The No.1 Website for Pro Audio
 Search This Thread  Search This Forum  Search Reviews  Search Gear Database  Search Gear for sale  Search Gearslutz Go Advanced
Two speakers or one speaker for ETF,REW etc?
Old 14th November 2008
  #1
Gear Addict
 
emreyazgin's Avatar
Two speakers or one speaker for ETF,REW etc?

This might be a stupip question, but I need to ask it to make sure..

Does anybody else also thinks that if we use 2 speakers playing the same sine sweep file, at a couple of frequencies, in relation to the measurment microphone and the distance between speakersa and the mic, there will be at least one significant null due to phase cancellation? Is this really a fact? Or does this might mean that one of my speaker cables is phase reversed by accident?
Old 14th November 2008
  #2
Gear Guru
One or Two

I am pondering that one myself at the moment. I have been talking with Chris of FuzzMeasure fame. His default is a single speaker which is certainly easier for my brain to handle. Working reality of course involves two. I think it is very important to measure at ALL significant listening positions (with two or both speakers if you follow me) , then use the Averaging function. Perhaps the end result would then be the same? I will do a test shortly.
DD
Sound Sound - Homepage
Old 14th November 2008
  #3
Gear Addict
 
emreyazgin's Avatar
I am doing some tests with REW, I am testing the response at the listening position one speaker at a time and left and right are quite different..When I do the test with both speakers on there is a null which wasn't in the single speaker tests..
Old 14th November 2008
  #4
Gear Addict
 
emreyazgin's Avatar
Also, even in a PERFECTLY treated room and even with single speaker, depending on your ears' distance from the speaker there will be a null at a certain frequency, right?...So, when doing tests with REW, ETF etc I think we must always measure the distance from the speaker to the microphone and omit the related null?..Any thoughts?
Old 14th November 2008
  #5
Gear Guru
 
Ethan Winer's Avatar
 

Lightbulb

In most pop music, bass instruments are panned to the center and come out of both speakers equally. So to me it makes more sense to have both speakers play, when measuring low frequencies anyway. But you should also check each speaker separately too.

--Ethan
Old 14th November 2008
  #6
Gear Guru
Test

I think it is important and seemingly overlooked, that measurements need to be taken at multiple locations. This might eliminate the difference between single speaker/multiple listen and double speaker/multiple listen.
I have taken to pointing the mic at the left speaker during a left speaker sweep. Then I move the mic a head width to the right, point at the right speaker and sweep. I tend to use at least 8-12 such sweeps.
DD
Sound Sound - Homepage
Old 15th November 2008
  #7
Gear Addict
 
emreyazgin's Avatar
Quote:
Originally Posted by Ethan Winer View Post
In most pop music, bass instruments are panned to the center and come out of both speakers equally. So to me it makes more sense to have both speakers play, when measuring low frequencies anyway. But you should also check each speaker separately too.

--Ethan
Thanks for the comment, but do you also agree that, at a certain frequency, due the distance between mic and speaker there will be a null?
Old 15th November 2008
  #8
Gear Guru
 
Ethan Winer's Avatar
 

Lightbulb

Distance has nothing to do with it except as relates to reflections. If you put a loudspeaker on top of a flag pole, then get a really long boom arm for a microphone, you will not measure peaks and nulls as you change the distance between the microphone and speaker. The only time peaks and nulls will appear is when the microphone gets close to the ground, and ground reflections interfere with the direct sound from the speaker.

Is this what you're asking?

--Ethan
Old 15th November 2008
  #9
Lives for gear
 
Rod Gervais's Avatar
 

Quote:
Originally Posted by Ethan Winer View Post
Distance has nothing to do with it except as relates to reflections. If you put a loudspeaker on top of a flag pole, then get a really long boom arm for a microphone, you will not measure peaks and nulls as you change the distance between the microphone and speaker. The only time peaks and nulls will appear is when the microphone gets close to the ground, and ground reflections interfere with the direct sound from the speaker.

Is this what you're asking?

--Ethan
Ethan,

first off - hi buddy...... hope you're doing great.

Unless I am mistaken he is referring to 2 speaker operation - and the fact that at equal distance from the listener - there will always be certain signals that can introduce either destructive (which is all he mentions) and constructive interference.

Because you have 2 separate transmissions of the same frequency - traveling the same distance and arriving at the same time (for everything panned other than hard left/right) this will occur, but only (obviously) for specific signals with corresponding lengths relating to that speaker distance. Which will probably
be pretty much in the high mid range.

To the original poster......... testing can provide a fairly wide variety of information - not the least of which is the ability to produce an analysis of the degree of accuracy between left and right speaker transmissions.

