Jump to content

Luke's Gjallarhorn/Othorn Discussion


lukeamdman

Recommended Posts

Here's what I posted at AVS:

 

_________________________________________

 

Ryan came over for almost 3 hours last night, and we spent more than half of that time taking measurements and discussing various methods for getting accurate results.
 
What Ryan was telling me is that the direct radiating sound from the speaker will always reach your ears first, and after that we'll hear any reflections.  That's hard to argue with, since the shortest path is a straight line, which reflections most certainly are not. 
 
At a very high level, his method, and please Ryan step in if I get any of this wrong, is to measure outdoors at 3-4ft~, get the response as flat as possible, and then leave the treble (1khz+) alone when in the room.  His reason is that our ears/brain mostly care about the direct radiating sound we hear first, while a microphone can't ignore that it hears absolutely everything in incredible detail.  The reflections the microphone pick up impact the response/sound in way humans ears do not.  
 
My approach was to measure from the seats where I'd be listening from, and doing this I got the sound dialed in pretty well.  However, on a few tracks here and there, there was "something" that I couldn't identify that had some room for improvement.  It sounded to me like a peak, and sometimes a null, was somewhere I needed to tackle, but the response didn't show any.  
 
I was skeptical of his idea, but this morning I put it to the test the best I could.  I can't move these behemoths outside, so I moved the mic in to about 2ft and used 8ms gating.  I got the response as flat as I reasonably could, and then played some of my reference material.  Immediately I thought it sounded better.  Every track, every clip.  I went back and forth from the old EQ settings to the new dozens of times, and I liked the new settings better every time no matter what I played.
 
Here's something else I noticed with the old settings.  The more I got the response to look good  from the MLP (center back row), the worse it sounded on the front row.
 
With the new settings, the only thing that seems to sound different is the bass/mid-bass from 30-300hz (I have a huge null at 30hz on the front row with the subs on and even with the main in full range and the subs off).  The treble now sounds great whether sitting in the front or back.    
 
Here's current EQ'd response at 2ft gated to 8ms
 
v11_zpseekn6aqa.jpg
 
Here's EQ'd vs raw:
 
v11vsnone_zpsgfaeueq8.jpg
 
 
 
I've also never experienced a stereo image this sharp/dead center/downright powerful. It's stunning.
Link to comment
Share on other sites

  • Replies 195
  • Created
  • Last Reply

 

Here's what I posted at AVS:

 

_________________________________________

 

Ryan came over for almost 3 hours last night, and we spent more than half of that time taking measurements and discussing various methods for getting accurate results.
 
What Ryan was telling me is that the direct radiating sound from the speaker will always reach your ears first, and after that we'll hear any reflections.  That's hard to argue with, since the shortest path is a straight line, which reflections most certainly are not. 
 
At a very high level, his method, and please Ryan step in if I get any of this wrong, is to measure outdoors at 3-4ft~, get the response as flat as possible, and then leave the treble (1khz+) alone when in the room.  His reason is that our ears/brain mostly care about the direct radiating sound we hear first, while a microphone can't ignore that it hears absolutely everything in incredible detail.  The reflections the microphone pick up impact the response/sound in way humans ears do not.  
 
My approach was to measure from the seats where I'd be listening from, and doing this I got the sound dialed in pretty well.  However, on a few tracks here and there, there was "something" that I couldn't identify that had some room for improvement.  It sounded to me like a peak, and sometimes a null, was somewhere I needed to tackle, but the response didn't show any.  
 
I was skeptical of his idea, but this morning I put it to the test the best I could.  I can't move these behemoths outside, so I moved the mic in to about 2ft and used 8ms gating.  I got the response as flat as I reasonably could, and then played some of my reference material.  Immediately I thought it sounded better.  Every track, every clip.  I went back and forth from the old EQ settings to the new dozens of times, and I liked the new settings better every time no matter what I played.
 
Here's something else I noticed with the old settings.  The more I got the response to look good  from the MLP (center back row), the worse it sounded on the front row.
 
