• 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

[MOD EDIT: This is a member contribution]
Hi all,

Lately I was playing a bit with digital room correction (DRC) SW to see if I can tame the bass in our living room with a pair of Revel M16 (no subs) so I thought it could be useful to share my findings. Some of my in-room measurements are available already in this post, and here are the quasi-anechoic measurements of our pair of speakers. As you can see, we get a bit of bloom under ~200Hz, which should be right around the Schroeder frequency of this room.

The three DRCs I tested were:
All of them offer free trial versions, which is what I used to test them out and compare.

Measurements were done with a Cross-Spectrum Labs calibrated Dayton EMM-6 measurement microphone and RME Babyface Silver edition soundcard.

Here's some of my quick thoughts on each one of them:

Dirac Live for Studio
Usability and UI is very nice in general - though I was having some issues with logging-in to the user account on the Processor component (VST plugin) and was also getting some errors when loading saved projects. I don't like the fact that you can't edit (or even see) the target curve in the plugin (Processor component), and due to issue with loading saved projects it can be tricky to edit saved curves.
The measurement process was relatively straightforward and quick (actually the quickest of these three), and the resulting responses seem in line with what I was getting with REW.
I'd say there's really a lot of power in Dirac - it sounds great, and is really flexible in target curve customization (you can load target curves, edit them, add many break points, freely set range for the correction...).
Here's the correction curve I liked the most:View attachment 100557

IK Multimedia ARC System 3
Not bad UI and usability, though there are a few things I didn't like: e.g. when you load the mic calibration curve, there's no indication that it loaded correctly, and due to this one of my measurement attempt turned out incorrect. Another thing I don't like is that there are only 6 target curve break-points, and the SW applies some correction to FR even outside of the configured range (unlike Dirac) - so you should be careful how you set the breakpoints if you limit the range, as you will need a few of them out of the range set at 0dB to really limit the correction. This means you get really limited with how you can set the target. You can also only set the break points in between the +/-6dB of the reference the SW selected - which can also be limiting when we consider downward-sloping in-room response curves that we get when listening in far field.
However, there's some nice additional features to play with, like two types of filter phase (natural and linear) and 'virtual monitoring' target curves that attempt to mimic other speaker system's FR.
Measurement process was relatively simple, but requires more measuring points (21 in total, with 8 sweeps per point).
Sounds quite good to me too, but the target curve configuration is a bit limiting.
This is the curve I ended-up with (range a bit wider that Dirac, and maybe a bit less bass boost):
View attachment 100558

Sonarworks Reference 4 Studio edition
Maybe the best UI, stability and usability of the three. I enjoyed the gamified concept of the measurement process, but in practice it is quite tedious to do as it takes something like 37 measurement positions to do the calibration - so it's quite a lot of loud sweeping and chirping noises to endure :D
This is also probably the most complete SW package of the three - it offers systemwide application and plugin version, and on top of offering loudspeaker calibration it also provides a pretty large database of headphone equalization configs. They were obviously aiming for user-friendliness :)
What I don't like is that target curve customization is very limited (including almost non-existent range limiting). Because of this, most of the result sounded over-processed to me and I couldn't really find any that sounded close to what I thought was natural (and which I could get with the other two).
Here's the best I got:
View attachment 100559

Filter response comparisons
First thing I'd like to show is soundcard loopback frequency response with each DRC after calibration to compare the filter responses:
View attachment 100550
As we can see, Dirac Live and ARC 3 filters were limited to our problem area (45-210Hz) and I was surprised to see that the filters between them end up looking fairly similar. Reference 4 cannot be limited in the same way so here I'm showing 'Reduced' LF limit and 'Normal' HF limit. We can see right away that filters used by Reference 4 are less sharp in the LF range and I couldn't find a way to get a better match to the other two.
Note: the above is with 'Natural' phase and 'Sharp' filter type set in ARC3 - other settings resulted in worse match to Dirac Live (i.e. less sharp filters). For Reference 4 I used 100 Wet config to get most sharpness (although I used it with 60% as that sounded more natural).

