Home Theater Forum and Systems banner

Question about using REW to time align drivers and/or Subs to mains...

23K views 24 replies 3 participants last post by  rewjack 
#1 · (Edited)
Hello ,

I know that the program has has it's origins in the room tuning and set up of A/V systems and isn't really designed for speaker builder per say , So please forgive me if I'm way off the reservation here ........


I have been reading some of the threads here at hts where the user is aligning subs and deciding what delay to implement so that the subs and mains will acheive better summation.
I followed the link to the pdf about sub alignment that points out that a sub is very much a limited band pass and that its Impulse response cannot strictly be used in time alignment because it contains little high frequency information and is made up entirely of slower( long) sine waves.

The writer suggest using the group delay plots , but mine are no where near as sanitary as those in the pdf !!.


My question is as follows...


I am setting up a mixing desk and will only be monitoring in the nearfeild ( most times). I am measuring where my ears would be . I know that will be the "sweet spot " and that adjustment made to optimize there will/probably/might not sound good or the same elsewhere .....

Can I use a measurment of the woofers and/or SW's that is taken W/O HP filters ?? That is, a fullband impulse.

I know that when I put the X-over back on I will rotate the phase and things will change ... I just basically want to get an Idea of the REAL ACOUSTIC OFFSETS ( Time of travel) so that I can delay the high frequency driver an appropriate amount ...





When I ctrl click and use the right mouse key , I get distances that are wholly inrealistic ; I know that finding the real acoustic center of a driver is dam hard , ( I know my tweeters are no more that a few inches ahead of the woofs for instance , but at the moment REW is saying it's 5 " !!!!



I am using a very good audio interface W/ fast , proven asio drivers and always have the loopback in place and the check box filled for using the loopback .



Should I uncheck the " set T = 0 at IR peak" box ?? in preferences/analysis



I Am using LR xovers and have read that they lose some of their effectivness if the drivers aren't time aligned ...


Once again , what I need most for this paticular facet of my tuning , is the " Time of Flight " . Which Will allow me to calculate distance.

I'm using a rack unit that is for pro sound so I have no automation like an audisey unit or the like.

I will of course use flipped phase null checks too .

Any thoughts about this ??



Thanks for any input:clap:
 
See less See more
#2 ·
Strictly speaking, the impulse responses, and a time reference are all that is needed for time alignment. The woofer with crossover is producing an impulse response. Relative to tweeter, woofer impulse peak is much lower in peak amplitude, and response is spread much further out across time. Stated another way: both tweeter and woofer start to move when signal is applied starting at t=0. Tweeter moves much faster with peak appearing quickly, and woofer peak occurs significantly later. This is the primary feature of group delay. Knowledge of where real t=0 is based on filters used to create tweeter and woofer responses v differences observed for t=0 calculated from actual peak data may be used as basis for delay setting. I say "may be used" because in most cases acoustic responses of drivers deviates significantly from perfect crossover function.

Example: Three way with Linkwitz-Riley 24dB/octave crossovers at 80Hz, 2500Hz. Effectively the tweeter impulse is at t=0, amplitude 0db. The 80Hz-2500Hz bandpass has peak t=0.173ms, amplitude -19.45dB, and the woofer has peak t=5.75ms, amplitude -48.3dB.

In reality you are crossing sub to small monitor, and most likely in a relatively small room compared to desired crossover frequency. If I glean correctly from other posts, you have two Dayton 10" woofer based subs? Sealed or ported? And Alesis 1(ported), and Auratone (5C Super-Sound-Cube design?) monitors?

Relatively predictable crossover behavior results when drivers in system are used in within the region of their flattest response. Quite often a drivers natural roll off, or roll off as mounted is integrated with a lower order crossover filter to achieve better results, for example: A driver/box has natural roll off of 12dB/octave at 80 Hz is combined with 12dB/octave Butterworth filter to achieve acoustic result approximating 24dB/octave Linkwitz-Riley filter. This is common approach with HT AVR systems for integrating sub with mains. Sub by itself without crossover may have fairly flat response out to 200Hz or more, and when 80Hz Linkwitz Riley 24dB/octave low pass crossover filter is applied, it integrates well with the Butterworth/mains treatment.

Similarly with pro gear digital crossover mixed slopes may be applied, or additional EQ applied to responses of drivers to get acoustic responses that conform better to the ideal. Relative acoustic levels are then matched. And then time delay is sought, usually starting with measured distance, that produces flattest sum through crossover region, and deepest notch with one of drivers has polarity reversed.

