Many years ago, back in 1992 to be precise - the year Windows 3.1 was introduced, when the typical PC was running below 33MHz, and Mortal Kombat 1 was hot in the arcades - I vaguely remember reading this article from Stereophile and filed it in my memory banks as something from science fiction for another day (I heard my first MP3 compressed song soon after and thought it was amazing given the data compression!). In it, Peter W. Mitchell reportedly "heard the future" and was clearly impressed by the technical wizardry of what a "Digital Signal Processor" could accomplish. The power of the digital room correction was described as "mind-boggling". If you haven't thought much about digital room correction, I would highly recommend reading that article in its entirety and consider whether the arguments make sense to you. I don't recall reading anything further about AudioSoft or whether Snell eventually created this cartridge-based "black box" audio correction system, but as we fast forward to 2015, we are living the legacy of what was described.
These days, DSP techniques are commonplace. As digital has long supplanted analogue as the preferred method of encoding our audio and video media with high fidelity, it has all become data that we can examine and modify as we desire depending on the situation. Compress to MP3 for times we don't need highest fidelity? No problem... Resize from 1080P to 720P for the cell phone screen/storage? By all means!
Explicit uses for improving audio quality include audiophile offerings like DEQX and Lyngdorf (including McIntosh MEN-220), some products like the Devialet amps include their "Speaker Active Matching" correction filters, and the home theater crowd have been using Audyssey and the like for years.
It is interesting these days then that DSP room correction is primarily the domain of home theater and AV receivers, rather than high-end audio. In fact, reading certain "high end" audio forums suggest a stigma against DSP techniques to improve the sound of an audio system. Among audiophiles, it's not unusual to see suggestions around listening to tubes, turntables, and vinyl LPs as providing some kind of elevated audiophile experience, but I've ironically experienced the cynicism when you bring up the idea that digital room correction is important if not essential as a very desirable way of improving the high-fidelity experience.
I moved to my new home about a year and a half ago. When I first set up my sound room, I used the Behringer DEQ2496 to perform the "first measurements" to achieve a reasonably balanced frequency response in the listening sweet spot. While it did sound better, I knew that sooner or later, I was going to aim for the more powerful finite impulse response (FIR) filter way to do room correction; the likes of which introduced above by Mr. Mitchell to not just improve the frequency domain, but also the time domain.
This post will document what I did in my own sound room for stereo playback as a start a few months back. The goal is to achieve customized room correction and playback with my HTPC connected to the TEAC UD-501 DAC and also the ability to use this same filter with my Logitech Media Server to the Logitech Transporter if possible (note that this is limited at this point as I will discuss below).
I. Measurement Preamble & Subwoofer CalibrationSo, out came the calibrated Behringer ECM8000 microphone (note I bought this years ago, see here for other options):
|A view down the boom to the Behringer measurement mic...|
Sound room stereo sound system:
Digital source: HTPC --> TEAC UD-501 DAC USB, Logitech Transporter via ethernet
Preamp: Emotive XSP-1
Amps: Emotiva XPA-1L x 2 monoblocks
Speakers: Paradigm Signature S8 + SUB 1 powered subwoofer
Cabling: balanced XLR cables (Mogami, Monoprice), Corning USB3 optical + powered USB hub, DIY Canare speaker cables
Pictures of the sound room:
|A look at the back wall - combination Ikea Kallax (top) and Expedit (bottom, slightly larger) for the vinyl collection.|
Notice that I have kept the layout the way I use it in daily life. For example, I still have that leather coffee table in front of my sofa even though it's not ideal and will add other surfaces for sound to interact with. But I do enjoy the occasional beverage when listening to music. Where else to put a nice glass of red or a shot of Macallan 21? :-)
I turned the volume to the usual listening level - around 70-80dB SPL with white noise estimated with my digital Radio Shack sound level meter. Many months back, I realized that deep bass from my subwoofer caused audible rattle from the doors of the storage cabinets in my room. I had to put some Blu-Tack to dampen the door of the glass cabinet behind the right speaker so it didn't rattle down around 30Hz. It's things like this you'll have to keep an eye out for in your particular room.