Edit: adding the impulse response comparison from post #24 for completeness:
View attachment 100628

Here's full phase and magnitude diagrams (please disregard the 'hairiness' above ~4k for Dirac and ARC - it is an artefact of the way I did loopback testing of these two plugins and not something they do normally):

Dirac Live:
View attachment 100552
What I found interesting was the gradually rising phase response after the filter. Is this due to impulse-response and delay correction/optimization Dirac does?

ARC 3 with 'Natural' phase variant:
View attachment 100553
Notice no rising phase response, unlike Dirac Live.

ARC 3 with 'Linear' phase variant:
View attachment 100555
Notice that filters are less sharp, but phase response is smooth (P.S - this was range limited to ~900Hz and with a different target - here just used to illustrate phase behaviour).

Reference 4:
View attachment 100554

In-room response measurement example
Next here's an in-room loudspeaker response comparison of no DRC (averaged across 9 positions) vs ARC3 correction (averaged across 5 positions):
View attachment 100563
Note that the curves don't overlap very well in the mid and higher frequencies - they were done on separate occasions and the measurement positions and mic height naturally weren't the same between the two measurement runs - but they are roughly close.
Still, we can see how much the bass response got evened out by DRC - peaks got flattened and the dips got filled in, as we could only have hoped for :)

Sample in-room recordings

Lastly, I thought how to illustrate the audible differences, and decided it might be interesting to do in-room stereo recordings of playback without any DRC as well as with the three DRCs reviewed above.

In-room sample recordings were done with the classic mid-side stereo recording technique, using a combination of Dayton EMM-6 as 'mid' and Rode NT2a in figure-of-eight mode as 'side' at the listening position in our acoustically untreated living-room. I chose mid-side technique for two reasons:
  1. Use of omni mic for 'mid' should be good to record realistic low-end response in the far-field
  2. It is not critical to have a closely matched pair of microphones :)
Edit: Listening position is ~2,2m from each speaker and SPL was between 75-80 dB(C) (measured with a cheap SPL meter close to mics) for each clip.

The recordings were not post-processed, except summed for stereo listening and (LUFS) loudness matched.
Here's a picture of the mic setup:
View attachment 100560

The song/recording I used was one of the vary few I have distribution rights for as it was made and recorded by my band - so I guess what follows could also be considered shameless self-promotion. :D Anyway the source track can be found on most streaming services via this link.

The resulting in-room recordings in FLAC (44,1kHz/16bit) are uploaded here - I suggest to listen on headphones to avoid adding additional room effects to an already quite live recording. :) These recordings definitely don't convey the full in-room listening experience, but I still thought it was an interesting project and may provide some insight.

Conclusion
All in all, this was quite an interesting little project for me. My takeaway is that there are definitely some really good options on the market for DRC - out of these three my preference would be:
  1. Dirac Live 3 - sounds the best to me and offers most flexible target curve config. Natural sounding and with reasonable target configs I thought it was not destructive in any way. Not cheap, though.
  2. IK Multimedia ARC system 3 - in my opinion it can be configured to sound close to Dirac, but required a bit more fiddling. It may not be able to satisfy every requirement as far as target curves go though - this is my main gripe with it. However it is quite a cheaper option than the other two.
  3. Sonarworks Reference 4 Studio edition - I couldn't make myself agree with this one :) Sure, there are some nice features there (plus great systemwide version and some extensive headphone EQ options) so I'm sure it works great for many - but lack of filter sharpness, true correction range limiting and detailed target curve editing are for now deal-breakers for me.
There it is - hope some will enjoy the read and maybe even find bits of it useful! :)

EDIT: Some thoughts and measurements related to nearfield use (in my case with JBL LSR305) can be found in post #55.

EDIT (2020-12-28): Some thoughts and measurements using MathAudio Room EQ in a nearfield listening setup can be found in post #96, and measurement and comments on use of MathAudio Room EQ in my main system can be seen in post #104.