Flat response though may be highly influenced by room mode/reflection behavior, thus may also be viewed as influencing crossover behavior.

Andrew
 
#3 · (Edited)
Thank you very much for your reply Andrew!

My mains Are 3 elements ; and 8" dayton rs-225 , a Morel mid dome and a Aurasound 3/4 titamium dome tweeter. They are sealed as are the 10" Dayton subs......


I have made two discoveries since my post last night ( amazing what a nights sleep can do !!!)

First off , when one unchecks the '"Use Loopback as Timing Reference" tick box in analysis preferences , the "Set t=0 at IR Peak" is automatically unchecked also . so the are part and parcel with each other .....


More Importantly though , I found out why the distances don't seem to jibe;...


In this screen shot of the impulse window there is the woofer with and with out the LP filter engaged ...





The red or orange is the impulse made up of the full sweep and the purple is made up of signal only less than 1khz .

This is reflected in the speed or length ... the purple is much wider .

The program is lining up the peaks. ( A Eureka moment for myself:blink:) I just wasn't noticing the shift of the beginning into negative territory !!!!

The important thing ( to me at least ) is that the LP filter is inducing phase lag ( indicated by the red box ; ( notice how the begining of the purple got placed behind or, to the left of the zero mark !!.

A hp filter (like on a tweeter) is going to cause phase lead. When these two ( X-over induced lead and lag) aren't included in calculations , the distances are thrown under the bus and mangled!!!

I just need to take this into consideration when setting the delays for the tweeter ( the tweet and mid are so close I'm not going to worry about their relationship as much; for now at least ! ) and the mains to sub delay....:scratch:

The journey continues.....:T
 
#4 ·
Just to be clear: When working with timing, checking "Use Loopback as Timing Reference" and using loopback connection is what you want. With timing reference Trace Arithmetic features become useful in predicting behavior of individually measured system elements.

With minimum phase filters, such as standard Linkwitz-Riley filter, hp filter just has less group delay than lp filter. This is seen in my previous three-way crossover example. Also, in REW IR display, uncheck "Plot Normalized" to see relative amplitudes of passband amplitudes.

What are you using for cross over duties?

What is driver layout? And what are box dimensions?

What leads you to three-way plus sub(s) for monitoring duties?

Please tell me, you are using calibrated microphone?

How big is mixing space?

All sealed setup is going to make this easier.


Regards,

Andrew
 
#5 ·
Just to be clear: When working with timing, checking "Use Loopback as Timing Reference" and using loopback connection is what you want. With timing reference Trace Arithmetic features become useful in predicting behavior of individually measured system elements.

With minimum phase filters, such as standard Linkwitz-Riley filter, hp filter just has less group delay than lp filter. This is seen in my previous three-way crossover example. Also, in REW IR display, uncheck "Plot Normalized" to see relative amplitudes of passband amplitudes.

Hey ! ,
Tx again for the response

Finally will have a little more time this weekend to get back at it .
I'm going to try and take some measurments with the loop back engaged and the t=0 not engaged . Having the peak always at zero after calculation is not helpfull . I made a measurement with and without DSP delay and the mic right in front and the traces just end up on top of each other !!!.


I'm curious as to what the advantage of not normalizing is ?? Doesn't seem to change things in the time domain and the lack of amplitude just forces a zoom ... Maybe I'm missing something ( Highly likely:bigsmile:)

What are you using for cross over duties?

What is driver layout? And what are box dimensions?

What leads you to three-way plus sub(s) for monitoring duties?

Please tell me, you are using calibrated microphone?

How big is mixing space?

X-over = Carvin XD360
http://www.carvinguitars.com/products/XD360


Box dimensions are 18" h X 10" W X 10.25" D

Driver layout =




I've always had a taste for three band speakers ; even before I knew the difference between Mrs. Butterworths and a 24db BW !!!:dumbcrazy:




Yes , calibrated mic , file is in place .

Room is pretty big ; some 700 sq ft with a sloped vaulted ceiling , fire place in a corner . But no treatments ( not wife approved ) It's got a resonance at 35 and 70hz , it seems like a good layout as it's not a symmetric box or any thing like that . !


Anyhow , I'll post soon again , wish me luck!
 
