Flint: High Dynamics, Low Distortion, State of the Art Hi-Fi Custom Speakers (2018)

[Flint]

For the super audio geeks out there, here's the impulse response of these Statement 1 loudspeakers:


This was measured from the listening position, so there is more early reflection energy in the voltage curve than you'd see in a lab environment. That said, the overall impulse ends at 0.5mS and everything is damn near perfectly aligned, to within less than 0.1mS accuracy. The audible result is higher clarity and realism in reproduced music. Basically this is nearly as good a single driver system, but with the higher dynamics, lower distortion, and great bass impact of a large multi-way speaker system.

 

[TitaniumTroy]

http://www.vaporsound.com/speakers/perfect-storm/
I know a guy who had a speaker company, Vapor Audio. His top of the line model was my favorite speaker of the show at AXPONA a few years ago. Only cost $30k, then he got hooked up with a dealer network and I think it doubled. He had a personal tragedy happen, and I think has left the industry because of it.

He did comment on the forums that if he had to do it over again, he would have charged a lot more. He kinda he felt was working his ass off, but was not getting much financial return.

I think one of his main problems was the way he made his speaker cabinets, which was by slicing MDF into many pieces and then be CNC routed. Next they would be glued together to create a stiffer/better cabinet. This seemed super labor intensive, hence one reason for the longggggg time between ordering a speaker and delivery.

I met him a few times at the shows, really nice guy but didn't quite have the best business model.

 

[team_tsp]

I posted this chart in the Rocketman loudspeaker thread, but since there is a thread dedicated to these speakers, I wanted to put it here as well.

The chart below shows the measured distortion across the operating range of these "Statement" speakers. The measurement gear only measured distortion to about 10,000Hz, so I chopped off the higher frequencies.

View attachment 8672

If you convert the difference in dB between the primary tone (top curve) and the lower curves (harmonic components), the distortion above 100Hz remains below 0.1% up to the limits of the measurement gear. The one anomaly is the 3rd harmonic increase centered around 340Hz where the distortion reaches 0.17% - still incredibly low for any speaker generating that much output.

By the way, the simple math to convert the difference in dB to % can be based on these reference points: -20dB = 10%, -40dB = 1%, -60dB = 0.1%, and so on.

Speaking of which, this is the distortion present when the speaker is generating 88dB SPL at the listening position (which is about 96dB SPL at 1M).

In the bass, down to about 40Hz the distortion is well below 1% and only shifts above 1% below 40Hz at these SPLs. at 20Hz the distortion is 2.5% for the 2rd Harmonic (pleasing to the ears even order) and 1% for the 3rd Harmonic (harsh to the ears odd order).

This is a stunning performance! I don't think you will find many speakers in the world which can generate this level of output at such low THD figures. Of course, a large part of this is the Ellipticor midrange and tweeter drivers I chose which are available to anyone to build speaker from. So, I believe we will start seeing those drivers used in audiophile speakers soon.

Wow cool stuff! I've never seen a graph like this, so learning about how to convert difference in dB to distortion % is definitely interesting. I was wondering, how does the measurement device/software know that it's picking up on the 2nd and 3rd harmonics? (Also not to be a spelling bee, but note your "Hsrmonic" ) Does it actually "decipher" the harmonics from the fundamental tone being played at any given time?

For the super audio geeks out there, here's the impulse response of these Statement 1 loudspeakers:

View attachment 8675
This was measured from the listening position, so there is more early reflection energy in the voltage curve than you'd see in a lab environment. That said, the overall impulse ends at 0.5mS and everything is damn near perfectly aligned, to within less than 0.1mS accuracy. The audible result is higher clarity and realism in reproduced music. Basically this is nearly as good a single driver system, but with the higher dynamics, lower distortion, and great bass impact of a large multi-way speaker system.