Everything is (of course) determined by exactly how much effort you are willing to put into this.

The separate testing of left right systems with the exact same transmissions will tell you whether you have problems relating to acoustic symmetry between the left and right sides of the rooms - and also whether you have issues with the speakers themselves.

To produce a test for room qualities I would recommend that you do a series of tests using the same speaker to test the left/right sides of the room. Exacting accuracy with speaker placement is (of course) extremely important for this purpose.

Once you determine that your room is acoustically symmetrical (or you correct for what ever anomaly might exist) then you set the speakers up and test them separately left to right - any anomalies that now exist are within the speakers themselves. Should the speakers be inaccurate - assuming they are at least close to one another - you can correct one of them slightly with EQ to create a matched pair.

Your final tests should be the pair of speakers with low frequent testing included which will bring the final room with all modal activity into the picture.

I hope this helped.

Rod
Old 23rd May 2009
  #10
Gear Addict
 
emreyazgin's Avatar
Quote:
Originally Posted by Ethan Winer View Post
Distance has nothing to do with it except as relates to reflections. If you put a loudspeaker on top of a flag pole, then get a really long boom arm for a microphone, you will not measure peaks and nulls as you change the distance between the microphone and speaker. The only time peaks and nulls will appear is when the microphone gets close to the ground, and ground reflections interfere with the direct sound from the speaker.

Is this what you're asking?

--Ethan
Dear Ethan,

I am so sorry, I have just came across this post of mine and realized somehow I couldn't reply back, I felt really bad, I apologise for leaving your question without a reply..


This is also true for Rod, sorry about forgetting about this post.

I do remember reading your answers but somehow forgot to post back.

I still have the same problem now.Rod, yes that was exactly what I meant but in my case it is happening at a low frequency as well (around 80hz) when I turn both speakers on 82hz dissappears, and when I turn one of them off (either of them) it is louder..
Old 23rd May 2009
  #11
Lives for gear
 

Personally, I like to do both. I measure just left, just right, and both. That gives me a good idea of what is happening for each one in relation to boundaries and listening position. The 2 together lets me see how the 2 are interacting and potentially causing cancellations and peaks.

As you said, reality dictates that you'll be listening to both so that's what needs to be right. The single measurements are just extra data points to show you what's causing what.

Bryan
Old 23rd May 2009
  #12
Here for the gear
 
SuperSheep's Avatar
 

Any time two waves meet, there is interference. This interference can be constructive (in phase) or destructive (out of phase). You can picture sound waves as areas of low pressure and high pressure or if you wish, a sine wave. When areas of low and high pressure meet, or the peaks and valleys of sine waves, there will be destructive interference or a tendency towards null. Whenever areas of low or high pressure meet each other, or peaks or valleys of sine waves, there will be constructive interference or an increase in the peak/valley.

A single speaker creates sound waves that move outwards from the speaker. If this speaker were in a space with no reflections, there would be no secondary waves to interact with the direct waves from the speaker and no interference.

Two speakers mounted near each other will interfere with each others sound waves as now there are two sources of sound waves and hence interference is the result. Interference and hence peaks and nulls in the amplitude of the sound waves will occur and will be dependent on the frequency and the position around the speakers.

A single speaker in an enclosed room: The sound waves emitted by the speaker will reflect off of the wall surfaces and interfere with those sound waves coming directly from the speaker. This is a highly complex interaction that is dependent upon the frequency emitted, the distances traveled by the waves and the listening position.

Multiple speakers in an enclosed room: The sound waves emitted by the speakers will interfere with each other directly and with the reflected waves.

Every object in the room including the listener will tend to alter the characteristics of the sound waves even more as some objects will reflect waves, some absorb and this situation will create peaks and nulls in the room that are quite complex and nearly impossible to calculate.
Old 25th May 2009
  #13
Gear Guru
A Test

I promised to do this test a millenium or so ago. Finally here it is.
The Red curve represents L+R, both speakers driven.
The Blue curve is an RMS average of two measurements, L only driven and R only driven.
Two speakers or one speaker for ETF,REW etc?-comparison.gif

We can see that there is little difference between the two techniques until we get into quite High Frequencies. The L+R curve shows how the two speaker sources cancel at the single measurement point. The ears are not a single point however, so I reckon the RMS average probably represents what one would hear more accurately. Perhaps the ultimate technique for shooting a room would entail a stereo pair, even a dummy head. The original question would of course reappear, one speaker or two? I am in the habit of measuring using pairs of spots, about a head-size apart. I shoot the left speaker with the left spot and so on. I didn't describe that in my primer as I was not sure of any real benefit.
So much for Frequency Response., let's look at Modal. Genelec have a paper which suggests that several subs drive a room's modes more evenly than a single. On this basis, perhaps the L+R driven measurement gives a better view of the room's modes. Unfortunately my old version of FM does not allow for Waterfalls of the RMS average. I can say though, that the Waterfall of L+R looks a lot better than either L or R.
Conclusion? I suppose I will have to test with a stereo ear spaced pair, again with single and dual speaker dirive. However in the meantime I reckon that single drive with RMS averaging is a good choice.

