MiniDSP as Linkwitz Orion ASP

While I'm not building the Orions, Linkwitz' ASP is a great basis for Open Baffle Speakers XO. Looking at the stages I can see that MiniDSP satisfies most of its requirements, except for a few which I would like to confirm:

Here I assume the requirement is 4-channel Loudspeakers per side... so I would need:

2x MiniDSP kit Rev B
1x 4-Way Advanced Plugin
15v DC power source


The XO itself must be able to do:

Slopes (obviously) ... ok
Phase alignment/allpass filters ... ok
Notch filters ... ok
Channel level ... ok
Linkwitz Transform ... ok (thanks to the netherlands DIYers!)
Shelving Lowpass filter ... not sure how to do this ?

The other important question is what are the effects of 4-way DSP XO without staging/cascading the XO slopes as SL describes as "common crossover topology mistakes" here ?

linkwitzlab.com/frontiers_5.htm#V
 
What is your objective? To try and improve the Linkwitz design? Or duplicate and then have the option to bring other features to the table......like room correction, etc?

The miniDSP doesn't support an equivalent allpass filter, but it will implement a pure delay that should satisfy the requirement for phase-coherence....in a different way.
Shelving filters are easy to achieve with the miniDSP. Set frequency to the center of the shelf range, set attenuation/boost, and set Q=1 and you have first-order shelf with exactly the same characteristics as an analog filter.

I currently have my miniDSP combo setup to implement a close equivalent of my Phoenix ASP. Programming/testing took me an hour or so. There are some potential pitfalls regarding clipping, cascading of EQ sections, etc, etc, but it seems to be working fine.

An Orion ASP (DSP) with Thor sub-woofers should be realizable in much the same way.......but I haven't done it yet.

Cheers,

Dave.
 
Why wouldn't you cascade the sections? I am not familiar with the miniDSP but if it doesn't have flexibility to allow the output of the first filter to go to two different places, just dulplicate that first filter in the other path. Because it is digital, two identical filters are truly identical (unlike two "identical" analog filters).

A shelving filter is also used for bass and treble controls. A bass control should be a shelving low-pass.
 
Thank you for the answer regarding shelving lowpass filter. Then that section is answered.

Shelving Lowpass filter ... ok

The second question which is as important, is because in a multi-way situation the highpass of previous section has phase shift effect for the next stage (see SL's explanation). This is both logical and audible in practice ie. the cascaded filter is coherent-sounding. The only exception is where the xo point is a decade apart (sometimes possible in a 3-way, but not possible/practical in 4-way). In a 2-way this is normally not an issue.

I was hoping that the algorithm of MiniDSP considers this. I'm not sure what it means by "low the output of the first filter to go to two different places, just dulplicate that first filter in the other path".... could you please elaborate further?

My objective is simply to have an xo similar to Linkwitz ASP. But different xo point, notch, etc. I have done it on a breadboard and am happy, but simply do not have the energy to transform it to a PCB.
 
Last edited:
You know, I've wondered about the cascading/phase issue SL points out. On the one hand, he definitely knows what he is talking about. But in my current DSP setup, where I do XO and EQ in the computer, I don't cascade my filters. I measure/develop each section separately, and then combine them. When I've measured it all as a whole, it sums just as one would hope. So I don't know what to think about the issue.

Luckily, it would be easy for me to change it to a cascaded setup - I just haven't yet. Hmmm...
 
I only have the 4-way PEQ plug-in so can only speak to that one. You can't "re-route" internally the output of one filter back to the input of another. With this plug-in, once the signal is split, it stays split.

In the case of an analog signal processor you can implement an all-pass delay circuit in one branch and then a further delay in another branch when you split the signal again. In the case of the miniDSP you can't do that, so you'd have to add up all the delays required in each particular branch.

I'm not sure it's really an issue.


Here's a copy of the architecture for each of the miniDSP plug-ins:

MiniDSP - MiniDSP Pluggins

Cheers,

Dave.
 

minidsp

Member
2009-12-15 12:15 am
Hey Guys,

In answer to some of the general questions being asked:
- Shelving Lowpass filter: It's indeed been there for a while now (check plug-in datasheet/manual). Have a look at the video here and it will show you how you can toggle each EQ band from a Peak/Low Shelf/High shelf.

- As for the 4Way crossover, current implementation is as per diagram Davey pointed out on our website. it isn't the configuration as per Linkwitz. Maybe one day we'll come up with a configuration like this one. It's quite a redesign actually since it breaks all current structure we have in the code/UI so no idea if when we would have it... Would be worth to see the advantages though.. I can't answer that question since I've never tried myself this arrangement. It may mean nothing at all, but most of Live/Studio DSP 4 way processor I know off wouldn't have such configuration though... By no mean a reference to say that Linkwitz isn't right (he's a reference by all means), but just stating that 4way crossovers do work on a lot of products with a normal configuration... As anything, any products out there, there is always something else you can add/modify on a product...

My 2cts.
 
Thank you. To summarise:

Slopes (obviously) ... ok
Phase alignment/allpass filters ... ok
Notch filters ... ok
Channel level ... ok
Linkwitz Transform ... ok (thanks to the netherlands DIYers!)
Shelving Lowpass filter ... ok
Master volume control ... ok
Cascaded filter stage topology ... no


As explained previously the last feature I mentioned above is very rarely used. In fact I have never seen any commercial product doing it (except SL's ASP).
 

minidsp

Member
2009-12-15 12:15 am
Yes, your summary is correct. Just one detail though, we don't have all pass filters implemented. Only delay and polarity.