With the new settings, the only thing that seems to sound different is the bass/mid-bass from 30-300hz (I have a huge null at 30hz on the front row with the subs on and even with the main in full range and the subs off).  The treble now sounds great whether sitting in the front or back.    
 
Here's current EQ'd response at 2ft gated to 8ms
 
v11_zpseekn6aqa.jpg
 
Here's EQ'd vs raw:
 
v11vsnone_zpsgfaeueq8.jpg
 
 
 
I've also never experienced a stereo image this sharp/dead center/downright powerful. It's stunning.

 

I've always preferred no eq over any eq when taken from the MLP. Its amazing how we can lock into one persons dialog even in a room full of loud music or many people talking and block out the rest. To a certain extent, I guess its the same with direct vs reflected sound. I never heard of this method of close mic'd EQing used in a home theater environment. Thanks for sharing.

Link to comment
Share on other sites

Generally speaking, yes you want to EQ the speaker and not the room for high frequencies.  The reason is that any room-induced variation in high frequencies will occur substantially over small distances because the wavelengths are very small.  A 8kHz wave is about 1.7 inches, so even ignoring head position effects, the frequency response at each ear can be totally different.  At the same time, the brain uses information from both ears to assess the sound, so it hears more of the speaker despite the room.

 

Likewise, using measurements from multiple locations clustered around the MLP can make EQ work much better, even when the measurements are gated and taken near field.  I suspect that this is even more important for horns.  However with horns, one must be more careful in the choice of locations and weighting to avoid EQing things too bright at the MLP.  I'm finding that horns are even weirder than that.  Often, one sees a big on-axis  hole that appears in the top octave.  If this is EQed to anywhere near flat, the power response at those frequencies will be off the charts.  Because of individual differences, some listeners will be very irritated by it with certain content, whereas other listeners may not noticing anything wrong.  One of my cats has unusually sensitive hearing in the 17 kHz+ range and complains very strongly to me if I play certain music.   :(   Given what I have learned, My next EQ will roll that off more aggressively.

Link to comment
Share on other sites

I've found much the same things about EQing the treble and upper midrange. Typically with in room measurements you will see a few dips in the overall treble and a general downward tilt towards 20Khz. I have found that if you attempt to equalize for flat at the listening position, even with a very short gate time, or just fill in any holes in the response, it will become very bright sounding and fatiguing long term. Often leaving the speaker response natural will sound much better despite how ugly the measurement at the seats may look on paper, assuming the speaker is relatively smooth to begin with. If anything I'll only use small adjustments of a few dB maximum and very low Q or broadband if something needs to be adjusted.

Link to comment
Share on other sites

Some thoughts based on the above comments, but nothing specific.

 

When Luke and I are talking about eq'ing his speaker, we're talking about using PEQ in the dsp to achieve the speaker design objectives. These objectives include, for me at least, flat axial anechoic response. To get flat axial anechoic response Luke has to measure in the relative nearfield (about 2ft) and use the software to block out the first reflection in the impulse response. Or he would have to take it outside or something different.

 

The idea of using eq at the seats above 300 to say 500hz is a complex issue. Most people love to throw out the "well, I listen in my seat, not an anechoic chamber, so I measure at my seat". What they're failing to consider is that a typical measurement microphone and REW are measuring the steady state resultant of sound pressure at the tip of the microphone. Our ear/brain uses a much more complex mechanism to determine what information is in the sound field. This includes the shape of the outter ear, the length and diameter of the ear canal, the time and intensity difference between the ears, and much more. The shape of the outter ear is the easiest to understand. Take for example, a sound coming from behind you. The ear will alter the sound by diffracting off the rear of your ear and leave a very defined diffraction signature in the response. This cues the brain to recognize the sound is coming from behind. An omnidirectional microphone however has no way of providing that information (true, the mics we use are not perfectly omni, and the top octave could roughly reveal the sound is coming from off axis, but not nearly as precise or useful). This example of sound direction is just one example of many well understood ways the ear/brain combo hears these sounds differently than a microphone.