DD
Old 25th May 2009
  #14
Lives for gear
 
ciro's Avatar
 

Quote:
Originally Posted by bpape View Post
Personally, I like to do both. I measure just left, just right, and both. That gives me a good idea of what is happening for each one in relation to boundaries and listening position. The 2 together lets me see how the 2 are interacting and potentially causing cancellations and peaks.
Exactly.

Ciro
Old 25th May 2009
  #15
Gear Guru
Thought for the day...

I have been considering this topic a little further. I now think that single speaker drive is the one. In the case of two speakers, a single microphone will show vividly the cancellations caused by variations between the two paths. See my test in an earlier post. However this is not at all similar to a single ear. In case of the ear, one of the paths is blocked by the head. This would greatly diminish or eliminate the interference effects, which from the tests are at high frequencies, i.e. entirely blocked by the head. I believe that the single speaker to single microphone path measurement model is closer to this reality than dual drive.
DD
Sound Sound - Homepage
Old 26th May 2009
  #16
Lives for gear
 
Weasel9992's Avatar
 

The issue I have with testing through only one speaker is this: if we're looking to determine the total room response in a realistic way, then don't we have to have both speakers active? I mean, that's how you're going to use the system...you'll never use only one speaker at a time, so while that's interesting information (particularly when we're trouble shooting), it doesn't seem to me to be very relevant. My comments assume that the speakers have already been accurately positioned, and that we're testing for room characteristics not speaker location.

Frank
Old 26th May 2009
  #17
Gear Guru
One sided

Hi Frank, thanks for joining it. I am still tossing this about, but am leaning heavily towards single speaker drive. I note that all of the prescribed test methods for building acoustics, in the text books use single drive. My reasons are unique to our purpose though.
Let me try again. Only a single microphone will get the cancellation effect from two clear paths from the two speakers. This does not represent the ear well. Each ear will receive a single path only, due to the head obstructing the other path. If we were listening with one ear only, at the front of the head, with two unrestricted paths, then two speakers would be fine.
The Final Front Ear?

DD
Old 26th May 2009
  #18
Lives for gear
 
Weasel9992's Avatar
 

Quote:
Originally Posted by DanDan View Post
Hi Frank, thanks for joining it. I am still tossing this about, but am leaning heavily towards single speaker drive. I note that all of the prescribed test methods for building acoustics, in the text books use single drive. My reasons are unique to our purpose though.
Let me try again. Only a single microphone will get the cancellation effect from two clear paths from the two speakers. This does not represent the ear well. Each ear will receive a single path only, due to the head obstructing the other path. If we were listening with one ear only, at the front of the head, with two unrestricted paths, then two speakers would be fine.
The Final Front Ear?
I hear you Dan, and what you're saying makes sense. In practice I typically work on speaker at a time, then both just like Bryan suggested. The issue I'm raising is that there's more to it than just determining what/where the high mid/high frequency cancellations are. Room interactions across the frequency spectrum are much more complex than a single speaker would allow you to see. I'm question the simple utility of single-speaker testing, I suppose.

Frank
Old 26th May 2009
  #19
Gear Guru
Test

Hi Frank, did you see my test and earlier above? It strongly suggests the interactions of interest occur at high frequencies only. The modal and RT issues are somewhat different, Genelec suggest more subs is good. To fully test the modal response or to get RT of a room, multiple locations for a single speaker is the norm in ISO etc. However, I see little difference between L, R, and L+R here. In fact I reckon the best way to stimulate the room properly to get a decent waterfall and octave band RT's is to put the speaker in the corner on the floor.
DD
Old 26th May 2009
  #20
Gear Maniac
 