If I'm interpreting your impulse response chart correctly, this is amazing to say the least. I'm fairly familiar with the step response graph that Stereophile usually includes in their speaker measurements, and as such a couple questions. For example, if we take the Dynaudio C1 (pre-Platinum) step response, we see that the tweeter's output starts first (at about 3.75ms), and then hands off to the woofer at about 4ms.

https://www.stereophile.com/content/dynaudio-confidence-c1-loudspeaker-measurements

Regarding your speakers:
1) It looks like the drivers in your speakers kick off at different times as well. Is that an intentional design principle, or is it just because the tweeter is smaller/lighter and moves first?
2) Are we able to discern which driver is moving at which points in your chart (just out of curiosity)?
3) It looks like there is some type of pre-ringing, or something similar, before 0 (zero) ms. Is that the early reflection energy you mentioned or something else?

But as you mentioned, an overall response of just 0.5ms is quite amazing, especially when you compare it 3+ms of the C1

 

[Flint]

First, on Harmonic distortion measurements:

The test software "looks" for the harmonics for the test tone being played through the speaker. If the tone is 400Hz, then it "seeks" tones at 800Hz, 1.2kHz, 1.6kHz, and 2kHz (2nd, 3rd, 4th, and 5th Harmonics) and records their SPL. The test tone is a medium speed sweep to give time to the CPU to do all that processing. As the pitch increases, the filters in the software increases which frequencies it is listening for.


Second, Step response and Impulse response are two different things, but they show similar information. A step response is better at showing the arrival time of a signal from a driver. An impulse response shows frequency dependent phase characteristics as well as arrival time. The step response of a theoretical "ideal" speaker will be a perfectly vertical upward spike which then tapers downward for about 2.5mS. That's a simple thing to see.

An Impulse response cannot show the energy in that way with a multiway speaker because it is essential showing the waveform of each driver, and since the rise time for a 10kHz sine wave is about 0.025 mS, which is the center frequency of the tweeter's range. The rise time for a 800Hz sine wave is about 0.3125 mS, which is the center frequency for the midrange driver. Meanwhile the rise time for the woofer's center frequency, which is about 100Hz, is 2.5mS. So, there are different rise times depending on the frequency.

With my speakers, the waveform is starting at the same point pretty much across the spectrum and it is gone within about 0.5mS, which suggests the arrival times are damn near ideal.

I don't have tools which can capture perfect step response measurements. The most affordable of those starts around $800 for the hardware, alone.

The pre-ringing is from the software interpreting the sweep wave. Basically, the test tone is a 5Hz to 20,000Hz sweep which lasts about 0.2 seconds, so the software have to compensate for the sweep time to interpret the data in absolute time. There will always be pre-ringing present in this sort of measurement technique. Whereas, a Step Response measurement is made by an pulse wave which energizes all the drivers simultaneously and then interprets the results. In the case of the chart you posted, the delay time from when the test equipment sent the pulse wave, or chirp, to the time it is amplified and it reaches the microphone is about 3.8mS. My software adjusts the impulse response to compensate for the delay in arrival time and uses the captured signal as a trigger for the location of 0mS.

It can get hairy, but Impulse Response charts don't look the same as a Step response chart.

 

[Flint]

I did some reading, in my Impulse Response measurements there is pre-ringing because the system only measures up to 22kHz and my speakers generate significant output to frequencies well above that. The higher frequencies are ignored in the FIR calculation to represent the chart, but they are still present. The greater the output levels in the speakers, the greater the pre-ringing will appear.

 

[team_tsp]

First, on Harmonic distortion measurements:

The test software "looks" for the harmonics for the test tone being played through the speaker. If the tone is 400Hz, then it "seeks" tones at 800Hz, 1.2kHz, 1.6kHz, and 2kHz (2nd, 3rd, 4th, and 5th Harmonics) and records their SPL. The test tone is a medium speed sweep to give time to the CPU to do all that processing. As the pitch increases, the filters in the software increases which frequencies it is listening for.


Second, Step response and Impulse response are two different things, but they show similar information. A step response is better at showing the arrival time of a signal from a driver. An impulse response shows frequency dependent phase characteristics as well as arrival time. The step response of a theoretical "ideal" speaker will be a perfectly vertical upward spike which then tapers downward for about 2.5mS. That's a simple thing to see.