 

I should note that I'm not one of these people who is saying measurements are useless, we don't hear via mics, we hear with our ears. Those people are equally wrong for different reasons.

 

Last point about this. Below about 300hz, and certainly below 80hz, use all the LP room eq desired. This region is dominated by the LP frequency response. The size of the ear and distance between the ears simply can't decipher nuanced information at wavelengths that long. Depending on the size of your head, we're into 1/8 WL sizes. Even compared to the room, reflections are starting to show up at the ear on the same beat/phase as the direct sound. So this is why we can change our approach.

 

PS. I'm no expert in this area, take it with a grain of salt. I'm relying on the work of people who have actually done the experiments and are in the know. I'm an engineer, not a scientist. Engineers hijack the work of scientists to solve engineering problems. :D

 

PSS. All just random thoughts, nothing specific to Luke's speaker.

Link to comment
Share on other sites

Some thoughts based on the above comments, but nothing specific.

 

When Luke and I are talking about eq'ing his speaker, we're talking about using PEQ in the dsp to achieve the speaker design objectives. These objectives include, for me at least, flat axial anechoic response. To get flat axial anechoic response Luke has to measure in the relative nearfield (about 2ft) and use the software to block out the first reflection in the impulse response. Or he would have to take it outside or something different.

 

The idea of using eq at the seats above 300 to say 500hz is a complex issue. Most people love to throw out the "well, I listen in my seat, not an anechoic chamber, so I measure at my seat". What they're failing to consider is that a typical measurement microphone and REW are measuring the steady state resultant of sound pressure at the tip of the microphone. Our ear/brain uses a much more complex mechanism to determine what information is in the sound field. This includes the shape of the outter ear, the length and diameter of the ear canal, the time and intensity difference between the ears, and much more. The shape of the outter ear is the easiest to understand. Take for example, a sound coming from behind you. The ear will alter the sound by diffracting off the rear of your ear and leave a very defined diffraction signature in the response. This cues the brain to recognize the sound is coming from behind. An omnidirectional microphone however has no way of providing that information (true, the mics we use are not perfectly omni, and the top octave could roughly reveal the sound is coming from off axis, but not nearly as precise or useful). This example of sound direction is just one example of many well understood ways the ear/brain combo hears these sounds differently than a microphone.

 

I should note that I'm not one of these people who is saying measurements are useless, we don't hear via mics, we hear with our ears. Those people are equally wrong for different reasons.

 

Last point about this. Below about 300hz, and certainly below 80hz, use all the LP room eq desired. This region is dominated by the LP frequency response. The size of the ear and distance between the ears simply can't decipher nuanced information at wavelengths that long. Depending on the size of your head, we're into 1/8 WL sizes. Even compared to the room, reflections are starting to show up at the ear on the same beat/phase as the direct sound. So this is why we can change our approach.

 

PS. I'm no expert in this area, take it with a grain of salt. I'm relying on the work of people who have actually done the experiments and are in the know. I'm an engineer, not a scientist. Engineers hijack the work of scientists to solve engineering problems. :D

 

PSS. All just random thoughts, nothing specific to Luke's speaker.

 

I just know that subjectively I prefer no EQ'ing from the MLP above 500hz!  :)

 

Also, Mike is coming over tonight and I'm going to give him a blind test to see which he prefers.  It's not like the differences are night and day and it sounds like a completely different speaker or anything, but I'm interested to see what he thinks of the subtle changes. 

Link to comment
Share on other sites

It's not like the differences are night and day and it sounds like a completely different speaker or anything,

 

Yup. The LP measurements do provide a close approximation, so we'd expect to be fairly good using LP measurements. And also why many people enjoy their setups using LP eq.

Link to comment
Share on other sites

I just know that subjectively I prefer no EQ'ing from the MLP above 500hz!  :)

 

FWIW I used to prefer it without but now I prefer it with, the difference is I changed the EQ I use. Moral of the story ... never rule it out (as the technology available evolves)  :) It will be interesting to see whether this holds true once I switch to my own seos/ae setup though.

