Jventer

REW and acoustics - help needed please

163 posts in this topic

38 minutes ago, andyr said:

 

I probably did misunderstand, Dave.  :D

 

I can't seem to access your link (I'll try again when I get home)

 

The link is to this picture

 

 

 

TextbbokLR4[1].gif

Share this post


Link to post
Share on other sites
4 minutes ago, davewantsmoore said:

 

The link is to this picture

 

 

 

TextbbokLR4[1].gif

 

Thanks, Dave.

 

Sure, that shows the filter FRs and the overall FR of a 24dB L-R XO @ 2Khz.  And the associated phase curve.

 

But I don't get the connection between this diagram and your comment "The delay varies with frequency."?

 

Andy

Share this post


Link to post
Share on other sites
40 minutes ago, andyr said:

I can't seem to access your link (I'll try again when I get home) but I can understand that the delay varies with frequency.  So surely that means if you decide to delay the mains, say, by 6.5mS compared to the subs ... then this will only be correct at one particular frequency?  If this delay is 'correct' at 100hz - then the delay will be wrong when a 120hz note is playing?

 

Yes.

 

What needs to happen, is the change in phase (with frequency) for the subwoofer and the main, need to be the same.   Then .... when the drivers are aligned in time, they remain aligned at every frequency.

 

The picture I posted, there are actually 2 phases curves....  one over the top of the other (so it just looks like one) .... like you say, if those two phase curves were different .... then they could only be matched at one frequency (or over a limited frequency range).

 

44 minutes ago, andyr said:

how can the mic & REW help us get to this "right delay"?

 

 

Take measurements of your speaker(s)

See that they don't have the same delay (phase)

Ask ....  "why don't they have the same phase?"   (there is likely more than one reason)

Do something about it

Repeat

 

47 minutes ago, andyr said:

No, was not using the 'loopback timing reference'.  I'll have to read the manual more and then use this, next time I do more REW work

 

It's not essential, and I wouldn't bother at this stage  (it's chasing rainbows).

 

Assuming the answer to my next question (estimate delay button) was no .... then you need to click that, and things may become more clear   (if not, ask questions - as it's a very important part of the puzzle / concept).

 

 

52 minutes ago, andyr said:

I'm afraid i beg to differ

 

Knowing the transfer function of some existing filters is a reasonably simple piece of work.... which is nice that it's already done / easy for you .....  but it's really not what I was alluding to.

 

 

Share this post


Link to post
Share on other sites
12 minutes ago, andyr said:

Sure, that shows the filter FRs and the overall FR of a 24dB L-R XO @ 2Khz.  And the associated phase curve.

But I don't get the connection between this diagram and your comment "The delay varies with frequency."?

 

"Phase" (ie. a time lead, or lag), and "delay" are the same thing.

 

"Delay", ie. every frequency delayed equally by X milliseconds .....  would have a phase curve that was horisontal.  (ie. constant with frequency)

 

The phase (delay) chart shown is not constant with frequency   ie.   it is a graphics representation of the statement "the delay varies with frequency"

Share this post


Link to post
Share on other sites

 

34 minutes ago, davewantsmoore said:

 

"Phase" (ie. a time lead, or lag), and "delay" are the same thing.

 

"Delay", ie. every frequency delayed equally by X milliseconds .....  would have a phase curve that was horisontal.  (ie. constant with frequency)

 

The phase (delay) chart shown is not constant with frequency   ie.   it is a graphics representation of the statement "the delay varies with frequency"

 

OK, yes - that I can understand.  Thanks, Dave,

 

So if we didn't have a 24dB L-R XO @ 2Khz ... the phase line would be flat - right?

 

But as soon as we implement XOs, we must put up with the resulting phase differences with frequency - right?

 

So in my system, having these XOs:

  • sub LP / bass panel HP @ 100hz
  • bass panel LP / mid panel HP @ ~350hz
  • mid panel LP / ribbon HP @ ~3,500hz

. . . it is not surprising there is nowhere (in my system) where I have minimum phase! :D

 

40 minutes ago, davewantsmoore said:

 

Knowing the transfer function of some existing filters is a reasonably simple piece of work.... which is nice that it's already done / easy for you .....  but it's really not what I was alluding to.

 

 

That I'm afraid I don't understand.  :(  (What you were alluding to.)

 

To me, going active with Maggies is a simple exercise - I thought you said it was complex?

  • we know what each model's XO schematics are
  • so we know the component values and the driver resistances
  • so we can get (using lspCAD or similar ... or even an excel spreadsheet) the slopes and -3dB frequencies (-6dB in the case of L-R XOs)
  • so we can feed these into the miniDSP to get active versions of these original passive XOs.

