Jump to content

Active, DSP, REW, Xilica 101


rplace

Recommended Posts

9 minutes ago, Chris A said:

If you select a single ramping filter on a channel, like a high shelf or a low shelf, you'll notice that the phase doesn't return to zero.  This is also the effect of crossover filters--they shift phase between channels (which is usually not desirable).  That's why I have advocated PEQs instead of ramping filters (and REW's EQ facility doesn't handle ramping filters, either). 

 

Every answer yields 10 more questions. When I started down the zero order cross over I did it with PEQs. I thought the idea was go get rid of the named (Bessel, LR, Butterworth) crossovers for phase reasons discussed in numerous posts. Then I became aware of shelf filters in the Xilixa from @Rudy81. I decided to give that a go for no other reason than it was another thing to learn about on my Xilica. *IF* I am understanding you correctly the shelf filters mess with my phase which is what we were trying to avoid. Is that correct?

 

Some screen captures to follow. One using Shelf filters 1 low and 2 high. The other uses only PEQs. I tried to look at the individual wrap of phase for the HF and LF and pick a point where they overlapped and were within in 90 degrees....but I'm still trying to figure this all out.

 

Side Note, attempts to cross around 500-800 (Shelf and PEQ) always seem to give worse Phase/GD than attempts around 200-350 but I get a better spectrogram with the 500-800 range. Is it just a matter of striking a balance?

 

Should I give my "shelved" preset a switch over to PEQs?

 

Here is the relevant screen shots for the Shelf Filtering to attenuate:

Shelf1.thumb.JPG.d4d9e7ed38cb1d0c95ef00ad22d56cd0.JPG

 

Shelf2.JPG.aaca30ed5ed1cd5c6b43cb036b0d1e91.JPG

 

 

And here is the relevant screen shots for the PEQ attenuation:PEQ1.thumb.JPG.34ff5199c0f53451d3da419eb9c5c00b.JPG

 

PEQ2.thumb.JPG.a264f839d73a2c54ab70dab601b82211.JPG

Link to comment
Share on other sites

7 minutes ago, Chris A said:

While you're here, can you recall an instance where you used an all-pass to correct phase?  Or a ramping filter...

 

Yes, allpass compensation is necessary in any Linkwitz-Riley crossover having more than two passbands. See response from Greg Berchin (who is actually me) here: https://www.dsprelated.com/showthread/comp.dsp/112896-1.php

 

As for a ramping filter, not sure what you mean. Is that commonly called a shelving filter? If so, then I use them occasionally in place of peaked-highpass filters to boost bass, because they have better phase response.

  • Thanks 1
Link to comment
Share on other sites

1 hour ago, rplace said:

*IF* I am understanding you correctly the shelf filters mess with my phase which is what we were trying to avoid. Is that correct?

Yes, if you're looking for phase flattening.

 

1 hour ago, rplace said:

Is it just a matter of striking a balance?

That's been my experience.

 

1 hour ago, rplace said:

Should I give my "shelved" preset a switch over to PEQs?

Try them both ways and look as the resulting phase response overall.  I've seen instances where it might be advantageous to use shelving filters (which I referred to as ramping filters above because they really do ramp instead of creating shelves).

 

Chris

  • Thanks 1
Link to comment
Share on other sites

Please help me understand what is going on here. Back on page 1 I think we concluded that with all the phase wraps around 6K and above I needed some more treatments around the speaker being measured, correct? So today I moved some treatments around, moved the HF section of the speaker slightly and moved the microphone a bit. No real method to the movements just trying to get different readings at that high end where all the "problems" seem to be. I'm going to post 4 sets of measurements with quite dramatically different phase/GD/SGrams. 

 

My gut tells me if I was to listen to these 4 sets of measurements they would sound a lot alike. Do you agree? After all, no change in PEQs, Gain, etc. just small movements of mic/speaker/absorption.

 

Here is all 4 on top of each other. We can see the only real changes happen above 4K. In the next few posts I'll put the Phase, Group Delay and Spectrogram for each and hopefully we can all comment on what is going on and if it is even a concern.

