/Ethercon means that you need to use hardware which is compatible, and if you don't have such hardware, you need a breakout cable - which is hard to find, and negates the benefit of Speakon/Ethercon. I personally think Speakon/Ethercon is a brilliant solution and keeps cabling simple. I too have an 8 channel active speaker system, and I hate the tangle of cables on my floor. If I had a solution like yours, I would switch in an instant. Two cables is much easier to hide and make the room look tidy than 8.

But regarding commercializong it.. I never planned to. I do this only for my own sake.

In that case, that's fine ... you can make the product you like and not worry about what the market wants!

Re: snake oil, you won't find that kind of thinking with most people who are considering a DSP solution. But stubbornness is something else, it's common among objectivists as well.

 

[voodooless]

Cool project

Quick tip: add an ESP32 and hook up Sigma Studio via Wifi. Saves you the USB cable

 

[Sokel]

I would focus more on the DSP implementation and its rock solid stability and fail-safes and less in DAC quality.
I would however focus at the output stages and give them a good margin and at the ADC quality and clocks between it and the DACs.

As Keith said, horsepower is the key for freedom at whatever you decide to build down the road, both hardware and software.
You may also want to take a look at papers like this and threads like this about quantization noise, etc.

 

[DannerD3H]

I haven't played with Sigma Studio but last time I checked it out it seemed to have all the tools to make decent IIR filters

Well it also does FIR filters. Just like Fusion amps can. Not sure about how many filters, havent touched FIR yet. Maybe some day And as mentioned, I could always change the DSP core board if I desire more power. But I will stick to the ADAU chips for now until everything works and performs satisfactory. As I said, I really also want to use Sigma Studio. Im not that strong in software so I do not want to write software for a ARM CPU or mess with Pi's more than I have to. At least not in the near future

And thanks for the confidence! And yes it also about the learning. Already learned ALOT about audio and audio electronics I didn't know!

 

[DannerD3H]

The problem is market resistance

For sure.. Thats why I make sure the back panel of both amps and DSP have room for XLR connectors as well as using the seperate connector PCB's just in case som of my friends or others wanna try, but wont commit to ethercon. Then They just need a different rear plate and connector PCB's. DAC's and everything else will stay the same
I however, wont let that limit me from making the ideal system for my self or others who arent stuck in the past

In that case, that's fine ... you can make the product you like and not worry about what the market wants!

But I would still like to get inputs from what other people would like to see from a system like this in case I missed something that I would also benefit form
Would also be cool to sell a great DSP system, but it sure sounds like hard work xD

Re: snake oil, you won't find that kind of thinking with most people who are considering a DSP solution. But stubbornness is something else, it's common among objectivists as well.

You mean that people considdering DSP's arent falling for snakeoil or?

 

[DannerD3H]

Cool project

Quick tip: add an ESP32 and hook up Sigma Studio via Wifi. Saves you the USB cable

Uuuuh, that sure is interesting! Gotta try that. Thanks for the tip!

A shame he didn't implement the selfboot pin.. then I still need a physical switch to allow writing to the EEPROM.. Including that in the programmer would have been fire!

 

[voodooless]

Uuuuh, that sure is interesting! Gotta try that. Thanks for the tip!

A shame he didn't implement the selfboot pin.. then I still need a physical switch to allow writing to the EEPROM.. Including that in the programmer would have been fire!

Probably not so difficult to add this.

 

[DannerD3H]

You may also want to take a look at papers like this and threads like this about quantization noise, etc.

Those are some very interesting links, thanks! And sure, once the basic electronics and audio quality is in place I need to dig deeper in the DSP technical part of my filters.

I would focus more on the DSP implementation and its rock solid stability and fail-safes and less in DAC quality.

How do you mean? I haven't seen any signs of the DSP beeing unstable. And surely a set of proper DAC boards must provide an audible improvemen over these 3$ China DACs?
I know the DAC chip itself may not necesarily be as important as the output circuit though. Thats why I will start by making a PCM5102 based DAC and focus on the output circuit. Its much cheaper for me to experiment with instead of expensie AKM chips.

ADC quality and clocks between it and the DACs.

The Clock circuit and MCLK distribution is the main improvements I have to make for the DSP board once I make a respin on that. Right now I use a very affordable oscilator and the routing and distribution of the MCLK could be better based on what I know now. Especially when I start using DAC's and ADC that actually needs a good MCLK signal. But I think I need to get me an osscilloscope and learn to use it. Then look at how the current performance is before I try to optimize the clock lines.

 

[DannerD3H]

Probably not so difficult to add this.

No, probably not if you know what you are doing.. For now I cant. maybe when everything else is in place