Dirac Live

[rhbblb1]

OK. So I finally got around to setting up Dirac live, do measurements and create some filters. I have been using Trinnov, but wanted to simplify my signal path and am running USB into the MSB DAC, rahter than doing USB to AES/EBU conversion to be able to go through the Trinnov,

Measurements went fine. Then created three filters;
Filter one: +5db to 0db in 20Hz-200Hz range, flat in 200Hz - 1Khz range and 0dB to 5db in 1Kzh to 20Khz range - this is pretty much the same target curve I am using on my Trinnov
Filter two: +5db to 0db in 20Hz-300Hz range, no correction at all on anything above 300Hz
Filter three: flat from 20Hz to 20Khz.

Then kicked back, played some music, switching filters in real time and switching filter on/off to compare DRC with no DRC.

Huge disappointement. First, I honest to God barely noticed a difference between any of the filters, and no filters at all. I am 100% sure I could not pick out a preference I was doing this blind. On Trinnov, there is always a clear difference between engaging DRC and bypassing it.

Am I doing something wrong? I'm very puzzled.

The way it stands right now, I'll just forgo DRC completely, at least for 2 channel. Still have my Trinnov processor for multi channel.

You must be doing something wrong. Each filter should sound different. By the way, why did you not try the default filter?
I have been tweaking the default filter and miniscule changes are audible. I have used Ominmic measurements to confirm that I am actually listening to each filter as designed. Do you have a way of confirming you are listening to each filter. They will measure differently and should be audibly distinct.

 

[audioguy]

You must be doing something wrong. Each filter should sound different. By the way, why did you not try the default filter?
I have been tweaking the default filter and miniscule changes are audible. I have used Ominmic measurements to confirm that I am actually listening to each filter as designed. Do you have a way of confirming you are listening to each filter. They will measure differently and should be audibly distinct.

+1

As rhbblb1 noted, it is VERY easy to hear the differences. If you can hear them with Trinnov then you certainly should be able to hear them with DIRAC!

Did you hit the "optimize" button? Can you use OmniMic or XTZ or REW to show the before and after plots?

Did you look at the filter created by Dirac and that it will display? Did it appear to be the inverse of the room measurements that Dirac displays?

 

[amirm]

Dirac doesn't re-measure itself post correction?

 

[rhbblb1]

Dirac doesn't re-measure itself post correction?

It shows you the correction and the filter. However, I always like to confirm with a different measurement system.

 

[edorr]

+1

As rhbblb1 noted, it is VERY easy to hear the differences. If you can hear them with Trinnov then you certainly should be able to hear them with DIRAC!

Did you hit the "optimize" button? Can you use OmniMic or XTZ or REW to show the before and after plots?

Did you look at the filter created by Dirac and that it will display? Did it appear to be the inverse of the room measurements that Dirac displays?

Turns out I had not selected the Dirac as the output device in JRiver, so my output was bypassing the processor and I was hearing nothing. I got it working now and the processor is "streaming". So now it is back into the basement and another round of listening. I'll report back with real impressions.

Now, the real embarrasement would have been if I had reported a preference one way or the other, strictly based on placebo effect. Thankfully turns out I am not completely tonedeaf.

 

[edorr]

Dirac doesn't re-measure itself post correction?

Neither does Trinnov. Big deficiency in my opinion. I once used a SVS sub EQ unit that showed a (near flat) theoretical curve. Measurement with REW showed this curve was a complete fiction.

 

[audioguy]

Neither does Trinnov.

Neither did the SigTech or the TacT or Audyssey or Dirac. But in the case of the TacT, you do/did have the ability to have the system re-measure through the filter and get "real" results as opposed to the theoretical results which most systems display --- and which are totally bogus.

OmniMic or REW or XTZ are really important tools for any of these room correction solutions. And since many times the actual FR after correction can still use some work, using a measuring tool allows you to go back and adjust the target so that you get the results you would like.

 

[Roger Dressler]

When we tune a room with, say, Dirac, it samples 9 mic locations. When we check it with REW, do we use 9 positions? Is the mic in the same place, exactly? Is the averaging/weighting method the same? If the answer to any one of these questions is no, then the measured responses will differ.

That does not mean Dirac (in this example) needs any further tweaking. It means there's a difference in the method, thus a difference in the result. Which one do you want to believe?

In my case, I do not believe either one, unless it sounds right.

In a similar vain, it is not a deficiency to not run a post EQ measurement. What are the chances the mic is in the same places? Little to none unless you do as Keith Yates does, multiple mics and a multiplexer.

Even if we are talking about a Trinnov cal where the mic stays in one location, it is not necessary unless you think the measured response errors cannot be inverted. My presumption is that the displayed response shows only the correctable error, but that may be incorrect. No matter, if the actual final response as measured by REW shows variations, they may be best left alone. Trinnov and Dirac (and yes, Audyssey) apply some smarts to decide what should/can be corrected and what cannot. REW does not in its basic measurement routine.

