How to measure high frequency resolution?

[ppataki]

This question came up last year in a different thread and I thought to create a new one dedicated to it

I have been building many (close to 20 now) DIY projects in the last 5 years and one recurring topic was how to measure resolution in the high frequencies
I was mainly focusing on projects full range drivers and those tend to have a really bad treble response - one can EQ it but of course it will still not sound the same as having a dedicated tweeter - the "resolution" will be lower (I can't think of a better word)

The above is quite logical and obvious but still how could we measure that?

Somebody gave the idea that distortion in high frequencies shall give a clue about it
I think I can confirm that fully with my measurements!

These measurements were taken in the same room in the same position (MLP, approx. 240-250cm from the speakers) during the last 5 years and as projects they were all previously documented here on ASR

I am attaching the mdat file too but let me quickly explain what you see

Distortion (THD) measurements from 1kHz to 10kHz for the below speakers:

- Markaudio MAOP 11
- SB Acoustics SB20FRPC30-8
- Audio Nirvana 12" Classic Ferrite
- Murphy Corner Line Array using Dayton ND91 drivers (48 pcs)
- Fane 15-300TC
- Adam Audio T5V
- Adam Audio T5V converted to sealed with enhanced cabinet bracing and acoustic enhancements
- Sica 5,5 C 1,5 CP 5.5" coax driver

(all these projects can be found here on ASR, fully documented)

These measurements confirm what I and others heard when listening to these speakers - resolution and high-frequency 'information' was better for those with lower distortion measurements
This year I am planning to try a few more drivers - will report on those too from this perspective

Any comments, questions, please don't hesitate - I would love to hear experience from others in this topic

 

[staticV3]

This stepping pattern in your measurements looks unlike any THD measurement I've seen so far and in need of investigation - especially since it happens across multiple speakers.

Personally, I'd search for clues of "resolution" using frequency response, multitone distortion, and Null tests if the environment permits.

REW's new Cross Correlation feature could be useful here.

 

[Keith_W]

I do not think that "high frequency resolution" is an objective term, or even a subjective term that is highly correlated to an objective phenomenon, such as "sibilance" or "boominess". So my issue is that I am not sure what you mean.

Perhaps you may be referring to how bright a speaker sounds? Speakers with a less pronounced treble tilt (i.e. brighter sounding speakers) may seem to have more "resolution".

Perhaps it may have something to do with time smearing of the high frequencies? For example, phase distortion, high frequency ringing, or early arrival of high frequency reflections?

Perhaps the tweeter is not time aligned with the rest of the speaker?

 

[ppataki]

So my issue is that I am not sure what you mean.

What I mean is for example when listening to certain instruments with high frequency content, cymbals, etc. I hear more details with for example the Adam Audio T5Vs (with their ribbon tweeter) compared to whichever full range speaker of mine, or even the Sica coaxes for that matter - and they were all DSP-d so the phenomenon is not attributable to their out-of-the-box frequency response

 

[Keith_W]

I see. I just realised that you uploaded the .mdat so I had a quick look. I am not sure how much you can read into these measurements given that the noise floor is so high:

This one is for MAOP 11.

And this one is for T5V reflex.

It makes me wonder whether your neighbour was mowing his lawn or something. The noise floor varies quite a bit between the measurements.

I should also point out that distortion should be measured with your microphone 1m from the speaker with an SPL meter (in dBC) showing 75dB with pink noise. If the speakers are not outputting the same SPL, then some of them will show more distortion than others.

 

[Sokel]

This stepping pattern in your measurements looks unlike any THD measurement I've seen so far and in need of investigation - especially since it happens across multiple speakers.

Personally, I'd search for clues of "resolution" using frequency response, multitone distortion, and Null tests if the environment permits.

REW's new Cross Correlation feature could be useful here.

It's the "use harmonic freq as ref" option used probably,that's how the 2 two with the most prominent effect looks like without it:

 

[DWPress]

As far as speaker drivers and distortion goes - a single speaker just can't do it. A 5+ inch driver will typically reproduce content well between 300-13Khz but ask it to do 300 and 13K signals simultaneously and the lower vibrations on the cone interfere with the smaller vibrations needed for the HF content and distortion ensues. Speakers loose their linearity at a certain point and that's when directivity and off axis response begin to suffer while still spraying out noise = distortion plus other undesirable artifacts.

