Cardioid Speakers THD Comparison

[Curvature]

This is a comparison of THD at 96dB SPL average of current cardioid models. If you know of other current speakers and measurements, please post them in this thread.

A few notes on the chart above:

• GGNTKT M1: Data converted from %. Average SPL was 95dB.
• Sigberg Manta: Data converted from %. Average SPL was 97dB.
• Sigberg Saranna: Data converted from %. Average SPL was 96dB.

@sigbergaudio was helpful in providing better, clearer charts for his speakers. I am glad that he is among those manufacturers willing to have their work examined in detail.

Multitone distortion measurements are unavailable other than in the case of the C6C: https://www.audiosciencereview.com/...re-about-to-launch.54794/page-77#post-2345946

It should be well-known that anechoic THD doesn't reflect in-room results or audibility directly, and neither does multitone distortion. The limitations for both are to do with complex perceptual thresholds that depend on signal contents and equally complex changes in "echoic" circumstances: in-room SPL always receives reverberant and boundary reinforcement; music, because of running reverberation, is its own best masker; subs take over for the low frequencies and help decrease distortion in a well-integrated system. That said, regardless of their applicability, these numbers are always interesting and demonstrate how the technology is progressing.

Known cardioid speakers without sufficient distortion measurements:

• Genelec 8381A
• Genelec W371A
• GGNTKT M3

Source Charts

Spoiler

AsciLab C6C

Kii Seven

Kii Three

Mesanovic CDM65

Dutch & Dutch 8C

GGNTKT M1

Sigberg Saranna

 

[tifune]

Shyvaudio, although it couldn't withstand 96dB test:

SHIVAUDYO Point Zero Speaker Review

SHIVAUDYO Point Zero

www.erinsaudiocorner.com

 

[kemmler3D]

Thanks for putting this together. Ascilab coming in hot, and it's sort of surprising to see the D&D 8C (which is apparently universally beloved) with some of the highest distortion all the way up to the crossover. I guess it proves that THD is really not a dealbreaker if everything else lines up well...

I wonder how the S6C will compare, since one of the major benefits of the Purifi drivers is meant to be low distortion...

 

[Curvature]

All available 86dB THD measurements.

Source Charts

Spoiler

Kii Three

Kii Seven

Mesanovic CDM65

Dutch & Dutch 8C

AsciLab C6C

SHIVAUDYO Point Zero

 

[Curvature]

Thanks for putting this together. Ascilab coming in hot, and it's sort of surprising to see the D&D 8C (which is apparently universally beloved) with some of the highest distortion all the way up to the crossover. I guess it proves that THD is really not a dealbreaker if everything else lines up well...

I wonder how the S6C will compare, since one of the major benefits of the Purifi drivers is meant to be low distortion...

The Purifi benefit from the multitone distortion chart between 100Hz-1kHz seems around 10-15dB, when looking at the A6B vs. F6B. I don't know if we can reasonably expect that for the S6C, given how low the C6C was able to push distortion. @AsciLab

I am very excited to see new D&D models released. @Martijn Mensink Not that this will affect your plans at all, but I don't need most of the integration features that the 8C featured. I like the studio version, and would have bought it, but my living room requires that speakers be white. I like that AsciLab offers white as well as grilles. This is where we have come to: the measurements are an important guide, but aesthetics rule. My unfortunate vanity.

 

[Hayabusa]

View attachment 461899

All available 86dB THD measurements.

Source Charts

Spoiler

Kii Three
View attachment 461900

Kii Seven
View attachment 461901

Mesanovic CDM65
View attachment 461902

Dutch & Dutch 8C
View attachment 461904

AsciLab C6C
View attachment 461905

SHIVAUDYO Point Zero
View attachment 461906

wow, so big differences!
Nice work!
Would be nice to keep colors consistent..

 

[Curvature]

wow, so big differences!
Nice work!
Would be nice to keep colors consistent..

Little too much work for me. Attached are the relevant mdats. I encourage you to play with data, like applying smoothing or averaging. It's interesting.

 

[Arindal]

Known cardioid speakers without sufficient distortion measurements:

Would not really describe Genelec 8381A as a cardioid concept in the frequency bands which are sensitive to higher distortion.

Maybe some bigger MEG models could be added here, like RL901k or RL921k, as these were really the first ones to make cardioid concepts popular. THD is usually pretty low from these, except for the very low bands <50Hz.

 

[markstein]

TTh

Would not really describe Genelec 8381A as a cardioid concept in the frequency bands which are sensitive to higher distortion.

Maybe some bigger MEG models could be added here, like RL901k or RL921k, as these were really the first ones to make cardioid concepts popular. THD is usually pretty low from these, except for the very low bands <50Hz.

That would be of great interest, especially regarding the RL901K.

Unfortunately, measurements for ME Geithain speakers are extremely rare. I am not aware of any spinorama dataset e.g.

 

[Curvature]

Would not really describe Genelec 8381A as a cardioid concept in the frequency bands which are sensitive to higher distortion.

