New to Dirac - a couple of questions and a lot of graphs

[beanbag]

Hello folks,

 

I have a set of Martin Logan electrostats set up in a funny arrangement.

 

 

Normally I sit facing the desk, and occasionally I will turn sideways to face the speakers. I figured that with a crooked room setup like this, I would use Dirac to help "fix" the frequency response and maybe improve the imaging.

 

Question 1) When making the 9 position measurement, it seems that only one of them is at ear level, and all the others are either above or below that? Some of the measurement points will be over the surface of the desk. Is that ok?

 

OK, so I ran the program and got this for the freq response:

 

 

And this for the impulse: (not much to fix)

 

 

Then I checked the results in Room Equilization Wizard. Here is the frequency response:

 

 

Question 2) Umm, what happened to my supposed high frequency boost?

Q3) Why does Dirac show roll-off at 10 kHz, and REW shows it at 8 kHz?

Q4) It seems Dirac flattened things out below 150Hz, but then wasn't very successful at the higher frequencies. Was it supposed to do anything?

 

Here is the phase response. It seems like Dirac made it change more smoothly.

 

 

Q5) So what would it mean that the phase changes more slowly?

 

Dirac made one of my speakers down by 0.1 dB, and delayed by .06ms. That just means my mic wasn't exactly equidistant between the two speakers. To try to make a fair comparison, I turned off the attenuation and delay when comparing.

 

Most importantly the listening test. To be honest, I can't really hear the difference. The only thing I can tell is that if I turn on the filter, it makes a pop and things sound louder for a fraction of a second. When turning the filter off, I can't tell anything at all. Maaaaayyyybe sometimes an instrument or voice sounds a little louder or more forwards. I can't tell a change in imaging, timbre, clarity, or anything.

 

Q6) I guess it is a fact that the bass response is more smooth, so maybe if I listen to music with content below 150Hz I should notice something? Or what differences should I listen for?

 

I was mainly looking to improve imaging and "presence", like what I keep hearing in various audio reviews about feeling like you are in the middle of a performance, and can point out all the instruments, the "size of the voices", etc. I have never experienced that. On my system the "clarity" is already pretty good, but sometimes the imaging is a little off, like I can't quite tell where something is coming from. Another problem is that occasionally an instrument or voice drifts left and right depending on what frequency or noise it makes. (e.g. the "s" sound of a voice comes from a slightly different location than the vowels)

[wgscott]

Can you set your speakers up like this:

 

[beanbag]

Short answer no - there's too much crap in the way.

[flak]

Hello beanbag,

 

I think that the thread that you started is very interesting and that it deserves an answer as detailed as your questions and measurements ... I'll be back

 

Flavio

[Boris75]

I find this thread very interesting. I am looking forward to the continued discussion.

[flak]

Hello Beanbag

 

the first thing that came to my mind when I have read your post was that, even if you figured that your set-up is a crooked one, it actually is one of the very best... so I think it is interesting to see how useful is Dirac Live in these cases, but let me explain first what I think of your set-up.

 

The speakers and the listening spot are diagonally placed... while this may eventually be visually unattractive it has many (sometimes unexpected) advantages:

 

- the woofers are placed at different distances from the walls so they do not produce the same resonances at the same frequencies that would double their effect...

 

- the speakers are panels that are dipoles at higher frequencies... the direct front radiation will not be smeared by first reflections because the side ones are cancelled while the back radiation will be delayed and attenuated by the long travel so that our brain will almost ignore it

 

- the listening spot is placed in a way that will make the rear walls reflection bounce away from you instead of interfering with the direct radiation (as a parallel and near wall would do instead)

 

b.t.w impulse response of phase coherent planar speakers is better to start with

 

Also the room, if we do not take into account the small wall and half wall, actually does have pretty good ratios.

I imagine that the height is close to 8 feet so that the ratios are 1:1,75:2,5

This happens to be one of the almost ideal ones (if interested I can enclose details about room ratios)

 

Finally it looks like the room has passive treatment and bass traps.

 

The synergy of all the above factors may lead to think that there is no room for improvement, but your measurements (both the REW and Dirac ones) show that this is not the case.

We will go through them but it has to be said that they cannot be directly compared.

 

The Dirac measurement is an average of nine positions (FR changes significantly at high frequencies depending on position and mic orientation) and the smoothing is a different one (48 dBs in your case), it also seems to me that the REW measurement is the sum of two channels while Dirac measures them independently and the REW plot stops before Dirac does.

 

Anyhow all of the above is not that important because we can see what happened in both measurements, and I think they are consistent.

