Linear phase discussion

Discussion in 'Speakers' started by Serious, Jun 21, 2017.

  1. ultrabike

    ultrabike Measurbator - Admin

    Staff Member Pyrate MZR
    Joined:
    Sep 25, 2015
    Likes Received:
    8,960
    Trophy Points:
    113
    Location:
    Irvine CA
    I bet you can't tell exactly. In fact, I bet you got no clue where it is by just looking at the speaker's phase response. I can't tell shit either. Which is my point. There are other easier ways to determine that which do not involve the step response.

    Actually on the MR5 I think the tweeter is quite a bit behind, allegedly to align the responses.

    Really? Well we couldn't tell shit by looking at the phase response. I maybe pressing the matter here, but you sometimes say "I can't tell exactly" and the you say it "will make itself known in the phase response". You can't have both. The phase response is there and I personally can't tell shit. What's you approach to figure this out. What do I look for?

    It may have to do with a discontinuity in the frequency response. Possible the delay is not over multiple cycles or a full cycle even. I know that because the offset is about 360 degrees. IMO is a measurement artifact. But you can tell shit is not perfect. No way that's minimum phase either.

    Well that's just a bad example. The reason your phase is all jagged and all over the place is that the frequency response you took was not in anechoic or semi-anechoic conditions. There will be combing all over the place due to room interactions. I can't tell shit from that plot's phase response.

    In room measurements will behave like that. And the phase on in-room applications will be fucked all over. Because the speaker interacts with the room. Such is life.

    Your Superlux results are not like the KSC75 or a K812 results, because your Superlux is not a KSC75 or a K812.

    Almost every measurement I've seen on a KSC75 exhibits a severe null in the same area. It is these nulls that produce not only discontinuities in the magnitude, but the phase as well.

    Dynamic drivers are not minimum phase. You keep saying that, but those IR definitively exhibit zeros outside the unit circle. I've done this with my HD600. I'm not sure what else to tell you. By definition those headphones are not minimum phase.

    All drivers are pistonic. Planars and stats move a lot less, that's all. All drivers compress air volume mechanically. The larger the driver, the less it's got go move. Stats have a huge driver, so they almost don't move relative to a dynamic driver. Their theory of operation is also a little different. But you better believe it moves, or no sound will come off of it.

    The apparent excess phase likely comes from the discontinuity. IMO is a measurement artifact. Still, there is no way I'm calling that minimum phase.
     
  2. Serious

    Serious Inquisitive Frequency Response Plot

    Pyrate BWC MZR
    Joined:
    Sep 28, 2015
    Likes Received:
    2,594
    Trophy Points:
    113
    Location:
    near Munich, Germany
    No and it's not, which is why I was wondering why apparently sometimes nulls and peaks cause issues and sometimes the excess phase is still close to flat. Is it like with the speaker measurements where there might be a reflection that nulls the signal causing the excess phase?
    The same goes for the Superlux. Almost every measurement shows a severe null followed by a peak, but apparently in this case excess phase isn't affected.
    I meant the motion of the diaphragm. Technically every driver has modes, but depending on the material and stiffness there should be a region with (almost) no modes at all. In fact I think the HD800 has no breakup issues until 38.2kHz.

    I know, but you can still tell the overall phase of the speaker and see that it doesn't wrap. Excess group delay (disregarding the room effects) is very close to 0. But that's not the point at all.
    My point was that the phase response looks like that because of room effects, so I was trying to suggest that these discontinuities from the phase are maybe not the driver. Do you have free air measurements for those headphones? I still don't see why the excess phase looks like that. I see that it does in your plots, but why?
    I think it's most likely that that's just how the visualization works in this case. Without the unwrapping the excess phase would be close to flat before and after the dip. Look at the group delay. Is there some delay to the frequencies below the dip? I bet there isn't, which is really what I wanted to tell you with my example with the speaker measurements.

    Yes, the in-room measurements do that. But the phase of the speaker is still the phase of the speaker, whether it's in my bathroom or in an anechoic room. The phase may look terrible due to reflections and all sorts of room effects, but the underlying speaker response is still the same. You just can't get your data to show you that, because most likely even the most aggressive windowing will not remove all the reflections while giving you a good picture about the FR. This is why I measured my speakers at the edge of my garage, 3m above the ground, to get measurements that I could design my crossover with.

