ProSoundWeb Community

Please login or register.

Login with username, password and session length
Advanced search  

Pages: [1] 2 3   Go Down

Author Topic: Thinking outside the greybox  (Read 18563 times)

Hayden J. Nebus

  • Hero Member
  • *****
  • Offline Offline
  • Posts: 637
  • Richmond, VA
Thinking outside the greybox
« on: March 16, 2012, 12:50:57 PM »

I'm working on liberating my microwedges from the confines of their prescribed processors. Below is a trace of the LF output of the "blue" setting on a UX8800. I'm curious about the filter @ 2k, specifically why one would bother putting it there, >40dB down. Is it just there to keep the phase trace from ticking upward rapidly, i.e. in the negative direction?

http://postimage.org/image/vxri8ss17/

Edit: link to image. size issue i think.
« Last Edit: March 16, 2012, 12:55:14 PM by Hayden J. Nebus »
Logged

Tom Young

  • Global Moderator
  • Hero Member
  • *****
  • Offline Offline
  • Posts: 620
Re: Thinking outside the greybox
« Reply #1 on: March 16, 2012, 03:05:55 PM »

I'm working on liberating my microwedges from the confines of their prescribed processors. Below is a trace of the LF output of the "blue" setting on a UX8800. I'm curious about the filter @ 2k, specifically why one would bother putting it there, >40dB down. Is it just there to keep the phase trace from ticking upward rapidly, i.e. in the negative direction?

http://postimage.org/image/vxri8ss17/

Edit: link to image. size issue i think.

Looks like the low-pass output to me.

Unless I am mistaken, the microwedge is 2-way active. What you have measured here appears to be the LF output of the processor. Nothing appears to be wrong/questionable about this.

There should be another output from the processor which, when you measure it, will look like the high-pass band and would "sit" nicely on top of this curve.
Logged
Tom Young
Electroacoustic Design Services
Oxford CT
203-888-6217

Hayden J. Nebus

  • Hero Member
  • *****
  • Offline Offline
  • Posts: 637
  • Richmond, VA
Re: Thinking outside the greybox
« Reply #2 on: March 16, 2012, 04:33:02 PM »

Looks like the low-pass output to me.

Unless I am mistaken, the microwedge is 2-way active. What you have measured here appears to be the LF output of the processor. Nothing appears to be wrong/questionable about this.

There should be another output from the processor which, when you measure it, will look like the high-pass band and would "sit" nicely on top of this curve.

Tom,
Yes, this is an electrical measurement of the low pass output. I'm using this measurement as a target transfer function for migrating processing. I am pretty sure there's a filter along the slope of the low pass. Magnitude takes a good elbow at 2k5, and the phase trace has a bell shaped wobble at about 3k that makes me think there's a filter in play there beyond just the LPF, else the phase trace would continue rising.

For instance, I'm trying to drive my MW12 with an MX8750. Below is the response I was able to get with the 5 output EQ bands available on the 8750, overlayed on the TF of the UX8800 greybox. The LPF is a 4th order Linkwitz. Note the change in phase slope introduced by the LPF alone, and the linear nature of the magnitude slope, vs. the greybox. Just trying to figure out what it's there for.
http://postimage.org/image/nctl3vshh/ 

Logged

Langston Holland

  • Jr. Member
  • **
  • Offline Offline
  • Posts: 82
Re: Thinking outside the greybox
« Reply #3 on: March 16, 2012, 06:06:25 PM »

Quote
I'm working on liberating my microwedges from the confines of their prescribed processors.

That brings to mind someone trying to liberate themselves from a rope halfway up a mountain. :)

Actually, I'm 100% in favor of experimentation. The instruction provided can be expensive sometimes, but it sure is fun.

Quote
Below is a trace of the LF output of the "blue" setting on a UX8800. I'm curious about the filter @ 2k, specifically why one would bother putting it there, >40dB down. Is it just there to keep the phase trace from ticking upward rapidly, i.e. in the negative direction?

http://postimage.org/image/vxri8ss17/

The 2.5kHz notch you're referring to is not very clear because you're using too much smoothing. Look at the trace again w/o smoothing. Generally, I'd disable smoothing and probably averaging for non-acoustic (electrical) measurements so you don't miss something.

This is part of a "notched crossover", which is a largely under-appreciated tool in loudspeaker crossover design. I don't know who was initially responsible for their use at EAW (I've got a strong hunch), but I started noticing them a few years after the great Neville Thiele brought to light a practical implementation of them in a 2000 AES paper. Read the abstract at least a few times until you have a feel for it. Buy and download the entire paper if you dare.

Loudspeaker Crossovers with Notched Responses

Also helpful is a reply he gave during an interview with Voice Coil magazine in 2006:

Interviewer:

Could you describe your innovative crossover filter design, the Neville Thiele Method?

Thiele:

I prefer to describe them as notched crossovers. They have a notch symmetrically in the high-pass and the low-pass responses, to achieve a very steep rolloff rate immediately outside the passband. Beyond the notch, the response rises again, but remains respectively low. They are especially useful when the amplitude, or more treacherously the phase, response of one or both drivers is poor outside their useful band.