An Impulse response cannot show the energy in that way with a multiway speaker because it is essential showing the waveform of each driver, and since the rise time for a 10kHz sine wave is about 0.025 mS, which is the center frequency of the tweeter's range. The rise time for a 800Hz sine wave is about 0.3125 mS, which is the center frequency for the midrange driver. Meanwhile the rise time for the woofer's center frequency, which is about 100Hz, is 2.5mS. So, there are different rise times depending on the frequency.

With my speakers, the waveform is starting at the same point pretty much across the spectrum and it is gone within about 0.5mS, which suggests the arrival times are damn near ideal.

I don't have tools which can capture perfect step response measurements. The most affordable of those starts around $800 for the hardware, alone.

The pre-ringing is from the software interpreting the sweep wave. Basically, the test tone is a 5Hz to 20,000Hz sweep which lasts about 0.2 seconds, so the software have to compensate for the sweep time to interpret the data in absolute time. There will always be pre-ringing present in this sort of measurement technique. Whereas, a Step Response measurement is made by an pulse wave which energizes all the drivers simultaneously and then interprets the results. In the case of the chart you posted, the delay time from when the test equipment sent the pulse wave, or chirp, to the time it is amplified and it reaches the microphone is about 3.8mS. My software adjusts the impulse response to compensate for the delay in arrival time and uses the captured signal as a trigger for the location of 0mS.

It can get hairy, but Impulse Response charts don't look the same as a Step response chart.

I did some reading, in my Impulse Response measurements there is pre-ringing because the system only measures up to 22kHz and my speakers generate significant output to frequencies well above that. The higher frequencies are ignored in the FIR calculation to represent the chart, but they are still present. The greater the output levels in the speakers, the greater the pre-ringing will appear.

Gotcha, I didn't realize there was a difference between impulse and step response measurements. I slightly edited my previous post to consistently refer to Stereophile's chart as a step response.

Do both step and impulse response experiments send a full range signal (of some sort) to the speakers? That's interesting about the pre-ringing, so is that due to frequencies above 22kHz, or greater output levels, or both?

 

[Flint]

Step response tests send a chirp or DC signal. Impulse response uses a fullrange sweep.

 

[Flint]

I posted this chart in the Rocketman loudspeaker thread, but since there is a thread dedicated to these speakers, I wanted to put it here as well.

The chart below shows the measured distortion across the operating range of these "Statement" speakers. The measurement gear only measured distortion to about 10,000Hz, so I chopped off the higher frequencies.

View attachment 8672

If you convert the difference in dB between the primary tone (top curve) and the lower curves (harmonic components), the distortion above 100Hz remains below 0.1% up to the limits of the measurement gear. The one anomaly is the 3rd harmonic increase centered around 340Hz where the distortion reaches 0.17% - still incredibly low for any speaker generating that much output.

By the way, the simple math to convert the difference in dB to % can be based on these reference points: -20dB = 10%, -40dB = 1%, -60dB = 0.1%, and so on.

Speaking of which, this is the distortion present when the speaker is generating 88dB SPL at the listening position (which is about 96dB SPL at 1M).

In the bass, down to about 40Hz the distortion is well below 1% and only shifts above 1% below 40Hz at these SPLs. at 20Hz the distortion is 2.5% for the 2rd Harmonic (pleasing to the ears even order) and 1% for the 3rd Harmonic (harsh to the ears odd order).

This is a stunning performance! I don't think you will find many speakers in the world which can generate this level of output at such low THD figures. Of course, a large part of this is the Ellipticor midrange and tweeter drivers I chose which are available to anyone to build speaker from. So, I believe we will start seeing those drivers used in audiophile speakers soon.

I should correct my statements...

The data I shared on these amazing speakers' distortion measurements are not "total" harmonic distortion. Instead they show 2nd and 3rd harmonic distortion, which is still stunningly low. If you combine the the 2nd and 3rd harmonic levels, you get a higher combined level. Add the 4th and 5th harmonics and it is still higher. So, the real THD is higher than my claimed 0.02%.

Using my software I cannot manipulate the data after capturing it, so I need to rerun the measurement process to get the 2nd thru 5th harmonics and combine them to get the majority of the harmonic distortions as a percentage.