I tried the "Dirac Live Room Correction Suite".

 

This is commercial software, which sells for €487.50, which is about US$650, which is a bit more than I am willing to spend. I suppose if you think of it as replacing the need to purchase a box with an analogue equalizer, that isn't so unreasonable, but it is far more than I have ever paid for any software. Hence my observations are limited to the free trial, which gives you 2 weeks of uncrippled use.

 

There are essentially two pieces of software in this suite. The first one does room measurements, and the second is a virtual audio interface that applies the corrections.

 

I used a decent but sub-audiophile-grade Samson USB mic that I had purchased for recording podcasts, based on its flat profile between 20 Hz and 20kHz, and a favorable review in the NYT. I figured it was good enough to get an idea for how this worked.

 

The first thing to do is the room measurements. The software leads you through it, and it is pretty much idiot-proof, unlike the REW, which I still have yet to grok.

 

Here is what my room + speakers look like. Try not to laugh:

 

<p><a href="/monthly_2013_02/58cd9bc35a287_ScreenShot2013-02-14at6_07_26PM.png.f547de0c0dd0539ce10e83d3a8475a42.png" class="ipsAttachLink ipsAttachLink_image"><img data-fileid="28146" src="/monthly_2013_02/58cd9bc35a287_ScreenShot2013-02-14at6_07_26PM.png.f547de0c0dd0539ce10e83d3a8475a42.png" class="ipsImage ipsImage_thumbnailed" alt=""></a></p>

 

The light blue lines are the L and R averages of several measurement positions (individual measurements are the darker blue lines). These measurements are all from slightly different positions on the listening chair. The software guides you through it. Roughly speaking, this gives you some idea of the error of measurement with respect to positioning the mic.

 

Next, the software allows you to provide a correction. You can correct the entire spectrum, or just a part of it (say below 150 Hz). I did both. I made a bunch of filters, including a perfectly flat one, and one that slightly descends as the frequency gets higher.

 

The software corrects both amplitude and phase, which makes it a bit different from a parametric equalizer correction (which some people prefer).

 

Here is an example of one such correction:

 

<p><a href="/monthly_2013_02/58cd9bc3660d5_ScreenShot2013-01-19at12_59_26PM.png.3cf7530ad0b2e347bd0cf27dc7e1025f.png" class="ipsAttachLink ipsAttachLink_image"><img data-fileid="28147" src="/monthly_2013_02/58cd9bc3660d5_ScreenShot2013-01-19at12_59_26PM.png.3cf7530ad0b2e347bd0cf27dc7e1025f.png" class="ipsImage ipsImage_thumbnailed" alt=""></a></p>

 

You then can save the filter, and apply it. The filters you design are specific to your favorite audio interface (DAC, SPDIF converter or whatever) and you choose any and all sample frequencies available. It makes one filter for each of these.

 

The second piece of software functions as a virtual audio interface, soft of like Soundflower, except it isn't (thankfully) a kernel extension. So to use the Dirac filter, you select the Dirac audio interface in your playback software, and the filter has the real audio interface programmed into it. So all you have to do is start the Dirac software at login (or whenever you want it) and start playback.

 

One thing I immediately noticed is Audirvana only works with this when in stand-alone playback mode. It also is incompatible with exclusive access and direct mode playback, not surprisingly. I tried it with both Audirvana and iTunes.

 

Here is what the Dirac playback software looks like when running:

 

<p><a href="/monthly_2013_02/streamb_361x221.jpg.ce47c78c9f21c2f5e314853cf8cf8b80.jpg" class="ipsAttachLink ipsAttachLink_image"><img data-fileid="28148" src="/monthly_2013_02/streamb_361x221.jpg.ce47c78c9f21c2f5e314853cf8cf8b80.jpg" class="ipsImage ipsImage_thumbnailed" alt=""></a></p>

 

This gives you the ability to switch between up to four separate filters, and to turn the whole thing on and off.

 

I could not hear any difference between on and off until I added in a second filter. Then I could readily hear differences between my low-frequency and full-range filters, and hear differences between on and off for all the filters. I am not sure whether this is user error or a bug, but I suspect it is a bug. At first I dismissed it completely as inaudible, so it is kind of a bad bug for them to have in their software for those demoing it.

 

The difference with the low-frequency filter I made on/off is quite subtle. I guess I don't have an overwhelming bass problem to begin with (although the measurements sure aren't inspiring).

 

With the full-range filter, it definitely sounds different, and arguably better, with it on vs. off. However, I could easily replicate most of the audible difference, using the AU parametric equalizer available for free in Vox, just guessing the frequency and Q-value from inspection of the measurement plot.

 

There is also a discussion of this and other software here: The Well-Tempered Computer

 

It was also interesting to read the comments about the limitations of this approach by the guy who invented it: The Well-Tempered Computer

 

I would say that about 20% to 30% gets corrected. BTW it is still no more than 1/3 of the actual room problems.

 

So, I'm not sure what to do now. I might play around a bit more with REW and try to replicate (or not) my results, and perhaps make some filters for a future version of Audirvana, if that becomes a possibility.

 

This is a nice neat self-contained solution if you can live within the constraints imposed by the software.

 

I think the idea of using the computer to do room corrections for computer audio is an ideal goal, especially for those of us who use the computer as our only audio source component. I'm just not persuaded this is the realization of that promise.