 

Andy

 

 

 

Edited by andyr

Share this post


Link to post
Share on other sites
1 hour ago, andyr said:

So if we didn't have a 24dB L-R XO @ 2Khz ... the phase line would be flat - right?

 

 

Yes.   The amplitude and phase of the input signal is flat.
 

1 hour ago, andyr said:

But as soon as we implement XOs, we must put up with the resulting phase differences with frequency - right?

 

Yes, when you distort the amplitude response .... you also distort the phase response    (because amplitude and phase are just two views of the same thing)

 

1 hour ago, andyr said:

. . . it is not surprising there is nowhere (in my system) where I have minimum phase! :D

 

Yes....  but you have a lot more phase rotation in your data than will be caused by the filters.     There is many possible causes of this (and you likely have more than one of them going on).    This comes back to what we were saying previously about only measuring one thing at a time.     If you measure the combined response of two drivers  (which will usually have a phase rotation due to the summed crossover response) ..... then there's lots of things which are obscured from us.     The result is not "minimum phase" .... but we don't know why  (and we can't say  "EQ here would be bad" ..... because we don't know if we should be expecting the response to be minimum phase or not .... the non-minimum phase-ness, could be caused by something which doesn't preclude EQ).

 

 

I think there is potentially a really big misunderstanding going on in this thread.     EQ can be used everywhere.    There is (basically) no region of a room, or a speaker, where you cannot use EQ.

 

.... but if you take a measurement.   There are areas of that measurement where the data is "wrong" and you should not use that data to design your EQ.

 

 

1 hour ago, andyr said:

That I'm afraid I don't understand.  :(  (What you were alluding to.)

 

Exactly what I said.    Acoustics and filter design.

 

Being told what filters to use (is helpful, I appreciate, but) covers only a tiny fraction of it.

 

Share this post


Link to post
Share on other sites
1 hour ago, davewantsmoore said:

 

Exactly what I said.    Acoustics and filter design.

 

Being told what filters to use (is helpful, I appreciate, but) covers only a tiny fraction of it.

 

 

OK, for now, Dave ... let's focus on just one scenario - which I will describe thusly:

  • given Maggies are supplied with a passive XO (which consists only of filters)
  • given we know the slope & knee frequencies of these filters, and which way round the drivers are connected
  • then we can enter this data into the miniDSP

. . . and end up with the same filters - but delivered actively - as were delivered passively in the stock Maggie.  Which means the same phase behaviour with the active scenario as in the stock (passive) Maggie.

 

Do you agree with this or not?  (If not ... I would like to know your reasons.)

 

 

Thanks,

Andy

 

 

Share this post


Link to post
Share on other sites
1 hour ago, andyr said:

 

OK, for now, Dave ... let's focus on just one scenario - which I will describe thusly:

  • given Maggies are supplied with a passive XO (which consists only of filters)
  • given we know the slope & knee frequencies of these filters, and which way round the drivers are connected
  • then we can enter this data into the miniDSP

. . . and end up with the same filters - but delivered actively - as were delivered passively in the stock Maggie.  Which means the same phase behaviour with the active scenario as in the stock (passive) Maggie.

 

Do you agree with this or not?  (If not ... I would like to know your reasons.)

 

Yes, I do agree.

 

However, the next step 'evaluate the result', is something which will be difficult without knowledge.

 

... and so, as much as you're trying not to be ... this scenario is actually a prime example of what ghost4man was talking about.  He makes a very good point.    Even the most trivial 'taking a speaker active' exercise like yours, requires reasonably extensive knowledge in order to actually get a positive change - simply replacing passive filters, with multiple amplifiers and filters which produce the same transfer function, is not something that's expected to be a positive change by itself.

 

For example ....  your crossover points 350, 3500 show sharp changes in the response.    This looks like a significant problem to me (but I don't have enough info to be sure).   What can you say about them?

Share this post


Link to post
Share on other sites
1 hour ago, davewantsmoore said:

 

Yes, I do agree.

 

 

Thank you, Dave.

 

1 hour ago, davewantsmoore said:

 

For example ....  your crossover points 350, 3500 show sharp changes in the response.    This looks like a significant problem to me (but I don't have enough info to be sure).   What can you say about them?

 

 

Sorry, you've lost me?

 

My XO points are 100, 350 and 3500hz - this is between:

  1. sub & bass panel (which is an extra - not part of the stock Maggie XO)
  2. bass panel & mid panel
  3. mid panel & ribbon.

#2 and #3 are just the same as in the stock passive setup.  I included the sub/bass XO that I now have simply to explain that with all these XOs ... there ain't gonna be - given the phase shift shown in the XO diagram you posted - any area of my overall FR that has a flat phase response?

