• WANTED: Happy members who like to discuss audio and other topics related to our interest. Desire to learn and share knowledge of science required. There are many reviews of audio hardware and expert members to help answer your questions. Click here to have your audio equipment measured for free!

Comparison of DRCs: Dirac Live for Studio, IK Multimedia ARC System 3 and Sonarworks Reference 4 Studio edition

Great review, thank you. Any reasons why you did not consider Audiolense or Acourate?
No particular reason except that I didn't investigate those much :) From a quick look it seemed to me these two would require a bit more work to learn so I took the easy way :D I may still give them a look at some point though!

@dominikz What is the distance between the front face of your speaker and the wall behind it?
They're close to the back wall - maybe ~40cm / ~16" from front face to back wall
 
No particular reason except that I didn't investigate those much :) From a quick look it seemed to me these two would require a bit more work to learn so I took the easy way :D I may still give them a look at some point though!

Acourate I did not try, but Audiolense is fairly easy when using the same Recording/Playback device such as an RME product (I have an Adi 2 Pro).

Here’s an easy and excellent walk through by @mitchco how to do things: https://audiophilestyle.com/ca/ca-a...nd-room-correction-software-walkthrough-r682/

I also tried Dirac, nice - but I liked Audiolense way better.
 
So this raises a concern:

index.php


There is a ton of amplification of that "null" around 67 Hz. In my limited testing of the latest version of Dirac, I noticed the same. My Purifi ampifier ran out of amplification power and started to distort! Even my 1000 watt amplifier was struggling. Did you notice a problem like this? It only shows up if you turn up the volume.

I wanted to confirm this but have not had a chance.

Here are the impulse responses:
DRC filter comparison - IR.png

Interesting to see is that Dirac Live has the most pre-ringing and deviation vs direct loopback - not sure what to make of that, but I assume it's related to the gradual phase increase we saw in FR and the time domain corrections it does.
Reference 4 seems to lose some symmetry vs ARC 3 (but also its correction range was largest in these measurements as I couldn't limit it to LF-only like I did the other two).
 
Acourate I did not try, but Audiolense is fairly easy when using the same Recording/Playback device such as an RME product (I have an Adi 2 Pro).

Here’s an easy and excellent walk through by @mitchco how to do things: https://audiophilestyle.com/ca/ca-a...nd-room-correction-software-walkthrough-r682/

I also tried Dirac, nice - but I liked Audiolense way better.

Thanks for sharing - will give it a closer look!

From a first quick glance at the guide you linked though, it seems the correction filters in Audiolense are generated based on responses from only one measurement position - is this correct? I think Acourate may be like this too, if I understood the process from Archimago's post on it correctly.

Given how much bass response can change with relatively small spatial movements I hope to see DRC software do some weighting and averaging of responses at multiple positions.
 
Very interesting comparison :)
I have ordered some measurement mic to do some testing myself after reading this, so thanks for the inspiration!

Would it be an idea to compare REW also which is free software? (https://www.roomeqwizard.com/)
(Saw your notes on this too late, sorry. So it could work but is not as easy.)
I do use REW and RME built-in 3-band PEQ DSP to manually correct a nasty bass resonance I get at my work desk with JBL LSR305 in nearfield. But it is more manual work and relies more on user's experience to get right vs these automatic DRCs that guide you through the whole process and do the bulk of the work automagically :)
 
Thanks for sharing - will give it a closer look!

From a first quick glance at the guide you linked though, it seems the correction filters in Audiolense are generated based on responses from only one measurement position - is this correct? I think Acourate may be like this too, if I understood the process from Archimago's post on it correctly.

Given how much bass response can change with relatively small spatial movements I hope to see DRC software do some weighting and averaging of responses at multiple positions.

Correct, It’s a single measurement. But from my experience with Trinnov where you can do single point or multipoint I did not get better results with multipoint (btw, Audiolense is much more acourate for low frequencies in my experience: https://www.audiosciencereview.com/forum/index.php?threads/trinnov-altitude-jbl-sdp-75.13095/).