 

AllSPL.jpg.3b77680f50c3f99e54c172243ab09890.jpg

Link to comment
Share on other sites

Here are the 4 phase plots. I was initially trying to get the minimum number of wraps as well as get it as flat as possible. After getting my "best phase" I was trying to get a better Spectrogram, which I did, but at the expense of GD and Phase.

 

Phase1.jpg.6d5747a2254b60f7558449fbaf851e7c.jpg

 

 

 

Phase2.jpg.89eee1526bdc541168ce5ab63be4ac5f.jpg

 

Phase3.jpg.66223dbe1555de4b52feba469a064535.jpg

 

Phase4.jpg.3804cb7443ee2bccc0fc959b8a258e50.jpg

 

 

Link to comment
Share on other sites

Now you can see with the same order and same colors of graphs how the Group Delay got progressively worse as the Spectrogram which I'll show in the next post below got better. At least I think the SGram got better. You please tell me.

 

Humps start to show up around the same area the phase wraps increase above.

 

GD1.jpg.05cc414acafef82859b1ac637c635f8f.jpg

 

GD2.jpg.c6853db78db9931bd163907d181795e3.jpg

 

GD3.jpg.15f90dce5589d25f11e5bc77b4ef6356.jpg

 

GD4.jpg.66d0538548959059b8aa5711f75328a7.jpg

 

 

Link to comment
Share on other sites

Above the phase and GD got progressively worse, I think. But below you will see the Spectrograms get better. Why is this and does it really matter? If we are only making small changes like a small mic move or adding treatments temporarily for the sake of measurements and no real changes to PEQs, Gain, XO points, etc. By the time you add a second speaker and room issues are we really making things better or worse?

 

Couple of questions about Spectrograms. Is it desirable to have less Red as seen on page 1 of the AMT driver? I think @Chris A referred to that as smearing. Is smearing the width of the red color or the widening overall of any colors? My 2nd and 3rd SGrams below have very little red, but those shifts around 4 K visually tell me something is wrong but is it really? Can is just be that the reflections we identified earlier are making REW draw the dotted line in the wrong place and in the last SGram the actual sound, not the reflections are being tracked. If I am way off base here why the "problems" in the first three not having a totally vertical line near the top of the spectrum?

 

SG1.jpg.17f69d6f9fc4c621ee6daa00641e40fd.jpg

 

SG2.jpg.204571630c6f4c51ad18cd9ff574a0f5.jpg

 

SG3.jpg.910e747c37f82d51a581ee69bd8ab6d4.jpg

 

SG4.jpg.dfd7b7089d3c20716d6230fcfdda943e.jpg

 

 

 

 

Link to comment
Share on other sites

1 hour ago, rplace said:

My gut tells me if I was to listen to these 4 sets of measurements they would sound a lot alike. Do you agree? After all, no change in PEQs, Gain, etc. just small movements of mic/speaker/absorption.

I believe that most of what you're seeing that's difficult to explain, above, is really related to small movements in the microphone position, not substantive, audible differences.  If you look at the spectrogram you posted above, much can be explained:

 

SG4.jpg

 

If you look at the complexity of the spectrogram between 1.5 kHz and 8-10 kHz, you can see that small changes in the relative SPL vs. time due to lobing of the polar output of the driver will strongly affect where you place the time of maximum energy (i.e., phase), and therefore will even more affect the group delay (the frequency-based derivative of phase).  So phase and group delay tend to be noisy measurements based on what you can see in the spectrogram. 

 

When in doubt to what's happening with regard to phase and group delay like this particular instance, I look at the spectrogram, because it tells you the time-based story with respect to relative SPL and frequency, especially if it is complex (as it is in this case).