Speaking of subwoofer and bass, since my Paradigm Signature SUB 1 has built-in room correction DSP (implemented as parametric equalization), I ran the "Perfect Bass Kit" PBK-1 software to do some basic calibration. No need to talk much about this other than to show you that the software measures and adjusts the bass response. As you can see the calibrated curve in my room still isn't perfect, there is a dip around 65Hz which is partially compensated for with the "calculated" curve (I mentioned this back in the previous post last year as well):
Something else to consider is the DACs that will be used to play back the audio. One of the good things about the 3 DAC devices used in this post; the TEAC UD-501, Squeezebox Transporter, and the Creative E-MU 0404USB (used for measurements and in the calibration process) is that they're all of equivalent frequency response and general sonic quality. I set each one to utilize the default sharp linear phase filter to keep this consistent as well.
This is good because in means that I can measure the output from one to create the room correction filters, and the playback quality will be similarly accurate with the others. (In fact, both the E-MU 0404USB and Transporter use the same AKM4396 DAC chip.)
II. Acoustic PanelsIn principle, remember that it's always best to optimize the sound quality of the room in terms of the physical characteristics before all the fancy compensation. Although digital room correction can do a lot to compensate, it cannot do everything. For example, it's essential to heed recommendations about a decent sized room and avoid cubical configurations where each side is equal in length due to exacerbation of room modes. Make sure you follow "best-practice" guidelines around speaker placements (check out the noaudiophile.com online calculator for some suggestions including RealTrap, Cardas, Rule of Thirds methods).
My room has a slight taper up front, and measures approximately 15'x20'x8'. As described in my previous post on the GIK Acoustics Freestand, I have a couple of these to the sides to balance out the strong first reflection from my glass cabinets. It's good to see the improvement in reverb time (RT60) which is one of the parameters you can't fix with DRC. Here's a good article summarizing things.
With these panels in place, the sub calibrated with PBK, here is the frequency and impulse responses in the room (HTPC playback using TEAC UD-501 DAC, 1/12 octave smoothing). I highlighted the fact that these are 1/12 octave graphs; more smoothing like 1/6 or even 1/3 octave will make things look much smoother than the blemishes I'm revealing here!
|Frequency Response (1/12 octave smoothing)|
III. DRC DesignerWell, despite the inadequacies, let's see if we can try to correct some of this at the listening sweet spot by calculating the room digital correction filter. The easiest and least expensive way I know of is to download and give the free DRC Designer a try! The web page already contains excellent screenshots and getting the program running should not be difficult (versions for Windows and Linux). The calculations themselves are done with the excellent (free) DRC program by Denis Sbragion. Check out the documentation page for technical details and a taste of the mathematical complexity inherent in the technique.
DRC Designer provides an easy way to record the sweep tones (and impulse response data) necessary to feed the DRC program for filter creation. It also allows you to easily "draw" the target curve of the frequency response you want to see in the room. It's great to see that DRC Designer can record and process up to 96kHz (which is what I've used). Of course the beauty of digital room correction is not only to improve frequency response, it also should improve time-domain performance... After correction, one should see much better looking impulse response results for the room including suppression of "undesirable" effects like pre-echoing and interchannel delay.
Note that subjective preference plays a significant role here in that the target curve can be customized for what sounds best to you. I like the built-in "B&K3" setting and usually listen to this even though I will calculate the flat and "B&K1" filters as well as the Harman curve (BTW: You can read more about speaker tests from Sean Olive and his presentations for Harman on his Audio Musings blog.). For those who want more historical background, have a look at Brüel & Kjær's research on room response preferences back in 1974 presented at the AES - here's the paper. A nice demonstration of research correlating objective performance with subjective preference.
