DIY speakers have fuzzy imaging

[chrisk15]

Hi all - I recently built an active 3-way speaker system and am loving the overall sound, but I'm having some issues with the imaging and soundstage not being very solid. I've tried moving them to the typical equilateral triangle position with them toed in, flat against the wall, pulled way in sitting only a few feet away, higher/lower, etc. but nothing seems to help enough.

Woofers are 10" Aurum Cantus AC250-F1, midranges are Morel MDM 55 2" domes, tweeters are Aurum Cantus G2si ribbons. I have them crossed over at 24db/octave with a Linkwitz-Riley simulated filter. W-M 800Hz, M-T 3500Hz. One thing I noticed is that the midrange blends much better with less phase cancellation when it's inverted, even though 24db/oct should be a full 360 degree phase rotation. Either way, normal or inverted, the imaging isn't great. Vocals and center-mixed sounds don't sound like they're coming directly from a phantom center and either float around or have a much larger, "blobby" focus. I tested each driver individually for inverted phase compared to one another and they all sound like they're connected and configured correctly.

I have another set of 3-way DIY speakers that have three stacked spheres as the enclosure, with a traditional crossover, and their imaging is super precise no matter how they're placed. Wondering if my issue is due to the cabinet/baffle diffraction, room placement, driver distance, the active amp (Acoustas AC650), or something else.

Picture for reference:

Like I said, I've moved them out, wider, toed in, sat really close or far away, nothing really changes the poor imaging. Any advice would be great!

 

[kemmler3D]

Afaik unstable imaging is often caused by directivity errors, i.e. beaming / lobing that changes erratically by frequency. If you haven't taken measurements to make sure that directivity changes smoothly between the 3 drivers I would start there.

Also you didn't mention the crossover frequencies, but the placement and size of the mid relative to the woofer makes me suspect you could be crossing the mid too high, leading to beaming at the top of the woofer range and wide dispersion at the bottom of the mid, i.e. directivity error.

 

[Arindal]

Main causes of unstable localization is either interaural discrepancies between the channels or drivers, particularly those changing with frequency bands mostly responsible for localizing a phantom source. Interchannel tolerances between drivers of one and the same type, crossover tolerances, phase difference - these all contribute to this type of deterioration of imaging.

The other big issue is reflections in the room causing our brain to erroneously assume that a phantom source is coming not only from one place but several, which usually is a frequency-dependent phenomenon. These are mostly the directivity-related or room-induced problems with unstable localization, particularly in reverberant rooms and wide-radiating speakers.

As a first step I suggest to take anechoic measurements and compare the signal of the two loudspeakers. Anything exceeding +-1dB in a narrow band or +-0.5dB over a broader band, as well as drift in phase, partially inverted phase, or crossover frequency, is a potential cause of localization drift.

The other useful experiment is to listen to the speakers in a nearfield environment, by gradually reducing the stereo triangle´s distance. Preferably a well-damped or overdamped room. If the localization issues persist, as you have hinted, most likely interchannel differences or loudspeaker tolerances are at play.

No speculation from my side what is at play here, but I see a lot of potentials candidates, like reflective surfaces and broad directivity pattern of midrange and ribbon, use of ribbon tweeters with both potential tolerance issues and problematic impedance. What makes me stumble is the woofer and midrange dome being very far from each other with presumably a very high crossover freq. This is typically a source of lobing/interference issues and can contribute to phantom localization stability issues as well.

Good luck!

 

[chrisk15]

Afaik unstable imaging is often caused by directivity errors, i.e. beaming / lobing that changes erratically by frequency. If you haven't taken measurements to make sure that directivity changes smoothly between the 3 drivers I would start there.

Also you didn't mention the crossover frequencies, but the placement and size of the mid relative to the woofer makes me suspect you could be crossing the mid too high, leading to beaming at the top of the woofer range and wide dispersion at the bottom of the mid, i.e. directivity error.

Good point, I'll try messing around with the crossover frequencies. I edited the original post to include them; right now it's 800Hz and 3500Hz. I had them at 700 and 4500 and 5600 and the imaging didn't really change, just the tone. I was a little concerned because the Fs of the mid is only 380Hz so I didn't want to go too low but with 24dB/oct maybe it's not as big of a deal.

 

[chrisk15]

Main causes of unstable localization is either interaural discrepancies between the channels or drivers, particularly those changing with frequency bands mostly responsible for localizing a phantom source. Interchannel tolerances between drivers of one and the same type, crossover tolerances, phase difference - these all contribute to this type of deterioration of imaging.

