MiniDSP as Linkwitz Orion ASP

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
I've noticed in my system (where I do digital attenuation), it requires a lot of reduction. -40dB-ish. I've also measured an increase in THD with more and more attenuation. Maybe gainphile is really cranking it down?

Hello,

Following the question of Dave, we're also interested to hear how you're reaching higher levels of THD when using the attenuation.. It just doesn't line up to our measurements on the miniDSP and technically I'm not sure how I could explain it. (though doesn't mean it doesn't exist as we certainly don't pretend speaking the truth... :)

With any attenuation you perform, you'll get your signal close to the noise floor. i.e. making your SNR worse. As you may know, SNR specs is a different one than THD specs though... Now by my standards a 40dB attenuation is a LOT and you should be conscious that on ANY DSP platform, it's not a good idea to work that way. Digital gain is great to tune for minor tweaks, but with a grain of salt for what I explained above (SNR).

Now attenuation is required in a lot of cases (e.g. HF doesn't require as much amp power than Mids or LF) but instead of attenuating in digital domain so much, best is to match up your amplifier volume. i.e. keep what we'd call unity gain (feed a strong signal at input, strong signal at output to maximize the ADC/DAC stage) THEN attenuate at the amplifier itself. (with the volume). You'll see that your system will sound better.

This technique/procedure is what has been done in the ProA/V for many years and what we'd call "gain structure". Without proper gain structure, you don't optimize the full potential of your DSP system.

Hoping this explanation makes some sense, we're still interested in seeing these THD graphs/test conditions if any... It may bring up more food for thought on what is happening.

DevTeam
 
Last edited:
I recently ordered a mini-dsp and am eager to get it!

One question/observation though. It seems Bob McCarthy defines an all-pass filter differently than is being discussed here (see "Sound Systems: Design and Optimization" page 26). According to him, an all-pass lets us set arbitrary ranges of the frequency response and tune the delay within each band.

It seems the biquads allow one to define a function that determines the range and amplitude over that range, but I don't see any variable to set the delay of each individual biquad filter in the mini-dsp. I'm wondering what the mini-dsp people have to say about this??

Having a roughly flat phase response across the entire speaker would be a "nice thing to have." I'm assuming it is more important for line-arrays in pro audio where multiple speakers need to work in unison to send the same signal, or in subwoofers when various subwoofers are working together to produce a signal (and can have dramatically different phase response depending on if they are ported, sealed, etc.)

Having total control of the phase could also be useful to cancel out noises coming from behind the speaker to reduce unwanted room colorations? This would require a second driver behind the primary driver to control the beam, I think. It seems like the pro-audio people use phase to do things like this, from what I can understand from their books.
 
Last edited:
Hi gainphile!
That's a lot of cool OB projects you have built!
Now, I wanted to ask, what did you think about sound quality, when you switched from analog crossovers to digital? Notice, it's not about flexibility, ease of use, etc.
I have a DCX2496 and honestly, I think it sounds like crap, I much prefer the sound of regular crossovers with caps and coils. Granted, I only tried it with regular, boxed speakers. But I think digital processing just sounds bad, really bad...
Now I'm thinking to attempt something OB and was thinking to build some active crossovers with opamps. Or maybe it's worth to try and mod the DCX with transformers and all? I imagine an opamp based crossover would still sound better. There's also PLL crossovers of course with tube/fet buffers...
Just wanted to hear your opinion, as a person who tried both digital and analog active crossovers...
Thanks!
 
Last edited:
Account Closed
Joined 2001
Regarding the distortion claim: Here are some distortion measurements performed on my MiniDSP unit. The "loopback" plot is a baseline measurement showing the residual distortion of my testing system. The other plots should be self-explanatory. (You folks can draw your own conclusions from the results.)

Cheers,

Dave.
 

Attachments

  • loopback.png
    loopback.png
    144.9 KB · Views: 1,298
  • 0db.png
    0db.png
    133.2 KB · Views: 1,241
  • -10db.png
    -10db.png
    144 KB · Views: 1,228
Hello,

Following the question of Dave, we're also interested to hear how you're reaching higher levels of THD when using the attenuation.. It just doesn't line up to our measurements on the miniDSP and technically I'm not sure how I could explain it. (though doesn't mean it doesn't exist as we certainly don't pretend speaking the truth... :)


DevTeam

Well, I initially raised the issue and even soldered pots on the output stage as I was really convinced that is the case.