An important reason for making measurements with both main speakers running, as mentioned earlier in the thread, is the tendency for LF to be mixed equally into both in source material. When speakers placed symmetrically across the width (defining width as the axis the speakers are aligned along) fed with mono LF drive the room they prevent the excitation of odd order width modes, so a width mode that may appear problematic when measuring individual speakers may be a non-issue with typical source material. At HF, path length differences between the speakers and the mic generate severe comb filtering which, as Dan points out, is much more severe than we would ever hear as neither ear has an unrestricted path to both speakers when facing forward. Similar issues mean it is important to take measurements of subs both with and without the main speakers running, as again a mode that is excited by the sub alone may not be excited (or not as much) when the contributions of the main speakers are added in - more awkward for HT installations as the LFE will be directed to the sub only whilst LF content in mains that are not full range will be partially redirected to the sub per the crossover setting.
Old 27th May 2009
  #21
Gear Guru
Good Question

To the original Poster, hardly a stupid question eh? Thanks for bringing this up.
JohnPM, a very welcome, informative and further thought provoking post, thank you too.
DD
Old 27th May 2009
  #22
Lives for gear
 
Weasel9992's Avatar
 

Nice post John...it's great to have you here.

Frank
Old 4th February 2015
  #23
Gear Head
 

Quote:
Originally Posted by DanDan View Post
I promised to do this test a millenium or so ago. Finally here it is.
The Red curve represents L+R, both speakers driven.
The Blue curve is an RMS average of two measurements, L only driven and R only driven.
Attachment 122487

We can see that there is little difference between the two techniques until we get into quite High Frequencies. The L+R curve shows how the two speaker sources cancel at the single measurement point.
DD
What software were you using to do this response averaging, DanDan? I just tried this in REW the other day and it actually exacerbated the null in the highs, and dramatically so. Am I doing something wrong? I had the left and right responses open and used the (A+B)/2 command in the control panel.*EDIT: I just answered my own question; somehow I missed the big "Average Responses" button in the lower left corner of the trace display. Though this leads me to another curiosity: What does the (A+B)/2 command actually do if not average the responses? The results are drastically different between the two...

I also have a question related to the notion of assessing the modal region with one vs two speakers. I currently have my speakers arranged relative to the listening position in a way the yields the flattest modal region response when both speakers are driven. However, each individual speaker has quite a different modal response from the other. If I plan on applying corrective EQ to each speaker, should I apply it to the modal region as well (essentially creating a symmetrical left and right bass response), or leave things as is considering the summed response is pretty good? What are the pros and cons of trying to EQ the same modal response out of both speakers?

Last edited by MaximalC; 4th February 2015 at 03:19 AM..
Old 4th February 2015
  #24
Gear Guru
Aritmethic

My graph has both speakers running together. Due to the slight differences in distance from mic to tweeter (and some uneven room reflections) there is huge cancellation and comb filtering going on. Adding SPL graphs can hardly simulate this unless the difference in distance is stated. I note that FuzzMeasure, with Loopback Correction engaged, does Time Domain Combining. I haven't tested to see if Waterfalls averaged from L and R single speaker measures look similar to a WF generated with the two speakers running simultaneously.
The same 'I don't know' applies to your last question. DRC products that I am familiar with test driving one speaker at a time, but I don't know whether the do some form of averaging before generating filters or do they create individual speaker filters and then do some sort of averaging on them. These are IMO great questions,. PerhapsChris from FuzzMeasure, and/or JohnPM and or Mathias from Dirac Live, will comment.

DD
Old 4th February 2015
  #25
Gear Maniac
 

Quote:
Originally Posted by MaximalC View Post
What does the (A+B)/2 command actually do if not average the responses? The results are drastically different between the two...
The "Average" button is an SPL (magnitude only) average, the (A+B)/2 trace arithmetic is a vector average, which takes into account the phase of the response. To use trace arithmetic successfully the responses should be properly time aligned, there are fuller explanations of each in the REW Graph Panel help.

Quote:
I also have a question related to the notion of assessing the modal region with one vs two speakers. I currently have my speakers arranged relative to the listening position in a way the yields the flattest modal region response when both speakers are driven. However, each individual speaker has quite a different modal response from the other. If I plan on applying corrective EQ to each speaker, should I apply it to the modal region as well (essentially creating a symmetrical left and right bass response), or leave things as is considering the summed response is pretty good? What are the pros and cons of trying to EQ the same modal response out of both speakers?
It is usually best to apply the same corrections to both left and right unless you are specifically correcting a known difference between the speakers themselves. Applying different corrections to left and right can have an odd effect on stereo imaging.
Old 5th February 2015
  #26
Gear Guru
Quote:
correcting a known difference between the speakers themselves.
Also the two sides of a room can be different at LF. e.g. A brick side wall on one side a studded plasterboard on the other.
DD
Old 6th February 2015
  #27
Gear Head
 

Quote:
Originally Posted by JohnPM View Post
The "Average" button is an SPL (magnitude only) average, the (A+B)/2 trace arithmetic is a vector average, which takes into account the phase of the response. To use trace arithmetic successfully the responses should be properly time aligned, there are fuller explanations of each in the REW Graph Panel help.
Ahhh, thank you, that is helpful to know. That would suggest trace arithmetic wouldn't be a bad way to substantively validate time alignment then, no? Where superior time alignment would yield the mildest null in the highs? I realize this can also be done with ETC, but it's an interesting thought nonetheless.