EDIT (2020-12-30): Some of my thoughts on using JuiceHiFi Audiolense XO trial in a nearfield setup can be found in post #109, and some measurements in post #114.

EDIT (2021-01-12): More measurements with JuiceHiFi Audiolense XO trial in my living room setup, including comparisons with Dirac Live and MathAudio Room EQ as well as some blind listening tests can be found in post #125.

EDIT (2021-04-18): Measurements and my thoughts about REW-generated EQ filters for room correction can be found in post #205.

Can you please post the step response too from this measure ?
 
Can you please post the step response too from this measure ?
This thread contains dozens of individual measurements I've taken over a span of several months. It would take too much time and effort now to find and show step responses for all of them, sorry.

However some examples of measured step responses were already shown in posts #55 and #125.
 
@dominikz I mean only the step response of the Revel M16 you write in your first post. I wish not all measure systems, only without eq version and room is ok too, can you please post it ?. because this is good comparable with the Kali. what do you think which db per octave the LP Filter have ?
 
@dominikz I mean only the step response of the Revel M16 you write in your first post. I wish not all measure systems, only without eq version and room is ok too, can you please post it ?. because this is good comparable with the Kali. what do you think which db per octave the LP Filter have ?
The measurements in this thread will not be useful for what you're interested in, as early reflections cannot be separated from the direct sound (due to loudspeaker positioning). I will post the appropriate step response of M16 in the other thread.
I have not investigated the crossover slope of M16 (but it is passive so probably some low-order filter).
 
I'm probably missing something, but got a "cross reference" question. What if you correct with one software and measure with the other? Do they measure same things? Would cross checking corrections and measurements between the soft suits show the same things?
 
I'm probably missing something, but got a "cross reference" question. What if you correct with one software and measure with the other? Do they measure same things? Would cross checking corrections and measurements between the soft suits show the same things?
They all measure frequency/impulse responses (so the same thing), but may have slight differences in how they process the measurement.

E.g. some of these tools require you to measure with sweeps at multiple positions (also different between SWs) and some require you to measure with a sweep at only one position. With REW you can also use MMM and periodic pink noise to create a spatially average measurement.
For the SWs that average the sweep measurements there could be differences vs how they average and smooth the responses to create the reference from which they calculate filters and how they calculate the filters themselves. Some may also apply phase corrections on top of magnitude corrections (like Dirac), etc...

However, it seems that in practice a lot of what they do is pretty similar - as you can see from how well the filters calculated independently by each of them align between each other, e.g. here (there's also an example in OP):
Some measurements
Naturally I wanted to see what MathAudio actually did :) First I ran a loopback test when running MathAudio as VST plugin in EAPO, this was the result (compared to ARC3 and Reference 4):
JBL LSR305 MathAudio REQ comparison of filter response Left.png

In post #125 there is another example where I measured and created a correction with Audiolense and then measured the result with REW using the equivalent approach (single position sweep) - they look practically identical.

If you correct the response with one tool, and then measure the result with another tool you will definitely clearly see the effect. Hope this answers the question!
 
I test the new sonarworks demo. they name it now soundid. https://www.sonarworks.com/soundid-reference .This is the first automatic system that do diffrent left right speaker corrections that sound great also in zero latency mode. I verify this even more and play sine tones at diffrent frequency and i hear that the testones are on all note frequencies in the mid. this is a big step. this is the measure. all other automatic systems sound worse in the range from 70-200 hz(which need most correction) in stereo. they correct so that sound is more out of mid as without correction. dirac is unusable for me. in arc there is switch combined LR correction then stereo mid position can not get worser. but with the new sonarworks stereo position get also better and not worser

sonarworks soundid.jpg
 
A new Dirac feature:

 
A new Dirac feature:

Thanks - very interesting! Reminds me of something from one of my old posts in this very thread :D:
Just wanted to share a few of my random thoughts on automatic room correction that I was mulling over the past few days...