Maybe some bigger MEG models could be added here, like RL901k or RL921k, as these were really the first ones to make cardioid concepts popular. THD is usually pretty low from these, except for the very low bands <50Hz.

I disagree regarding the 8381A. Even if the cardioid function is restricted to the bass, it is still worth making the comparison. Readers just have to keep the design in mind when viewing the measurements, which we don't have.

I should have remembered the MEG speakers. Like @markstein said, there are no comprehensive measurements, and I certainly don't count the manufacturer's smoothed H2 and H3 results as useful.

However, I did extract them for illustrative purposes.

Edit: Adding H2+H3 gives the following curve.

 

[Martijn Mensink]

Nice work, @Curvature.

Just a heads-up: please take into account that not all of those measurements are done the same way. For instance, some are referenced to the speaker’s output level, others to the input signal. That makes a difference, especially if there's a dynamic high-pass filter or compression involved. To compare meaningfully, you really want to match output levels.

Distortion is always an interesting topic, but also a bit tricky. Whether it's audible or actually a problem depends on a bunch of things: the type of distortion, how much there is, where in the frequency range it falls, and how much the distortion is masked by the signal itself. That masking can be quite significant with music.

Also worth noting that as you increase playback level, speaker distortion rises, but so does the ear’s own distortion. The ear's distortion tends to mask what the speaker is doing. Looking at it the other way around, we're most sensitive to distortion at low levels, but most good speakers have very little distortion there, particularly of the most annoying kinds. You can sometimes hear low levels of distortion with pure tones or specific test signals, but with music it's usually not an issue with well-designed speakers played at typical levels.

The 8c’s are built to sound great up to levels most people would call very loud. But there's always something louder. The 8c's trade a bit of cleanness at very high SPLs for overall performance. Most people who’ve heard them will say they play loud and clean for their size. But if you regularly need very high SPL, to be fair, there are better options.

As for the white finish, this is probably the closest available right now. A new model is coming later this year.

 

[Curvature]

Just a heads-up: please take into account that not all of those measurements are done the same way. For instance, some are referenced to the speaker’s output level, others to the input signal. That makes a difference, especially if there's a dynamic high-pass filter or compression involved. To compare meaningfully, you really want to match output levels.

These are not very accurate comparisons for those exact reasons and others. There is no replacing the original data and all of the contextual information that comes with. These charts are my is best effort at compiling what we have.

 

[Arindal]

Distortion is always an interesting topic, but also a bit tricky. Whether it's audible or actually a problem depends on a bunch of things: the type of distortion, how much there is, where in the frequency range it falls, and how much the distortion is masked by the signal itself.

Fully agree, it is pretty difficult to interpret THD charts or evolve useful predictions from them.

One additional aspect: As distortion is usually occurring in frequency bands where some sudden and elevated peaks in the music happen, it is also pretty important how fast distortion is increasing relative to the level, in which frequency bands it is elevated and if it is coincidental with compression or resonances. Higher THD might be completely harmless if is almost stays constant with increasing level, while low THD @96dB in a critical band might hide the fact that the driver in question will puke or compress if you add +2dB of input level.

A static distortion chart at 96dB SPL does say basically nothing. It might give some hints which frequency bands should be investigated using proper test material, but that is it.

Really appreciate the advantages of cardioids when it comes to room integration, but my experience with available models is they all have their ´weak octave´. If you push them too hard there, you notice the limits pretty quickly at higher SPL. With MEGs, this is usually the lowest band <50Hz.

I disagree regarding the 8381A. Even if the cardioid function is restricted to the bass, it is still worth making the comparison. Readers just have to keep the design in mind when viewing the measurements, which we don't have.

Did not mean to say it is no cardioid, but it does not have a cardioid causing any typical weakness in a relevant frequency band.

 

[roland{at}GGNTKT]

As @Martijn Mensink said, it's not easy to compare different measurement methods and setups. In our case, even the speakers (software) preset can make a big difference. I can contribute THD for Model M1 (@85 dB) and Model M3 (@90 dB), all made in an anechoic chamber using WinMF (we use both, Klippel NFS and "real" mesurements).

The question remains, what is audible with music. There was a nice test on the Klippel website, but I think they removed it...

 

[Curvature]

As @Martijn Mensink said, it's not easy to compare different measurement methods and setups. In our case, even the speakers (software) preset can make a big difference. I can contribute THD for Model M1 (@85 dB) and Model M3 (@90 dB), all made in an anechoic chamber using WinMF (we use both, Klippel NFS and "real" mesurements).

The question remains, what is audible with music. There was a nice test on the Klippel website, but I think they removed it...

I tried but don't think I can extract the results of H2-H5. At minimum I would need to extract each curve individually and them sum them before subtracting from FR for a relative THD value. The distortion measurements are a little too complicated. I appreciate the you posting the charts nonetheless. Even eyeballing them is useful.

 

[neutrio]

@Curvature
Thanks for your effort. It's an interesting work.

Your graph can give us the trend of THD vs frequencies.

However this can't be a fair absolute picture without the exact same conditions, hardware, room, placement...