Quote:
It is usually best to apply the same corrections to both left and right unless you are specifically correcting a known difference between the speakers themselves. Applying different corrections to left and right can have an odd effect on stereo imaging.
Are you meaning with regard to the modal region exclusively, or to the entire frequency range? If it is the latter, wouldn't the stereo image already be skewed by the differing reinforcements and cancellations due to the asymmetry of the room (as mine is)? For example, my right channel response (including the room) features a 5 dB greater peak at 600Hz that is not found in the left channel. Am I best to leave it alone as opposed to notch out that portion of the right channel with EQ? That would be quite a departure from how I had always understood the purpose of DRC, but then again I probably still have much to learn.

Thank you for the info!
Old 6th February 2015
  #28
Gear Maniac
 

Whether to try and address individual channel response anomalies depends on what caused them. If they are due to reflections from surfaces that are quite close to the speaker or listener there might be benefit in trying to address them, if not your ear/brain will distinguish between the direct sound from the speaker and the later contributions from room surfaces and attempts to apply correction will alter that direct sound and could easily make things worse rather than better. It is quite a complex area though, whether individual reflections are perceived separately from or as part of the direct sound depends on a range of factors one of which is the material being reproduced at any moment (including what comes just before and just after it). Keeping left and right adjustments the same is least likely to make things worse.
Old 7th February 2015
  #29
Gear Head
 

Quote:
Originally Posted by JohnPM View Post
Whether to try and address individual channel response anomalies depends on what caused them. If they are due to reflections from surfaces that are quite close to the speaker or listener there might be benefit in trying to address them, if not your ear/brain will distinguish between the direct sound from the speaker and the later contributions from room surfaces and attempts to apply correction will alter that direct sound and could easily make things worse rather than better. It is quite a complex area though, whether individual reflections are perceived separately from or as part of the direct sound depends on a range of factors one of which is the material being reproduced at any moment (including what comes just before and just after it). Keeping left and right adjustments the same is least likely to make things worse.
Sounds like an open and shut case then... Seriously though, it would appear that the most conservative, do-no-harm approach would be to restrict frequency response corrections to regions that have comparable deficiencies in both channels. Anything beyond that may be best left to trial and error, if at all (at least as a starting point). Thanks again!
Old 7th February 2015
  #30
Gear Guru
Symmetry

If one speaker of a pair had a slightly less efficient woofer or LF amplifier, I would readily turn up the LF control on it. Similarly if one speaker is next to a solid brick front wall, reinforcing the LF, while the other was at a window and thus not gaining a similar boost, well I would boost the weak one or turn down the stronger. i.e. Make them more equal by applying unequal adjustments.

Mathias of Dirac Live has shown me some rather spectacular results using multiple channels with individual filters on every one. Note they do high end systems in luxury cars and aircraft. 14.2 even!

One could look at the simple magnitude, or move on to include phase, when doing additions or averaging. But what of the different response caused by the different positions in the room. e.g. one speaker is in a modal null, the other not, or in a peak? Or the two speakers are in the nulls of the second Width mode?
FuzzMeasure does offer timed domain combination but I wonder does it or can any calculation take account of these location factors.

DD
Post Reply

Welcome to the Gearslutz Pro Audio Community!

Registration benefits include:
  • The ability to reply to and create new discussions
  • Access to members-only giveaways & competitions
  • Interact with VIP industry experts in our guest Q&As
  • Access to members-only sub forum discussions
  • Access to members-only Chat Room
  • Get INSTANT ACCESS to the world's best private pro audio Classifieds for only USD $20/year
  • Promote your eBay auctions and Reverb.com listings for free
  • Remove this message!
You need an account to post a reply. Create a username and password below and an account will be created and your post entered.


 
 
Slide to join now Processing…
Thread Tools
Search this Thread
Search this Thread:

Advanced Search
Similar Threads
Thread
Thread Starter / Forum
Replies
beerbarrel / Gearslutz Secondhand Gear Classifieds
1
JB872 / Studio Building / Acoustics
3
emad / Studio Building / Acoustics
12
Kris75 / Studio Building / Acoustics
1
kissingonstilts / Live Sound
18

Forum Jump
Forum Jump