 

Chris

  • Thanks 1
Link to comment
Share on other sites

57 minutes ago, Chris A said:

believe that most of what you're seeing that's difficult to explain, above, is really related to small movements in the microphone position, not substantive, audible differences

 

So if I quit one day with what I feel is a good combination of settings, with say a 500 Bessel XO. Then I come back a week later to tackle a 800 PEQ only (zeroth order) set up and I've done nothing to my system and get different phase/GD/SGram at those high frequencies before I change anything, should I just ignore it and move the mic around a bit till I'm in the ballpark of where I was? I've got painters tape on the ground for both speakers and mic stand. Measure with a tape measure to the same distance for the mic but there is no way to be 100% repeatable. It always seems like the 20-4K range is pretty consistent but around 4-6K I can often get different measurements. I've started to ignore the rapid phase wraps up high blaming reflections of the room or the horn itself. Is that reasonable?

 

1 hour ago, Chris A said:

When in doubt to what's happening with regard to phase and group delay like this particular instance, I look at the spectrogram

 

Does the left legend going from 20 to 70 and blue to red represent SPL? Meaning that large triangle of green around 6-7ms from 400-4K is just that ~50dBs of sound in that frequency 6ms after the computer sent the sound to my speakers and the mic measured it?

 

Would the perfect Spectrogram be a vertical line with what ever color of SPL your pushing out around it in a small capsule? No "volcano" shape top to bottom no matter the color? 

 

SGTriangle.thumb.JPG.afd1f8102a9efa6cdd6c6158f4d884ca.JPG

Link to comment
Share on other sites

1 hour ago, rplace said:

should I just ignore it and move the mic around a bit till I'm in the ballpark of where I was?

Unless you've got an anechoic chamber standing by for your tests, the solution I've found is to mark the location of the microphone stand, and the position of the microphone on the stand is not moved (i.e., the microphone and stand in my service stays assembled and ready to use off out of the way in-room at ear height), setting up to place the microphone well within an inch of where I had it in prior measurements on prior occasions.  I have done this many times successfully.

 

1 hour ago, rplace said:

I've got painters tape on the ground for both speakers and mic stand. Measure with a tape measure to the same distance for the mic but there is no way to be 100% repeatable. It always seems like the 20-4K range is pretty consistent but around 4-6K I can often get different measurements.

This may be due to the HF drivers/horns that you own.  I have no issues if I get within the inch tolerance that I mentioned above, but the K-402 horns have very consistent coverage and well behaved time-based behavior by comparison. 

 

1 hour ago, rplace said:

I've started to ignore the rapid phase wraps up high blaming reflections of the room or the horn itself. Is that reasonable?

Yes, because you can verify nothing really changed in the spectrogram plot.

 

1 hour ago, rplace said:

Does the left legend going from 20 to 70 and blue to red represent SPL? Meaning that large triangle of green around 6-7ms from 400-4K is just that ~50dBs of sound in that frequency 6ms after the computer sent the sound to my speakers and the mic measured it?

Yes.  The scale is dBSPL (i.e., no correction for "A" or "C" weightings). 

 

1 hour ago, rplace said:

Would the perfect Spectrogram be a vertical line with what ever color of SPL your pushing out around it in a small capsule

Not a spike, but rather a spire with concave sides, because the period of the sound waves is increasing logarithmically as you traverse downward on the spectrogram plot toward lower frequencies.

 

Chris

  • Like 1
  • Thanks 1
Link to comment
Share on other sites

5 hours ago, rplace said:

 

My gut tells me if I was to listen to these 4 sets of measurements they would sound a lot alike. Do you agree? After all, no change in PEQs, Gain, etc. just small movements of mic/speaker/absorption.

 

 

 

This has been my experience.  Move the mic 6" and all bets are off.  The whole measurement effort changes.  I have a 5-way system and it is painstaking.

 

I resorted to only measuring 3' from the horns long ago.  Placing the mic anywhere out in the room yielded inconsistent results.........and settings derived from out in the room did not sound anywhere near as good as the near field measurement/settings plus some minor tweaks using the graphic on the Xilica from the listening position.  I have had satisfaction doing this.

  • Like 1
Link to comment
Share on other sites

On ‎12‎/‎10‎/‎2019 at 9:42 PM, mark1101 said:

 

I resorted to only measuring 3' from the horns long ago

 

I've assumed, perhaps incorrectly, this is the norm around here. In case it is not obvious all my measurements are 1M/3'. I measure there and listen at the listening position. My comments/observations were more to hopefully drive discussion and learning.

