• WANTED: Happy members who like to discuss audio and other topics related to our interest. Desire to learn and share knowledge of science required. There are many reviews of audio hardware and expert members to help answer your questions. Click here to have your audio equipment measured for free!

DIY a decent active crossover on a budget

cyruz

Active Member
Joined
Sep 13, 2020
Messages
100
Likes
83
Location
Italy
Hi guys,

recently I spent a long time looking for a multichannel DAC or active crossover with a low price tag (~500€).

MiniDSP stuff price is very honest but the quality of the internal DAC is only subpar. Okto devices are wonderful but well over my budget.

I evaluated a solution including a RME Digiface USB and several SMSL Sanskrit 10th. This solution comes with a significant quality, considering also RME drivers, but I will miss the remote control and the total budget will still be over.

I'm now considering DIY starting from a MiniDSP Shark board. The idea is to assemble the following setup:

USB to I2S converter
https://a.aliexpress.com/_vFvoet

MiniDSP Shark
https://www.minidsp.com/products/minidspkits/minisharc-kit

4x DAC board with I2S input
https://a.aliexpress.com/_vXfjLJ

Would this be a valid setup? What do you think about that specific DAC board? Any advice?
 

mdsimon2

Major Contributor
Forum Donor
Joined
Oct 20, 2020
Messages
2,478
Likes
3,316
Location
Detroit, MI
I have a similar setup (DIY miniSHARC + 4x AES out), there are definitely a few things to look out for.

Do you have any experience working with I2S? Do you have a scope (100+ MHz) that can verify I2S signal integrity? Working with I2S is finnicky and unless you are implementing a known design it is very likely you will run in to issues that require troubleshooting.

Just because those DACs use a 9038Q2M chip does NOT mean they will be well implemented, in fact for the price I would probably assume they are not.

You may be able to get away with splitting the MCLK, BCLK, LRCLK signals from the miniSHARC to multiple DACs without a buffer or you may not. A buffer is a good idea for these projects but adds cost. How many DACs are you planning on using? 4?

Have you looked at the nanoDIGI? That plus 4x of a decent ~$100 DAC with SPDIF input (Schiit Modi 3+, Khadas Tone Board, SMSL Sankskrit 10th MK2) would put you near 500€ and would definitely work. If you needed a USB input you could add something like a Topping D10S or a miniDSP USBStreamer for ~$100.

Michael
 
OP
C

cyruz

Active Member
Joined
Sep 13, 2020
Messages
100
Likes
83
Location
Italy
Thanks for your feedback Michael, I just finished reading your thread in the MiniDSP forum.

I have no experience with I2S and no specific tools. I see from your experience that it's not easy stuff to deal with. I was wondering, doesn't the Shark board have 8 separated I2S outputs? Why buffering the I2S signal?

I would like to use 4 DACs. The problem is that is the only board I found with a decent chip, accepting a Master Clock, with a low price tag. I could eventually go higher, sacrificing one board and increasing the budget a bit (only 6 outputs).

I looked at the nanoDIGI but I don't like to have so many boxes around for this setup. Any idea for a DIY solution with the nanoDIGI?
 
Last edited:

mdsimon2

Major Contributor
Forum Donor
Joined
Oct 20, 2020
Messages
2,478
Likes
3,316
Location
Detroit, MI
Thanks for your feedback Michael, I just finished reading your thread in the MiniDSP forum.

I have no experience with I2S and no specific tools. I see from your experience that it's not easy stuff to deal with. I was wondering, doesn't the Shark board have 8 separated I2S outputs? Why buffering the I2S signal?

I would like to use 4 DACs. The problem is that is the only board I found with a decent chip, accepting a Master Clock, with a low price tag. I could eventually go higher, sacrificing one board and increasing the budget a bit (only 6 outputs).

I looked at the nanoDIGI but I don't like to have so many boxes around for this setup. Any idea for a DIY solution with the nanoDIGI?

For each I2S DAC you will need to provide MCLK, BCLK, LRCLK and I2S data (plus corresponding grounds) from the miniSHARC. The miniSHARC only has single MCLK (pin 9), BCLK (pin 21) and LRCLK (pin 20) outputs so you will need to split these signals. MCLK (24.576 MHz) and BCLK (3.072/6.144 MHz) are RF signals that are sensitive to geometry and loading so you can run in to issues when splitting these without a buffer. MCLK is this most problematic in this regard because of the higher frequency but not all DACs require this. The old Curryman DACs for example did not require a MCLK signal and I know folks were successfully using at least 3 Curryman DACs with no buffer using very short wires soldered directly to the miniSHARC.

Something like the McFIFO / McDualXO will act as a buffer and provides 4 separate outputs for MCLK, BCLK and LRCLK so there is no need to split these signals.

