Debugging an imaging issue - Martin Logan Ethos

[tengiz]

II’ve been troubleshooting an audio issue in my setup, and after fixing one problem (thanks to help from the forum), I’ve run into something else.

When I listen to a particular vocal track, one specific note — F4, around 349 Hz — sounds wrong. The voice gets blurred or smeared, and the image actually shifts slightly to the left just for that note. It only happens when the singer pushes their voice a bit — it’s not there when the passage is sung more softly.

It’s not a general EQ issue or just a volume problem. The rest of the track sounds fine, and the vocals are usually locked in the center. This only happens with that one note and when it's forced by the singer.

Here’s what I’ve tried:

Turning off room correction (Audyssey) — the issue is still there.
Trying different speakers, with or without correction — the problem goes away completely.

Interestingly, the note in question lines up almost exactly with the crossover point (350 Hz). I replaced the electrostatic panels about 4 years ago, but the woofer amps and crossover components are still original.

On one hand, it might be related to how the woofer and panel are blending — maybe something has drifted over time. But since the issue only happens when vocals are more “pushed,” and not just at higher volume, it feels like it’s more about how the harmonics are handled in that moment. That suggests it’s not just a non-linearity or aging issue in the crossover components. So replacing the crossover boards or amps may not actually solve it.

Honestly, I’m close to moving on from the electrostatics. I tried a potential replacement — the KEF R5 Meta — and it worked beautifully. But I’m still hesitant because I really like the shallow MartinLogan Motif X center I have, and KEF’s R-series doesn’t offer a center channel that shallow.

If there’s still a chance to salvage the setup and keep the electrostatic speakers, I’d appreciate any practical tips. I could probably EQ around this one note, but I’m worried there are more hidden problems, and I’ll end up chasing them one by one.

Thank you!

 

[Heinrich]

II’ve been troubleshooting an audio issue in my setup, ...

Do you have access to measurements?

 

[MarkS]

When I listen to a particular vocal track

If the problem is with a particular vocal track, it's very likely an issue with the recording, just as it was with Abba recordings you wrote about earlier.

 

[Arindal]

The voice gets blurred or smeared, and the image actually shifts slightly to the left just for that note.

Does this problem stay the same, tending towards the left direction, in case you exchange left and right speakers? In this case, reflections and room acoustic problems might be the main root.

If the direction reverses and after swapping the speakers, you have a tendency towards the right side, I would rather investigate the speakers.

349Hz is not localizable to a degree that only changing the interchannel balance of the fundamental note will significantly alter the perceived direction and stability of the phantom source. It is advisable to also look at increased distortion (affecting multiples of this frequency), resonances or other flaws. In case you have a problem with only one speaker, checking polarity of all four drivers might be a good idea.

 

[BenB]

I made you two files to help you diagnose and understand the problem.

Noise_Left_Right_Center_00350.wav has band limited noise centered at 350 Hz. It plays for 3 seconds in the left channel, then 3 seconds in the right channel, then 12 seconds in both channels. The level should be the same to you no matter the location. When it's playing in both channels it ought to image at the center. Please listen and let us know if the levels are consistent, and where the image (phantom source) is.

Transients_Left_Right_Center_00350.wav has transients with a limited bandwidth centered at 350 Hz. (These transients are just the minimum phase filters I designed to filter the noise). The transient plays in the left channel 3 times, then plays in the right channel 3 times, then plays in both channels 6 times. The level and character of the impulse should be the same regardless of where it is playing from. It shouldn't be smeared or lose it's edge/impact when playing in the center. The transient files are of questionable usefulness, but I figured I'd include them.

I put both wav files in a zip archive to conform to forum rules.

 

[tengiz]

Do you have access to measurements?

Do you mean some "official" detailed measurement or Spinorama like data? - as far as I know - there is none.

Or are we talking about being able to do some basic measurements? That I certainly can do.

 

[raindance]

Measure your speakers in room from the listening position. Either there's a mismatch between speaker components at that frequency or there's a room issue (something asymmetrical). But I'll bet it's a beamwidth or amplitude issue.

I had similar issues with Maggies. It was down to poor pair matching.

 

[Heinrich]

Or are we talking about being able to do some basic measurements? That I certainly can do.

Thanks, that's more than I could expect. Let's see ...

 

[tengiz]

This is the RMS average I took this morning across 15 positions around the main listening spot. With the leather couch (which has high-profile backs), the frequency response above 0.5 kHz becomes much more uneven compared to when I remove the sofa entirely. But that measurement wouldn’t reflect the normal listening environment.

The green trace is the left channel — taken after I switched the left and right speakers. And sure enough, there’s a ~3–4 dB difference at and around 359 Hz with 1/12th octave smoothing, which reduces to ~2–3 dB with psychoacoustic smoothing.

But now it’s barely audible — I think it still pulls slightly off-center (to the right, after I switched the speakers), but if I didn’t already know it was there, I probably wouldn’t have noticed it at all.