 

Dirac Live treats the spectral colouration in a room after it has performed a time-domain correction and it is unique in correcting the impulse response in a relatively large listening area... let's start by that, the easiest way to do it is looking at the impulse response.

 

In you post you suspect that your system may not need improvement ("not much to fix")... so let me say what we have to look at in an impulse response.

 

An ideal response would have an infinitely thin vertical line above zero... so nothing below it, and you can see that there is a substantial improvement after correction... it should be as high as possible, and you can see that it is twice as high after correction... and nothing after, you can see the ringing just after the pulse that disappears after correction.

 

We said that DiracLive treats the spectral colouration in a room after it has performed the time-domain correction

It also makes sure not to overcompensate. This is done by paying attention to the variations across different measurement positions, not just the average.

 

So let's now look at the frequency response... you can see what happened in both measurements.

 

As you correctly pointed out there is a substantial improvement at the low frequencies, a dip slightly over 100 Hz is filled for something like 9 dBs and other dips are improved as well.

As far as frequency response at the higher frequencies is concerned there is an improvement also, if you compare the target to the average it is at most points very close to the average, and where they are different there is a difference as well. This is most obvious at the dip at 2 KHz.

 

Also the correction has to be selective as we do not want to correct for ripple that it is not a property of the frequency response that we perceive.

(things get complicated but you can find a document about it here:

http://www.dirac.se/media/12044/on_room_correction.pdf)

 

Finally you mention the fact that "to make a fair comparison, I turned off the attenuation and delay when comparing"

While I understand that the intent was a laudable one I have to say that doing so is not a good idea...

the speakers may have a frequency response or another depending on target curve, but in any case the sources have to be perfectly correlated both in frequency response and time of arrival.

 

So when you say that "occasionally an instrument or voice drifts left and right depending on what frequency or noise it makes. (e.g. the "s" sound of a voice comes from a slightly different location than the vowels)" you are right, but it may depend on the fact that if one frequency sounds one or two dBs louder on a speaker, and you then switch to a frequency that sounds one or two dBs louder on the other, the apparent source will move sideways.

Again you will find a more detailed discussion in the document that I mentioned before.

 

Now this is probably an excessively long post already but I have not addressed the point that you hear a very small or no difference at all.

The possibility I can think of is that the filter may seem to be on, but it isn't... is the "streaming" light lit in the DAP?

 

Thanks for your time

Flavio

[beanbag]

Hi flak,

 

Thanks for taking the time to respond.

What is that picture you posted?

Can you go back and answer Q1 and 5?

 

As for turning off the gain and delay, they were only .1dB and .06ms, or about 2cm worth of listening position. The first time I ran Dirac, I had my mic even more off center, and the delay was about .23 ms, or about 8cm. When I did the A/B comparison back then, I first noticed a difference in sound quality. Then I realized that voices were being pulled over to one side. Then when I noticed what the delay was and turned it off, I went back to hearing almost no difference. So it is not "fair" to do a listening test where one setting assumes your head is in one position, and another setting has your head elsewhere.

 

Yes, I am sure the Dirac filter is running because I can purpose mess up the delay and gain setting and notice a difference.

 

One difference I think I can notice is that under Dirac, some bass notes don't ring for as long, and bass notes have a little bit more "texture". What I mean by this is that notes from an organ pipe, which have a low fundamental frequency plus higher buzzing harmonics, have a little more of this "buzz" under Dirac. I am not sure if it is simply a matter of equilization, or whether Dirac fixed some aspect of the phase relation between the fundamental and harmonics.

 

Dirac did do a good job of cleaning up the low frequency decay times.

 

But at such low frequencies, it is also kind of hard to notice.

[flak]

Hi Beanbag,

 

thanks for sending me your .mdat file with your REW measurements,

I'll extract the relevant images to answer your questions and I'll post my comments tomorrow.

 

Flavio

[flak]

Hi flak,

 

Thanks for taking the time to respond.

What is that picture you posted?

Can you go back and answer Q1 and 5?

 

Hi Beanbag,

 

the picture that you see at the bottom of my previous post was about room ratios... I had removed it when I realized that the post was becoming too long but it "survived" as an attachment.

The source is an interesting document about room dimensions for small listening rooms from the University of Salford (acoustics.salford.ac.uk) with Excel files:

Room Sizing Tutorial | Acoustics, Audio and Video | University of Salford

 

Question 1... it is ok to have measurement position over the desk if you expect to be listening in the area. The measurement points should be in positions that represent the expected listening area

 

Question 5... interpreting the phase plots that you have posted is not easy as the phase is wrapped, it is easier if unwrapped (REW allows that) but comparing the group delay plots is even better.