You can purchase the nanoDIGI without a case (nanoDIGI 2x8 K) for DIY purposes. That plus 4x Khadas Tone Boards which also come without a case could be a good single box DIY solution if mounted in a larger case. The only thing that would be a bit annoying would be the need to run short SPDIF cables between the nanoDIGI and the DAC outside of the case.

A lot of this comes down to how much uncertainty / troubleshooting you want to deal with. You could definitely try your originally proposed solution with the understanding that it may not work. The problem with this is you may end up spending a lot more money than you originally planned getting it work.

The other thing I should say is I personally think the impact of a DAC on overall sound quality is very overrated. For example when comparing an OpenDRC-DA8 (miniSHARC + miniDAC8) which has a SINAD of ~78 dB to my DIY miniSHARC + Okto DAC8 pro which has a SINAD of ~118 dB, the only obvious difference is the completely silent noise floor on the Okto.

Michael
 
OP
C

cyruz

Active Member
Joined
Sep 13, 2020
Messages
100
Likes
83
Location
Italy
The other thing I should say is I personally think the impact of a DAC on overall sound quality is very overrated. For example when comparing an OpenDRC-DA8 (miniSHARC + miniDAC8) which has a SINAD of ~78 dB to my DIY miniSHARC + Okto DAC8 pro which has a SINAD of ~118 dB, the only obvious difference is the completely silent noise floor on the Okto.

Thanks for all the explanation. Your journey for this building is enlightening.

Regarding last statement, I agree, specially if we consider the distortion of the last two element of the audio chain. But after all we on ASR, so we want the best numbers we can get.

I'm considering the solution with the nanodigi as well. What makes me wonder though, it's how the Khadas boards will manage the clock based on the clock passed through the SPDIF... Will they be aligned?
 

mdsimon2

Major Contributor
Forum Donor
Joined
Oct 20, 2020
Messages
2,478
Likes
3,316
Location
Detroit, MI
Definitely getting outside my area of expertise with clocking but I will give it a shot.

The Khadas boards use ESS DACs that each have an internal ASRC and therefore each will be clocked from its own internal oscillator. That being said you would have had the same issue with the I2S DACs that you originally proposed. In practice this difference is probably something like +/- 1 sample (+/-0.01 ms at 96 kHz) and inaudible.

AKM DACs do not have an internal ASRC but if the DAC manufacturer implements an ASRC elsewhere in the system you would have the same issue. If they do not have an ASRC my understanding is that they will use a voltage controlled oscillator (VCO) + phase locked loop (PLL) to sync with the incoming SPDIF stream. This will result in worse performance from a jitter perspective compared to an ASRC but should be clock synced. For example the Schiit Modi 3+ uses an AKM DAC and based on the jitter measurements from Amir's review I imagine it uses a simple PLL.

If you want to guarantee clock sync you really need an 8 channel DAC, unfortunately these are rather rare and typically expensive. An interesting option that you could use with the miniSHARC is the DIYINHK 8 channel DAC.

https://www.diyinhk.com/shop/audio-...l#/fifo_option-null/dac-dac_pcb_with_dac_chip

But again I've never seen any measurements on it and the implementation may not be great. It also requires soldering.

To me all of this points to the value of products like the 4X10HD and OpenDRC-DA8 even if the DACs are not SOTA. :)

Michael
 

Razorhelm

Active Member
Joined
Jan 31, 2020
Messages
160
Likes
316
you might want to take a look at my post about my diy active crossover:

https://www.audiosciencereview.com/...-crossovers-via-rephase-and-camilladsp.19910/

I used camilladsp (a program which runs on windows or linux) and a relatively cheap sound blaster sound card (you could replace this with any multichannel dac your chosen os supports).

The advantage of this method is even a raspberry pi has way more processing power than an all in one solution so there is little limit on what filters you can use and you can use any multichannel dac you like and upgrade it later if you want.

edit: have just seen your focus is more on getting a good dac than the crossover implementation, there is a way to join multple dacs into one audio interface in linux but its beyond my ability!
 
Last edited:

dwkdnvr

Senior Member
Joined
Nov 2, 2018
Messages
418
Likes
698
I'm considering the solution with the nanodigi as well. What makes me wonder though, it's how the Khadas boards will manage the clock based on the clock passed through the SPDIF... Will they be aligned?

Maybe I'm being too cavalier about clock management, but if all 4 SPDIF source signals are being generated from the same master clock (which they are in the NanoDigi), I don't see how the DACs would have a clocking problem. Sure, the ESS dacs are 'asynchronous', but they are slaves to the SPDIF signal - their design spec would be to maintain 'sync' to the incoming signal. There may be some opportunity for misalignment at the master-clock pulse level, but on a macroscopic level it's shouldn't be a concern.
Similar setups have been in use in studio situations for a while - multiple spdif/aes or ADAT signals, and if there is a chance of meaningful clock misalignment it would have shown up there and a product like the NanoDigi would basically be useless right from the start.

