Announcement

Collapse
No announcement yet.

HiFiCompass measurements - How to interpret and what matters?

Collapse
X
 
  • Filter
  • Time
  • Show
Clear All
new posts

  • HiFiCompass measurements - How to interpret and what matters?

    There are questions regarding Step response, CSD cleanness, and the Energy Time Curve (ETC) and frequency response. What matters and makes something state of the art? Let's assume THD & IMD are equivalent.

    I'm not just interested in what's better but what's audible to human ears. Giving a correct weight to measurements and not seeking overkill.

    1) Step Response
    Click image for larger version

Name:	StepResponse2.png
Views:	445
Size:	109.6 KB
ID:	1483274
    Click image for larger version

Name:	StepResponse1.png
Views:	415
Size:	83.7 KB
ID:	1483275
    So here we have two different step responses. I have three questions regarding these two graphs.
    • Different drivers have a significantly greater peak at the beginning of the step response. What does this indicate? Dynamics?
    • For the slope of the decay is it better to to fall rapidly or at a steady pace as in the second graph?
    • The first graph decays mostly by 2ms and is pretty much dead by 3ms. While the second graph decays at 4ms. What does this indicate and how is it audible?
    2) Waterfall - CSD cleanness
    Click image for larger version

Name:	Waterfall2.png
Views:	416
Size:	416.1 KB
ID:	1483276
    Click image for larger version

Name:	Waterfall1.png
Views:	420
Size:	399.9 KB
ID:	1483277
    Let's ignore the breakups at 3.5khz and 9khz and assume they are properly dealt with in the crossover.
    • Notice how much of a difference there is in the cleanliness of the graph from 2khz and under. If we have state of the art HD & IMD what significance does this cleanliness play? How will the human ear interpret this?

  • #2
    3. Energy-Time Curve (ETC)
    Click image for larger version  Name:	ETC1.png Views:	0 Size:	96.9 KB ID:	1483279
    Click image for larger version  Name:	ETC2.png Views:	0 Size:	119.0 KB ID:	1483280
    Here are two state of the art midranges and their ETC graphs. Exceptionally clean waterfall graphs, identical step response graphs yet their Energy-Time Curve's could not be more different.
    • How much does the slope of decay matter? The second graph is quite consistent.
    • What is the significance of a quick decay. How does this translate to better sound to the human ear?
    4. Frequency response vs VoiceCoil test bench
    Click image for larger version  Name:	Wavecor2.png Views:	0 Size:	150.3 KB ID:	1483281
    Click image for larger version  Name:	Wavecor1.png Views:	0 Size:	43.2 KB ID:	1483282
    Here are the WF120BD03 and WF120DB04 drivers measured.Same driver different impedance
    • The Voicecoil testbench tends to have a lot of frequency response irregularities beneath 1000hz which do not exist on ZaphAudio or HifiCompass. Why is this?
    5. What measurement(s) should be prioritized if HD is already exceptionally low? If you had to pick between these three tweeters which would you pick and why? Let's assume all are crossed at 3k 4th order with an active crossover with DSP to correct for frequency response.Based on measurements alone Is the lowly BG neo3 really that close to the Raal? Is the Viawave closer to a dome than a ribbon to the human ear or every bit the equal to the Raal?

    Comment


    • #3
      To relate things to the subjective language often used: A drivers ability to start and stop making sound as and when that sound starts and stops per the signal provided is an important aspect of how clean/clear/tight/detailed (etc) it will sound re-producing music. Conceptually if a driver is still making a sound after the sound has actually stopped and a new sound is supposed to be started, then this will lead to what is often referred to as smearing/muddiness/lack of clarity (etc). It comes out in some reviews where the comment is made that a speaker tends to loose a bit of control/dynamics with more 'complex' material (its not the only reason of course, there many many other factors).

      To what degree this is audible/problematic vs not is probably a hot debate but you would be surprised at what the human auditory system is capable of detecting. I personally find the waterfall or step response is about all that I can be practically concerned with. However, its often after the fact as this data isn't that readily available, for the range of drivers on offer, to aid in the up front selection. If it is, great, I would just be avoiding anything that had some obvious standout issues like ringing etc.
      Constructions: Dayton+SB 2-Way v1 | Dayton+SB 2-Way v2 | Fabios (SB Monitors)
      Refurbs: KLH 2 | Rega Ela Mk1

      Comment


      • #4
        Step response of individual driver I find is quite useless. It is just a worse way to view frequency response. Smooth step response = smooth frequency response. Long decay of step response = low frequency extension. They are directly related.

        Step response can be beneficial for complete multi-driver system however, providing some insight to timing errors between drivers and filter orders. Here's an example:
        Click image for larger version

Name:	project3 Impulse, ETC, Step.png
Views:	385
Size:	93.1 KB
ID:	1483296

        What is "better" or "worse" mostly depends on your intended use, response on the intended baffle with crossover in place much more important. Of course long ringing energy storage needs to be worked on, and drivers with constant response slope, flat or not will be much easier to work with than anything with sharper lumps/shelves in the response.

        AudioXpress/Voice Coil doesn't measure on an infinite baffle, but rather small baffle usually. You will find the detail of this in the words of the AudioXpress reviews.
        "I just use off the shelf textbook filters designed for a resistor of 8 ohms with
        exactly a Fc 3K for both drivers, anybody can do it." -Xmax

        Comment


        • #5
          Great way to help understand step response and ETC of a single driver is to use VituixCAD. View menu - > impulse response. Then just check the "step" box to view the step response.