The group delay describes the delay of each frequency so this should optimally be constant for all frequencies.

 

 

I have extracted the above image from your data file, as you may notice even if your listening room has already been passively treated the general behaviour is better with the Dirac correction.

 

Two other interesting plots that I have extracted from you data file are the waterfalls, an ideal waterfall would show a nice smooth response with very even and quick decay:

 

 

 

Again I think that the improvements by Dirac Live are visible (and more important audible) even in your already good listening room.

 

Good evening

Flavio

[beanbag]

Thanks for the reply.

 

Regarding the measurement positions for the mic, how big of a "cube" do I need to make? (If it represents the expected listening areas, then that is pretty small because my chair doesn't move that much) Only the initial measurement position is actually at ear level. All the other measurements are either above ear level, or below. That doesn't make sense to me. For example, one of the position is behind, below, and to the left of the regular listening position. That makes it 6" above the level of the desk. (If you look at my diagram where the chair faces towards the speakers, and therefore away from the desk.) No, I am not going to put my head there.

[flak]

Thanks for the reply.

 

Regarding the measurement positions for the mic, how big of a "cube" do I need to make?

 

cut and paste...

"Avoid making measurements in a too small space. Even for the “Chair” listening environment, it is important to spread out the microphone positions in a sphere of at least 1 meter of diameter. A too small space will result in over-compensation that will sound very dry and dull"

 

Ciao, Flavio

[beanbag]

cut and paste...

"Avoid making measurements in a too small space. Even for the “Chair” listening environment, it is important to spread out the microphone positions in a sphere of at least 1 meter of diameter. A too small space will result in over-compensation that will sound very dry and dull"

 

Ciao, Flavio

 

OK, maybe laterally this makes sense, but I still don't get why we need to take measurements above and below ear level.

[sdolezalek]

The original question was posed with regard to Martin Logan electrostats, mine is more specific to Magnepan's, but both are dipole designs. In a recent review of the Magnepan's, the reviewer included the following advice:

 

"Magnepans are dipolar speakers that produce sound to the front and to the rear, but not to the sides of, their thin panels. As a result, Magnepans tend to interact less with rooms than conventional piston-type loudspeakers typically do. They also exhibit very different volume fall-off characteristics as the listener moves farther away from the speakers (with Maggies, volume falls off at 1/x, where x is the distance between the listener and the speakers, where with conventional speakers volume falls off at 1/X2).

 

Given factors like these, the fact is that Magnepans tend not to respond well when used with most automated speaker setup/room EQ systems (most of which were designed with the characteristics of conventional speaker systems in mind) so that Magnepan strongly advises against using them."

 

My guess is that the reviewer was referring to the EQ systems that Pioneer, Denon, etc put in their receivers and not necessarily Dirac. However, it made me wonder whether Dirac has any way of compensating for these differences in dipoles? Does it treat box and dipole speakers the same way?

[flak]

OK, maybe laterally this makes sense, but I still don't get why we need to take measurements above and below ear level.

 

In the "chair" model we can think of a one meter sphere (the measurements one) floating in a box with walls at distances of the same order of magnitude, the floor and the ceiling are treated the same way of the lateral walls.

In the next post we'll discuss why nine measurements and why those measurements are used in a way that is different from the usual one.

 

Flavio

[Fitz Music]

In the "chair" model we can think of a one meter sphere (the measurements one) floating in a box with walls at distances of the same order of magnitude, the floor and the ceiling are treated the same way of the lateral walls.

In the next post we'll discuss why nine measurements and why those measurements are used in a way that is different from the usual one.

 

Flavio

 

I'm confused about your use of the term "sphere". All the Dirac material I've seen illustrating the chair mode shows a rectangular solid, the ends of which, after the establishment of the central head position, determine the remaining eight measurements. Or is it a cube, distorted by the artist's perspective? Or does it matter, so long as the long side of the rectangular solid is at least one meter? What is a good distance choice above and below the head position?

[flak]

I'm confused about your use of the term "sphere". All the Dirac material I've seen illustrating the chair mode shows a rectangular solid, the ends of which, after the establishment of the central head position, determine the remaining eight measurements. Or is it a cube, distorted by the artist's perspective? Or does it matter, so long as the long side of the rectangular solid is at least one meter? What is a good distance choice above and below the head position?

 

My bad... may be I should have not quoted the word "sphere", the intended meaning is that the minimum distance of the mic from the central head position should be cm.50.