However since then I used the digital attenuation and could hear no degradation even to -18db.

I guess I was "hearing things". Today I run the MiniDSP as it is with no analogue pots.
 
DSP using DCX2496

A Japanese builder had done the Digital implementation for Orion transfer function using DCX2496. Should be a breeze to convert to MiniDSP? The issue with MiniDSP's HF notch filter inconsistency may be minimal (or may even be fixed already).

Orion DSP (3.2)


An externally hosted image should be here but it was not working when we last tested it.



Key frequencies according to DCX2496:

Tweeter
The next high-pass Linkwitz-Riley4 (LR 24) 1.44KHz

Midrange Subsequently
The following low pass Linkwitz-Riley4 (LR 24) 1.44KHz
The next high-pass Linkwitz-Riley4 (LR 24) 92Hz

Woofer Finally

The following low pass Linkwitz-Riley4 (LR 24) 92Hz

EQ

89Hz-high pass shelving filter set to 172Hz (58) 5/6/11 ORION Revision 3.3
Designation of the ASP circuit is 89Hz-SHP was of 172Hz, set f1 = 88Hz results in a measurement by ARTA.
1.Gain-6dB, f1 = 88Hz, Slope 6dB/oct, Type Low Shelving

1450Hz-low-pass shelving filter set to 2123Hz (54) 7/26/10 ORION Revision 3.2
1.Gain-3.3dB, f1 = 2.11kHz, Slope 6dB/oct, Type High Shelving


Midrange equalization.

High-pass characteristic may be determined by the width of -6dB/oct baffle dipole characteristics first. To compensate for the lack of it because the bass. Then crush the dipole characteristic peaks. To handle all of the units also as a characteristic peak near 5KHz.

20Hz-161Hz low-pass filter set to shelving
1.Gain-3dB, f1 = 161Hz, Slope 6dB/oct, Type High Shelving
2.Gain 15dB, f2 = 20Hz, Slope 6dB/oct, Type Low Shelving

478Hz-parametric equalizer gain of 10dB Q = 1.6

4.77kHz-parametric equalizer gain of 20dB Q = 2.8
(4.77kHz-in can not be set once and 20dB-8.9dB Q = 2.8-stacked in two stages 11.1dB Q = 1)


Woofer equalization.
Dipole dipole equalization characteristics, the two drivers to set the equalization optimization Qts.

Set the low pass filter 20Hz-shelving of 286Hz (dipole equalization)
1. Gain-8.3dB, f2 = 284Hz, Slope 6dB/oct, Type High Shelving
2. Gain 15dB, f1 = 20Hz, Slope 6dB/oct, Type Low Shelving

Set the low pass filter 20Hz-shelving of 110Hz (equalizing driver)
1. Gain-15dB, f2 = 109Hz, Slope 6dB/oct, Type High Shelving


I don't know why M-T xo is 92hz. I thougt Orion is 120hz ?
 
Account Closed
Joined 2001
Creating a proper transfer function for past or present Orion iterations with a DCX2496 (or similar) is not difficult (although some folks have screwed it up.) :) I first did this back in 2005.

The potential problems are not in creating the transfer function.......they're possible signal/noise issues, voltage levels/clipping, dynamic range, volume control placement, etc, etc.

This is old news in any case.

Cheers,

Dave.
 
Creating a proper transfer function for past or present Orion iterations with a DCX2496 (or similar) is not difficult (although some folks have screwed it up.) :) I first did this back in 2005.

The potential problems are not in creating the transfer function.......they're possible signal/noise issues, voltage levels/clipping, dynamic range, volume control placement, etc, etc.

This is old news in any case.

Cheers,

Dave.

I guess the real question is whether someone has actually run the Orion with the MiniDSP. There's lots of reasons something might not work, but until it's tried, you don't know for sure.

I suspect the MiniDSP would work well.
 
Account Closed
Joined 2001
Sorry that I didn't make that clear in my posting. Of course I've tried it.

I'm aware of a number of folks who've done it. In PM conversations I got the impression a few of them did it correctly and a few of them didn't. They are not close to me geographically so I couldn't audition any of those systems first hand.

Your suspicion is perfectly valid. Why not try it and judge for yourself?

Cheers,

Dave.
 
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.