Link to comment
Share on other sites

I am using FIR filters for EQ now, and I am getting outstanding (subjective) results using EQ up high that's remarkably fine grained.  My speakers are the Hsu Research HC-1 with horns.  They exhibit a fair amount of fine-grained variation (or ripple) in their native response, much of which is consistent between different angles.  While the SEOS horns may exhibit less variation like this due to its low-diffraction, extended round-out shape, I've seen similar ripple in its measurements.  With my room treatments, the steady response at the seats is almost identical to near field and/or gated measurements.  Subjectively, the high frequency response is far from anechoic as the diffusion adds a lot of late arriving high frequency energy that is not really seen by the mic.  IIRC, I applied 1/24th octave smoothing to the filters as a sanity check but otherwise relied on a large number of measurements to distinguish between room and speaker issues.  The result is subjectively superior to both the native sound and that using Audyssey MultEQ XT.

 

It's not perfect.  One of my tweeters has a nasty break-up resonance that appears as a high Q dip and peak at around 7 kHz and again at around 12.5 kHz.  Unfortunately, my current EQ tries to fix the dip and ends up exaggerating the breakup instead.  When I first discovered it, I thought I had found a glitch in my AVR because the anomaly appears to be acausal, but after further investigation, I narrowed the problem down to the speaker and determined (using multiple sweeps at different levels) that the problem is highly non-linear as well.  This fact completely contradicts the use of any linear correction methods.  More recently, I've learned that such break-up resonances at high frequencies are quite common with compression drivers, even some of the more expensive ones, so I may be fortunate that my other tweeters (using titanium diaphragms) don't appear to exhibit this, at least at the 75 dB or so I use for my sweeps.  As such, I'll be looking into ways of excluding correction of certain narrow frequency regions that are manually specified to be break-up areas.  It may also be possible to automate detection of these trouble areas by analyzing sweeps at multiple sound levels and looking for non-linearity.

 

I could try to get the tweeter replaced as I imagine it would be quite inexpensive, but I'm actually looking into upgrading to a DIY SEOS-based solution.  My current speakers also have some unwanted crossover behaviour, and my calculations indicate that I can't adequately solve this even using a fully active FIR filtered solution.   The crossover is simply too high for the required C-to-C spacing.  I can at least say that these speakers are designed very well considering their size, but without a larger horn and lower crossover as offered by the SEOS, the vertical coverage is necessarily very constrained.  A big challenge for this build will be to find the right compression driver offering extension from 1 to 20 kHz and as little break-up as possible, without the cost of esoteric materials like beryllium.  I'll take all the good luck I can get.

Link to comment
Share on other sites

FWIW I used to prefer it without but now I prefer it with, the difference is I changed the EQ I use. Moral of the story ... never rule it out :)

 

Good point.  

 

I always trust my ears above anything else, and I'm willing to test anything to find out what my preference is.  In this case I was surprised by the result.  

Link to comment
Share on other sites

 

 

We eventually stripped the tweeter of all dsp and took a nearfield (about 0.5m) with a 8ms gate time. This was cause we kept hearing some glare in the sweeps that we narrowed down to the 2-3khz range. We saw that there's a very high Q shelf at 3500hz about 7db tall

Tux, how do you put a gate on the mic?  Luke, did you happen to save the LP before EQ and after EQ after your last tweak session with Tux?  I see your nearfield before and after here in post 126 http://data-bass.ipbhost.com/index.php?/topic/314-lukes-gjallarhornothorn-discussion/?p=9037

 

Why did you guys choose to not address the dip at ~14k?  Tux, did you use the close mic, gated mic technique specifically because Luke has compression drivers or do you do this for regular dome tweeters as well? 

Link to comment
Share on other sites

Tux, how do you put a gate on the mic? 

 

Tux, did you use the close mic, gated mic technique specifically because Luke has compression drivers or do you do this for regular dome tweeters as well? 

not tux but will offer a reply anyway.