|Harman Synthesis Curve - see here for discussion.|
|A custom curve loosely based on the general shape from Harman...|
Here's the "B&K3" curve that's built into DRC Designer as a default which I normally use:
To keep things simple, I basically used the E-MU 0404USB as both playback and 96kHz recording device with DRC Designer using the ASIO driver (single-ended output to the Emotiva XSP-1 preamp and recording from the Behringer measurement mic with 48V phantom power through the E-MU's input channel 1 - left channel):
Running the room measurement through DRC results in the creation of FIR filters as .wav files in 32-bit floating point, 96kHz format. Notice the convenience of being able to generate filters of various strengths - I normally just use the "normal" filter:
IV. DRC Filter in JRiver DSP playbackArmed with the .wav filters from DRC Designer, we now put the FIR filter into JRiver Media Center for playback through the excellent 64-bit "DSP Studio". Most of the time, I'll use JRiver running on my HTPC connected by USB to my TEAC UD-501 DAC. It's all quite straightforward, just point the JRiver convolution DSP to the correction file and turn it on. When playing music, JRiver will calculate how fast it's able to run the DSP - I've noticed that on my system, it will stutter or slightly 'crackle' if the results is <4x realtime (notice in the picture below, I'm achieving 7.6x)...
You could have fun also with the upsampling DSP. With the processing speed of my HTPC's Pentium G3220 processor, the most I can do is upsampling to 192kHz plus using convolution DSP for the 96kHz filter (complex 130k taps created by DRC Designer). Upsampling with room correction to 352/384kHz can be done with a faster machine but my fanless, underclocked to 2.8GHz, and undervolted G3220 will stutter under this load. I suspect experimenting with the filter length will help. (Feeding 384kHz into the TEAC DAC will turn off the internal upsampling digital filter so all digital filtering will be done by the computer.)
These days, I mostly use eos (other options include Gizmo / JRemote) as my Android controller app for playing music off the HTPC running JRiver.
For me, the main feature added to JRiver 20 has been the ability to apply DSP processing to a DLNA stream; great for those with hi-res capable DLNA renderers. As a Squeezebox user, this allows me to use Whitebear Media Server to act as a bridge between JRiver and the Squeezebox system. In doing so, music can stream from JRiver, utilize its DSP processing, and the "room corrected" audio data sent over to my Transporter in the soundroom. Remember to tell JRiver to apply DSP to the network stream (Tools --> Options --> ... Add or configure DLNA servers... --> Audio-Advanced --> ... DSP Studio...).
Sadly, I would not get the Squeezebox system to play either 24-bit or even 16-bit streams if I set the "Format" (in the DLNA Servers panel) to one of the PCM options... The only streaming option I can see that works are unfortunately the MP3 variants. Clearly this is suboptimal. I can confirm that the DSP is doing its job, but clearly transcoding everything to MP3 is not good. I see from this discussion with the developer of Whitebear that "Squeezeboxes don't play 24 bit audio natively". Alas, I haven't even been able to stream 16-bit PCM. Has anyone had any success with lossless transmission to their Squeezebox running the JRiver DLNA DSP setting with Whitebear or even the LMS UPnP/DLNA plugin?
As a side note, if I were running LMS in Linux, I would have no hesitation trying out Brutefir for the Squeezeboxes. Would love to know if anyone has ported this to Windows.
V. ConfirmationSo, does this all work? And did the sound change for the better (at least at the measured sweet spot)?
Using Room EQ Wizard, applying the DSP in JRiver 20 on my HTPC and playing back through my TEAC UD-501 DAC, this is now what the room frequency response looks like (using the "normal" strength filter based on the "B&K3" target curve, again 1/12 octave smoothing):
Here's the left channel before & after applying the DSP... Note that the "with DRC" amplitude is different but I tried to keep it reasonably comparable to the "no DSP" curve:
And here's the impulse response after applying the room correction filter (right & left channels overlaid):
Notice that the initial impulse for both speakers are time-aligned now. As you can see, early first reflections are still there and these need to be dealt with in physical room treatments as suggested above (like removing the coffee table). And here is the before & after for the left channel again:
Clear improvement in both frequency response down to 15Hz and cleaner impulse response results.
Remember, I used the E-MU 0404USB for initial playback and recording of the sweep tone, but for verification purposes, the REW measurements above were done with the TEAC UD-501 playback with filters applied in JRiver. As I mentioned above, I know that the TEAC DAC, the E-MU 0404USB, and Transporter all have very similar frequency response and utilize sharp linear phase digital filters. As a result, the calculated filter should be usable interchangeably with the DACs without concern as demonstrated here.