The other big issue is reflections in the room causing our brain to erroneously assume that a phantom source is coming not only from one place but several, which usually is a frequency-dependent phenomenon. These are mostly the directivity-related or room-induced problems with unstable localization, particularly in reverberant rooms and wide-radiating speakers.

As a first step I suggest to take anechoic measurements and compare the signal of the two loudspeakers. Anything exceeding +-1dB in a narrow band or +-0.5dB over a broader band, as well as drift in phase, partially inverted phase, or crossover frequency, is a potential cause of localization drift.

The other useful experiment is to listen to the speakers in a nearfield environment, by gradually reducing the stereo triangle´s distance. Preferably a well-damped or overdamped room. If the localization issues persist, as you have hinted, most likely interchannel differences or loudspeaker tolerances are at play.

No speculation from my side what is at play here, but I see a lot of potentials candidates, like reflective surfaces and broad directivity pattern of midrange and ribbon, use of ribbon tweeters with both potential tolerance issues and problematic impedance. What makes me stumble is the woofer and midrange dome being very far from each other with presumably a very high crossover freq. This is typically a source of lobing/interference issues and can contribute to phantom localization stability issues as well.

Good luck!

Thanks! I'll try to take more measurements and get some understanding of the directivity. This room is rather reflective and pretty large (23'x25' or so) with a lot of hard surfaces, but because the other speakers sounded so precise I figured the room couldn't be 100% of the issue. The other ones use an AMT instead of a true ribbon, but also a 2" dome midrange and a 6.5" woofer. That said, the woofer and mid are much closer together. Time to do more testing

 

[chrisk15]

Sorry for the triple post, but I'm ashamed to admit that sitting about 3 feet away from the speakers, toed in and equidistant with no furniture between, the imaging sounds excellent. Some "stereo surround" songs that use phasing for imaging really pop now, just like my other speakers. Interesting how they seem to be less affected though, I guess the directivity and baffle diffraction is just better because of the driver spacing/small spherical enclosures.

 

[voodooless]

I would stop making changes until you have measurement gear. Otherwise you’re basically blind.

First off, did you do a baffle-step compensation?

Online Baffle Step Calculation

Baffle Step Calculation

www.micka.de

As for a guesstimate on the crossovers to get smooth directivity: try 600 and 2400 Hz.

 

[Cote Dazur]

View attachment 503644

Like I said, I've moved them out, wider, toed in, sat really close or far away, nothing really changes the poor imaging. Any advice would be great!

One very important factor for imaging is having tweeter at ear level, speaker on the floor will never image well.
Away from the front wall will also help.

Speaker Placement Guide: Get The Best Sound From Your Stereo | Best Audiophile Speakers

Elac - The life of sound | Best Audiophile Speakers

elac.com

Your speakers are probably correct.

 

[DanielT]

I would stop making changes until you have measurement gear. Otherwise you’re basically blind.

First off, did you do a baffle-step compensation?

Online Baffle Step Calculation

Baffle Step Calculation

www.micka.de

As for a guesstimate on the crossovers to get smooth directivity: try 600 and 2400 Hz.

Exactly. Check the basics. Do both speakers measure the same?
Measure them separately in the same position, with the same conditions.
Do the speaker drivers measure the same?

A bit far-fetched perhaps, but still; is the signal into the speakers the same? For example, let's say you have an amp with a balance control and it's a little off, that could cause problems.

Then we have this thing about the room's influence, but that's already been addressed in the thread:

Main causes of unstable localization is either interaural discrepancies between the channels or drivers, particularly those changing with frequency bands mostly responsible for localizing a phantom source. Interchannel tolerances between drivers of one and the same type, crossover tolerances, phase difference - these all contribute to this type of deterioration of imaging.

The other big issue is reflections in the room causing our brain to erroneously assume that a phantom source is coming not only from one place but several, which usually is a frequency-dependent phenomenon. These are mostly the directivity-related or room-induced problems with unstable localization, particularly in reverberant rooms and wide-radiating speakers.

As a first step I suggest to take anechoic measurements and compare the signal of the two loudspeakers. Anything exceeding +-1dB in a narrow band or +-0.5dB over a broader band, as well as drift in phase, partially inverted phase, or crossover frequency, is a potential cause of localization drift.

