Bi-amping the CSS 3TD-X with DSP

[S Moore]

Using a MiniDSP to bi-amp the CSS 3TD-X with intention of removing parts of the passive crossover. See the 3TD-X review.

Here are some notes to avoid this thread from going off-topic with tangential arguments.

Note 1. My intention of bi-amping is to eliminate passive components using DSP. Upgrading the quality of the passive components (air core inductors and high quality capacitors) is expensive. I already use DSP to implement EQ and room correction. Since I'm already using DSP, it makes sense to continue exploiting features and benefits. I do not plan on removing the DSP from my system, so please don't argue that passive sounds better than DSP or that I'm not a purist or whatever.

Note 2. I already have a MiniDSP with 4 outputs and 4 channels of amplification. That means bi-amp. Of course, tri-amp would be nice, but that's a future expansion. "A bird in the hand is worth two in the bush". Sure, I can dream of more DSP channels and more amplifiers, but I wish to use what I have today without buying more stuff. Be content with what I have today instead of dreaming about spending more money tomorrow.

The passive crossover can be seen on the CSS website. I can bi-amp two different methods.

• Biamp between the bass and the midrange/tweeter. Then I can remove all of the bass elements and the HP (High Pass) midrange passive elements.
• Biamp between the bass/midrange and tweeter. Then I can remove the LP midrange and all of the tweeter elements.

 

[staticV3]

I guess the obvious question: why not triamp?

Hypex FusionAmp FA123 with on-board SMPS

The FA123 has two 125 Wrms Ncore® amplifiers which can be bridged to form one 250 Wrms amplifier. Additionaly, a 100 Wrms Ncore® tweeter channel is added.

www.hypex.nl

 

[S Moore]

It took 13 minutes to go off-topic. The topic is bi-amplification, as indicated by the title and content of the OP.

I guess the obvious question: why not triamp?

The obvious answer is because I have 4 channels of DSP and 4 channels of amplification. I figured it would be beneficial to gain something with the gear I have instead of giving up altogether because I don't have 6 channels of DSP and 6 channels of amplification.

 

[Ktacos]

If one only has the 4 out minidsp, better to keep the filtering from woofer to mid passive so you can do the mid to tweeter active. This would let you dial in some delay for the tweeter and overall have more linear integration where hearing is more sensitive.

Suggestion of going all active is very much on topic, not sure what you mean by off topic, and you never stated that you are limited to 4 channels, minidsp is a brand name and they make several different models with various I/O counts. You didn't state which DSP you had.

 

[S Moore]

I'm getting a Jule Fidelity triple binding post and putting the passive elements in a project box outside of the speaker enclosure. This will allow me easy experimentation (it's much easier for project development that fussing around with the pinouts of a SpeakOn connector). The retains me the right to easily revert back to the passive crossover if I move the speakers elsewhere or give to one of my boys.

Attached is a first-draft crossover proposal.

 

[Ktacos]

You'll wanna measure the drivers to develop the filters, prob a little early to settle on slopes and notches.

 

[S Moore]

I'm not going to re-engineer the existing passive crossover. I want to remove the passive parts that the DSP will replace and leave the parts still needed. All of the DSP work will be done later. But the hardware part needs to be finalized.

 

[Ktacos]

I understand what you're trying to do, but if you don't intend to develop new filters for the mid LP+tweeter HP there's pretty much no reason to do what you're doing. Have you developed filters for a speaker before this? You'll probably need to at least sim the mid to make sure the high pass passive network is behaving fine as you generally can't remove remove sections of the passive network and leave others in and get a predictable response. That 3 ohm resistor is after the high and low pass and likely needs those there to work right. Might have to change it or take it out.

 

[Kerry Armes]

We are actually working on an active version of the speaker right now. You can try this as a starting point. The last node on the tweeter is at 10,000 Hz. I should also mention that the amps on the tweeter and mid have 2.5 dB more input gain than the woofer amps so you will have to scale accordingly based on the amps you are using.

