[NeilBlanchard]

Greetings,

I am a recent convert to using an active crossover - I was very skeptical, but my speakers were able to be biwired, so the passive crossover was easy enough to bypass. It requires a second stereo amp, and I had that as well. My Linaeum LT1000 speakers are 2-ways, so this is a lot easier than 3-ways, for sure.



I am using a miniDSP 2x4 HD, which is not expensive at $205, and you'll need a $75 UMIK-1 calibrated mic, too. We can discuss the process to setting this up - it was a couple of days work, and pretty easy. You can set up each of the 4 presets to different crossover settings and do quick comparisons; and then adjust, and then compare, and so on. Then you can sweep (or pulse) each driver and get EQ filters for each driver, and then you can even out the speaker's response for even smoother response.

Once you get things where you want them, you can then use the 3 additional presets to compensate for old / lousy recordings, to boost bass etc.

The main thing I hope to discuss, is to understand how an active crossover is better / worse than a passive crossover. The biggest question I have is about whether an active crossover has better phase coherence - or not - than a well designed passive crossover.

My thinking is that the 2x4 HD is an ADC/DAC - which obviously has its effects - but once the music signal is zeros and ones - it would seem that filtering can be done without affecting the phase of the output on the other side of the DAC. Can someone enlighten me on this, please?

The stock crossover in the LT1000 has an iron core choke - and getting this out of the circuit - was YUGE. Additionally, the amp has as simple a load as possible - it is connected directly to the woofer's voice coil. There are far fewer dynamic interaction effects going on.

The ability that an active crossover has to make fine tuned changes like shifting the crossover points by 5-10 Hz at a time, until things gel. I should mention that the Linaeum TLS tweeter is crossed in at 440Hz, and I have the Dayton Audio RS225P-8A crossed out at 400Hz. I had it at 410Hz for a while, and vocals were just a bit too congested. I will live with 400Hz for a while, and then maybe try it at 390Hz.

And the driver filters are amazing  - you can aim for perfectly flat response, if you want to. The TLS tweeters are about 10dB louder than the woofer, so even though it rolls off a lot above 10-12kHz, it can be made nearly perfectly smooth. I have gone with a "house curve" which is a mild Harman curve. Flat between 200Hz and 1kHz - below 200 rises 1dB/octave (so +3dB at 20Hz) and it drops by 0.5dB/octave above 1kHz )so that is -3dB at 20kHz).

Here's what I measured:

[click on image for a larger version]
The horizontal lines are 2dB apart. The 2 dips are floor reflections.

[richidoo]

The main thing I hope to discuss, is to understand how an active crossover is better / worse than a passive crossover. The biggest question I have is about whether an active crossover has better phase coherence - or not - than a well designed passive crossover.

The biggest advantage of active crossover is not the filtering, but in the direct connection of the amplifier to the load without any reactive components in between, resulting in more efficiency, less distortion, and above all, much better electrical damping of the driver. It's the improved damping that sounds better in a passive/active comparison.

Well designed passive crossover can have excellent phase coherence through the crossover band. But parts tolerances make that more difficult than DSP crossover. Some mfgs test parts and xo function before assembly, then measure final performance and listen also. But not all.

My thinking is that the 2x4 HD is an ADC/DAC - which obviously has its affects - but once the music signal is zeros and ones - it would seem that filtering can be done without affecting the phase of the output on the other side of the DAC. Can someone enlighten me on this, please?

That is incorrect. Phase is affected by any EQ applied to the signal as with all minimum phase filters, whether analog or digital. But the phase coherency is always perfect in DSP, while physical parts tolerances make it more difficult to achieve in passive filters. In either case there is always prescribed phase shift after the filter. Described in "poles," or "orders," each 90 degree rotation of phase is a pole. 1st order filter is 90 degree phase shift after the filter. 2nd order is 180 degrees, 4th order filter is 360 degree phase shift, etc. Understanding of phase is required to achieve good imaging and accurate tone through the xo band.

