Emotiva XMC-1 Processor with Dirac Room EQ Review

[RichB]

For transparency's sake let me say first that I may be biased in favor of Dirac Research.

Another measurement system would closer replicate the Dirac's graph that is representative of the average of the nine measurements, Dirac graphs are accurate... b.t.w. they have been validated by indipendent third parties but I cannot post the link because I don't have enough posts.

Each REW measurement is an average of the 9 Dirac positions. REW sweeps used the Java drivers (due to Windows 10 issues) and both speakers. That is a difference. Still, the curves within REW were stable and can be reasonably compared.

REW is accurate also... but I have serious doubts about that REW graph if it is a full bandwith one (that's the reason I conjectured that the correction had been limited to the low frequencies)
First I doubt that Rich would be satisfied listening with the garbled response above 400 Hz of that REW graph BEFORE the Dirac correction.
Second I doubt that the AFTER correction graph is representative of a proper Dirac correction, if any.

In a few words no creative tech guys moved from Germany to Sweden
may be something went astray in that REW measurement.
Flavio

I do not believe that REW are off for the higher frequencies. I plan on trying more experiments in a week for so.

- Rich

 

[RichB]

Maybe a couple of the creative tech guys from Germany made the move to Sweden after they finished designing the VW emissions control software!

Seriously, I bet another measurement system like Omni-mic would closer replicate the REW measurements.

The Salon2's were on order for a long time. In the interum, I used the Studio2's as the mains.

These measurements show the Studio2's (I did not have the Salon2's at the time) taken with the OmniMic2 to compare Reference Stereo (aka Pure/Direct) and Dirac.

The Dirac default target curve is flatting the bass but providing a boost between 80 Hz and 300 Hz. If you believe the OmniMic2 there is no dip to be corrected. This is the way it sounds too. I did not trust either program above 80 Hz which is why I used the Dirac curtain to limit correction to 80 and below.

- Rich

 

[flak2]

Hi Kurt,

yes, I'm affiliated with Dirac Research and yes, I also think that Rich knows what he is doing.
After making that clear do you think that Rich is listening with full satisfaction in a room with the response shown above 400 Hz in this BEFORE CORRECTION graph?
Furthermore do you think that Dirac Live applies no correction at all (measurement tolerances aside) as shown in this AFTER CORRECTION graph?

So please understand why I have serious doubts about that graph and I think that something may have gone astray...

Flavio

P.S. I've just seen that Rich has been posting while I was writing... I'll read his comments now.

 

[PENG]

Hi Kurt,

yes, I'm affiliated with Dirac Research and yes, I also think that Rich knows what he is doing.
After making that clear do you think that Rich is listening with full satisfaction in a room with the response shown above 400 Hz in this BEFORE CORRECTION graph?
Furthermore do you think that Dirac Live applies no correction at all (measurement tolerances aside) as shown in this AFTER CORRECTION graph?

So please understand why I have serious doubts about that graph and I think that something may have gone astray...

Flavio

P.S. I've just seen that Rich has been posting while I was writing... I'll read his comments now.

I can't believe the Salon2 would yield graphs like that regardless of the room. I would have to assume that something went wrong in the process.

 

[RichB]

REW has been around a long time and has proven itself capable if the person running it knows what they are doing (and I believe Rich does).

Regardless, I hope some effort is made to determine why there is a conflict.

I had other look over my files. I will be trying the REW ASIO drivers in a bit and I have not given up.
As a result of this unpredictability, I now own an EMM-1 (Emotiva), OmniMic2, UMIK-1, and UMM-6

You are right in that Dirac certainly should have no reason to do anything different than what they claim. I can't see any extra cost to doing EQ right over doing it wrong. I could, however, see Dirac adding some extra smoothing to the results curve to make it look better. Businesses have a pretty solid history of that.
I have seen this with other EQ packages and would not be surprised if Dirac has, at a minimum, chosen a smoothing rate to make their results look pretty.

I have been following other Dirac threads and Dirac predicted results are a bit optomistic

- Rich

 

[RichB]

I can't believe the Salon2 would yield graphs like that regardless of the room so I have to assume something is wrong.