IMHO one of the main issues with practically all current automatic room EQ software solutions is their default target curve. As you probably noticed, it is typically flat (Sonarworks, MathAudio, ARC System) or close to it (slight tilt in Dirac Live).

My problem with this is that I feel many users will at least initially just use the default target, and it doesn't seem likely to me this would fit many (perhaps even most) real-world scenarios. Of course, in my tests I did find that the flat target may work OK in nearfield with flat on-axis response loudspeakers, but I'm not sure how widespread that scenario is in reality.
Different loudspeakers will have different on-axis response, and especially if listening in farfield, louspeaker dispersion as well as room characteristics factor-in and one needs to adapt at least the tilt of the curve to get natural sounding results.

So I wondered why none of the DRCs I tested try to automatically generate a target curve shape based on the measured in-room response?
E.g. perhaps they could do a linear regression of the measured in-room response above the transition frequency to build a flat but tilted target better matched to the room and loudspeaker.
Another option would be to simply follow loudspeaker in-room response above the transition frequency, and use e.g. 1dB per octave bass boost below it. I'm sure there are other/better options that one could use.
Sure, some loudspeakers may measure so uneven that it could be very difficult to estimate a reasonable target curve - though probably in that case a flat target also wouldn't perform much better.

Any thoughts on this?

Seems to me like it could be a simple way to provide better out-of-the-box results for many casual users.

Anyway, IMHO looks like a potentially very nice usability development of an already good automated DRC solution!
 
Yes, was always wondering why Dirac etc. had a fixed recommended target at the LP and didn't adapt it, some years ago was playing with the idea of designing a DRC product that would also take loudspeaker nearfield and room reverberation measurements to do so.
 
Yes, was always wondering why Dirac etc. had a fixed recommended target at the LP and didn't adapt it, some years ago was playing with the idea of designing a DRC product that would also take loudspeaker nearfield and room reverberation measurements to do so.
This is what I do for a long time and have also implemented on loudspeakers.audio : with some informations about the room, loudspeakers and what you get from measurements, target and corresponding correction are automatically calculated.
 
Such a great comparison dominikz! Thank you!
Also your test song is great, great voice!! I enjoyed listening to it through your room especially :)

Not sure if it's been brought up already but Sonarworks Reference looks like it now has a fully adjustable correction EQ.
Blessings & thanks..
 
Such a great comparison dominikz! Thank you!
Also your test song is great, great voice!! I enjoyed listening to it through your room especially :)

Not sure if it's been brought up already but Sonarworks Reference looks like it now has a fully adjustable correction EQ.
Blessings & thanks..
Thanks a lot for the kind words, and for the tip on Sonarworks functional development - sounds interesting!
 
I just bought sonarworks, or should I say soundID. I mostly chose it over other options because I really need the headphone EQ and need to be able to switch between speaker and headphone a lot for production work.

I'd love to see if the new version has made improvements.
 
I just bought sonarworks, or should I say soundID. I mostly chose it over other options because I really need the headphone EQ and need to be able to switch between speaker and headphone a lot for production work.

I'd love to see if the new version has made improvements.
My main gripe with Sonarworks Reference 4 was limited target curve customizability; and specifically there was no way to limit the correction just to the low frequencies. In case this facility was added I'm sure I'd be much happier with it!
 
My main gripe with Sonarworks Reference 4 was limited target curve customizability; and specifically there was no way to limit the correction just to the low frequencies. In case this facility was added I'm sure I'd be much happier with it!

Yeah that's a feature they've added in the newer version. I have mine just working from 450hz down.
 
Yeah that's a feature they've added in the newer version. I have mine just working from 450hz down.
Sounds good, thanks! I might give it a try again at some point. :)
 
GREAT thread!

Regarding how to position microphones for room analysis, it isn't quite clear. Traditionally, for room analysis it was always said to use a 90 degree position, pointing at the ceiling and 0 for loudpseaker analysis. But I remember Sonarworks and I'm seeing ARC 3 now demonstrating to point it in the direction of the speakers. Can somebody clarify this?
 
Back
Top Bottom