 

[S Moore]

I obtained an 8-channel MiniDSP. I'll implement this over the weekend for some preliminary impressions. What's the frequency under the mouse? It looks like a 0 or 6. With -5dB, is the tweeter getting screechy up top? Wow 48dB filters, that's steep!

 

[Kerry Armes]

I obtained an 8-channel MiniDSP. I'll implement this over the weekend for some preliminary impressions. What's the frequency under the mouse? It looks like a 0 or 6. With -5dB, is the tweeter getting screechy up top? Wow 48dB filters, that's steep!

It's set at 10,000 Hz. No, the tweeter is not screechy at all. The 5 dB PEQ there is because there is a peak in the tweeter response that people don't like the look of on a graph. I actually think it sounds better without this. Please note that the mid and tweeter amps are not the same input stage gain as the woofer amp so you will have to adjust the gain in the A1 and A2 filter block to match the gains on your amp.

 

[S Moore]

WOW. My first impressions are very good.

(Next day edit: my impression did not change with a second listening. The active crossover improvement is very significant and listening is much more enjoyable. Because of the complexity of the setup, it is quite difficult to arrange an A/B test between passive and active crossovers without a lot of fussing about.)

It is clear I strongly prefer the active xover over passive. I also found the sound a lot better with the speaker about 8-10cm off the floor. I could not hear this with the passive crossover, but as soon as I dialed in the active crossover, it became immediately apparent. I put some 4x4 lumber underneath and the sound improved. I guess it's time to build some stands with sorbothane dampers. Not running subwoofers or Dirac room correction for my first impressions. Focusing on just the 3TD-X. I tried it with and without the 10kHz shelving filter and my initial preference is without. My benchmark is Triangle Acoustics, which are known for their aggressive presentation, so this figures.

I'm running identical Fosi V3 amplifier with OPA2604 opamps on tweeter and midrange (I had a whole tube of these in my junk box) and another on the woofers. I have a EL84 push-pull amplifier I might try on the tweeter once everything settles in.

I'll continue to work and post results. The next step is to ensure loudness calibrations of each channel for my amplifiers. As Kerry said, his settings were calibrated with different amplifiers. I plan on using REW and UMIC-1. If I switch the tweeter to my tube amplifier, I'll have to go through the recalibration process again.

Once I have it dialed in, and on 8-10cm stands (probably fabricated with some cheap lumber initially), my next step is to integrate subwoofers with Dirac. My first idea is to crossover around 80Hz at LR 24dB and adjust slope upon listening and measurements. I will probably disable the 25Hz PEQ bass notch when in subwoofer mode.

Question about using "safety" capacitor in series with the tweeter. I'm making a patch cable by splicing a Solen 12uF FastCap (I found some in my junk box) into a speaker cable. This might be cutting it too close to 2700Hz. The datasheet is really hard to read. But 12uF is what I have on hand. The 48dB LR crossover point helps because it'll cutoff really quickly even if the capacitor starts to intrude.

The safety capacitors are being considered to protect from hummm and clicks and pops as I'm configuring and validating the setup and playing with amplifiers.

 

[S Moore]

Update.

Once I got everything dialed in, I added the subwoofers. I have twin towers that stand behind the main speakers each having 2 x 7" active drivers (front facing) and 2 x 10" passive radiators (side facing) tuned to 26Hz. Since the subwoofer towers sit behind the mains, I delayed the mains by 2.4ms for time alignment. The subwoofer towers are in-phase with mains. I initially used 80Hz 24dB LR crossover for both subwoofer and mains but gradually lowered the mains to 40Hz, so there is an octave of overlap between the subs and mains. It sounds better this way.

I bought some landscaping paver stones that have approximately the same footprint as the 3DT-X. I washed the stones and painted them with polyurethane, put felt feet on the bottom and top and using them for 8cm speaker stands. They are heavy! I'll probably swap out the felt feet between the stone and 3TD-X with some sorbothane pads for additional isolation.

The miniDSP Flex HT has multichannel Dirac that can be programmed for independent control of tweeter, midrange, bass, and subwoofer. I ran a few sweeps for room correction and didn't like the results. It appears I have a lot of homework to extract the benefits of Dirac with this system. It's not a turnkey optimization.