A cube with a 1 meter side will be OK as well as larger, but not smaller, parallepipeds of various proportions.

 

Flavio

[flak]

Hello sdolezalek,

 

your question is another interesting one... the reviewer is correct in saying that the side reflections of panels are minimized by cancellation at certain frequencies because of the interaction of the front and back radiation, as well as he is right in saying that because of their height (and floor and ceiling reflections) their volume fall off is more similar to a linear source than a single point source.

 

So in a panel we have a dipole behaviour with a back radiation which is as strong as the front direct radiation.

It's worth noting that the back radiation will travel a relatively much longer path because of the toe-in so that it will bounce on a few walls before reaching the listening point.

 

As a result the back radiation will not only be attenuated but also delayed in its arrival time vs. the direct radiation.

While the brain will discriminate the delayed reflections and subjectively attenuate their contribution (Haas effect) this does not happen in a frequency response measurement which completely removes the time domain information (the time of arrival does not affect the measurement)

 

So how can Dirac Live discriminate the late reflections from the direct radiation?

This can be done thanks to its unique algorithm that makes use of nine measurements.

Those are not used as usual to make an average but they are used to identify what changes, and what does not change, in different positions.

 

Because the rear reflections will have bounced on different walls their measured response is different in different positions... and they have not to be corrected.

In a few words Dirac Live will properly correct both the conventional speakers and the planar ones.

 

Ciao, Flavio

[sdolezalek]

Dirac Live will properly correct both the conventional speakers and the planar ones.

 

Ciao, Flavio

 

Flavio: Thank you for the response. Your answer makes sense and I will try it. One additional question: Does the demo version of Dirac allow me to test a 3.1 (R+L+center+Subwoofer) or 7.1 surround system or do I need to purchase the full version in order to test that type of system?

[flak]

Flavio: Thank you for the response. Your answer makes sense and I will try it. One additional question: Does the demo version of Dirac allow me to test a 3.1 (R+L+center+Subwoofer) or 7.1 surround system or do I need to purchase the full version in order to test that type of system?

 

Having to purchase the full 7.1 version in order to test it would be unreasonable, the demo version can be used both in stereo and multichannel.

 

If you use a home theater PC you may have a look at the following instructions on connecting the PC to an HDMI input on the receiver (e.g. AUX or PC):

http://diracdocs.com/PCtohomecinemareceiver.pdf

 

Good listenings,

Flavio

[beanbag]

edit: never mind

[Boris75]

I just downloaded the trial version for Mac, which I haven't installed yet because the environment is too noisy. There is too much traffic today, including frequent honking, in the street where I live.

 

My problem is that I am not sure I am going to have an opportunity (which requires available time on my side and quietness in the street) to test Dirac before a couple of weeks.

 

My question is: does the trial period start when the software is installed or when it is ordered from the website?

[flak]

Hello Boris,

the trial period starts when it is installed

 

Flavio

[Boris75]

Hello Boris,

the trial period starts when it is installed

 

Flavio

 

Hi Flavio,

 

Thanks for your message. I had time on my hands and a quiet environment this morning, so I took the plunge. My first surprise was that, with the UMIK-1 on Mac OSX, quite a high sound level is required, even with the input gain at its maximum. This made me fear for my tweeters, and this also convinced me that my concern for low-level background noise was probably misplaced. Since repeating these potentially tweeter-damaging test noises at quite high SPLs nine times frightened me quite a bit, I measured only my sofa (while initially I thought that I would also measure the chair in front of my computer).

 

The main outcomes were:

• Imaging improved significantly thanks to Dirac,
  
• Sound remains very clear (Dirac does not muffle anything)
  
• My tweeters still work, and
  
• Dirac revealed that my subwoofer is not functional anymore.
  

 

Dirac did nothing to my sub: the frequency response below 70Hz is flat on all measurements. What it did is that it revealed that it does not work anymore. It is quite a bit embarrasing that I did not realise this for myself even if, in retrospect, I understand why, lately, I felt that some of my favourite organ CDs lacked oomph. Well, it's quite clear why now. Still, the ability of the ear-brain system to reconstruct what is missing keeps astonishing me.

[beanbag]

You can unscrew the cap on the umik-1 and flip some toggle switches to increase the internal gain of the mic. That's what I did.

[flak]

You can unscrew the cap on the umik-1 and flip some toggle switches to increase the internal gain of the mic. That's what I did.

 

Yes, my understanding is that there are dip switches inside the mic... and I heard that the mics have been delivered in the past with a setting that was lower that the present one.

 

Flavio