Linear phase filters can affect EQ while maintaining flat phase response. but requires much more powerful CPU and there's a time latency to perform the calculation, although that's not a big deal for playback, but a very big deal for recording/monitoring. Essentially each digital word is "convoluted" by an impulse response file that describes the changes desired, aka FIR, or Finite Impulse Response. IIR, or Infinite Impulse Response is the kind of minimum phase  DSP you have in the miniDSP, which has same phase shift as all analog filters, but with very low latency and small CPU demand, thus far lower price. Programs like AudioLense, Acourate, rePhase, can create the impulse filters used in the FIR convolution. Programs like BruteFIR, JRiver, Roon can host the convolution DSP when you provide the filters. DEQX is an all in one speaker management preamp which calculates FIR filters and does active FIR convolution internally.

The ability that an active crossover has to make fine tuned changes like shifting the crossover points by 5-10 Hz at a time, until things gel. I should mention that the Linaeum TLS tweeter is crossed in at 440Hz, and I have the Dayton Audio RS225P-8A crossed out at 400Hz. I had it at 410Hz for a while, and vocals were just a bit too congested. I will live with 400Hz for a while, and then maybe try it at 390Hz.

If you are using LR4 filters, phase alignment of the electrical input signal occurs when the DSP crossovers are set to the same frequency, which describes the F6 frequency on the HP and LP filters. If you like it better at alternate frequencies then you're playing with  other factors than phase alignment. Most likely room acoustic effects, or the acoustic phase shift fromthe drivers themselves. Running a tweeter that low might have significant acoustic phase shift especially if it's horn loaded, etc. Measurements can be helpful in seeing what's happening and why. A great book is a "Measuring Loudspeakers" by Dan D'Appolito.

And the driver filters are amazing  - you can aim for perfectly flat response, if you want to. The TLS tweeters are about 10dB louder than the woofer, so even though it rolls off a lot above 10-12kHz, it can be made nearly perfectly smooth. I have gone with a "house curve" which is a mild Harman curve. Flat between 200Hz and 1kHz - below 200 rises 1dB/octave (so +3dB at 20Hz) and it drops by 0.5dB/octave above 1kHz )so that is -3dB at 20kHz).

Which curve you like depends on the directivity of the speaker, the furnishings of your listening room, and what kind of music you like, how you like to listen and your hearing ability. But DSP EQ certainly makes such changes easy for different occasions.

The hardware quality of the active crossover has a big effect on the sound quality despite the DSP itself being essentially perfect. Analog i/o stages, jitter and overall system complexity are what prevent audiophiles from going active. It's difficult and expensive in time and money to make a "high end" multichannel active system to rival the refinement of a good high end passive system.

[JohnR]

Actually, the 2x4 HD that the OP is using has FIR fliters and can therefore implement linear phase crossovers. You don't need a particularly long filter for a two-way.

[richidoo]

https://www.minidsp.com/products/minidsp-in-a-box/minidsp-2x4-hd

Wow, that's impressive, for only $205USD

[NeilBlanchard]

Thanks for your responses - I will have to try 4th order Linkwitz Riley filters. I have not tried it, mostly because there are so many possibilities. I didn't mention that I put a new 22uF cap on the tweeter - that is the value used in the stock crossover - so that there is something to protect the tweeter, in case I click something in the software that I shouldn't. The spec says the crossover point is 400Hz; though I measured the acoustic crossover at 485Hz. So, with the cap, I should do 3rd order LR on the tweeter and 4th order on the woofer - and I can try various frequencies.

I did originally try setting it at 800-900Hz, to try to avoid the peak of second harmonic distortion that I am measuring in the tweeter, that is in the 400-500Hz region. I don't know what is causing that. And when I dropped the crossover to 440Hz - the speaker sounded significantly better. Having 5+ octaves coming from the tweeter is amazing, and I can't imagine anything better.

The Linaeum TLS (true line source) tweeter is not your ordinary tweeter:



The tweeter is the monopole version and is 6" high, and it is open behind; though there is a ~1/2" thick piece of felt visible behind the curved leaves. The cabinet is a transmission line. It has very wide horizontal dispersion, and quite limited vertical dispersion - this is the biggest weakness in an otherwise stellar driver.