Now, there IS a major problem with DAC alignment if an asynchronous protocol like USB is used. I suspect this is where the "multiple DACs have timing problems" idea originated, and it's a valid concern in that context. Trying to 'combine' multiple independent USB DACs into a single logical device is probaly not going to work very well for very long. When the DAC is slaved to an incoming clock signal though, this shouldn't be a concern.
 

dwkdnvr

Senior Member
Joined
Nov 2, 2018
Messages
418
Likes
698
you might want to take a look at my post about my diy active crossover:

https://www.audiosciencereview.com/...-crossovers-via-rephase-and-camilladsp.19910/

I used camilladsp (a program which runs on windows or linux) and a relatively cheap sound blaster sound card (you could replace this with any multichannel dac your chosen os supports).

The advantage of this method is even a raspberry pi has way more processing power than an all in one solution so there is little limit on what filters you can use and you can use any multichannel dac you like and upgrade it later if you want.

Yeah, I've thrown in some discussion on this idea into the 'open source streamer' thread - CamillaDSP on a Raspberry Pi seems to be an interesting idea. My disappointment is that investigating the state of soundcard / audio interface support on Linux reveals a pretty desolate landscape. Linux supports UAC2 class compliant devices, so you'd think things would be pretty solid - and for 2-channel DACs it actually is. But, it turns out that 'class compliance' can mean different things, and there are actually relatively few multi-channel devices that work reliably in Linux. The Okto DAC8 Pro appears to work fine, so there is at least one 'high end' solution available, but it only has AES/EBU inputs that all have to be sync'd to the same clock, so input management is limited. The Motu AVB series is ostensibly 'class compliant', but the current state is that they don't work under Linux - or, they 'work', but experience channel-remapping resets and occasionally decide to emit blasts of noise - not ideal in general and makes use in an active crossover impossible. The Focusrite Scarlett series does seem to be functional and reliable, but they are performing at the ~100dB sinad level rather than at the 110dB level of the Motu units - maybe nothing to worry about, but it does feel like a bit of a compromise.

If you only need 4 output channels the Motu M4 + RPi 4 seems to be a viable, cheap, and likely very good setup, though. It's really too bad they didn't produce an M6, though.

edit: have just seen your focus is more on getting a good dac than the crossover implementation, there is a way to join multple dacs into one audio interface in linux but its beyond my ability!
this is almost certainly a bad idea, for reasons outlined in my prior post. USB Dacs these days are the clock masters, and having multiple master clocks in a system isn't likely to work reliably. It's possibly OK for testing purposes, but with something like the Motu M4 available for so cheap, I'm not sure there is much reason to consider this direction.
 

Razorhelm

Active Member
Joined
Jan 31, 2020
Messages
160
Likes
316
It really is a shame there's no Motu M6, would have bought that in a heartbeat!

As the Motu AVB works on windows maybe you could put Camilladsp on a fanless windows mini pc?
 

abdo123

Master Contributor
Forum Donor
Joined
Nov 15, 2020
Messages
7,425
Likes
7,941
Location
Brussels, Belgium
All MiniDSP products have a SINAD around 90 dB and above.

With 99% of content being CD quality, why on earth would you do this.

It’s literal torture for no audible benefits at all.
 
OP
C

cyruz

Active Member
Joined
Sep 13, 2020
Messages
100
Likes
83
Location
Italy
you might want to take a look at my post about my diy active crossover:

https://www.audiosciencereview.com/...-crossovers-via-rephase-and-camilladsp.19910/

I used camilladsp (a program which runs on windows or linux) and a relatively cheap sound blaster sound card (you could replace this with any multichannel dac your chosen os supports).

The advantage of this method is even a raspberry pi has way more processing power than an all in one solution so there is little limit on what filters you can use and you can use any multichannel dac you like and upgrade it later if you want.

edit: have just seen your focus is more on getting a good dac than the crossover implementation, there is a way to join multple dacs into one audio interface in linux but its beyond my ability!

For the processing part I will use VST, as I already have it setup for downmix and Dirac. Sadly a soundcard won't make the cut because I use a mini pc.

Maybe I'm being too cavalier about clock management, but if all 4 SPDIF source signals are being generated from the same master clock (which they are in the NanoDigi), I don't see how the DACs would have a clocking problem. Sure, the ESS dacs are 'asynchronous', but they are slaves to the SPDIF signal - their design spec would be to maintain 'sync' to the incoming signal. There may be some opportunity for misalignment at the master-clock pulse level, but on a macroscopic level it's shouldn't be a concern.
Similar setups have been in use in studio situations for a while - multiple spdif/aes or ADAT signals, and if there is a chance of meaningful clock misalignment it would have shown up there and a product like the NanoDigi would basically be useless right from the start.