 

Just for the sake of argument, lets say I have a great Spectrogram, nice and vertical. But I have a lot of phase wraps between 6-10K and everything else WRT phase is perfect. I move the mic an inch or two up/down or left/right. Now my phase wraps at 6K are better but the Spectrogram has a hiccup along the way. Nothing else changed. The sound at LP will be what it is, because nothing on the DSP side changed.

 

So which of the two are the correct measurement? I can repeat either the next day by putting the mic in just about either spot. This has to be a mic/room/speaker interaction not a change in what the speaker is actually reproducing. I'd like to fix "everything" and get a perfect SGram and Phase....maybe not possible with my gear, knowledge, experience.

Link to comment
Share on other sites

2 hours ago, rplace said:

 

I've assumed, perhaps incorrectly, this is the norm around here. In case it is not obvious all my measurements are 1M/3'. I measure there and listen at the listening position. My comments/observations were more to hopefully drive discussion and learning.

 

Just for the sake of argument, lets say I have a great Spectrogram, nice and vertical. But I have a lot of phase wraps between 6-10K and everything else WRT phase is perfect. I move the mic an inch or two up/down or left/right. Now my phase wraps at 6K are better but the Spectrogram has a hiccup along the way. Nothing else changed. The sound at LP will be what it is, because nothing on the DSP side changed.

 

So which of the two are the correct measurement? I can repeat either the next day by putting the mic in just about either spot. This has to be a mic/room/speaker interaction not a change in what the speaker is actually reproducing. I'd like to fix "everything" and get a perfect SGram and Phase....maybe not possible with my gear, knowledge, experience.

 

That's when I have a beer and start listening to the music.  Close enough for me.  I don't sweat that stuff.  If I hear something wrong I make changes.

 

Move the mic around the room and look how crazy it gets.  :)  So what's right and what's wrong?  There comes a point where you have to use your ears and operate your equipment to get its best sound.  Audio is both engineering and artistry..............start using the artistry part of audio.  Measurements are good but not everything.  Make it sound the way you like.

 

  • Like 2
Link to comment
Share on other sites

One thing you could do is reduce the delay on the 402 horn until you can detect it.  In other words figure out your threshold of hearing on these settings.  How little a change can you hear?  The answer will tell you when you are splitting hairs.  This is another value of the tools.  Figuring out when you are close enough.  There is a lot of reflected sound in your room.

 

I have not done this due to time.  I may play with stuff like this over the holidays as I have a nice vacation coming.

 

When the delays are close enough, what I always found most difficult and what mattered most in the sound to me was the EQ on the 402. I could easily hear being off by as little as a db in the 1-5K range (listening to music) and would make fractions of a db change in the eq sometimes.  When Klipsch settings are not available (such as Xilica/TAD) I always aimed for the response curve of the Jub that is posted on this forum.  I printed out a few copies and keep them handy.  I speak about this as in the past because I don't fool with it very much and when I set things up and check it........it usually hasn't changed.

  • Thanks 1
Link to comment
Share on other sites

I haven't followed this thread closely, but reading an occasional post I see many mentions of "phase wrap" and efforts to eliminate it. Understand that phase wraps are an artifact of the way we typically express phase angle -- between -180° and +180°. When a phase angle passes through -180° on its way to -181°, it wraps-around to +179°. The same thing happens in the other direction, too. It is not indicative of a problem. It is just an indication that the phase angle is increasing or decreasing with frequency.

 

Understand also that time delay causes phase shift. Imagine a perfect loudspeaker with perfectly flat phase response, and a perfect microphone measuring it. Move the microphone farther away by exactly one wavelength at 20kHz (about 5/8 of an inch), and you will see the phase change from a flat line to a downward sloped line, passing through -180° at 10kHz then wrapping around and returning to 0° at 20kHz. Move it by a total of two wavelengths at 20kHz, and you will see the phase pass through -180° at 5kHz, 0° at 10kHz, -180° at 15kHz, and back to 0° at 20kHz. So some of the phase wraps being observed are caused by the measurement apparatus -- the microphone isn't in exactly the same place "today" as it was "yesterday".