After 300 Hz, the Dirac measurements seem reasonable. Relatively little was done at higher frequencies.
I have reason to doubt the 100 to 300 Hz measuremnts and correction applied.

Post 22 shows the OmniMic2 measurents of the Studio2's as the mains before and after Dirac.
That seems more like it.

- Rich

 

[PENG]

Hi Kurt,

yes, I'm affiliated with Dirac Research and yes, I also think that Rich knows what he is doing.
After making that clear do you think that Rich is listening with full satisfaction in a room with the response shown above 400 Hz in this BEFORE CORRECTION graph?
Furthermore do you think that Dirac Live applies no correction at all (measurement tolerances aside) as shown in this AFTER CORRECTION graph?

So please understand why I have serious doubts about that graph and I think that something may have gone astray...

Flavio

P.S. I've just seen that Rich has been posting while I was writing... I'll read his comments now.

After 300 Hz, the Dirac measurements seem reasonable. Relatively little was done at higher frequencies.
I have reason to doubt the 100 to 300 Hz measuremnts and correction applied.

Post 22 shows the OmniMic2 measurents of the Studio2's as the mains before and after Dirac.
That seems more like it.

- Rich

I know, but the 300 Hz to 20K part shouldn't look like that at all. You seem some of my REW graphs right? They are with a few dB, yours are like +/- 10 dB and sort of oscillate.

 

[RichB]

I know, but the 300 Hz to 20K part shouldn't look like that at all. You seem some of my REW graphs right? They are with a few dB, yours are like +/- 10 dB and sort of oscillate.

Agreed. They should not look like that.
The Dirac measurements of that range should not look like that either

I went through a lot of mic's and runs of REW and Dirac and the results were stable and not aligned.

Also, the OmniMic2 had a different opinion.

- Rich

 

[flak2]

Agreed. They should not look like that.
The Dirac measurements of that range should not look like that either

I went through a lot of mic's and runs of REW and Dirac and the results were stable and not aligned.

Also, the OmniMic2 had a different opinion.

- Rich

Well, the OmniMic2's graph should not look like that either... it is not weird as the REW one but as you correctly noticed it shows a 10 dB boost from appx. 100 to 300 Hz while the Dirac default target applies a downward sloping curve instead.

Sorry for being pesky...
Flavio

 

[RichB]

Well, the OmniMic2's graph should not look like that either... it is not weird as the REW one but as you correctly noticed it shows a 10 dB boost from appx. 100 to 300 Hz while the Dirac default target applies a downward sloping curve instead.

I've never found this kind of result...
Flavio

I'd like to get to bottom of this and am open to suggestions.

- Rich

 

[PENG]

I'd like to get to bottom of this and am open to suggestions.

- Rich

I can only comment on REW, check all your settings, and mic cal file selection etc.

 

[RichB]

I can only comment on REW, check all your settings, and mic cal file selection etc.

Tried 4 Mic's with their own cal files.
10 measurement sessions.

I sent my .mdat files out for review.
Windows 10 is new...

All I have left is to send my Mic's out for another calibration. I sent contact request to cross-spectrum but nobody's answering.

Next week I will try to get the ASIO drivers to work.
I really don't want to buy any more mic's

- Rich

 

[PENG]

Tried 4 Mic's with their own cal files.
10 measurement sessions.

I sent my .mdat files out for review.
Windows 10 is new...

All I have left is to send my Mic's out for another calibration. I sent contact request to cross-spectrum but nobody's answering.

Next week I will try to get the ASIO drivers to work.
I really don't want to buy any more mic's

- Rich

If you send your file to

http://www.hometheatershack.com/forums/rew-forum/35311-rew-v5-tips.html

John M, the REW author will likely answer your questions. Did you use the Umik mic?

 

[RichB]

If you send your file to

http://www.hometheatershack.com/forums/rew-forum/35311-rew-v5-tips.html

John M, the REW author will likely answer your questions. Did you use the Umik mic?

I have been in contact with John and he compared the cal files. The .mdats I sent him had been smoothed so they were not as helpful as they could be.