 

[audioguy]

Trinnov and Dirac (and yes, Audyssey) apply some smarts to decide what should/can be corrected and what cannot. REW does not in its basic measurement routine.

Yes and no. Not every room will sound the best with the identical target curve and I am also of the camp that says "trust but verify". I will say that Dirac (when compared to Audyssey) does give a measured response that is much closer to the desired target than Audyssey.

But Dirac WILL try to correct what appears to be a null and provide a filter with an excessive (and audible) bump in that area that needs to be dealt with and modifying the target at that area addresses that issue.

 

[Kal Rubinson]

When we tune a room with, say, Dirac, it samples 9 mic locations. When we check it with REW, do we use 9 positions? Is the mic in the same place, exactly? Is the averaging/weighting method the same? If the answer to any one of these questions is no, then the measured responses will differ.

That does not mean Dirac (in this example) needs any further tweaking. It means there's a difference in the method, thus a difference in the result. Which one do you want to believe?

In my case, I do not believe either one, unless it sounds right.

I agree but that does not mean that a different measurement paradigm, applied consistently, isn't informative. I tend to check the operation of Audyssey/Trinnov/Dirac with single position measurements before/after at the MLP with XTZ. It tells me what is different at that spot. I can do another spot or two if my curiosity tells me to. I do not use multiple positions for these measurements although XTZ permits it because of the imprecision of getting the same spots for the before/after.

 

[Roger Dressler]

Yes and no. Not every room will sound the best with the identical target curve and I am also of the camp that says "trust but verify".

I did not say anything about the target curve. That's part of what comes out of the "trust but verify" process.

I agree but that does not mean that a different measurement paradigm, applied consistently, isn't informative. I tend to check the operation of Audyssey/Trinnov/Dirac with single position measurements before/after at the MLP with XTZ. It tells me what is different at that spot.

Totally agree this is valid and useful in understanding what the correction is doing. I also like to measure the electrical response of the EQ. Just to see.

I was simply offering a possible explanation why the responses seen with REW after Dirac/Trinnov are done is not necessarily an indication they need changing. But if indeed an excessive (and audible) bump results as Audioguy found, that would be something to verify with REW and take measures to correct.

 

[amirm]

Neither does Trinnov. Big deficiency in my opinion.

I agree. That's why the system I use does show post EQ measurement . Here is an example:

The JBL Synthesis ARCOS uses an 8-microphone measurement system. As such, one leaves the mics there allowing the system to preform remeasurements for multiple positions easily. Above is it trying to blend the subs with each channel -- a process that is darn near impossible to do for multiple channels and subs.

 

[audioguy]

I agree. That's why the system I use does show post EQ measurement . Here is an example:

The JBL Synthesis ARCOS uses an 8-microphone measurement system. As such, one leaves the mics there allowing the system to preform remeasurements for multiple positions easily. Above is it trying to blend the subs with each channel -- a process that is darn near impossible to do for multiple channels and subs.

Interesting target curve. If that is the curve that Harmon determined that most users liked, I would not be in that group, at least for music. I do like a rising low end but not so dramatic. Mine looks just like the Dirac standard curve for music but a little hotter in the low end (from 40Hz south) for movies.

 

[amirm]

Interesting target curve. If that is the curve that Harmon determined that most users liked....

They did indeed . Here are the results of double blind tests which was done with music by the way:

RC1 and RC2 are the prototypes of this system (one optimized for a single seat, the other for more). As you see the sloping down curve is the preferred one.

, I would not be in that group, at least for music. I do like a rising low end but not so dramatic. Mine looks just like the Dirac standard curve for music but a little hotter in the low end (from 40Hz south) for movies.

Clearly there is some room for variation here. Can you post the curves for Dirac? I can't find them in their manual.

 

[edorr]

I finally got around to listening to Dirac Live (I did the calibration a few weeks ago) and here are my observations. I created four target curves:

1. +5db at 20hz sloping down to flat at 300Hz flat through 2000hz then sloping down to -5dB at 18,000 hz - This is very similar to my target curve in Trinnov
2. +5db at 20hz sloping down to flat at 300Hz - no correction beyond 300 Hz
3. The Dirac house curve (standard generated)
4. Flat curve

First of all, the difference betwen the four curves and no DRC were quite dramatic. Curve (2) and (4) were immediately eliminated. Somehow (2) muffled the sound, and (4) just did not sound right. Betweeen (1) and (3) was a close call but I preferd (1).

The interesting comparison was between no DRC and curve (1). Preferences appeared to vary depending on reportaire, but most importantly I believe recording quality. There was a very distinct difference in particular in imaging of the vocals (for example, the voice of JJ Cale on his "Live" album changes completely!), and mid bass. Surprisingly, not so much on low bass, which is where most problems happen according to the measurement. In the end I ended up preffering curve (1) over no DRC. it appears this setting was more unforgiving but I believe it got me closer to what is actually on the recording.