I wish I could figure out how to set up the FIR. And I am pretty sure the analog portions of the miniDSP 2x4 HD are not super high quality. I have suggested to Schiit Audio that they should make a quality unit that does similar things, as a standalone, and possibly as part of a preamp.

I stand corrected on the phase coherence. I am very much a newb. As I said - I was a skeptic of inserting a digital device into the mix. But, hearing the results made it undeniable - if I can make a speaker sound better than stock, in just a few days, I can only imagine how much a practiced designer could do. If only to model a crossover, to then build a passive one based on what you hear and measure.

The driver filters are much more specific than what can be done with a passive crossover. And dialing in a virtually flat response, or with whatever room curve you want, is just icing on the cake.

These speakers are the best I have ever heard. They seem to have a "flavor" and character - and then you play a different recording, and it is quite different. I listen to a very wide variety of music, and the system is letting me hear it with a clear and open way.

[Doublej]

Am I reading your graph correctly, your speakers are 82dB efficiency?

I'll again crow (tomato shield up) for Meadowlark Audio's site. www.meadowlarksings.com. There are some good articles on this topic, though not in the level of technical detail that is being discussed here.

[NeilBlanchard]

No, that is just the volume it happened to be with the mic about a meter away on the listening axis. They are probably about 89-90dB efficient, as that is the rated efficiency of the woofer. The tweeter is about 94-95dB, and I had to pull it down with the driver EQ filter.

[NeilBlanchard]

I changed the crossover settings from LR2 and the woofer starting at 400Hz and the tweeter at 440Hz - over the LR4 and both drivers at 440Hz. The fact that I can do this - is amazing all in itself. With the 22uF capacitor on the tweeter, I set the actual setting for the miniDSP 2x4 HD for the tweeter to LR3, and the woofer to LR4.

The change to the sound is VERY noticeable; though it was not a big change tonally. It is a very big change to the sound stage - it is now much "bigger" and more of a whole. It is wider, and a bit "taller" and the speakers essentially disappear. The center image is much "stronger" than it was, and while it still shifts as you move from side to side - it doesn't evaporate as easily as it used to.

I am still gathering things in, and I was not expecting this kind of change.

I will experiment with a range up and down in frequency, and see if I hit a sweet(er) spot.

[NeilBlanchard]

Another use for the miniDSP 2x4HD is with a pair of speakers, and a subwoofer. This young man Technico made an excellent guide to using the miniDSP 2x4HD - which I found to be invaluable. He is really sharp, and he has a LOT of information in these videos:

https://www.youtube.com/watch?v=ihOMKQ3i2OM
https://www.youtube.com/watch?v=RZNH4vxgz9A

There is a LOT that you can do, using this device, and the UMIK-1. If I am not mistaken, he is using a pair of Spica TC50 speakers. And if you have 3-way or 4-way speakers you can either use two 2x4HD or a 2x10HD that miniDSP also make.

On my testing, I tried 400Hz, 420Hz, 440Hz, and 460Hz; all with the LR4 slopes. The differences were subtle, but I focused on vocals, and what I heard with the crossover set to 400Hz, was more details, but less warmth. At 460Hz, there was a bit more body and warmth, but slightly less clarity.

Long story short, I ended up with the 440Hz, that I had on the tweeter, previously. It may be that this blends best with the 22uF capacitor. To reiterate, I am using a LR3 on the tweeter, which combines with the cap, yields the same as a LR4 slope. And yes, there are no issues with the sound at the crossover point, that I can hear.

[brj]

Just a note... active crossovers can be analog as well as digital.  By using an analog line level crossover, I avoid extra ADC/DAC conversions.  Of course, analog crossovers implement only IIR filters (which are rarely linear phase), and typically of no more than 4 poles, but this isn't an issue for me, as I only needed 2nd and 3rd order filter slopes.

(I do implement driver correction digitally on my source prior to my DAC, however.  I keep meaning to implement an all-pass filter in my source as well in order to see if I'm even able to detect non-linear phase, but I never quite seem to get around to it.  The system sounds so good that I just never seem to be able to make that a priority...)

Enjoy!