#7 ·
The subs are powered by a Crown XTi 2002 ; built in DSP. I knew that since I couldn't go hog wild with room treatments, I'd want to have lots of adjustment capabilities and I don't really think that all the recent crop of automated eq's are going to do much except deal with some early reflections or peaks that are obvious.. The algo's are so new ....... I guess I wanted to do it myself...

Anyhow , My lack of placement options brings it's issues and I wanted maximum flexibility. For instance ;


I have a suck out at 120hz that I know is because of the 1/4 wave boundary effect and I can't have the desk anywhere else. So I'm going to try ( on the list !) crossing over at that frequency and leaving a gap ( under lap) between mains and sub in hopes that there will be less anti-phase in that region and a flatter response in the nearfeild ...... probably will cause other problems though ......


Will report on (hopefully) my PROGRESS soon !!!



Thanks for your kind interest in my little mad scientist project !!!
 
#11 · (Edited)
Tri-amped indeed....

Been working on my Crossovers ; they were a little ragged. I kinda went charging into my SW build so it was unfinished buisness!! I like the way the tweeters sound , but they turned out to be marvelous fun with a little bit of eq required ( had 4 bands of fully parametric) Took several iterations to get them right !! I also wouldn't go with the rs-225's again, but that's only because I now know I'm using subs and I went with them them for the extension .. at least these boxes can stand on their own in the bass department now but, I'd probably go with a paper cone knowing what I know now !!




I've been doing a little research and there seems to be allot of ideas about how to execute the time alignment; I've come up with a few Ideas myself and have done searches around here and read allot of threads !


So here are the mdats...

View attachment middy.mdat
View attachment tweet.mdat
View attachment woof close 1 inch.mdat


what's your approach to using REW for driver time alignment Andrew??

Many thanks for your kind interest :T
 

Attachments

#12 ·
P.S.

The mids and tweets are phase flipped ( not in the mdats however) and the level of the tweet is a litttle low in the mdat too....... The mid and the tweet are sealed in their own Assembly's ( and the Morel had a slope less than 12 !!) so a 12db electrical combined w/ the acoustic and some eq was nessecary to massage the crossover regions into shape . These are sounding very good now , I'm interested to see if time align will make them even better ( I can adjust the dsp delay at quarter inch increments... ) . It's very important to me because I want good dispersion and as little lobing as possible in the nearfield ; As I mentioned , untreated room :foottap:....
 
#13 ·
Moral/Aura assembly: How is this set up? Describing "...12dB electrical combined with the acoustic..." suggests that passive components are involved with tweeter and mid. Please clarify.

The page I'm on has: Sub with own amplifier and crossover. Carvin XD360 controls RS-225, Morel, and Aura. Some sort of passive components are connected to Morel and Aura.

What are you using for amplification of RS-225, Morel, and Aura drivers?

What are you using for microphone? Cal curve rules out ECM8000.

For speaker this size and focus on nearfield use, a single microphone location for alignment may work best. You appear to be targeting crossover points in mains at 1000Hz and about 4kHz.

A practical approach: With speakers at working locations and looking at them from listening position at mixing desk, microphone should be placed on sight-line to point midway between mid and tweeter. Since this appears to be very close to sight-line to point midway between mid and woofer, a compromise between these two is possible.

Once relative location of microphone to speaker is decided, speaker should be moved clear of walls, furniture, mixing desk, on stand that preferably raises speaker midway between floor and ceiling. Microphone should be placed along predetermined sight line at distance of 3-4ft. Closet thing to microphone after speaker should be floor or ceiling. Same full range sweep is used to measure each driver. Resultant impulse responses are gated to exclude floor/ceiling reflection, roughly 3-4ms. Blackman-Harris4 windows are my preference.

To make windows work in REW using loopback timing reference, all three impulses must be equally moved closer to t=0: Chose the tweeter impulse response, and press "Estimate IR Delay", note the value that is calculated before applying to tweeter measurement, then manually apply same offset to mid and to woofer.

Switch to All SPL tab, and you should have nice gated response display of all three drivers.

Typically, when doing this, I will start with a set of measurements at low level with no crossovers set for drivers, or with temporary crossover points expanding driver overlap. Then I equalize each driver flat through crossover regions, and adjust levels so drivers have proper balance. Desired crossover points are then used for fresh set of measurements using above procedures.