Folks, now this is significant audible difference; and a change for the better. The frequency response is now much smoother across the audible spectrum. The difference is especially notable in the lower frequencies - for example, just listening to the amplitude consistency with the bass line at the start of Rebecca Pidgeon's old-school audiophile demo track "Spanish Harlem" (off The Raven, 1994) there is a clear evenness to the amplitude of each note. Likewise, there are clear benefits to modern dance/electronica/pop like the Black Eyed Peas "Boom Boom Pow" (from The E.N.D., 2009) where the plunging bass remains tighter and more controlled. Furthermore, as you can see, the frequency response now is similar to the "B&K3" target curve used which has a gentle gradient in the treble range which sounds less harsh and improves long-term listening especially with poor or "tinny" masterings.
Sure, you can accomplish a similar feat with graphic and parametric EQ, but well implemented DRC is also about optimizing the ability of the sound system's imaging and soundstage rendering (time domain characteristics). With good recordings like say the Ray Brown Trio's Soular Energy (DVD-A rip 24/96) or Al Di Meola et al.'s Saturday Night In San Francisco, the speakers literally disappear into the soundstage and instruments have an improved dimensionality about them horizontally and in apparent depth. Placement of instruments sound more defined and consistent in space, timbral quality more natural (or at least more true to the recording quality). While already good, QSound effects like Roger Waters' Amused To Death or his live In The Flesh become even more convincing in simulating the 360-degree surround. Check out HAL 9000's voice ("Stop Dave...", "I'm afraid Dave...", "My mind is going... I can feel it...") on "Perfect Sense", disk 2 of In The Flesh.
Even with mono recordings, the precision of the central image becomes more clearly defined in space. Channel amplitude balance overall has improved through the audible frequency range. I can imagine some people not liking this level of "focus" if they've been listening to channel asymmetry and liking it all along!
VI. ConclusionsLet's go back to where we started with Peter Mitchell's Stereophile article:
"In my judgment, acoustic environment correction is the single most important advance in audio since the CD - perhaps the most important advance since the advent of stereo."Yes, I concur. In the last number of months of switching back and forth between DRC-enabled and DRC-disabled listening, there is no question where my preference lies. In the face of the significant improvement to fidelity that can be achieved, that people would spend hundreds if not thousands on things like audio cables, various power supply / outlet tweaks, and supposed jitter reduction devices which in my opinion at best may make minimal audible difference (if we even accept that they do make any difference at all!) is rather amazing. Performing room measurements like these do take some time and require some technical know-how (carefully done, no digital clipping...), so it's not as straight forward as a "plug and play" purchase. I must say though that DRC Designer was very easy to use and provides a good start. I'd certainly be tempted to purchase something like the miniDSP products if a box that just "did the job" is what's needed although there will still be some room measurement work using a PC.
Although obviously DSP cannot fix everything wrong with a room's acoustics, I do believe it is potentially beneficial no matter how great or expensive your hardware is already. Remember, the most important piece to the audio system really is your room and given that every room is different, DRC provides an extremely powerful method to take that into account... Think of it as the icing on top - and it can truly be a delicious addition to the cake.
DRC Designer has been impressive already (major kudos to Denis Sbragion and Alan Jordan), but given what I've heard so far, I went ahead and purchased AudioVero's Acourate. Lots of positive comments on effectiveness and quality of support (there are of course some other software options like Dirac Live, Audiolense, MathAudio Room EQ). Going over some of Mitchco's articles over the next while will be fun :-). Even more challenging, getting multichannel correction going with playback through the HDMI output to the AV receiver (once I get my Onkyo TX-NR1009's faulty HDMI board repaired under "extended warranty")...
This should be a good autumn season with stuff to experiment and play with. I hope the discussion in this post presents an example of a "real life" sound room with all the limitations as we make our domestic places more conducive to high-fidelity playback... With the ultimate goal of maximizing enjoyment.
As always... Hope you're all enjoying the music!