Summary:

You can have a 6th order (36dB per octave) filter rolloff rate with the same effective time domain behavior as a 4th order (24dB per octave) filter. This can make a huge improvement in certain applications.

Quote
For instance, I'm trying to drive my MW12 with an MX8750. Below is the response I was able to get with the 5 output EQ bands available on the 8750, overlayed on the TF of the UX8800 greybox. The LPF is a 4th order Linkwitz. Note the change in phase slope introduced by the LPF alone, and the linear nature of the magnitude slope, vs. the greybox. Just trying to figure out what it's there for.

http://postimage.org/image/nctl3vshh/

You did a great job copying the low pass output for the MW12 using the MX processor. You've matched the phase throughout the important region of crossover, thus it will sum properly with the high pass of the UX. Now you'll need to repeat the process for the high pass in the MX. The UX messes with the time domain of the high pass in very nice ways that the MX can't, but this magic occurs above the crossover region with the MW12, thus you should be able to do what you're proposing.
« Last Edit: March 16, 2012, 06:10:17 PM by Langston Holland »
Logged
God bless you and your precious family - Langston

Langston Holland

  • Jr. Member
  • **
  • Offline Offline
  • Posts: 82
Notched Crossovers
« Reply #4 on: March 16, 2012, 06:55:42 PM »

Actually, my interpretation of this situation is wrong, though it probably shouldn't be.

EAW may have goofed, or missed an opportunity, or neither. :)

Background:

To explain why, we first need to see how a notched crossover works. Dr. Thiele outlined several classes of notched crossovers, but the one that is most useful has a 6th order magnitude response while exhibiting the phase behavior of a 4th order Linkwitz Riley filter. Seems too good to be true, but it isn't. The additional phase rotation of the 6th order filter class is pushed far enough into the cutoff region that it doesn't contribute to the loudspeaker system's overall phase rotation because the passband's output is virtually zero at that point.

Below is an example using of a 6th order 1kHz low pass (green) and a 4th order LR low pass (purple):



The notch filter's 3kHz bump at -36dB will contribute almost nothing to the high passband. Let's do a little Syn-Aud-Con decibel addition for correlated sound. Assume the high passband output is 100dB at 3kHz, that will make the low passband 100dB - 36dB = 64dB.



Substituting 100dB for L1 and 64dB for L2, we get a combined output of 100.14dB. That means if you fed the loudspeaker system a 3kHz tone and switched the low passband on and off, it wouldn't make any difference.

That makes the 6th order notched crossover an exciting thing IMO.

Conclusion:

Looking back at Hayden's low pass filter measurements of the UX and the overlapping magnitude of 4th order LR from the MX makes it clear that EAW simply added a notch filter to a 4th order LR low pass. In this case the UX is not employing a Thiele notched filter with its much higher overall rate of attenuation. Concerning the original question about the purpose of the notch filter, I can only think of two:

1. It was an attempt at the low pass portion of a Thiele notched crossover. (seems quite unlikely)
2. It was an attempt to correct for an acoustic magnitude and/or phase problem in the low passband at 2.5kHz. This would be worthless that far down the filter skirt's cutoff. (seems quite unlikely)

I'm going to measure one of my UX's tomorrow and verify that Hayden's measurements are accurate so that the trouble I get into is justified. They certainly look good.
« Last Edit: July 19, 2012, 03:01:11 AM by Langston Holland »
Logged
God bless you and your precious family - Langston

Hayden J. Nebus

  • Hero Member
  • *****
  • Offline Offline
  • Posts: 637
  • Richmond, VA
Re: Thinking outside the greybox
« Reply #5 on: March 16, 2012, 07:08:32 PM »

That brings to mind someone trying to liberate themselves from a rope halfway up a mountain. :)

Actually, I'm 100% in favor of experimentation. The instruction provided can be expensive sometimes, but it sure is fun.

The 2.5kHz notch you're referring to is not very clear because you're using too much smoothing. Look at the trace again w/o smoothing. Generally, I'd disable smoothing and probably averaging for non-acoustic (electrical) measurements so you don't miss something.

This is part of a "notched crossover", which is a largely under-appreciated tool in loudspeaker crossover design. I don't know who was initially responsible for their use at EAW (I've got a strong hunch), but I started noticing them a few years after the great Neville Thiele brought to light a practical implementation of them in a 2000 AES paper. Read the abstract at least a few times until you have a feel for it. Buy and download the entire paper if you dare.

Loudspeaker Crossovers with Notched Responses

Also helpful is a reply he gave during an interview with Voice Coil magazine in 2006:

Interviewer:

Could you describe your innovative crossover filter design, the Neville Thiele Method?

Thiele:

I prefer to describe them as notched crossovers. They have a notch symmetrically in the high-pass and the low-pass responses, to achieve a very steep rolloff rate immediately outside the passband. Beyond the notch, the response rises again, but remains respectively low. They are especially useful when the amplitude, or more treacherously the phase, response of one or both drivers is poor outside their useful band.