  • Thanks 2
Link to comment
Share on other sites

15 hours ago, mark1101 said:

 

That's when I have a beer and start listening to the music.  Close enough for me.  I don't sweat that stuff.  If I hear something wrong I make changes.

 

Move the mic around the room and look how crazy it gets.  :)  So what's right and what's wrong?  There comes a point where you have to use your ears and operate your equipment to get its best sound.  Audio is both engineering and artistry..............start using the artistry part of audio.  Measurements are good but not everything.  Make it sound the way you like.

 

 

Looks like the best advice I've seen on any topic in a while.

 

I'd say "use the measurements to get a handle on what you're hearing instead of making what you're hearing being the result of measurements" if that makes any sense.

Link to comment
Share on other sites

21 hours ago, rplace said:

Just for the sake of argument, lets say I have a great Spectrogram, nice and vertical. But I have a lot of phase wraps between 6-10K and everything else WRT phase is perfect. I move the mic an inch or two up/down or left/right. Now my phase wraps at 6K are better but the Spectrogram has a hiccup along the way. Nothing else changed. The sound at LP will be what it is, because nothing on the DSP side changed.

 

So which of the two are the correct measurement? I can repeat either the next day by putting the mic in just about either spot. This has to be a mic/room/speaker interaction not a change in what the speaker is actually reproducing. I'd like to fix "everything" and get a perfect SGram and Phase....maybe not possible with my gear, knowledge, experience.

What you're describing is the limitations of how phase/group delay is measured.  If you look closely at your spectrogram posted above...

 

1541633003_Richsspectrogrammarkup.thumb.JPG.ddfd91ee9c968e8daab9e5fff2aa86f6.JPG

 

In the 1-6 kHz region, you actually see two different times that there is high SPL: one at close to 0 ms and another as much as 0.8-0.9 ms behind the first event (depending on frequency, shown in the vertical direction).  If you slightly change anything about how the measurement is done in that frequency band, you will see dotted line move due to this dual-impulse event.  The way REW measures phase and takes the derivative of the phase vs. frequency curve to obtain the group delay vs. frequency is highly sensitive to this dual impulse nature of this particular driver/horn. 

 

Most drivers/horns don't have a secondary pulse that's anywhere near the SPL as the first impulse, so REW's calculation of phase/group delay will produce the same results regardless of small changes in lobing or microphone position, and you will see a consistent phase/group delay response from measurement to measurement.  The driver response shown above is ambiguous in terms of phase and group delay because there are actually two events occurring, the second of which is almost as loud as the first, and in fact does overtake the initial pulse of SPL depending on microphone placement, showing the three dimensional polar lobing issue of the driver/horn combination.

 

Also note that REW calculates phase in a left-to-right, bottom to top fashion as shown in the spectrogram plot.  If there is any disruption or upset in the calculation of phase vs. frequency, REW will simply add  360 degrees of phase lead in order to "back up" to a phase lag as the frequency increases during a measurement calculation issue, and you see that a lot in the REW measurements of phase.  The phase seems to jump forward by 360 degrees as the frequency increases when you have an issue with a measurement.  This is why I recommend using a lot of absorption in and around the driver/horn areas--to reduce nearfield early reflections as much as possible.  This results in much better phase vs. frequency curves and group delay vs. frequency curves (an inherently noisy calculation at any time). 

 

So to answer your question: which is correct phase?  None of them, because the notion of a single phase curve is not possible when you have two impulses occurring, as your spectrogram shows.  And small changes in anything during the measurement will cause the phase/group delay calculations to move around, sometimes dramatically.

 

The reason why I recommend looking at the spectrogram is because it doesn't really change much, even though the phase and group delay calculations might be moving around a lot.  The spectrogram plot will tell you what the issue is: in this case, that you've actually got two phase events, and REW can't always decide which one is the correct one vs. frequency.