I think @mitchco did write something about multipoint vs single point in one of his articles, respectively the listening area covered by a single point Measurement in Audiolense.
 
Here are the impulse responses:
View attachment 100627
Interesting to see is that Dirac Live has the most pre-ringing and deviation vs direct loopback - not sure what to make of that, but I assume it's related to the gradual phase increase we saw in FR and the time domain corrections it does.
Reference 4 seems to lose some symmetry vs ARC 3 (but also its correction range was largest in these measurements as I couldn't limit it to LF-only like I did the other two).

Keep in mind that those impulses are loopback impulses, not measured from the speakers. Speakers are typically a minimum phase system, and it looks here like Dirac may be trying to shift the speakers towards more of a linear phase response (hard to tell though what it's doing without impulse response measurements in-room).
 
Keep in mind that those impulses are loopback impulses, not measured from the speakers. Speakers are typically a minimum phase system, and it looks here like Dirac may be trying to shift the speakers towards more of a linear phase response (hard to tell though what it's doing without impulse response measurements in-room).

I've published in-room impulse responses for Dirac. It does a remarkable job of cleaning those up.
 
So this raises a concern:

index.php


There is a ton of amplification of that "null" around 67 Hz. In my limited testing of the latest version of Dirac, I noticed the same. My Purifi ampifier ran out of amplification power and started to distort! Even my 1000 watt amplifier was struggling. Did you notice a problem like this? It only shows up if you turn up the volume.

I wanted to confirm this but have not had a chance.
You should ***NEVER*** correct dips that are the result of room dips. They are cancelations waves so more power only means more cancelations and no improvement. You can chop off any bass peaks. I always follow the curve quite deep down , say at most +6db but preferably less.

Nice review. I wonder how much you smoothed the graphs? With Dirac I have very different results depending upon the room and set. Sometimes a full range correction worked best getting a lot more focussed sound at the listening position. Sometimes just correcting the room modi until the Schroeder frequency works better. I would try both.
 
You should ***NEVER*** correct dips that are the result of room dips. They are cancelations waves so more power only means more cancelations and no improvement.
The in-room measurement clearly shows improvement.
 
Keep in mind that those impulses are loopback impulses, not measured from the speakers. Speakers are typically a minimum phase system, and it looks here like Dirac may be trying to shift the speakers towards more of a linear phase response (hard to tell though what it's doing without impulse response measurements in-room).
Thanks! Unfortunately I didn't do any in-room REW measurements with Dirac to take impulse response from :(
But in Dirac Live GUI you can see a comparison of IR before processing and estimated changes afterwards - this is what it showed in my case with a full range correction:
Dirac Live - fullrange FR target curve.PNG

Dirac Live - IR.PNG

So if that diagram is to be trusted, looks like Dirac does change the shape of the IR quite a bit and also tries to align the two speakers in the time domain.

In comparison, looking at REW before and after measurements, ARC3 doesn't seem to do much to the IR:
In-room FR comparison for a single speaker at a single measurement position - no vs ARC3.png

In-room IR comparison for a single speaker at a single measurement position - no vs ARC3.png


You should ***NEVER*** correct dips that are the result of room dips. They are cancelations waves so more power only means more cancelations and no improvement. You can chop off any bass peaks. I always follow the curve quite deep down , say at most +6db but preferably less.

Nice review. I wonder how much you smoothed the graphs? With Dirac I have very different results depending upon the room and set. Sometimes a full range correction worked best getting a lot more focussed sound at the listening position. Sometimes just correcting the room modi until the Schroeder frequency works better. I would try both.

Thanks! Filter responses are unsmoothed and in-room average responses are very lightly smoothed to 1/24 octave - just to remove some hairiness.
As far as full-range correction goes, through these tests I found it only works well for me if the SW allows me to configure the target curve to follow the overall natural loudspeaker response above the Schroeder frequency. If not, then I found I prefer to limit the correction range below it.
As you can see in the in-room measurement example before-and-after above, looks like DRC was able to fill-in this dip in the response somehow. Given that it used multiple measurement points to establish filter coefficients, I assume the SW does some clever processing to decide which dips to try and boost.