In All SPL tab Trace Arithmetic is used to sum driver responses. Summed responses should be flat through crossover regions. Then A-B math is done for Tweeter with Mid, and Mid with Woofer. When properly timed these should result in deep notch (36dB and deeper are possible, 24dB is very good, 15dB is a cop out).

In impulse response tab impulses may be shifted. Small shifts are applied, new sum/differences are assessed. Usually with difference trace some manner of notch is seen, and even single sample shifts +/- changes result. If notch gets deeper, keep going in that direction until notch contracts, otherwise start shifting in other direction. Keeping track of which impulse is being moved, and how far, allows tweaks to crossover.

Alternate of course is tweaking delay on crossover, remeasuring, assessing, tweaking etc. For tweeter and mid alignment for example, phase of one driver is reversed, either with crossover, or by reversing leads of one driver.

Here is example with your posted mdat, using tweeter and mid. Based on experience I bumped timings around, since microphone wasn't fixed location, tweaked tweeter level, and did the A+B and A-B math. Here notch is about 40dB:

Text Line Plot Slope Diagram


If this can be hacked up, then I'm sure with practice and pointers you can get really good results.

Similar techniques are applied for sub to mains, more on this later.

Regards,

Andrew
 
#16 ·
O.K. So I've been digging into this ..

Text Line Plot Slope Diagram


I'm going to get this set up like you described to get rid of the floor bounce ... My ceiling is sloped here so it will get it's bounces diffracted off too . I found a way to get the boxes about 5' up and I'll get the mic around 3' back .

Above I messed around with the 1" measurements which have the filtering in the measurements . Aligning the peaks doesn't work because visually the slower impulse ( mid after tweet ,woofer after mid ) is always shown as further back than it actually is because of the "electrical" lag induced by the filter. you really want them both compressing or refracting in tandem so it's not good to go to far with the delay ... If you forget about that you're actually putting the driver that was ahead behind !!!!

The only question I have is about this...

Typically, when doing this, I will start with a set of measurements at low level with no crossovers set for drivers, or with temporary crossover points expanding driver overlap. Then I equalize each driver flat through crossover regions, and adjust levels so drivers have proper balance. Desired crossover points are then used for fresh set of measurements using above procedures.
I'm wondering why you to take the dips out of the x-over regions ? will nulls be created solely on the timing/phase differences ? will this be exaggerated without the filters turning down the conflicting areas where both drivers are putting out the same frequencies ?? after you put the filters back in , won't the frequency/timing variables that were maximized in the absence of the filters be thrown off by the re-introduction of the filters phase lag and lead ???

I can't wait to get back to measuring and making the final tweaks so I can listen to these with the proper time alignment in place also !!


Thanks for your insights ; you've helped immensely!:T
 
#14 · (Edited)
Hello Andrew,


What I meant was that I applied the additional 12db HP for the bottom side of the mid and tweet in the dsp of the Carvin X360. I'm sorry if using the term electrical inferred that I was using analog components .. semantics!!

Obviously I'm trying to implement the classic 24db LR configuration because of is polar radiation's favorable lack of nulls ( or at least where it does have them , they are in a good place for them to be !)

I now have a first hand experience as to why it is good practice to have drivers which maintain flat response for an octave or two ( at least ) beyond the x-over point ! I think I would be in big trouble if it weren't for parametric eq. The mid and the tweet had the same resonance at 4.8khz so I have some baffle edge diffraction going on ( I know grills are bad , but we have curtain crawlers here so it was a must !)

My point is that I know that these x-overs aren't perfect!...

The mains are driven by an ART Sla2 and a Sla4. Not real fancy amps but they do seem of a better league than so many of the plate amps ( especially those that don't even have a substantial transformer ... But I confess I'm no expert in these matters of electronic amplification) This gives me the 6 channels I need in a fairly compact package .

My measurement Mic is a Dayton EMM-6 , and I'm using the calibration file that accompanied it .


I'm not as adept at using the deeper functions of REW as you are , so I may stick to the iterative measure , adjust, measure routine that I have fallen into . I will need to climb the learning curve for a bit here I'm afraid:crying:

I understand that your method is very well thought out in that it only requires one set of measurements to use . Since small positional discrepancies in Mic placement can lead to huge changes in the resultant measurements . I have to set up and break down when I measure so I never really can repeat the same exact mic position.