For all the benefits of point source speakers single driver or coax, a good 3-way can still shine in many of the measurements by allowing multiple drivers to do what they do best in a limited frequency range AND achieve a higher SPL without distortion.

 

[ppataki]

It's the "use harmonic freq as ref" option used probably,that's how the 2 two with the most prominent effect looks like without it:

View attachment 425647

You are right, the stepping pattern is gone now, which is great (the differences are still there)

It makes me wonder whether your neighbour was mowing his lawn or something. The noise floor varies quite a bit between the measurements.

Yes, agreed - I was also scratching my head about that.... all the measurements were taken with absolutely zero (external or internal) noise sources - I have always been ensuring that ever since I have done measurements (started around 12 years ago). So I have really no idea what caused those weird noise curves.

If the speakers are not outputting the same SPL, then some of them will show more distortion than others.

Unfortunately all these measurements were gathered from my archives from the last 5 years or so and they are not exactly level matched - so I agree it might not be the the best exercise in this shape and form

However I need to add that the MAOP 11 and the Sica are indeed level matched so we can still take a look at those
The Fane 15", the Audio Nirvana 12" and the SB 8" are also pretty well aligned until about 1kHz - the differences above 1kHz are due to the fact of them being full range drivers and they have screwed up response either here or there (or pretty much everywhere for that matter )

 

[AnalogSteph]

I was mainly focusing on projects full range drivers and those tend to have a really bad treble response - one can EQ it but of course it will still not sound the same as having a dedicated tweeter - the "resolution" will be lower (I can't think of a better word)

Well... yes you can EQ it, but that doesn't mean the treble dispersion isn't wonky when the driver breaks up. Comparing on-axis with MMM measurements may yield some surprises. Arguably you even need breakup; pure pistonic action would generally result in diffuse sound becoming rather dark - but obviously breakup tends to be a bit messy and chaotic, so smooth dispersion is anything but guaranteed.

Also, as mentioned, IMD absolutely is a thing in fullranges. Not only is Doppler distortion a potential issue, they tend to have limited excursion to begin with. By the time they are big enough, you'll be running into treble extension and dispersion issues. (So kinda like traditional coaxes but even worse.) Short of going line source with a bunch of BMRs or something, you're quite screwed. And this is why fullranges aren't mainstream. People are happy when they can get away with a mere 2 ways thanks to the power of waveguides and high-excursion drivers. (Case in point e.g. March Audio Sointuva assuming no assembly issues, DXT-MON or the trusty Behringer B2031s.)

 

[Avp1]

This question came up last year in a different thread and I thought to create a new one dedicated to it

I have been building many (close to 20 now) DIY projects in the last 5 years and one recurring topic was how to measure resolution in the high frequencies
I was mainly focusing on projects full range drivers and those tend to have a really bad treble response - one can EQ it but of course it will still not sound the same as having a dedicated tweeter - the "resolution" will be lower (I can't think of a better word)

The above is quite logical and obvious but still how could we measure that?

Somebody gave the idea that distortion in high frequencies shall give a clue about it
I think I can confirm that fully with my measurements!

These measurements were taken in the same room in the same position (MLP, approx. 240-250cm from the speakers) during the last 5 years and as projects they were all previously documented here on ASR

I am attaching the mdat file too but let me quickly explain what you see

Distortion (THD) measurements from 1kHz to 10kHz for the below speakers:

- Markaudio MAOP 11
- SB Acoustics SB20FRPC30-8
- Audio Nirvana 12" Classic Ferrite
- Murphy Corner Line Array using Dayton ND91 drivers (48 pcs)
- Fane 15-300TC
- Adam Audio T5V
- Adam Audio T5V converted to sealed with enhanced cabinet bracing and acoustic enhancements
- Sica 5,5 C 1,5 CP 5.5" coax driver

(all these projects can be found here on ASR, fully documented)

View attachment 425600


These measurements confirm what I and others heard when listening to these speakers - resolution and high-frequency 'information' was better for those with lower distortion measurements
This year I am planning to try a few more drivers - will report on those too from this perspective

Any comments, questions, please don't hesitate - I would love to hear experience from others in this topic

What was sound pressure level for your measurements? If it was 72.5dB, then all results are rather poor. You should do measurements at 90-95dB