So with that settled I kicked back and did some listening with Dirac engaged. It then occured to me that the Dirac in bypass mode does not sound as good as not loading Dirac at all (which was what I previously did). So now I need to go back to the drawing board to compare Dirac with filter (1) with taking Dirac completely out (instead of just disengaging the filter in Dirac).

Interesting stuff. Working from memory, I believe the difference with engaging and bypassing Dirac were more pronounced than doing the same in Trinnov, which appeared to be more subtle.

When I have my lynx card in, I will repeat the whole experiment in multi channel.

 

[dallasjustice]

Very interesting Edorr. I have some quesitons about DIRAC vs. Audiolense and I would appreciate a fuller comparison at some point as well.

Does DIRAC create FIR filters for the Jriver convolution engine like Audiolense? Is DIRAC still limited to 96k or can it reach 192K?

I finally got around to listening to Dirac Live (I did the calibration a few weeks ago) and here are my observations. I created four target curves:

1. +5db at 20hz sloping down to flat at 300Hz flat through 2000hz then sloping down to -5dB at 18,000 hz - This is very similar to my target curve in Trinnov
2. +5db at 20hz sloping down to flat at 300Hz - no correction beyond 300 Hz
3. The Dirac house curve (standard generated)
4. Flat curve

First of all, the difference betwen the four curves and no DRC were quite dramatic. Curve (2) and (4) were immediately eliminated. Somehow (2) muffled the sound, and (4) just did not sound right. Betweeen (1) and (3) was a close call but I preferd (1).

The interesting comparison was between no DRC and curve (1). Preferences appeared to vary depending on reportaire, but most importantly I believe recording quality. There was a very distinct difference in particular in imaging of the vocals (for example, the voice of JJ Cale on his "Live" album changes completely!), and mid bass. Surprisingly, not so much on low bass, which is where most problems happen according to the measurement. In the end I ended up preffering curve (1) over no DRC. it appears this setting was more unforgiving but I believe it got me closer to what is actually on the recording.

So with that settled I kicked back and did some listening with Dirac engaged. It then occured to me that the Dirac in bypass mode does not sound as good as not loading Dirac at all (which was what I previously did). So now I need to go back to the drawing board to compare Dirac with filter (1) with taking Dirac completely out (instead of just disengaging the filter in Dirac).

Interesting stuff. Working from memory, I believe the difference with engaging and bypassing Dirac were more pronounced than doing the same in Trinnov, which appeared to be more subtle.

When I have my lynx card in, I will repeat the whole experiment in multi channel.

 

[edorr]

Very interesting Edorr. I have some quesitons about DIRAC vs. Audiolense and I would appreciate a fuller comparison at some point as well.

Does DIRAC create FIR filters for the Jriver convolution engine like Audiolense? Is DIRAC still limited to 96k or can it reach 192K?

Dirac has its own engine. You install the Dirac runtime and select this as the output device in JRiver. You then select your DAC driver as the output device in the Dirac runtime engine.

Come to think of it, this could be a problem for me since I need to use the Jriver volume control for my MCH setup once I have a lynx card. Not sure if this would work selecting Dirac as the output device in JRiver.

Dirac supports up to 96/24.

I believe Nyal is also (planning to) fiddle with this stuff, so he may be a good source.

 

[dallasjustice]

I believe Nyal is also (planning to) fiddle with this stuff, so he may be a good source.

I tried calling Nyal on Friday to talk about this stuff and he was already on his bike on some remote mountain. Whenever he gets off of his bike and back to work, I plan on peppering him with as many dumb questions as I can remember.

 

[edorr]

I tried calling Nyal on Friday to talk about this stuff and he was already on his bike on some remote mountain. Whenever he gets off of his bike and back to work, I plan on peppering him with as many dumb questions as I can remember.

Good luck. I was surpised about how much the various filters changed the sound - more so than what I recall when playing with Trinnov. This is not (just) about dealing with some room modes in the low frequencies.

 

[dallasjustice]

Good luck. I was surpised about how much the various filters changed the sound - more so than what I recall when playing with Trinnov. This is not (just) about dealing with some room modes in the low frequencies.

That's what I found when I bought audiolense 4 years ago. Bernt helped me set it up. I used the recommended gear. I applied the first curve and it sounded like all of the life was sucked from the recording. So, I eventually quit and haven't really messed with room correction since then. I used the Jriver parametric eq with some limited success. Looking back on my previous experience, I believe my expectations were too high. I think DRC can be useful for certain limited purposes but it's not going to fix all of the room problems. I think an "all of the above" approach is best. I am just interested in DRC again because I want to use a crossover/delay and integrate a couple of subs into my room. I think I can do that without killing the high frequencies. We will see.