I'm going to have to give this some thought and see what I can do.:scratch:

( I have a much greater appreciation for the acoustician after all this !)


As always, many thanks for your kind interest !:T
 
#15 · (Edited)
Where the nulls are ....................

This ( the dark blue) is from the spot where my ears would be when sitting at my desk ( not the three bands; they are 1 inch ).... the band levels were set a little hastily

Text Line Slope Plot Design


It looks like the actual woof to mid xover point is a little lower that 1khz the null for the mid to tweet is a bit disturbing ( not in the right place !! .... tweeter level is high though.... )


Looks like I've got more work to do ! :yikes:


Will post again when I make more progress, thanks again !!!
 
#17 ·
Equalization to flat is in region of target crossover, not the temporary expanded range crossover. I find it much easier to do this, rather than trying to match slope of a crossover filter.

Pic looks like you've got the idea! And yes, especially at high frequency, it is possible to delay whole wavelength apart. Impulse responses resulting from this are fairly easy to tell apart.

Cheers,

Andrew
 
#18 · (Edited)
It took some doing , but I kept at it and found what seems to be a good combination of factors for these drivers. I've learned a ton with this project , so even though these speakers are far from perfect , the experience has been great!. I have a much greater respect for baffle edge diffraction as it seems to have had allot of effect on what were seemingly smooth drivers....

Text Line Plot Slope Pattern


Green Text Line Plot Slope



I ended up using all 24db bessels at 1khz and 4 khz. I had been trying to use a 12db filter in conjunction with the acoustic slope on the low side of the mid ( as it is a sealed unit) but as this was not a true 12db it didn't add up and the results were something like 19 or 20 db!

I also learned that whilst 21 micro increments of dsp delay are quite sufficient around 1 khz , up at 4 khz it is an interval which gets a bit chunky !!

The bessel and .210 ms delay got the biggest null. the mid is delayed .063 ms.

these distances (.084" and 2.80") also make sense since the woofers basket is around 3.40 inches deep and the mid is a really deep little soda can shape so it's almost as deep as the woofer ! the tweet is very shallow so it needed the most delay.

Blue Cobalt blue Feather Electric blue Font


Text Line Plot Slope Font



The Bad part about putting the boxes up 5 feet plus and 7 feet away from any walls is that even though I managed to get them flat to plus or minus approximately 2 db without smoothing (with the quasi anechoic methods); ........ It sure was humbling when I then measured them back at the desk !!! OUCH :help:

Rooms are cruel !!:hissyfit:


SO I have one question as I head onto the SW integration ;


I have noticed that the tweeters frequencies seem the be given the most weight when the IR delay is calculated in the impulse tab/display window. IS this of any great effect when trying get the most accurate phase number (degrees) at my SW crossover frequency ?????

my woof on the mains are about -118 degrees around the crossover frequency ( preliminarily 70-90hz)

I've been looking at this article http://www.rythmikaudio.com/phase1.html


I know that measuring the subs is allot harder because of the room effects at the frequencies they operate at . One thing is for sure , they are not going up in the air like with the mains ; way to heavy !!!

How do I make sure that my Subs and mains don't have those nasty dips at cross over like were shown in those measurements at the Rythmic site ..... Guess I'll find out !!!:ponder:



Thanks!:D
 

Attachments

#21 ·
Jacques,

I follow above link, and read along Bob McCarthy's blog. I'm not sure what hardware Bob is using, apparently some type of real time FFT analyzer. FFT goes both ways between frequency domain and time domain. His setup appears to capture frequency domain info, likely at a limited number of frequencies, and to see impulse response in time domain FFT is used. Limited number of frequencies limits resulting impulse response to only approximation of real system impulse response.

Bob is spot on here:

The procedure is simple. measure them both individually, view the phase and adjust the delay until they match. You have to figure out who is first and then delay the leader to meet the late speaker. This will depend upon your speaker and mic placement.
This is exactly what you have been attempting to do with your setup, and is process I guided Flatfinger to. Flatfinger's small sealed monitor type speaker and big space made it relatively easy to set speaker/microphone up so that early reflections could be gated out while keeping needed information for crossovers at 1kHz and 4kHz.

Flatfinger's above post shows that he's now familiar with how corrupting early reflections can be, including those created by speaker baffle, or by his mixing desk.