 

Chris

Link to comment
Share on other sites

18 hours ago, Edgar said:

Understand that phase wraps are an artifact of the way we typically express phase angle -- between -180° and +180°. When a phase angle passes through -180° on its way to -181°, it wraps-around to +179°. The same thing happens in the other direction, too. It is not indicative of a problem. It is just an indication that the phase angle is increasing or decreasing with frequency.

 

Interesting, and very informative. Thanks! I thought any time I saw the quick "wrap" or vertical jump from -180 to +180 that was in fact a "problem" I needed to address if I could.

 

Are we saying that we need to know more about the two phase plots below (orange and green) to determine which one is better, or maybe that neither is necessarily better than the other? I was under the impression that less wraps = better. Note that there is nothing I was able to do PEQ or XO location to remove that one in the green plot around 2500K. Unlike the last yellow plot where I was able to play with the amount of PEQ, and move the XO point around to get the 225 one to go away.

 

1WrapPhase.jpg.6eb4169e8cb4dd561bd05e4698eeac5e.jpg

 

ExtraWrapPhase.jpg.642c26b076d4788d114ed82212187929.jpg

 

PEQWrap.jpg.106d4c929734b00bf04f7192bc0b7abe.jpg

Link to comment
Share on other sites

5 hours ago, glens said:

Looks like the best advice I've seen on any topic in a while.

 

I'd say "use the measurements to get a handle on what you're hearing instead of making what you're hearing being the result of measurements" if that makes any sense.

 

I've been trying to tell myself that on a regular basis. I did spend a lot of time listening this weekend and I'm liking what I am hearing. Flip side is I seem to have the personality that is constantly saying but what if it could be a tiny bit better.

 

3 hours ago, Chris A said:

The way REW measures phase and takes the derivative of the phase vs. frequency curve to obtain the group delay vs. frequency is highly sensitive to this dual impulse nature of this particular driver/horn. 

 

Most drivers/horns don't have a secondary pulse that's anywhere near the SPL as the first impulse, so REW's calculation of phase/group delay will produce the same results regardless of small changes in lobing or microphone position, and you will see a consistent phase/group delay response from measurement to measurement

 

Okay, more good info. When we discussed this a week or so back I came away with the impression that the extra energy we saw in the Spectrogram that you have circled was a reflection of the sound. Meaning it was not coming from the actual driver/horn but the back and or side wall and arriving  .8 to .9ms later. I must have misunderstood your explanation.

 

3 hours ago, Chris A said:

So to answer your question: which is correct phase?  None of them, because the notion of a single phase curve is not possible when you have two impulses occurring, as your spectrogram shows.

 

Ahhhh, light bulb moment :emotion-55: it was not a binary question I was asking. I was thinking at one point I had the mic "right" and the other "wrong" but could not determine which was correct with such a small change.

 

FWIW, the sound was never terrible back in September/October. Even then it was very good. I'm happy with the improvements and changes I'm hearing. I'm also having a bit of fun learning something new. But as I'm sure you can all tell, I don't know what I don't know. And for every question answered there are several more that pop up.

 

Thanks again to everyone for all the help/support.

 

1 hour ago, glens said:

As a refresher, what are the horn/drivers we're seeing here?

 

Oris 150 horns with Audio Nirvana Super 8" Neodymium drivers. 2-way system, the bass bins are 15" open baffle H frames.

 

Horn Info

Driver Info - Terrible web site you have to hunt around a bit to get info

 

rps20191207_150923.thumb.jpg.0dea1372bdb12143f29df2ad84960eca.jpg

Link to comment
Share on other sites

Join the conversation

You can post now and register later. If you have an account, sign in now to post with your account.
Note: Your post will require moderator approval before it will be visible.

Guest
Reply to this topic...

×   Pasted as rich text.   Paste as plain text instead

  Only 75 emoji are allowed.

×   Your link has been automatically embedded.   Display as a link instead

×   Your previous content has been restored.   Clear editor

×   You cannot paste images directly. Upload or insert images from URL.

×
×
  • Create New...