 

To me this is not "a problem" ... it is simply the result of wanting to use so many XO points.

 

1 hour ago, davewantsmoore said:

 

... and so, as much as you're trying not to be ... this scenario is actually a prime example of what ghost4man was talking about.  He makes a very good point.    Even the most trivial  'taking a speaker active' exercise like yours, requires reasonably extensive knowledge in order to actually get a positive change - simply replacing passive filters, with multiple amplifiers and filters which produce the same transfer function, is not something that's expected to be a positive change by itself.

 

 

Mmmm, all I can say is that (having gone 3-way (analogue) active about 15 years ago) all I have done is swap digital filter pairs for the same analogue filter pairs that I've been using for years.  But with the cross to digital filters, I added subs.  Which makes it even harder to have any 'minimum phase' regions.

 

Yes, obviously some knowledge about filters and lspCAD is required ... but I have this.  Ozzie - as he's a relative 'newbie' to Maggies and active XOs - does not.  (But I'm sure he would dispute this.)

 

Andy

 

Share this post


Link to post
Share on other sites

 

 

On 4/19/2017 at 1:24 PM, davewantsmoore said:

 

TextbbokLR4[1].gif

 

On 4/19/2017 at 1:32 PM, andyr said:

But I don't get the connection between this diagram and your comment "The delay varies with frequency."?

Andy

As Dave said, non-zero negative phase (phase lag) is delay - let's leave "phase lead" alone for a second.

Take the graph above and look at 1kHz - the brown phase line is approx -90 degrees.

To convert this to time delay the formula is

time delay = phase / (360 x F)

So in this case time delay = 90/(360 x 1000) = 0.25 ms

 

Now look at 10kHz - the brown line is approx (30 - 360) = -330 degrees (remember we're looking at a wrapped phase plot so subtract 360 degrees as it's after 1 vertical line)

time delay = 330/(360 x 10000) = .092 ms

 

@andyr - does that make it easier to understand that the delay varies with frequency?

 

All IIR filters have variable delay with frequency. A linear phase FIR filter has constant delay for all frequencies.

 

The nice thing about an LR4 Xover is that High Pass and Low Pass responses are always in phase - as Dave said the brown line is the phase response of both HP and LP - so although the delay varies with frequency, the delay is the same for each.

The same can't be said for a 2nd order Butterworth Xover for example (hence the bump or dip in FR at Xover).

It's because the drivers of an LR4 are in phase that you get a perfectly flat FR after summing.

 

The benefits of the LR4 Xover requires you've achieved an "Acoustic LR4" and not just an "Electrical LR4" within the Xover - Linkwitz discusses the need for time alignment of non-coincident drivers etc etc, and unless the drivers (or in your case panels) are flat several octaves away from the Xover point, you'll never achieve close to an Acoustic LR4 response.

These are some of the complexities Dave was alluding to in designing Xovers.

 

cheers

Mike

Share this post


Link to post
Share on other sites
On 4/19/2017 at 0:05 AM, Jventer said:

Follow up. I have done 2 quick sweeps with a pair of JPW booshelves. (The graph is not 100% as the JPW's were done to get 83 dB whilst before I was trying to hit 85dB.)

The JPW look totally different at about 150hz but then expexperience a drop/suckout  depending on position at about 150 to 180hz

Also note the difference between left and right speakers at about 160hz whilst the left is low the right drops double as low.

There is something going on between 150 and 200hz and I am still stumped.

 

JPW and prev.jpg

I feel your frustration

 

On 4/19/2017 at 11:25 AM, davewantsmoore said:

There isn't a simple answer to this.

 

agreed

And my personal experience doesn't help a lot in your case - since I've been measuring my speakers and room, I've always done speaker measurements outdoors, then corrected the speaker response to flat based on the outdoor measurement, then moved the speakers into the room and worked on room correction (knowing that the baseline speaker response outdoors was flat).

I've also never dealt with large floorstanders like yours with multiple woofers - I run a "simpler" system with one driver covering 40 - 300Hz.

 

In your case we have an additional "degree of freedom" where we don't know the "anechoic" response of the speakers, so being able to tell what is room related and what is speaker related is much harder - but even outdoor measurements will have reflections in the measurement down low.

 

Keep posting and I'll keep trying to assist - but if it get's too tiresome you could consider contacting someone like @Paul Spencer who does this stuff (room measurement and consulting) commercially, and he will possibly be able (with your assistance) to do some remote measurements - a (likely long) phone call, internet and tools like Team Viewer make remote measurement possible - just a thought and I have no idea if Paul has tried remote measurement...