Flatfinger has also learned that with fixed delay step of one sample is coarse when trying to tune alignment at 4kHz, and relates well to your situation. At sample rate 48kHz, 1 sample represents about 21 microseconds. One cycle of 4kHz sine wave is 250 microseconds. (21microseconds/sample)/(250microsends/cycle) x 360degrees/cycle = 30.24degrees/sample. As Flatfinger says, this is "chunky".

A 4kHz Linkwitz-Riley 48dB/octave crossover filter is generated and its high pass and low pass components are imported to REW. Phase display shows perfect matched phase, summed response is flat, and A-B produces super deep notch:

Text Line Plot Slope Parallel


and:

Text Line Parallel Slope Number


When a single sample delay is added to tweeter impulse response and new A-B generated this results:

Text Line Parallel Font Screenshot


As calculated above, HP filter now has 30 degree phase shift relative to LP filter. The phases sum to 15 degrees at 4kHz. The A-B notch now is only about 12dB deep:

Text Line Plot Parallel Diagram


This shows great relationship of relative phase angle and depth of reverse null notch as alignment indicators. In the above pic, 30 degree error doesn't do much to the apparent frequency response. The impulses look nearly identical, and perceivable audible effects are not likely distinguishable in blind A/B testing.

Real driver performance includes lots of phase issues. Phase plots of driver impedance indicate that the greater the differences in driver size, moving mass, voice coil inductance, etc. the greater the differences of slopes phases will be through crossover region:

Text Line Plot Font Parallel


Jacques Mid and Tweeter have phases with move divergent slopes. Ripple is primarily noise in measurement. Flatfingers Mid and Tweeter stay closer together in phase over greater range. Dome Mid is much smaller that Jacques' Mid. Microsecond delays were used to get all phases close to 0 degrees at 4kHz.

Resultant notches after adjusting phase with delay and adjustment of amplitudes; gates <3ms:

Text Line Plot Diagram Wave



I continue quote of Bob from above:

I say this is simple - but in reality , it is quite difficult to see the phase response down here. Reflections corrupt the data – it is a real challenge. Nonetheless, it can be done. It’s just a pain.
Yes, at low frequencies become somewhat of a pain

But I get to his passage:

Did I mention that all speakers (as currently known to me) are time stretched? So this means something pretty important. The simplistic single number derived from an impulse response can not be used to describe ANY speaker known (to me) especially a subwoofer.
I'm not sure what he's referring to with "simplistic number".

The important point is that in time domain, complete system response is fully described by the impulse response, also the system is fully described in the frequency domain by the amplitude and phase for each frequency component, and finally the frequency domain and time domain are translated with each other using FFT or Discrete Fourier Transform (DFT).

Mathematically Fourier transforms are implemented via convolution.

Regards,

Andrew
 
#22 ·
I'm not sure what he's referring to with "simplistic number".

The important point is that in time domain, complete system response is fully described by the impulse response, also the system is fully described in the frequency domain by the amplitude and phase for each frequency component, and finally the frequency domain and time domain are translated with each other using FFT or Discrete Fourier Transform (DFT).

Mathematically Fourier transforms are implemented via convolution.

Regards,

Andrew
Indisputable basic points. Your knowledge is far , a couple of light years, ahead of mine. I'm sure you've noticed, that I'm a kind of nasty beginer. Though I'm here to learn and share, and your help is greatly appraciated. On my side I need to explain this to my fellow friends down here.:nerd:

This is why I want to get a little deeper in Bob's explainations for transition between large cone speaker towards smaler ones e.q subwoofer/main. What I've interpreted is that the large 15" surface won't react equally from close to center and the more to the edge position. He says that an impulse response derived from the FFT will give a mean value which is unbalanced in favor of the HFs meaning that it could be missleading. This is why he is suggesting not to use IR for alignment in this particular cases. It looks more like a combination of sample distribution and, I'm not sure if he is saying this, maybe physical properties of large cones Vs small ones??.
This is what I think I've understood,:doh:
Bob said:
The sub is NOT flat (duh!!) so there is a tradeoff game that goes on in the analyzer. As we lose energy (frequency rising) we gain data points (liner acquisition) so the most likely place the peak will be found is in the upper areas of the subwoofer range and/or somewhat beyond, before it has been too steeply attenuated. If you have a subwoofer that is similar to 100% of the speakers I have measured in the last 26 years, then one thing is certain: the response at 30 Hz is SUBSTANTIALLY behind the response we just found at its upper region.
This is unfair for Bob, because I's just a small part of his text.