    The steeper the crossover, the larger the overall excess phase shift. The center of the excess phase shift should be the crossover frequency, or at least close to it. The step response clearly shows the delay and polarity of the drivers, so that definitely helps too with figuring out the design. You can't have a 2nd order crossover with both drivers in the same polarity, or if you do you'll get a null. Still, distance to the driver will influence the phase response. With a perfect design you should be able to look at the graphs and know the crossover. But in the real world things are never perfect. One may use a crossover with different slopes for the two drivers to account for some delay, etc.

    For example:
    http://www.troelsgravesen.dk/Ellam-Discovery-15.htm
    LR4 had -180° phase shift at the crossover frequency. It's not exact because this is the phase response, not the excess phase and FR isn't flat.
    LR2 was close to -90° at the crossover frequency.

    Or look at the discussion in the comments here:
    https://www.stereophile.com/content/wilson-audio-specialties-alexx-loudspeaker-measurements
    Once you know how the step response looks like and how many drivers there are you can assume what order the crossover is.

    I realized I read the phase wrong. 360° total phase shift, so 4th order crossover. Manual says 3kHz. Hmm, based on the plots I'd have expected it to be lower. Maybe there's some delay or something. I'm not a computer.
     
    Last edited: Jun 22, 2017
  3. ultrabike

    ultrabike Measurbator - Admin

    Staff Member Pyrate MZR
    Joined:
    Sep 25, 2015
    Likes Received:
    8,960
    Trophy Points:
    113
    Location:
    Irvine CA
    Many headphones exhibit peaks and nulls. Why? Because they do. Single driver or not. And that's going to poop all over the phase response.

    You are giving lessons on cross-over phase shifts, and how it's so obvious in the frequency phase response. We are on page two of this thread and you still have not showed me how I can figure out the cross-over frequency location and the filter order based on a semi-anechoic MR5MK3 measurement I made.

    There are no room interactions or crap combing on the data.

    Don't give me Troelsgravens and Stereophile articles to read. Tell me the crossover frequency and crossover filter order based on my MR5MK3 plots.
     
    Last edited: Jun 22, 2017
  4. Serious

    Serious Inquisitive Frequency Response Plot

    Pyrate BWC MZR
    Joined:
    Sep 28, 2015
    Likes Received:
    2,594
    Trophy Points:
    113
    Location:
    near Munich, Germany
    Yes. the question was not why, but why sometimes these peaks and nulls cause an excess phase shift and sometimes they don't. That the phase won't look nice should be obvious, but why is it sometimes still close to minimum-phase and sometimes it isn't?

    The room interactions don't even matter as you can still see the phase response, especially in the upper midrange. As I said, probably LR4 and my estimate was around 2.5-3kHz. Doesn't have to be more accurate than that.
     
  5. ultrabike

    ultrabike Measurbator - Admin

    Staff Member Pyrate MZR
    Joined:
    Sep 25, 2015
    Likes Received:
    8,960
    Trophy Points:
    113
    Location:
    Irvine CA
    @Serious, in the case of the headphones, the excess phase is IMO a measurement artifact. The phase jumps like hell due to the null. The algorithm that unwraps the phase is likely unable to distinguish between multiples of 360 degrees with discontinuous data. The unwrapping works probably well for situations where the phase is continuous. Otherwise all bets are off.

    As far as your guess on the crossover frequency for the MR5MK3, I think you are wrong. And as far as the filter, you are assuming it's an LR. I don't know. Could be a Bessel for all I care and those behave differently. Assuming your guess is right and it's an LR filter, I do not think it's an LR4.

    However, how did you came up with your estimates. Please explain.

    (BTW, I can tell you where I think the cross-over is, and what I think the filter is with some data and explain my rationale).
     
    Last edited: Jun 22, 2017
  6. Serious

    Serious Inquisitive Frequency Response Plot

    Pyrate BWC MZR
    Joined:
    Sep 28, 2015
    Likes Received:
    2,594
    Trophy Points:
    113
    Location:
    near Munich, Germany
    That's what I thought regarding the unwrapping. So you think the excess phase is a measurement artifact? So does this mean you do agree that headphones are close to minimum phase systems? Or just above and below the dip/peak?