I would’ve never guessed that something like this could be audible. Even before I switched the speakers, it was subtle — I could hear it, but it was hard to make it consistently repeatable.

 

[tengiz]

I made you two files to help you diagnose and understand the problem.

Noise_Left_Right_Center_00350.wav has band limited noise centered at 350 Hz. It plays for 3 seconds in the left channel, then 3 seconds in the right channel, then 12 seconds in both channels. The level should be the same to you no matter the location. When it's playing in both channels it ought to image at the center. Please listen and let us know if the levels are consistent, and where the image (phantom source) is.

Transients_Left_Right_Center_00350.wav has transients with a limited bandwidth centered at 350 Hz. (These transients are just the minimum phase filters I designed to filter the noise). The transient plays in the left channel 3 times, then plays in the right channel 3 times, then plays in both channels 6 times. The level and character of the impulse should be the same regardless of where it is playing from. It shouldn't be smeared or lose it's edge/impact when playing in the center. The transient files are of questionable usefulness, but I figured I'd include them.

I put both wav files in a zip archive to conform to forum rules.

Thank you! I tried these tracks - everything works as expected - the phantom center is where it should be, just a little fuzzy, not in "perfect" focus.

 

[raindance]

In my current system my center channel has an anomaly (a steep bump) around 700-800 Hz. I simply tell Audyssey to correct up to 1kHz for the center and then up to 500Hz for the mains and it cleans it right up.

The issue you have here would be fixed by Audyssey if you band limit using the app to 500Hz.

 

[tengiz]

I tried the tone signal generator with harmonics, courtesy of REW. I can clearly hear that without harmonics, the pure 359 Hz tone pulls to the right — it’s not “laser focused,” since the frequency is relatively low, but it definitely and reliably shifts rightward. As soon as I add harmonics one by one, starting with 2× at 718 Hz, each one localizes dead center while remaining perfectly resolvable. So, it aligns well with the averaged measurement.

 

[BenB]

This is the RMS average I took this morning across 15 positions around the main listening spot. With the leather couch (which has high-profile backs), the frequency response above 0.5 kHz becomes much more uneven compared to when I remove the sofa entirely. But that measurement wouldn’t reflect the normal listening environment.

The green trace is the left channel — taken after I switched the left and right speakers. And sure enough, there’s a ~3–4 dB difference at and around 359 Hz with 1/12th octave smoothing, which reduces to ~2–3 dB with psychoacoustic smoothing.



But now it’s barely audible — I think it still pulls slightly off-center (to the right, after I switched the speakers), but if I didn’t already know it was there, I probably wouldn’t have noticed it at all.

I would’ve never guessed that something like this could be audible. Even before I switched the speakers, it was subtle — I could hear it, but it was hard to make it consistently repeatable.

If I'm understanding correctly, you measured both speakers in the same position, right? If so, these measurement represent sample-to-sample differences in response. It's easy for the room to make more than a 3-4 dB difference at 359 Hz, so when they are played in different positions the sample-to-sample differences can either be exaggerated, diminished, or pushed entirely in the opposite direction. I'm assuming your room isn't entirely symmetric. If it's not, it may very well be that stereo imaging works better when the sample-to-sample differences and the room differences counteract each other. That would happen with the speakers placed in one configuration but not the opposite. So if you have something that works really well, you could just stop here.

If you want to know more, you could try a few more things:
You could play the noise track I gave you at a fairly high (but reasonable) volume, and listen carefully to each speaker in the extreme nearfield. Try to see if you can identify any resonances that are excited on one but not the other.
Similarly, you could go even more narrowband and use a tone generator to excite a resonance that exists in one but not the other:

Online Tone Generator

Simply enter your desired frequency and press play to hear your tone. It's free, simple and easy to use.

onlinetonegenerator.com

You can play with your computer's left - right balance if that helps, since the tone generator is mono.

I do have a whole series of noise and transient files spanning the entire audible range if anyone is interested in those. I made them for helping tune a car stereo using DSP, since those are very messy environments where imaging is very hard to dial in. They generally don't seem as useful for home audio. My stereo images center or very, very close to center across the entire band above my subwoofer crossover point.

Edit: I see you beat me to the tone generator, but I still recommend the extreme nearfield listening.

 

[tengiz]

In my current system my center channel has an anomaly (a steep bump) around 700-800 Hz. I simply tell Audyssey to correct up to 1kHz for the center and then up to 500Hz for the mains and it cleans it right up.

The issue you have here would be fixed by Audyssey if you band limit using the app to 500Hz.

I’ve found that, at least in my setup, this can be a tricky balancing act. Audyssey tries to “level” the filters to maintain a consistent overall loudness for each channel. As a result, when the correction range is limited, it may compensate for the gain difference above and below the cutoff frequency by adjusting the overall loudness below the cutoff. In my case, this led to unintended effects similar to adding a shelf filter.