The other useful experiment is to listen to the speakers in a nearfield environment, by gradually reducing the stereo triangle´s distance. Preferably a well-damped or overdamped room. If the localization issues persist, as you have hinted, most likely interchannel differences or loudspeaker tolerances are at play.

No speculation from my side what is at play here, but I see a lot of potentials candidates, like reflective surfaces and broad directivity pattern of midrange and ribbon, use of ribbon tweeters with both potential tolerance issues and problematic impedance. What makes me stumble is the woofer and midrange dome being very far from each other with presumably a very high crossover freq. This is typically a source of lobing/interference issues and can contribute to phantom localization stability issues as well.

Good luck!

Well said.

 

[chrisk15]

I tried crossing the woofer and midrange from 600 all the way down to 250Hz (at lower volume) and it didn't really alter the imaging or soundstage much, other than maybe bringing it a bit higher in elevation, which makes sense. What did help was shrinking the setup to about half the normal distance between and to the speakers, which also involved bringing it about 2.5' from the front wall. Not perfect but much better. And turns out the other speakers also suffer from the same issue when placed at the same height and distance... who would've thought

Unfortunately the right wall is much closer than the left, and has a fireplace, so I can't really add treatment. I'm kind of stuck with the speakers/TV where they are so I suppose I'll have to live with a less than ideal soundstage except when I'm by myself and can pull them out.

I did EQ to compensate for baffle step since it's an active setup, and the individual 1m measurements are relatively flat with no big spikes or dips after shortening the IR windows. I still want to go ahead with directivity measurements for their own sake, will be nice to dial them in a bit more regardless.

 

[Devnull]

I've got to ask... How come you didn't flush mount the mid and tweeter? They're not designed for recessed mounting. Even surface mounted would be better than having them recessed playing through smaller than optimal openings with hard square reflective edges.

At minimum I would correctly mount the mid and tweeter. Move the mid closer to the woofer. Redo your box calculation, those boxes are HUGE, those woofers need around a 30 liter box for sealed.

 

[chrisk15]

I've got to ask... How come you didn't flush mount the mid and tweeter? They're not designed for recessed mounting. Even surface mounted would be better than having them recessed playing through smaller than optimal openings with hard square reflective edges.

At minimum I would correctly mount the mid and tweeter. Move the mid closer to the woofer. Redo your box calculation, those boxes are HUGE, those woofers need around a 30 liter box for sealed.

They actually are flush mounted now, I wasn't (and still am not) done with the build. I have to grout the tile and clean up some stuff, but the drivers are flat with the surface of the baffle now. Grout filling in the gaps will probably help some too.

Unfortunately I can't really remount the woofer, that would require a complete redo of the enclosure and tile. I'm happy to have the sound quality suffer a tiny bit for the aesthetics on these, just wanted to optimize what I could. Also, the box volume was pretty much exactly what was calculated for a Qtc of 0.5 for that driver (97 liters). Minus the driver volume it's probably around 90 liters. I could've gone smaller, and I actually did have them in a smaller enclosure for an older build, but I'm pretty happy with the bass out of the current setup.

 

[Arindal]

because the other speakers sounded so precise I figured the room couldn't be 100% of the issue.

Such phenomena are usually very dependent on both directivity and speaker placement, so they might appear with one speaker and not with the other.

The other ones use an AMT instead of a true ribbon, but also a 2" dome midrange and a 6.5" woofer. That said, the woofer and mid are much closer together.

The latter geometry plus lesser difference in driver diameters as well as different crossover frequencies, are combined a pretty solid cause for different localization properties. 10" + 2" far apart from each other, crossed over at 600Hz, as well as 2" + ribbon with a high crossover frequency, are a recipe for almost unpredictable lobing, vertical interference and localization instability issues as well as directivity anomalies.

but I'm ashamed to admit that sitting about 3 feet away from the speakers, toed in and equidistant with no furniture between, the imaging sounds excellent.

3ft is really close for such a concept which is not exactly acting as a pointsource, so I would assume you can rule out the loudspeaker-inherent causes such as inter channel tolerances and crossover issues. Would take that as an indication that room reflections are the root cause of the problem, with maybe some directivity issues contributing to the reflections.

Unfortunately the right wall is much closer than the left, and has a fireplace, so I can't really add treatment.