To give a bit more insight into my room, these are responses at 7 measurement positions with variable smoothing, taken without any room correction (left speaker only):
Revel M16 in-room measurement without any DRC.png

And this is 5 measurement positions with ARC3 correction applied (left speaker only):
Revel M16 in-room measurement with ARC3.png
 
Last edited:
Thanks for doing this comparison, it's very interesting indeed.

I recently bought a MiniDSP SHD box and have found that Dirac seems to get pretty good results, subjectively and from REW measurements afterwards. I did find taking the calibration measurements a bit of a faff though, even though I used the 9-point version only. You are obviously a lot more patient ! Dirac definitely does apply some boost in places for my setup, but it lowers the level overall so has headroom to do that. There's one fairly significant dip that it doesn't seem to try to address, so either it's too deep, or somehow it's worked out that it's a cancellation that can't actually be boosted.

I found it worked well just doing room correction for the bass response below 210Hz, and then I modified the target curve to give more of a "Harman" shape, so Dirac is actually operating up to much higher frequencies as well, and the overall balance is much nicer.

I've also tried doing the process manually using REW, and trying to avoid boosting at all, but it's just not possible in my room to get a good result that way. One advantage of REW is that you can see directly what the filters are, which you can't with Dirac.

My current feeling is that Dirac gives a 80-90% "good" solution, and you could spend a very long time messing about with tweaking filter generation in REW to get the last 10-20%.

FWIW here's the uncorrected and final Dirac result from my room:
dirac.jpg
 
Last edited:
Dominikz: Let me join the chorus of thank-yous for undertaking and sharing this. It is interesting to me (if a little head-spinning in some of the technicalities) because I just did room correction with a cheap measurement mic and Room Equalizer Wizard (REW) to create a WAV file to load into the Convolution filter in Roon. So here is my ignorant question:

How do REW-generated filters implemented through a player like Roon or Jriver compare to the approaches in your review? Are the three you reviewed simply more automated, or is there more functionality?
 
Last edited:
one of the options to utilize it system wide would be to use wdm driver from say jriver acting as vst host

If you have a quality soundcard with loopback, I'd recommend Element. It's a free VST host. I route all my system sound to an unused output on my RME Fireface (ADAT 7/8), enable the loopback which then makes the ADAT 7/8 output available on the ADAT 7/8 input. Then I set up Element to use those inputs, split them into two separate linear phase EQs. Output of one of those route to analog 1/2 out (monitors) the other analog 3 out (subwoofer) so I could do manual correction. This way I've got system wide correction with no added latency except for what the EQ is adding (~27ms) but you can only use this with pro audio interfaces with lots of outputs.

Before that I was trying "VoiceMeeter Banana" which is as hacky as the name sounds, required admin privileges to run and never once recovered from sleep properly - but that would work even without a pro interface. Maybe jriver would have been better.

I actually just wanted to do Dirac instead of manual correction, but I couldn't get the plugin to work at all and my trial expired without me ever getting a chance to try it. My impulse got a lot cleaner even with manual correction but I really wanted to see what Dirac would have done..
 
I do use REW and RME built-in 3-band PEQ DSP to manually correct a nasty bass resonance I get at my work desk with JBL LSR305 in nearfield. But it is more manual work and relies more on user's experience to get right vs these automatic DRCs that guide you through the whole process and do the bulk of the work automagically :)
Other than the automatic nature of these, do you get any audible difference between say DIRAC and REW? Other considerations?
 
Dominikz: Let me join the chorus of thank-yous for undertaking and sharing this. It is interesting to me (if a little head-spinning in some of the technicalities) because I just did room correction with a cheap measurement mic and Room Equalizer Wizard (REW) to create a WAV file to load into the Convolution filter in Roon. So here is my ignorant question:

How do REW-generated filters implemented through a player like Roon or Jriver compare to the approaches in your review? Are the three you reviewed simply more automated, or is there more functionality?