For the meantime, the system (without Dirac room correction) is sounding absolutely fantastic. I had some people over for some "first impressions", including a sound producer. He's accustomed to my extremely analytical and aggressive Triangle Loudspeakers setup and has been disappointed with the stock (passive crossover) 3TD-X. That opinion changed straight away with the active setup. The system is beating many of the far more expensive Focal and B&W systems found around here and is on par with my Triangles. Definitely beats the pants off the SVS Ultra Evolutions I listened to last week in Vegas at the CES in terms of transparency and imaging (but the SVS wins in the bass department, having 4 x 8" subwoofers integrated into the towers per side). I have yet to develop an impression against some of the larger Sonus Faber offerings. One of my buddies is a Magnapan fanboy so I hope to make a Magnepan comparison. I also need to drag out my Linkwitz Phoenix system, but that's a big project. It's still running on the first miniDSP that Siegfried configured many years ago (edit: 20 years ago).

The total system cost (approximate numbers to make the math easy) is $2500 for the 3TD-X (without coupons). I already had a miniDSP Flex but needed to upgrade to the Flex HT. I chose the Flex HT over the Flex 8 because of the multichannel Dirac. I sold my Flex on eBay at a slight loss and purchased the Flex HT, resulting in an exchange cost of about $300. Now I have 8 output channels. I'm using Fosi V3 amplifiers about $100 each on Amazon. Then I had to purchase 4 more speaker cables using Amazon TNP 12 Gauge about $25/each for another $100. Then you need a UMIK-1 for another $100. Since this is a DIY project, you will need to perform manual tuning. From scratch (having to buy everything new), the whole setup would cost around $4500. You can buy a lot of things for this amount of money!

For me, I already had a bunch of Fosi V3 sitting around and already had a couple speaker cable runs. I figure most people reading this forum already has a collection of equipment. My cost was the 3TD-X, upgrade the MiniDSP Flex to Flex8, a couple more speaker cables, and purchase one more Fosi V3 for a total cost about $2600 (factoring in the "Black Friday" coupon at CSS). I'm also excited about using the new hardware setup to ressurect my Linkwitz Phoenix/Beethoven system!

The upgrade path is clear. The SVS Ultra Evolution towers had wonderful bass. How do I reproduce that bass with the 3TD-X? I tried my giant Dayton Ultimax 18" subwoofer with the 3TD-X, but it was a poor match. Therefore, I need to rethink my tower subwoofers. Following the SVS approach, two towers with 4 x Dayton UltimaxII 8" each will probably be the ticket (for a total of eight drivers). Or maybe the Epique. Another project, another day (when I have money again), and lots to think about. Until then, I'll just keep tweaking these 3TD-X.

 

[Kerry Armes]

WOW. My first impressions are very good.

(Next day edit: my impression did not change with a second listening. The active crossover improvement is very significant and listening is much more enjoyable. Because of the complexity of the setup, it is quite difficult to arrange an A/B test between passive and active crossovers without a lot of fussing about.)

It is clear I strongly prefer the active xover over passive. I also found the sound a lot better with the speaker about 8-10cm off the floor. I could not hear this with the passive crossover, but as soon as I dialed in the active crossover, it became immediately apparent. I put some 4x4 lumber underneath and the sound improved. I guess it's time to build some stands with sorbothane dampers. Not running subwoofers or Dirac room correction for my first impressions. Focusing on just the 3TD-X. I tried it with and without the 10kHz shelving filter and my initial preference is without. My benchmark is Triangle Acoustics, which are known for their aggressive presentation, so this figures.

I'm running identical Fosi V3 amplifier with OPA2604 opamps on tweeter and midrange (I had a whole tube of these in my junk box) and another on the woofers. I have a EL84 push-pull amplifier I might try on the tweeter once everything settles in.

I'll continue to work and post results. The next step is to ensure loudness calibrations of each channel for my amplifiers. As Kerry said, his settings were calibrated with different amplifiers. I plan on using REW and UMIC-1. If I switch the tweeter to my tube amplifier, I'll have to go through the recalibration process again.