He normalized the mic traces into this graph:

Time to try something completely different

- Rich

 

[flak2]

The REW curve is wavy most probably because of comb filtering due to the impulse response having been gated, the full impulse response has likely not been captured.
Of course we have to measure in the same nine measurement positions and then average the response because the magnitude response changes with position and even the few centimeters between our ears make a difference... as a result what we want to correct is the common general behaviour, the tonal balance.

When you average, you have to do the average in the frequency domain, a spectral RMS average.
Any single position shows a lot more variations than the average response... so we have to measure in many positions and average the result to obtain anything really useful.

The Dirac Live filter is mixed-phase, which means it has a built-in latency.
This makes it possible for the system to ”prepare” the speaker for the music that is coming next.
So, when we measure the response after correction, we must make sure to capture the whole impulse response.

In my opinion out of those measurements the accurate one is the Dirac Live's one.
Flavio

 

[PENG]

The REW curve is wavy most probably because of comb filtering due to the impulse response having been gated, the full impulse response has likely not been captured.
Of course we have to measure in the same nine measurement positions and then average the response because the magnitude response changes with position and even the few centimeters between our ears make a difference... as a result what we want to correct is the common general behaviour, the tonal balance.

When you average, you have to do the average in the frequency domain, a spectral RMS average.
Any single position shows a lot more variations than the average response... so we have to measure in many positions and average the result to obtain anything really useful.

The Dirac Live filter is mixed-phase, which means it has a built-in latency.
This makes it possible for the system to ”prepare” the speaker for the music that is coming next.
So, when we measure the response after correction, we must make sure to capture the whole impulse response.

In my opinion out of those measurements the accurate one is the Dirac Live's one.
Flavio

If you are right, then Rich's room definitely needs EQ, even at above 300 Hz.

 

[RichB]

The REW curve is wavy most probably because of comb filtering due to the impulse response having been gated, the full impulse response has likely not been captured.
Of course we have to measure in the same nine measurement positions and then average the response because the magnitude response changes with position and even the few centimeters between our ears make a difference... as a result what we want to correct is the common general behaviour, the tonal balance.

When you average, you have to do the average in the frequency domain, a spectral RMS average.
Any single position shows a lot more variations than the average response... so we have to measure in many positions and average the result to obtain anything really useful.

The Dirac Live filter is mixed-phase, which means it has a built-in latency.
This makes it possible for the system to ”prepare” the speaker for the music that is coming next.
So, when we measure the response after correction, we must make sure to capture the whole impulse response.

In my opinion out of those measurements the accurate one is the Dirac Live's one.
Flavio

With REW, I measured the same positions and averaged them using REW.
In any, case Dirac correction between 100 and 300 Hz produced a very boxy sound.

- Rich

 

[RichB]

If you are right, then Rich's room definitely needs EQ, even at above 300 Hz.

I will be remeasuring with the OmniMic2 software and with REW using ASIO.
For REW, I still have the tape marks on the floor to use the same positions.
In my view, than if this is not good enough, than there is a problem with REQ accuracy.

I prefer to have solid REW measurements to compare the before and after.
In Post 22, you can see th Dirac default curve measured against Reference Stereo.
In this configuration, Stereo Reference sounded much better than the Dirac default target curve.

- Rich

 

[flak2]

With REW, I measured the same positions and averaged them using REW.
In any, case Dirac correction between 100 and 300 Hz produced a very boxy sound.

- Rich

Well, if your Dirac measurements have been properly done then the range between 100 and 300 Hz is at least 10 dBs lower than the rest... if you consider that a much less "boxy sound" that's OK, I never debate on listening opinions being them so subjective.

Flavio

 

[PENG]

I prefer to have solid REW measurements to compare the before and after.
In Post 22, you can see th Dirac default curve measured against Reference Stereo.
In this configuration, Stereo Reference sounded much better than the Dirac default target curve.

- Rich

That's true, I missed that. I have taken numerous measurements with REW and minidsp Umik-1 and have never come across anything close to what you've got with yours.