Would say this calls for a concept with more or less constant directivity and rather narrow horizontal dispersion. Something like a midrange-capable spherical horn, or coaxial+horn like this:

A 2" midrange dome in a broad baffle is basically the opposite of that, as the driver acts as omnidirectional over a broad frequency band, and the baffle is boosting the sound level to the sides, which are exactly the angles causing such side-wall reflections the most.

 

[AnalogSteph]

Would have to agree. A wide-directivity speaker like that is generally calling for a very plush decor, 1970s style. I still have some slight doubts about the vertical as well, but if thing are much the same with lower XO that's probably not the main problem.

For more challenging acoustics, I might be pursuing something that looks more like a Neumann KH420.

 

[DanielT]

Thanks! I'll try to take more measurements and get some understanding of the directivity. This room is rather reflective and pretty large (23'x25' or so) with a lot of hard surfaces, but because the other speakers sounded so precise I figured the room couldn't be 100% of the issue. The other ones use an AMT instead of a true ribbon, but also a 2" dome midrange and a 6.5" woofer. That said, the woofer and mid are much closer together. Time to do more testing

..much closer together..plus speaking of loobing. 1/4 of the wavelength is usually said to be ideal regading c-c distance. Now for practical reasons, for example between mid and tweeter it is difficult to physically achieve. So kimmosto, the creator of:

VituixCAD Loudspeaker simulator

..has this to say:

Minimum c-c is 1.0 x wave length and maximum about 1.4 x wave length at XO frequency assuming that design is conventional uni-directional box (not open baffle) with phase matching (acoustical 4th order) slopes. Good and quite flexible initial/design value for c-c is 1.2 x wave length at XO, giving smooth combination of power and early reflections i.e. balanced sound without significant power dip at XO due to bump in DI and dip in vertical early reflections. In other words, this concept aims lobe nulls to directions which are the least significant for power response and vertical early reflections - and listener sitting in sweet spot of course.

Some help with lobing

I think I understand the concept but can't figure out the solution. "Lobing occurs with any two drivers producing the same frequency range in parallel (identical signal) where the distance between the drivers is greater than 1/4 wavelength of any frequency they are reproducing." So if I am...

www.audiosciencereview.com

Note ...with phase matching (acoustical 4th order) slopes..in that case.

 

[DanielT]

@chrisk15 had you measured the respective drivers/speakers to see if the FR is the same? Unfortunately, this is what it can look like if you're unlucky.
@ppataki's pair of brand new ribbon tweeters:

3D printed classic 3-way budget speaker project using SB and Dayton drivers

A small update: I have decided to upgrade the Dayton dome tweeters (just out of curiosity) to these HiVi Swan RT1.3WE ribbon tweeters I am extremely happy with the Dayton tweeters, they sound awesome, but I am just curious what (if any) a ribbon tweeter can add to the listening experience They...

www.audiosciencereview.com

Without similarity and symmetry then fuzzy imaging can be the result.

 

[Juhazi]

Make sure that you have similar driver polarities on both sides, sounds like that is the cause. Look at measured step response. Mistake in dsp or wiring, it happens...

 

[Arindal]

Make sure that you have similar driver polarities on both sides, sounds like that is the cause.

That would be the mother of what I call interchannel discrepancies and can surely cause all sorts of funny localization-related effects. Although, as the OP was reporting that in close proximity stereo listening (like 90cm listening distance) localization stability was excellent, I would almost certainly rule out this explanation. Polarity issues are more likely to get more dramatic in perception in a nearfield scenario.

Look at measured step response.

Except from very obvious mistakes, like one whole channel or a midrange or tweeter in reversed polarity, I always found it difficult to extract any useful information from step response graphs (even inverted subwoofer or defective subwoofer lowpass is not always identifiable). Do you happen to have some deeper insides or hacks, or were you referring to such obvious flaws?

 

[chrisk15]

I took some measurements of each driver with the others muted so the responses are isolated:

They seem pretty well matched to me, especially since the tweeters both had their ribbons replaced (by me ). Excuse the rising top end on the tweeter, I had been playing with EQ. Here's a full range from about 0.5m, tweeter level, speakers against the wall:

From listening position (~3m), speakers against the wall, without tweeter EQ:

 

[audio_tony]

I have to grout the tile and clean up some stuff,

I thought the cabinets looked as though they were tiled. That's creating a hugely reflective surface surely?

That won't be good for imaging at all, as there will be all manner of reflections, not only from the tiles, but the uneven surfaces.

As a starter, I would apply a test covering of something absorbent to the baffle and then listen again.