Once I have it dialed in, and on 8-10cm stands (probably fabricated with some cheap lumber initially), my next step is to integrate subwoofers with Dirac. My first idea is to crossover around 80Hz at LR 24dB and adjust slope upon listening and measurements. I will probably disable the 25Hz PEQ bass notch when in subwoofer mode.

Question about using "safety" capacitor in series with the tweeter. I'm making a patch cable by splicing a Solen 12uF FastCap (I found some in my junk box) into a speaker cable. This might be cutting it too close to 2700Hz. The datasheet is really hard to read. But 12uF is what I have on hand. The 48dB LR crossover point helps because it'll cutoff really quickly even if the capacitor starts to intrude.

The safety capacitors are being considered to protect from hummm and clicks and pops as I'm configuring and validating the setup and playing with amplifiers.

View attachment 421713

I would go with around a 20-30 uF

 

[S Moore]

The tweeter has TONS of space below 2700Hz, especially with the 48dB crossover slope. I've been playing with some lower frequencies. Is this something you are considering in coordination with your active crossover development partner?

 

[S Moore]

I've been running the midrange:tweeter crossover at 2kHz 48dB LR, looking at ways to improve soundstage. The loudspeakers create audible "cues" that localize the speaker on the soundstage. Part of this might be the physical nature of the front baffle but a portion might be recoverable with electronic tuning.

The passive crossover is 12dB/octave at 2.7kHz. This results in -18dB attenuation when the frequency reaches the tweeter's resonance frequency of 865Hz.

Active crossover at LR 48dB/octave at 2kHz results in -52.8dB attenuation at 865Hz. Because of the steep slope, almost no low frequency signal reaches the tweeter, even when the corner frequency is lower. So the tweeter is being stressed less at 48dB/2000 than the passive crossover 12dB/2700.

I configured a couple MiniDSP slots with 48dB/2.7kHz (your suggested configuration) and 48dB/2kHz (my test configuration). I had someone randomly switch between the configurations every couple seconds (I could not see which one was selected because I put a piece of paper over the display) while I played a variety of genres and test tracks. I almost universally preferred 48dB/2kHz in a true blind listening test.

My first hypothesis is the wavelength of 2.7kHz is 12.7cm, very close to the 11cm diameter (3127Hz) of the LDW6 midrange driver. That means -2.64dB of 11cm signal is reaching the midrange driver, resulting in edge diffraction from the surround and conical directivity ("beaming").

The wavelength of 2kHz is 17.2cm, significantly larger than the 11cm LDW6 diameter. At 2kHz, -30.7dB of 11cm signal is reaching the midrange, a massive reduction. This avoids the frequency range where beaming and surround diffraction become problems.

2kHz might be too low, but a worthy experiment. With DSP, experimentation is "free".

 

[ThatSoundsGood]

I've been running the midrange:tweeter crossover at 2kHz 48dB LR.

The passive crossover is 12dB/octave at 2.7kHz. This results in 18dB attenuation when the frequency reaches the tweeter's resonance frequency of 865Hz.

I think it's even lower than 18db since there are some resistors in there too.

Active crossover at LR 48dB/octave at 2kHz results in 52.8dB attenuation at 865Hz. Because of the steep slope, almost no low frequency signal reaches the tweeter, even when the corner frequency is lower.

This is true, but the knee of a 48db filter gives a lot of energy at the lower range where it's being used and doesn't necessarily mean that it will be more protection than a less steep filter (even though this seems counter intuitive). Either way, it's plenty of protection for it. I've taken it fairly low and it works just fine. You might have more distortion but you won't blow it up unless you're really running an enormous amount of power into it. I've always found myself being limited by the woofers long before this tweeter even when it's crossed as low as 1800Hz (2nd order) since it has some additional attenuation to match the woofers.
Here's an article describing why steeper filters aren't always better or more protection for the tweeter. It changed my mind about this subject.