 

Reviewing your graphs I don't think you've done a series of on-axis measurements on 1 speaker at different distances.

Unfortunately large tower speakers like yours with multiple woofers in a D'Appolito config is trickier.

I think the Xoxer between mid and woofer is 125Hz on your speakers.

We're trying to find an issue <300Hz so let's ignore the tweeter response

Do a series of on axis measurements in line with the midrange from close miked (nearly touching the driver) to as far back as possible (still on axis).

We "should" see a smoother result close miked, with more peaks and dips that shift in frequency as you move the mike further away.

 

cheers

Mike

 

 

1 person likes this

Share this post


Link to post
Share on other sites

@almikel @davewantsmoore

Thanks for responses. I am hoping to do some more (proper) measurements during the weekend or Anzac day. I will measure outdoors, on axis and off axis and see how it goes.

Share this post


Link to post
Share on other sites
12 hours ago, almikel said:

 

As Dave said, non-zero negative phase (phase lag) is delay - let's leave "phase lead" alone for a second.

Take the graph above and look at 1kHz - the brown phase line is approx -90 degrees.

To convert this to time delay the formula is

time delay = phase / (360 x F)

So in this case time delay = 90/(360 x 1000) = 0.25 ms

 

Now look at 10kHz - the brown line is approx (30 - 360) = -330 degrees (remember we're looking at a wrapped phase plot so subtract 360 degrees as it's after 1 vertical line)

time delay = 330/(360 x 10000) = .092 ms

 

@andyr - does that make it easier to understand that the delay varies with frequency?

 

 

Yes it does, Mike - thanks.  (Even at a quick glance.)  I'll look at it closer, tonight.  Might have some more Qs, though. :)

 

12 hours ago, almikel said:

 

These are some of the complexities Dave was alluding to in designing Xovers.

 

cheers

Mike

 

OK - understood.  In my case, I'm not interested in designing XOs which are better than what Magnepan supplied - simply in reproducing with a miniDSP, what the passive XO delivers.

 

Andy

 

Share this post


Link to post
Share on other sites

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!


Register a new account

Sign in

Already have an account? Sign in here.


Sign In Now

  • Recently Browsing   0 members

    No registered users viewing this page.

  • Similar Content

    • By robmid
      I just picked up a virtually unused REL Quake sub for $175 and thought I would give it a go with my LS50's.
       
      The aim was to enhance the bass without interfering with the natural attributes of the LS50's.
       
      I have found that using REW setup to measure in room humps and bumps and doing a sweep with the sub OFF then another sweep with the sub ON works well. After a bit of tweaking the sub level, moving the LS50's and sub around everything suddenly came together. A crossover of around 35hZ seems to stop any muddiness or bloat around 80hZ in my room.
       
      I would assume the REL T-5'S would be better but the Quake has really given my LS50's some real weight without destroying the magic. I have tried a REL T-Zero. It improved bass but didn't have quite enough bottom end weight for me.
       
      Green trace is LS50's with sub OFF and orange with sub ON.
       
       
    • By robmid
      My listening space is relatively small and it isn't practical to move furniture around and it is an awkward space with one corner on the left which is partly enclosed by the side of a cabinet. The right side of the room opens into a void with a high ceiling so I get no reflection from that side. The left side has sliding glass doors and curtains so it sounds different with the curtains closed.
       
      I have found that my KEF LS50's respond dramatically to positioning and my REL T-Zero sub crossover frequency and level are very difficult to optimise because the LS50's are so responsive that introduction of extra bass can have a dampening effect on their sound.
       
      I was unable to get the balance right after messing around for months trying to balance my sub and position the LS50's optimally.
       
      I knew there were an enormous amount of REW users but I didn't think it applied to me because I couldn't make physical change to my room. WRONG!
       
      I installed REW on my Mac and used the inbuilt mike but the readings weren't very informative. I didn't want to spend hundreds of dollars on a calibrated mike but as luck would have it, Loud Speaker Kits had Dayton Audio iMM-6 mikes for around $25 so I splurged.
       
      After a some experimenting and variable results I got the hang of it. I took readings without the sub and was able to optimise the LS50's bass response by positioning alone. I couldn't get rid of a dip around 400 hZ but I don't notice any deficiencies.
       
      This gave my my benchmark for the main speakers.
       
      Optimising the sub suddenly became simple because I was able to adjust the crossover frequency to eliminate peaks then adjust the level to give a nice smooth balance between the LS50's and Sub which now pressurises my room nicely despite the unconventional shape of my room.
       
      Now I have an amazing sounding system with gutsy, clean musical bass that has become an extension of the LS50's plus I don't have to get up all the time to fiddle with my sub any more.
       