          1. Start with default flat response driver, you will see the impulse and step is perfect.
          2. Add in active high pass filter and low pass filter, view step response and you adjust the driver frequency range.
          3. add in PEQ bands to similate peaks and dips in the response.

          Click image for larger version

Name:	Screenshot_20220221_222042.png
Views:	384
Size:	217.0 KB
ID:	1483298
          "I just use off the shelf textbook filters designed for a resistor of 8 ohms with
          exactly a Fc 3K for both drivers, anybody can do it." -Xmax

          Comment


          • #6
            1. Step Response - Nothing more than frequency response. The longer the decay the greater the lower frequency capability.
            2. Waterfall graph - Not sure if a super clean CSD is a greater indication of a drivers ability to start and stop making sound.
            3. Energy Time Curve - Not sure if a quickly decaying ETC is a greater indication of a drivers ability to start and stop making sound. I would assume ETC is a much greater indicator.
            4. Frequency response irregularities - Result of Audioxpress measuring on a small baffle.

            Hard cone 3rd and 5th order harmonics
            I do have one additional question regarding suppression of metal & ceramic cone harmonic distortion in the critical 1kz to 3khz band. In the HifiCompass review of the Bliesma 3" aluminum and beryllium domes it stated that "At frequencies of 2.2 and 3.7 kHz for AlMg and 2.9 and 4.7 kHz for Be we can see bursts of the 3rd and 5th harmonics. These bursts are due to the voice coil current harmonics amplified by the main resonance of the membrane. With proper crossover design they can be significantly neutralized. This is exactly the case when passive loudspeaker filters can have advantages over active filtering."

            And here is a ceramic cone distortion graph with before and after filtering:
            Click image for larger version

Name:	Accuton harmonics.png
Views:	360
Size:	91.3 KB
ID:	1483387

            So looking at metal cones such as the Seas Excel or Dayton RS-180 can the 3rd and 5th order harmonics above 1.5khz be filtered away with a passive crossover and look as clean as a paper cone in this range? Or will the breakup region always rear its ugly head as harmonics no matter how well suppressed?
            Click image for larger version

Name:	SeasHexadym.png
Views:	336
Size:	253.9 KB
ID:	1483388

            Comment


            • #7
              Not just any passive filter, but properly designed to address the resonance. Keep checking the Purifi site blog/tech post section, I’m told they will be providing some design detail on specifically this topic as it applies to the recently released Aluminum cone models.
              "I just use off the shelf textbook filters designed for a resistor of 8 ohms with
              exactly a Fc 3K for both drivers, anybody can do it." -Xmax

              Comment


              • #8
                https://purifi-audio.com/wp-content/...otchfilter.pdf
                "I just use off the shelf textbook filters designed for a resistor of 8 ohms with
                exactly a Fc 3K for both drivers, anybody can do it." -Xmax

                Comment


                • #9
                  Simply out of curiosity about passive vs active suppression of the breakup modes. Why do passive filters have an advantage? Assuming one dials in an appropriate notch in DSP.

                  Edit: Nevermind. I needed to actually read the Purifi paper.
                  Francis

                  Comment


                  • #10
                    Should a driver that costs that much need that many crossover parts to make it usable?

                    Comment


                    • #11
                      Usable or optimized for best performance? I’m sure it’s “usable” with whatever basic filter you throw at it, but you may not experience all it has to offer. How much do those components cost in the grand scheme of a high end speaker? This article applies to all hard come drivers with sharp breakups peaks, not just this particular model from Purifi. Intention of the post was so that we can learn something from it and apply that knowledge to our own designs.
                      "I just use off the shelf textbook filters designed for a resistor of 8 ohms with
                      exactly a Fc 3K for both drivers, anybody can do it." -Xmax

                      Comment


                      • #12
                        Originally posted by dcibel View Post
                        Usable or optimized for best performance? I’m sure it’s “usable” with whatever basic filter you throw at it, but you may not experience all it has to offer. How much do those components cost in the grand scheme of a high end speaker? This article applies to all hard come drivers with sharp breakups peaks, not just this particular model from Purifi. Intention of the post was so that we can learn something from it and apply that knowledge to our own designs.
                        Yeah, the series vs parallel notch business is very interesting.
                        Francis

                        Comment


                        • #13
                          Originally posted by devnull View Post

                          Should a driver that costs that much need that many crossover parts to make it usable?
                          If you like what hard-cone drivers do, you're stuck with that sort of thing. Obviously, that's a matter of taste. I like the "see-through" clarity I get from my W18s.
                          Francis

                          Comment


                          • #14
                            If this works out so well for hard coned drivers why doesn't anyone do this for tweeters? Notch the breakup at 24-30khz, instead beryllium and diamond domes are chased.

                            Comment


                            • #15
                              Well, I'm not chasing Be or diamonds personally ;). There's the marketable factor, most consumer will understand difference in materials, but completely glaze over if you try to teach proper filter design... I see many Be tweeters also have a flatter dome, so overall dispersion characteristic is altered, and many will have arguments about other diaphragm sound characteristics that exist beyond simple material hardness and breakup node.

                              And yes, some people do notch the ultrasonic resonant mode in tweeters, and there's nothing preventing you from doing the same. Design some filters and measure / listen for yourself and make your own conclusions, that's the power of DIY.
                              "I just use off the shelf textbook filters designed for a resistor of 8 ohms with
                              exactly a Fc 3K for both drivers, anybody can do it." -Xmax

                              Comment

                              Working...
                              X