Crossover Slopes and Protection for Speakers: Does a Steeper Filter Provide More Protection? - Audiofrog

Someone asked recently about choosing a crossover slope and asked, specifically, if using a 48dB/octave crossover would provide better protection than a 24dB/octave slope. I usually recommend 24dB/octave Linkwitz Riley […]

www.audiofrog.com

I configured a couple MiniDSP slots with 48dB 2.7kHz (your suggested configuration) and 48dB 2kHz (my test configuration). I had someone randomly switch between the configurations every couple seconds (I could not see which one was selected because I put a piece of paper over the display) while I played a variety of test tracks. I almost universally preferred when the 48dB 2kHz slot was engaged in a true blind listening test.

My first hypothesis is the wavelength of 2.7kHz is 12.7cm, very close to the diameter of the LDW6 midrange driver (resulting in edge diffraction of the surround and directivity).

The wavelength of 2kHz is 17.2cm, significantly larger than the LDW6 diameter to avoid beaming and surround edge diffraction. 2kHz might be too low, but a worthy experiment. With DSP, expermentation is "free".

I would think that crossing it lower might have more to do with the off-axis/ in-room response being a smoother downward slope and that's why you prefer it. But that's just a guess without seeing any measurements.
I'm interested in your progress on all of this. Seems really cool.

 

[S Moore]

This is true, but the knee of a 48db filter gives a lot of energy at the lower range where it's being used and doesn't necessarily mean that it will be more protection than a less steep filter (even though this seems counter intuitive).

Yea, I was briefly tempted to integrate the energy being delivered to the tweeter (area under curve) which the article from Audiofrog does with his eyes (he looks at the difference between the red and green lines). I decided against doing it because the thermal affect on the voice coil changes with decreasing frequencies. The difficulty is compounded by the tweeter's resonance frequency at 865Hz, under which the thermal effects will become even more pronounced.

I was wondering why Kerry chose 48dB. See my first response to Kerry was a surpise at that steep slope! I'll configure 24dB/2kHz and see what I think. I doubt if I will hear any difference. It'll still keep the 11cm away from the midrange.

I would think that crossing it lower might have more to do with the off-axis/ in-room response being a smoother downward slope and that's why you prefer it.

Please explain.

 

[Ktacos]

You're going to want to measure the drivers in cab or maybe see if CSS can toss you their data to develop some filters. Slapping various orders of filters on a driver isn't going to really get you anywhere or inform you of much.

I generally just measure and linearize my drivers first, that way I'm left with as close to an ideal acoustic response for the filters to do their work.

The loudspeakers create audible "cues" that localize the speaker on the soundstage. Part of this might be the physical nature of the front baffle but a portion might be recoverable with electronic tuning.

I'd wager most of your problems are from your filtering. Really need to be utilizing data, it will tell you what's wrong. Not developing filter tasks in something like virtuixcad is just guessing.

 

[ThatSoundsGood]

Yea, I was briefly tempted to integrate the energy being delivered to the tweeter (area under curve) which the article from Audiofrog does with his eyes (he looks at the difference between the red and green lines). I decided against doing it because the thermal affect on the voice coil changes with decreasing frequencies. The difficulty is compounded by the tweeter's resonance frequency at 865Hz, under which the thermal effects will become even more pronounced.

I was wondering why Kerry chose 48dB. See my first response to Kerry was a surpise at that steep slope! I'll configure 24dB/2kHz and see what I think. I doubt if I will hear any difference. It'll still keep the 11cm away from the midrange.

It's more complicated than that though. It's really about utilizing the filters to create a flat response on-axis and downward trend off-axis with a gradual and consistent narrowing of the dispersion while having the drivers play within their distortion limits.

Please explain.

My guess is that the mid woofer is narrowing below the crossover point between it and the tweeter causing a bump up at the crossover frequency in the estimated in-room response. This generally happens with non-waveguide tweeters, but can still be at least partially mitigated with a good crossover. I can guarantee that Kerry knows this and is calculating for it in whatever he is doing. But again, for me, it's guess without seeing any actual measurements.