    I'm no expert, really just a guess. Phase seems to shift 360°, so 4th order. LR because flat FR. Crossover frequency is probably different and it could all just be different.

    I'd be interested in your findings. What does the step response look like? The specs do seem to say 24db/octave crossover at 3kHz, though.
     
  7. ultrabike

    ultrabike Measurbator - Admin

    Staff Member Pyrate MZR
    Joined:
    Sep 25, 2015
    Likes Received:
    8,960
    Trophy Points:
    113
    Location:
    Irvine CA
    You could say headphones are mostly minimum phase systems across the spectrum.

    The phase shift argument IMO is not easy to follow because the drivers are not aligned. Again, the MR has the drivers offset allegedly to help with driver integration.

    However, if instead of following the phase plot, one follows the spec, well that's another story. You wouldn't be reading the phase. You would be reading the owners manual.

    The way I did it was with close mic response (go down to post #4):

    http://www.superbestaudiofriends.org/index.php?threads/mackie-mr5-mk3.2457/#post-64598

    I kind of said there around 3 kHz, but one could argue that it's more like 2 kHz. I could have just disengaged the tweeter, but I didn't do it. You (or the manual) are not far off. They do claim a 24dB/octave in the user's manual, which is LR4. But to me it almost looks like an LR2 and I'm not even sure about the woofer.

    Anyhow, in the end it seems to work out well. Reading integrated phase responses taken at 1 m, to me look like a difficult way to figure out all this cross-over stuff. Just do a close mic deal or something like that.
     
    Last edited: Jun 22, 2017
  8. ultrabike

    ultrabike Measurbator - Admin

    Staff Member Pyrate MZR
    Joined:
    Sep 25, 2015
    Likes Received:
    8,960
    Trophy Points:
    113
    Location:
    Irvine CA
    I'll see if I can give you the step response or the impulse response later tonite.

    BTW, I think a close mic measurement will give you an idea of the cross-overs. But I don't think it would give you an idea of how the phase response behaves with the finished product. For that you do indeed measure frequency response at tweeter level about 1 meter away (standard) and use as anechoic conditions as you can afford.

    I agree with you (and it makes sense) that there will be some offsets because of the two way design vs. say a single driver. But I don't think one can just eye ball the phase response and figure out the filter characteristics and all that stuff. At least I can't.
     
    Last edited: Jun 22, 2017
  9. Serious

    Serious Inquisitive Frequency Response Plot

    Pyrate BWC MZR
    Joined:
    Sep 28, 2015
    Likes Received:
    2,594
    Trophy Points:
    113
    Location:
    near Munich, Germany
    Lol, I read the owners manual after my wild assed guess. Nearfield measurement might play a role here, but technically if anything it should make the filters look steeper than they are, at least based on my experience. The tweeter does look like 24db/octave to me. And with a higher woofer level to get the FR to look more like what it is in the measurements, I'd say that looks more like 2.5kHz! Closer to what I guessed than what the manual says. But it could still be that the manual is right. The woofer ... still not sure what's going on there. The slopes are probably asymmetrical for a reason and at least the woofer seems to be well-behaved so no nasty breakup stuff.

    Lol, yes it is. My point mainly was that crossover filters/step response/IR/phase response/time delay/FR are all interrelated (which I know you know) and that you could sort of guess one from the other. Or probably use software to guess it for you, if you don't have any other data.
     
  10. ultrabike

    ultrabike Measurbator - Admin

    Staff Member Pyrate MZR
    Joined:
    Sep 25, 2015
    Likes Received:
    8,960
    Trophy Points:
    113
    Location:
    Irvine CA
    I take the above back to some extent.

    You may have something there with the high and low frequency 360 degrees deal. Would have to see more data points and comparisons between single and 2-way speakers. It would only be an eye balling, very rough order of magnitude kind of deal though.

    Again, one may need a lot more data to back that up.

    I bet there might be cases where it doesn't fully hold up. As in weird ass speaker boxes. Crazy Omni-directional designs. Line array type of deals... Or even simple 2-way systems were one driver and the other are not aligned, or even some crazy ass driver that has some weird behavior.
     
    Last edited: Jun 22, 2017

Share This Page