Summary:

You can have a 6th order (36dB per octave) filter rolloff rate with the same effective time domain behavior as a 4th order (24dB per octave) filter. This can make a huge improvement in certain applications.

You did a great job copying the low pass output for the MW12 using the MX processor. You've matched the phase throughout the important region of crossover, thus it will sum properly with the high pass of the UX. Now you'll need to repeat the process for the high pass in the MX. The UX messes with the time domain of the high pass in very nice ways that the MX can't, but this magic occurs above the crossover region with the MW12, thus you should be able to do what you're proposing.

Thanks for the info. I look forward to your revelation later.  here's what I was able to put together this afternoon. green traces are the 8750 and blue traces are the 8800 "blue" greybox.  smoothing is off.

Mod Edit: removed missing image placeholder.
« Last Edit: March 24, 2012, 03:49:11 PM by Langston Holland »
Logged

Hayden J. Nebus

  • Hero Member
  • *****
  • Offline Offline
  • Posts: 637
  • Richmond, VA
Re: Notched Crossovers
« Reply #6 on: March 16, 2012, 09:02:42 PM »

Actually, my interpretation of this situation is wrong, though it probably shouldn't be...

Got to get to a son's soccer game - I'll post later tonight. VERY interesting revelation on the notched crossover coming up. EAW may have goofed, or missed an opportunity, or neither. :)

The concept makes sense and is clever. We've probably all done something similar (one-sided, with lower order filters) on a channel strip to thin a boomy vocal.

The greybox TF does not show behavior like that, so I see what you're saying. Whatever the filter is, it's not very deep, nor is there one symetrically opposed on the other band.

Logged

Langston Holland

  • Jr. Member
  • **
  • Offline Offline
  • Posts: 82
Re: Thinking outside the greybox
« Reply #7 on: March 16, 2012, 11:01:55 PM »

Quote
...here's what I was able to put together this afternoon. green traces are the 8750 and blue traces are the 8800 "blue" greybox.  smoothing is off.

Turning smoothing off helped. Additional detail in the notch area can be had using a standard 16K (or higher) FFT, though for acoustic measurements you'll want to stick with Rational's 48th octave MTW.

I do not understand why the phase traces overlap on the high pass transfer function for the non-focused MX and focused UX outputs. The traces should overlap with the low pass since it probably doesn't include the FIR coefficient based time correction (focusing).

I've got a couple of UX8800's at the shop that I'll measure to tomorrow with the MW12 preset and see what the deal is. TBC.
Logged
God bless you and your precious family - Langston

Iain.Macdonald

  • Full Member
  • ***
  • Offline Offline
  • Posts: 213
Re: Thinking outside the greybox
« Reply #8 on: March 17, 2012, 05:22:30 PM »

For those who had a "brain event" when reading Langston's reply on Theile filters, which are similar to  elliptical filters, check this Wikipedia link, which might help. These may, or may not, have linear phase designed in. Also the grey box might have some tricks incorporated. The OP really needs to measure the box/driver to get a better understanding of what's going on. Polar measurements before and after, are a must! 

Quote from: EAW Manual Pg 26
Often, audio system designers, technicians, or end-users attempt to create loudspeaker processing from
scratch using one of the several, excellent measurement systems publicly available. However, the systems
and methods cannot duplicate what is involved for EAW engineers to create Greybox processor settings:
• Loudspeaker data far greater in quantity, types, precision, and accuracy (especially polar data).
• Use of proprietary analysis tools developed specifically for loudspeaker data.
• Use of proprietary Gunness Focusing algorithms for the processing.
• Extensive listening tests performed by a wider range of people over a much greater time period
and under a much greater range of conditions.
• Controlled, laboratory conditions.
• Extensive knowledge in the application of loudspeaker data analysis and signal processing.

Iain.

Logged

Uwe Riemer

  • Newbie
  • *
  • Offline Offline
  • Posts: 43
Re: Notched Crossovers
« Reply #9 on: March 17, 2012, 10:15:33 PM »

...
1. It was an attempt at the low pass portion of a Thiele notched crossover. (seems quite unlikely)
2. It was done to correct for an acoustic magnitude and/or phase problem in the low passband at 2.5kHz. This can be very usefull even that far down the filter skirt's cutoff. (seems quite likely)*

*Mod: italicized words changed by Uwe to help make his point.
...

only some minor corrections  :)

compare this http://www.eighteensound.it/index.aspx?mainMenu=view_product_simple&pid=248 ( Curves )
it is probably not the same driver but, but maybe similar

Uwe
« Last Edit: March 18, 2012, 08:03:09 PM by Langston Holland »
Logged

ProSoundWeb Community

Re: Notched Crossovers
ยซ Reply #9 on: March 17, 2012, 10:15:33 PM ยป


Pages: [1] 2 3   Go Up
 



Site Hosted By Ashdown Technologies, Inc.

Page created in 0.032 seconds with 21 queries.