As you have indicated
Barleywater said:
Real driver performance includes lots of phase issues. Phase plots of driver impedance indicate that the greater the differences in driver size, moving mass, voice coil inductance, etc. the greater the differences of slopes phases will be through crossover region:.
This is going in the same direction.

For these reasons I pay more attention to the frequency domain for subwoofer-woofer transition.
I find this easyer not to focuss to much on IR for this particular position.

If we come back at the free work you have done for us.
A 4kHz Linkwitz-Riley 48dB/octave crossover filter is generated and its high pass and low pass components are imported to REW. Phase display shows perfect matched phase, summed response is flat, and A-B produces super deep notch:
I'd like to be able to generate this A-B notch. Can we do it in REW?
It seams quite usefull and conclusive to work with it.

For main/tweeter "XO" creation, I used only IR response, since phase plots where not obvious. Your A-B process would have been helpfull.
For woofer/main, it was IR first then phase matching at XO point by changing slope and/or moving frequency position and again delays. A play of foreward-backward manipulations.
Resultant notches after adjusting phase with delay and adjustment of amplitudes; gates <3ms:
Can you give more details on amplitudes adjustments and more advices.

regards
jacques
 
#23 · (Edited)
Hey Jacques !,

I'm no expert , and I just try and be tenacious with this difficult subject matter ! I found the step response window to be a little more intuitive, and this article ( which starts with allot of infos about the polar dispersion of butterworth crossed drivers - but is very informative about time align in the second half under the heading " looking in time") to be a great help in conceptualizing getting the three bands to work together. ( It helped me ; after I read it more times that I care to admit !!!!!:D)

http://customanalogue.com/elsinore/elsinore_17.htm#Explanation%20of%20angle%20of%20radiation


after I used the time delays to get the deepest nulls ( and therefore the best phase relationships at the crossover hand off between drivers) I can't say that the Impulse window could have given me an intuitive way to arrive at the delays ..... you can't just line up the peaks and if you look at the before and after relationship of the three drivers impulses the tweeter is of course not so far out front , but I can't imagine that I would have guessed or arrived at the present relationship between the three by using that graphic display alone .

Also , I read a few articles ( Mostly about sound reinforcement in big venues) that said to let the sub rip at full unfiltered frequency response ( to get the higher , faster traveling freq's) and use the "first arrival " ( NOT peak) of the impulse to get you into the ball park ; but then it moved onto other methods to fine tune it from there .

Best regards!
 
#25 · (Edited)
Also , I read a few articles ( Mostly about sound reinforcement in big venues) that said to let the sub rip at full unfiltered frequency response ( to get the higher , faster traveling freq's) and use the "first arrival " ( NOT peak) of the impulse to get you into the ball park ; but then it moved onto other methods to fine tune it from there .

Best regards!
Thanks for this article flatfinger. I'll read it this week for sure, looks good. I'm still looking for a compromise between woofer and mains. I still think that transition from a closed 400mm cone woofers to 215mm open baffle mids is tricky. Two world.
Concerning position of peacks for sub, I agree with you, not have to be aligned, as it shows on ETC plot, see pict. The "XO" at 40Hz is matching allright, and "IR" response "FIRST ARRIVAL" to.

I saw that you want to get closer to the wall? I don't know much yet about digital filters, but I can tell you that getting closer than 1.2 meters (arbitrary distance from many friends and I experiences) to the wall will erase the best of the soundstage 3D.

regards
jacques
 

Attachments

#24 · (Edited)
Andrew ,

I'm going to work on the placement of the boxes on the desk ( specifcally , moving them closer to the wall) and then go through set up ( toe in ,ect.) and try and make some improvements in the early reflections department .

I have a serious 112khz suck out that is the quarter wave / Allison effect and I'm hoping that my boxes are big enough that I can get away with pushing it up higher ... we will find out !!

When I find the best configuration I'll post some insitu graphs.

P.S. I only had then flown one at a time and for sweeps only ....They do sound very good ( perhaps after all this toil I might be biased though ! :devil:)

The thing I'm most pleased with is how punchy things are ! there isn't a hint of hangover so when sounds are supposed to stop they stop !! Snare drums are fantastic! I am glad that I went sealed .
 
This is an older thread, you may not receive a response, and could be reviving an old thread. Please consider creating a new thread.
Top