It’s possible to turn the leveling off, but doing so may cause the overall loudness to become uneven across channels.

My go-to solution has been to leave bass management in place but avoid correcting the main speakers — using what they call “L/R Bypass” mode. I still apply filter limits in the “Reference” and “Flat” Audyssey modes, but it took quite a bit of trial and error to find the right combination of cutoff frequency, limiting options, and custom biquad filters that would let Audyssey handle the bass without affecting the frequency response of the speakers above 300 Hz.

In hindsight, it was probably a mistake to set the effective cutoff frequency that low, since it left the area around 359 Hz uncorrected.

 

[tengiz]

If I'm understanding correctly, you measured both speakers in the same position, right? If so, these measurement represent sample-to-sample differences in response. It's easy for the room to make more than a 3-4 dB difference at 359 Hz, so when they are played in different positions the sample-to-sample differences can either be exaggerated, diminished, or pushed entirely in the opposite direction. I'm assuming your room isn't entirely symmetric. If it's not, it may very well be that stereo imaging works better when the sample-to-sample differences and the room differences counteract each other. That would happen with the speakers placed in one configuration but not the opposite. So if you have something that works really well, you could just stop here.

No, this isn’t a sample-to-sample variation. I simply swapped the speakers and took RMS-averaged room response measurements at and around the main listening position with each speaker in place. As far as I can tell, this setup reflects the actual listening experience as closely as possible.

The room is highly symmetrical — not perfect, but close. There’s a small slanted corner on one side of the back wall, a small window on the other, and a door that isn’t positioned symmetrically with respect to the speakers. However, the door is covered with diffusers, and the wall opposite it has the same diffusers placed in a symmetrical location.

To me, the frequency response looks mostly even between the left and right channels — except around 359 Hz. That makes it highly likely the issue lies in the difference between the speakers, rather than any asymmetry in the room, I believe.

 

[BenB]

No, this isn’t a sample-to-sample variation.

...

That makes it highly likely the issue lies in the difference between the speakers, rather than any asymmetry in the room, I believe.

This looks like a contradiction. What do you think sample-to-sample variation means? You have two samples of a manufacturing yield that don't appear to match perfectly. That implies sample-to-sample variation, which you deny. I'm scratching my head here.

 

[tengiz]

...



This looks like a contradiction. What do you think sample-to-sample variation means? You have two samples of a manufacturing yield that don't appear to match perfectly. That implies sample-to-sample variation, which you deny. I'm scratching my head here.

My apologies — I misunderstood what you were saying. I thought you were asking whether I had done a sample-to-sample variation test — meaning, measuring one speaker, then placing another speaker in exactly the same spot and orientation to keep the room contribution identical, so that any differences would most likely be due to sample-to-sample variation.

Sorry for the confusion.

 

[tengiz]

I tried the tone signal generator with harmonics, courtesy of REW. I can clearly hear that without harmonics, the pure 359 Hz tone pulls to the right — it’s not “laser focused,” since the frequency is relatively low, but it definitely and reliably shifts rightward. As soon as I add harmonics one by one, starting with 2× at 718 Hz, each one localizes dead center while remaining perfectly resolvable. So, it aligns well with the averaged measurement.

well, it probably was not a totally valid experiment - I mean with pure tone + harmonics. It's too 'holographic' without any irregularity like frequency/phase/amplitude instability/variability.

 

[Heinrich]

well, it probably was not a totally valid experiment - I mean with pure tone + harmonics. It's too 'holographic' without any irregularity like frequency/phase/amplitude instability/variability.

I would expect it to be hard to localise a pure tone, even if there were harmonics added. As you have and understand REW pretty well, there is a mode that generates 'tone bursts'. Those should be localised more reliably. I used them lately to investigate my own ears in regard to symmetry (headphones).

 

[tengiz]

I would expect it to be hard to localise a pure tone, even if there were harmonics added. As you have and understand REW pretty well, there is a mode that generates 'tone bursts'. Those should be localised more reliably. I used them lately to investigate my own ears in regard to symmetry (headphones).

Right — I should’ve known better: with a continuous tone, localization is basically an illusion . It’s fascinating to listen to the superposition of correlated monochromatic steady-state sound “holograms” in the room, but the spatial cues they produce are mostly meaningless.

And yes, thank you — I had seen the "Burst" tab in the tone generator, but since it didn’t offer direct control over harmonics, I initially dismissed it. However, the window option turns out to be good enough for dialing the harmonics up or down.

Even at 360 Hz, and even using the window that suppresses harmonics the most, localization still seems to be present — at least for short bursts of around 20 cycles. But again, I’m not sure how reliable that perception is.

That said, I think my issue is resolved — first by simply swapping the left and right speakers, and then by tweaking the auto-correction filter’s cutoff frequency to cover the extend beyond F4–F#4.

I really appreciate everyone’s input and help!