      Everybody NEEDS to make REW part of their system even if they can't modify their listening space.
       

       
       
    • By robmid
      Hi all.
       
      I am using REW 5.6 on a Macbook.
       
      The graph looks to be ok but I thought the trave should be sitting near 0dB but traces are sitting around 50db. I have configured the sound card, adjusted speaker level, mike input level and dBFS Level is sitting on -12dB and I have adjusted that but it is the same as adjusting the amp level.
       
      I can calculate dB variations but I am wondering if there is a config that allows me to do that apart from adjusting Data Offset for each reading.
       
      I am using a calibrated Dayton Audio iMM-6 for measurements.
       
      Can anyone point me in the right direction please?
    • By Jonesy_SA
      *Please Read Disclaimer at end before posting*
       
      Hi all
       
      I’m pretty excited with what landed on my desk today; 2 x 1 inch tweeters and 2 x 5" woofers.
       

       
      These are OEM from discontinued products, most likely manufactured by Tymphany/Vifa, in turn specifications such as TS and FR are unavailable; which partly explains my purchase. I have assembled speaker kits before but thought it would be fun to measure some drivers myself in order to produce a 2-way bookshelf system from scratch.
       
      Driver data as provided by seller:
      Tweeter

      Tweeter: 25mm Soft Dome, 40 Watts – 120 Watts, 8 Ohm Nominal, 25.5 Hz – 20 KHz, Diameter: 94 mm, Cutout Size: 70mm 
       
      Woofer

      Data 50 Watts RMS 100W peak, 4 Ohm, 88 dB, 54Hz, 1" Voice Coil, Poly doped, treated paper cone. Double magnet. Dimensions: 148mm frame, Cut out size: 120mm
       
      Having trawled Google images I'm pretty confident these came paired in a bookshelf with the following specs:
      Crossover Point 1.9kHz, Nominally 6 ohms, Bass reflex rear vented, Dimensions 295mm high x 175mm wide x 240mm deep, Volume 7 litres internal.
       
      With that said my plan is to approach from a clean slate and not replicate existing designs. I have never measured TS or FR or designed a crossover so I’m hoping you knowledgeable folk can jump in and assist.
       
      I am unable to find a good reference explaining how to measure drivers and implement in a system, however my understanding of the process is: 
      (1) Measure TS parameters of woofer, 
      ARTA:
      http://zobsky.blogspot.co.uk/2008/01/simple-loudspeaker-measurement-jig-for.html
      REW:
      http://www.avsforum.com/forum/155-diy-speakers-subs/1340740-lilmikea-s-diy-impedance-measurement-jig.html
      However this article raises issues with TS data and enclosure design.
      http://www.troelsgravesen.dk/ports.htm
      (2) Design and build enclosure from TS parameters 
      (3) Perform frequency response testing of drivers mounted in enclosure (I think I need to obtain the impedance response of the drivers as well)
      (4) Create crossover from data
      (5) Enjoy 
      (If you have a relevant link please post it – links above are indicative of how I will approach that aspect)
       
      Two software suites come to mind when looking into the above:
      REW
      http://www.roomeqwizard.com/ ... REW manual: http://www.roomeqwizard.com/help/help_en-GB/html/index.html
      ARTA
      http://www.artalabs.hr/
      There are others such as SpeakerWorkshop, SoundEasy etc however they are not free and do not offer lite versions. With that said, I believe most suites use the same algorithms for calculating TS, and REW is apparently quite powerful for FR analysis(?). 
       
      My aim is to update the first few posts with milestones and references to the relevant posts while the entire thread will show chronological order and discussion. 
       
      Feel free to jump in and assist :-)
       
      Cheers 
      Jonesy
       
      #Disclaimer - I will not post the OEM brand details and would ask respondents follow suit. They are a great company that I have purchased from before and I think its fantastic end of line stock is released onto the market and not made landfill; however they may not like their brand name popping up in a DIY thread from Google searches. I would also encourage those that wish to build a speaker kit perhaps look at Loud Speaker Kit (LSK). Their bang for buck drivers are most likely a better option and their kits are well designed#
    • By Cafad
      OK, so I've downloaded Room EQ Wizard and I'm looking forward to spending many hours playing with it, using it to evaluate my new lounge acoustics and getting very confused about room frequencies and nodes and all that great stuff, but...
       
      I need a microphone to play with it properly and since I have such a broad range to choose from (from $12 to over $200, and that's just on ebay) I was just wondering, does the microphone matter much?
       
      Or, to put it another way, which microphone should I buy to share my future analysis trials and tribulations with me.