Now, there IS a major problem with DAC alignment if an asynchronous protocol like USB is used. I suspect this is where the "multiple DACs have timing problems" idea originated, and it's a valid concern in that context. Trying to 'combine' multiple independent USB DACs into a single logical device is probaly not going to work very well for very long. When the DAC is slaved to an incoming clock signal though, this shouldn't be a concern.

Yeah that's what I was guessing too, my concerns were related to the performance gap with multiple dacs driven by a master clock. Anyway you got a strong point regarding the existence of a device like the nanodigi.

All MiniDSP products have a SINAD around 90 dB and above.

With 99% of content being CD quality, why on earth would you do this.

It’s literal torture for no audible benefits at all.

Are you talking about the active crossovers? I don't remember them showing great SINAD numbers.
 
OP
C

cyruz

Active Member
Joined
Sep 13, 2020
Messages
100
Likes
83
Location
Italy
It really is a shame there's no Motu M6, would have bought that in a heartbeat!

As the Motu AVB works on windows maybe you could put Camilladsp on a fanless windows mini pc?

Yeah i think so. My only concern regarding these audio interfaces is that they miss a remote control.
 

abdo123

Master Contributor
Forum Donor
Joined
Nov 15, 2020
Messages
7,425
Likes
7,941
Location
Brussels, Belgium
Are you talking about the active crossovers? I don't remember them showing great SINAD numbers.

You need to understand what these numbers mean. Like i said, 99% of content requires (theoretically) 96 dB of dynamic range.


All MiniDSP products cover that, unless you’ll be using them with really sensitive speakers (horn .etc) and you’re concerned about hiss then you’re fine with one of their product.
 

Razorhelm

Active Member
Joined
Jan 31, 2020
Messages
160
Likes
316
Yeah i think so. My only concern regarding these audio interfaces is that they miss a remote control.

Yes it is a pain, I used an old home theatre amp so that I had a single point for volume control with a remote but they do not meet your measurement requirements, sadly avr tend to measure pretty poorly .

Maybe you could get an IR receiver for the computer and tie it to the volume control?
 

mdsimon2

Major Contributor
Forum Donor
Joined
Oct 20, 2020
Messages
2,478
Likes
3,316
Location
Detroit, MI
All MiniDSP products have a SINAD around 90 dB and above.

With 99% of content being CD quality, why on earth would you do this.

It’s literal torture for no audible benefits at all.

I think this is a good point. The 4X10HD seems to offer everything you want in one box with good DAC performance. If you do not need the extra processing power of the miniSHARC I do not see much of a practical reason to DIY (unless you just want a DIY project for fun).

Another interesting product is the C-DSP 8x12 v2 for $500. It has a newer generation SHARC processor compared to the miniSHARC, decent DACs and more output channels than the 4X10HD or miniSHARC. miniDSP now offer an OLED display/remote for $50 which solves one of the major drawbacks of the 4X10HD, a lack of display. It can also be upgraded to use Dirac. The only drawbacks I see with this product are that it cannot be used to implement user defined FIR filters (current software limitation, not a hardware limitation), it only has unbalanced outputs and the inputs are a bit limited compared to the 4x10HD.

Michael
 

levimax

Major Contributor
Joined
Dec 28, 2018
Messages
2,349
Likes
3,462
Location
San Diego
Another solution, although not popular on these boards, is analogue active crossovers fine tuned with "Rephase". Depending on the equipment you all ready have this can be cheaper and easier and still "transparent". I have built these and they work well https://sound-au.com/project09.htm. You can save a lot on DAC's and crossover programing and don't need a expensive and somewhat under powered (computationally) "MiniDSP" box yet still have all the fancy phase and EQ options with Rephase.
 
OP
C

cyruz

Active Member
Joined
Sep 13, 2020
Messages
100
Likes
83
Location
Italy
unless you just want a DIY project for fun

This is the real point. Being this a DIY subsection, even for a little improvement, I'm considering DIY, specially if I can contain the budget.

Maybe you could get an IR receiver for the computer and tie it to the volume control?

Yeah I guess that can make for it, but you'll be controlling your volume in software. It's worth considering, although I think I set my mind on the DIY Nanodigi solution.

Another solution, although not popular on these boards, is analogue active crossovers fine tuned with "Rephase". Depending on the equipment you all ready have this can be cheaper and easier and still "transparent". I have built these and they work well https://sound-au.com/project09.htm. You can save a lot on DAC's and crossover programing and don't need a expensive and somewhat under powered (computationally) "MiniDSP" box yet still have all the fancy phase and EQ options with Rephase.

Cool, it's an interesting read, but I'd like something more flexible, with a bit less tinkering.
 
Top Bottom