 

On the former, this is using the window controls in REW et al. There are no doubt a load of posts on this over the web but I attempted to write a laymans guide to what the window controls, using REW as a worked example, do in this post. Perhaps it will be useful though I'm not sure if you're asking what is meant by "a gate" or whether you're asking about the mechanics of how to do it using REW. If the latter, the actual options are available from the "IR Windows" button on the main toolbar. 

 

On the latter, this is the difference between correcting the speaker vs correcting the room (or correcting the rooms view of the speaker depending on how you look at it). 

Link to comment
Share on other sites

Tux, how do you put a gate on the mic?  Luke, did you happen to save the LP before EQ and after EQ after your last tweak session with Tux?  I see your nearfield before and after here in post 126 http://data-bass.ipbhost.com/index.php?/topic/314-lukes-gjallarhornothorn-discussion/?p=9037

 

Why did you guys choose to not address the dip at ~14k?  Tux, did you use the close mic, gated mic technique specifically because Luke has compression drivers or do you do this for regular dome tweeters as well? 

 

Here's a good article I found on gating in REW:

 

http://www.minidsp.com/applications/acoustic-measurements/loudspeaker-measurements

 

 

 

New settings in red and the old settings in green (both at 2ft with 8ms gating)

 

beforeandafter_zpsmgbsftwy.jpg

 

 

Unfortunately since I last ran a measurement from the LP, all the furniture in the room has changed.  

Link to comment
Share on other sites

Tux, how do you put a gate on the mic?  Luke, did you happen to save the LP before EQ and after EQ after your last tweak session with Tux?  I see your nearfield before and after here in post 126 http://data-bass.ipbhost.com/index.php?/topic/314-lukes-gjallarhornothorn-discussion/?p=9037

 

Why did you guys choose to not address the dip at ~14k?  Tux, did you use the close mic, gated mic technique specifically because Luke has compression drivers or do you do this for regular dome tweeters as well? 

 

Seems like how to do it has been said. The term "gate" is confusing, because in PA usually that's a device that compresses noise less than a set SPL. This is something that limits the FFT length the software processes. It's built right into REW, HolmImpulse, Arta, SoundEasy, etc.

 

I think the 14khz dip might be a throat diffraction. When I moved the mic off axis, the null got deeper. Somewhere off axis I think it might fill in. So it's tricky to fix this. Based on the quick and dirty measurements I did there, I think it would be rather safe to fill it in, provided the DSP has the headroom. But I suggested Luke looks into the Geddes foam plug to try and clean it up sans dsp. Probably not that audible, but may provide a bit of "air" or something. It's important he gets it where he likes it in the critical region 500 to 8000 and then he'll probably check out the foam plug and dsp.

 

Just thinking about it, it could also be an artifact of the coaxial. Didn't think of that until now. There is a noticeable change in the throat profile about 1" from the CD exit, so the throat probably is responsible.

Link to comment
Share on other sites

Cool, I never noticed that you could customize the window time in REW.  I'll have to mess with that for sure.  Wow, that foam plug thing is wild.  Luke, are you going to get a couple and see what they're about?  Seems like a lot of complicated physics stuff goes on with the horn design in the higher registers. 

Link to comment
Share on other sites

Cool, I never noticed that you could customize the window time in REW.  I'll have to mess with that for sure.  Wow, that foam plug thing is wild.  Luke, are you going to get a couple and see what they're about?  Seems like a lot of complicated physics stuff goes on with the horn design in the higher registers. 

 

I'm going to make a couple of the foam plugs to see what they do.  Should be interesting.  

Link to comment
Share on other sites

Sorry I didn't make it down on Monday. Still planning on coming down tomorrow though. I am looking forward to seeing if I tell which one is which. Have you changed anything since you got the new furniture? Also, have you incorporated the center yet? Or, are you waiting until you finish the platform? Looking forward to tomorrow Luke!!

Link to comment
Share on other sites

Well jeepers...apparently the new EQ settings weren't very subtle to Mike.  I blind tested him and it literally took him about 3-4 seconds into the first song to say which he thought sounded better.  Needless to say we both like the new settings better.

Link to comment
Share on other sites

Archived

This topic is now archived and is closed to further replies.


×
×
  • Create New...