Glad it was interesting! :) I personally find REW very useful, and as I mentioned in the post quoted by @MediumRare, I use REW + manual PEQ to do room correction at my workdesk nearfield setup.
However I was having trouble finding a good EQ setting for my living room stereo setup, which is why I started looking into automated DRCs. If you're not very technically oriented (or don't want to use lots and lots of time measuring-EQing-remeasuring) in my opinion it is very much worth investing in good DRC. E.g. With Dirac Live I was able to get very good sounds very quickly.

Other than the automatic nature of these, do you get and audible difference between say DIRAC and REW?

For my nearfield setup, I still feel I get better results with both Dirac and ARC System 3 (didn't test Reference 4 there yet) than I get with manual EQ - with simple manual EQ there's still some variation in loudness depending on which note is played in the bassline (i.e. still some peaks and dips :)), whereas with DRC I perceive notes in the bassline to be similar in loudness. Note that I use only 3-band PEQ in the desktop setup, as that's what I have in RME built-in DSP, whereas DRCs use many filters and in some cases also do additional phase / time-domain corrections.
Also, note that this is a non-blind subjective experience without any objective data to back it up - so please take it with a grain of salt.

Certainly someone more experienced with manual room EQ, with some patience and more EQ bands could get it far better than I have!

Anyway, I will probably stick to manual EQ for the most part on my desk, due to convenience it offers in my setup through RME TotalMixFX interface.

I'll try to do a short summary of my thoughts on using these three DRCs in the nearfield setup too - I feel the use-case is different enough to warrant a closer look.
 
If you have a quality soundcard with loopback, I'd recommend Element. It's a free VST host. I route all my system sound to an unused output on my RME Fireface (ADAT 7/8), enable the loopback which then makes the ADAT 7/8 output available on the ADAT 7/8 input. Then I set up Element to use those inputs, split them into two separate linear phase EQs. Output of one of those route to analog 1/2 out (monitors) the other analog 3 out (subwoofer) so I could do manual correction. This way I've got system wide correction with no added latency except for what the EQ is adding (~27ms) but you can only use this with pro audio interfaces with lots of outputs.

Before that I was trying "VoiceMeeter Banana" which is as hacky as the name sounds, required admin privileges to run and never once recovered from sleep properly - but that would work even without a pro interface. Maybe jriver would have been better.

I actually just wanted to do Dirac instead of manual correction, but I couldn't get the plugin to work at all and my trial expired without me ever getting a chance to try it. My impulse got a lot cleaner even with manual correction but I really wanted to see what Dirac would have done..

I also had a issues getting VoiceMeeter Banana / Element to work with Dirac, but thought the approach with VB-Cable and Pedalboard2 as suggested by IK Multimedia here was quite simple to implement as a systemwide solution, and it worked with all three DRC VST plugins for me (including Dirac). It does add considerable delay, though, so may not be suitable for anything too interactive :)
 
However I was having trouble finding a good EQ setting for my living room stereo setup [using REW], which is why I started looking into automated DRCs.
This is the part I'm not getting. I would describe the process I went through with REW as automated: it took my measurements (averaged over the listening field) to generate 20 filters in each channel, targeting a predetermined curve. When you say you use REW manually, I assume REW is not generating the filters for you, but rather you are typing in specific values for specific frequencies or ranges. Or is there a level of automation in these commercial DRC systems that REW doesn't have?
 
Other than the automatic nature of these, do you get any audible difference between say DIRAC and REW? Other considerations?
Let 10 people create a DRC filter using REW and you will get 10 different results. So for sure an REW filter will sound different from Dirac. Not to mention Dirac uses mixed phase filtering, performs time-domain corrections and uses a proprietary algorithm to process multiple measurement points.
 
Back
Top Bottom