All pass filters can however easily be implemented with a simple Biquad filter in the Advanced mode... Just need to punch up the numbers...

Hope this makes sense,
 
The theory of cascading filters is:
Lowpass and highpass filter sums correctly in a 2 way system. In 3-ways they don't - unless the xover frequences are more than a decade apart - In the mid channel phase lag from one xover will affect the phase at the other xover. The solution is to make 2 perfect 2-ways systems and cascading them - consider mid/high as the high part in a 2-way system.

In miniDSP you can only do that in the advanced mode, where you can duplicate the HP part of the low freq xover in the LP and the HP part of the high freq xover - as mentioned by someone else.
 
...
I was hoping that the algorithm of MiniDSP considers this. I'm not sure what it means by "Allow the output of the first filter to go to two different places, just dulplicate that first filter in the other path".... could you please elaborate further?
...

Note how in Linkwitz's crossover, the input for the high pass of the mid is not the input signal per se, but the output of the high pass for the woofer. That is because the woofer high pass filter will still cause altered magnitude and phase response even at the mid's high pass frequency.

Instead of using the output of the woofer high pass as the input to the mid high pass, you can duplicate the woofer high pass filter before the woofer high pass filter. That will have exactly the same effect. Like this:
 

Attachments

  • crossover.GIF
    crossover.GIF
    7.4 KB · Views: 1,836
In miniDSP you can only do that in the advanced mode, where you can duplicate the HP part of the low freq xover in the LP and the HP part of the high freq xover - as mentioned by someone else.

Yes. Can MiniDSP do this... and how? This is exactly what I want to hear! That is similar to cascaded filter stages.

That will make MiniDSP perfect (well maybe not yet .. I'll reserve that for hires ADC/DAC :))

Thanks for the drawing also!



Edit: I just re-read the 4-Way advanced Plugin manual. Are you referring to using "PEQ" to achieve this?
 
Last edited:

minidsp

Member
2009-12-15 12:15 am
To further comment Macboy post, you can indeed build this exact configuration but only by using the miniDSP 4way Advanced plug-in allowing custom Biquad programming.

In advanced mode, you basically have access to 8 biquad per channel for the crossover only (see back of datasheet of 4way advanced plug-in for example). These biquad are typically used for the crossover section (LPF + HPF) in the basic mode. Each biquad is able to perform a 2nd order filter so that's how we can go up to 8th order on the LPF (4 biquad) and 8th order on the HPF (remaining 4). Makes sense?

Now to build the configuration shown by MacBoy, you simply need to calculate each filter biquad coeff separately and load the Biquad settings to each section as per the diagram.... Eventually allowing you to have that exact configuration.

I apologize if I confused a couple of you guys earlier on when saying that we couldn't do it. I meant we couldn't do it in the basic mode that is pre-configured for HPF + LPF per output. In the advanced mode (Biquad programming), you really have a lot more flexibility than people actually realize.

If 8 biquads (up to 16th order) are not enough filter cascading, you can always do the same with the remaining 6 biquads used for PEQ... Up to 14 x 2nd order filters should be enough for you guys, no ? ;-)

Finally, I'm sure that one of you will say: Calculating biquads is too complex... If you have high school math knowledge, you'll see that it's a real simple task. :) Just google RBJ (Robert Bright Johnson) EQ cookbook and you'll find a wealth of knowledge.

Soon enough, we'll figure out a way to embed all this into a nice online calculator...

Hope this helps clarify the Advanced Biquad concepts,
 
gainphile,

I'm not sure why this (cascading filter topology) might have been a show-stopper for you. I don't really think it's a concern. As mentioned by hhenne, when xover frequencies are at least a decade apart, (which they are in the Orion) and then you factor in the acoustic responses of the drivers this effect disappears into the mud. (More so with higher slope crossovers.)

From a theoretical standpoint...analyzing just electrical responses on paper it looks good, but in the real world most probably a non-issue.

Linkwitz designs his ASP's using this approach more from a practical standpoint because it requires less components to implement and the analog topology makes it the obvious way to do it.

minidsp,

Thanks for the clarification/update. Good info.

Cheers,

Dave.
 
Last edited:
Hi Davey, indeed it would have been a showstopper for me, because I have such experience with my xo. Furthermore, in the future I want to develop a 4-way system in which separating frequency a decade apart would not be achievable.

At that level I believe it can be a real differentiator for MiniDSP product.

As for analog xo, it is actually a 'sacrifice' to implement cascaded filters because that means more noise is injected into the signal (signal passes more op-amps).
 
Last edited:
Ah, you're one of those op-amps-are-horrible guys. :) I don't agree that it's a sacrifice (at least from a noise standpoint) to implement cascaded filters in an analog crossover.

Anyways, if noise is a concern you may want to re-think your decision. While the miniDSP outputs are less noisy than any DSP-based unit I've auditioned (quite a few) they are still a good deal more noisy than the outputs of a typical analog, active crossover.

My Phoenix ASP, for example, is inaudible (with my ear right next to the drivers) hooked via a power amp with 30db voltage gain and 93db sensitive speakers. That's low noise. :)

I'm not sure your priorities are the same as mine, but it doesn't matter. All good fun.

The miniDSP guys have done a nice job with these units.

Cheers,

Dave.