SUMMER MUSINGS: Room EQ Preferences - Vintage Spendor SA1-like curve, Hang Loose Convolver (HLC), and the increasing relative importance of software.

Getting ready for an evening of digital filter evaluation!

I mentioned a few month back in my Spendor SA1 (1976) review/measurements that I really enjoyed the sound of these classic speakers!

While it would not be possible to exactly replicate their sound - these are complex devices with all kinds of unique properties including distortion idiosyncrasies and unique radiation patterns - to some extent, we can try to replicate the frequency response via DSP.

These days, we have the technology available to modify the sound we hear in the soundroom without great difficulty; the trick is to make sure the sound has improved, at least subjectively, when we add things like EQ or applied other signal processing.

Although some (typically more traditional) audiophiles will speak of maintaining "purity" of sound with minimalistic set-ups, if one is a contemporary technologically astute music lover and audiophile, especially one already using complex streamers, computer gear, networked libraries, I don't believe there should be any fear in going all the way using high-quality DSP to shape and adjust what we hear. In fact, from the frequency response perspective, we can "emulate" almost whatever we might want depending on the flexibility of the software used.

For fun, recently, using the Spendor SA1 (1976) measurements, what I did was extract the "predicted in-room" frequency response for those speakers and basically traced out what an idealized curve might look like as shown in the graphic below:

Using my previously captured room log-sweep (I just used the one I measured here since nothing much has changed in my room) with Acourate, I can then redesign the target room frequency curve to something that looks like the Spendor's:

Unlike the actual Spendors which are not full-range speakers, with my Paradigm Signature S8 + subs combo, I can keep a flat frequency response to 20Hz. I used Acourate's "PeakNDip" function in combination with the usual "Basic" adjustments. Added a "HF Roll-Off" up top.

From there, I created a simple FIR DSP filter to conform to that general frequency response (as per usual procedure). And here's the resultant filter simulating the pattern we see from the Spendor compared to my native un-DSP'ed measured response at the listening position:

For completeness, here's the time-domain step response post-DSP as well:

Note that this is a more idealized step response than what we get through the passive filter network from the Spendor thanks to Acourate's excessphase correction. When experimenting with Acourate (and related software), depending on your room and speakers, you'll notice that there's only so far you can go when tweaking the filter parameters so as not to have too much pre-ringing or group delay artifacts. No worries if there are slight anomalies, just make sure to listen and you're satisfied with the sound! ;-)

Even if one's speakers measured anechoically flat typically at 1-2m, the in-room "steady state" frequency response will naturally "tilt" downward in the treble region on account of frequency-related directivity at the listening position. In the graph above, we can see this natural tendency with the Paradigm Signature S8's actual "In-Room" curves.

As such, I usually will create filter variants with different amounts of "tilt" in the high frequencies. Through a high-fidelity system, compressed modern pop and rock recordings/productions are often harsh in the high frequencies, so for those, I will often listen with something like a -6dB tilt as opposed to classical or jazz where I might only apply -3dB.

For this "Spendor SA1-like" filter, I've created -3dB and -5dB variants to tame the treble as you can see here:

Based on subjective listening, I prefer the -3dB variant most of the time. For the harshest overly-bright pop or sibilant content (like say Cowboy Junkies' The Trinity Session, "Misguided Angel"), I will happily dial in the -5dB variant.

As discussed here and elsewhere, empirical research looking at listener preference "target curves" in home theaters and small rooms typically show a bass bump which I also find satisfying so long as it doesn't go overboard (see this Figure 14 from Toole, 2015):

For the average listener, there's a preference of about +5dB from around 30Hz to 70Hz. Interesting but probably not surprisingly, the "untrained" listener likes a much higher bass boost even above +10dB (and more treble as well to create that "U" shape)! Since we're all sophisticated music lovers and audiophiles here, let's not just aim for "average" ;-). I'll add +3dB bass boost to the "-3dB tilt" version for added "euphonia" - here's the frequency response design in "Target Designer" and simulated sweeps using the filters:

Now, typically I would then pop these DSP settings into Roon presets:

And away I go to listen to the sound of the "new" settings with the Spendor SA1-like frequency response characteristics.

While simply using Roon or the JRiver built-in DSP will work, insightful audiophiles know that human sensory memory where we hold on to much of the details/nuances (echoic memory for audio) is actually very short; usually a "buffer" on the order of <5 seconds. I think many audiophiles, especially the "Golden Ears" find this fact hard to believe (while at the same time arguing against controlled listening tests). While I don't doubt that major differences can be remembered and declared to be "obvious", this time duration is usually too short to allow "definitive" judgements to be made about more subtle variations. Unless one has fast fingers, going into menus and changing listening settings can take longer than this, plus it distracts one's attention which is also important for focused evaluation (listening requires both sensory acuity and cognitive attention).

Enter Mitch Barnett's Hang Loose Convolver (HLConvolver/HLC) (US$129 for license), available for Windows and MacOS (works on the new ARM CPU models). I was able to "beta test" and provide feedback to Mitch on the software over the last year - thanks for the opportunity, Mitch!

Have a look at the software "operations guide" to get an idea of how it works. Installation is clear and takes a few minutes to get up and running. There are different ways of operating the program including a standalone app used in combination with VB-Audio Virtual Cable allowing processing of loopback audio from programs like Roon, HQPlayer, or even system wide. For audio software that is compatible with Steinberg's VST3, there's a plug-in that can be used in JRiver 27.0.52+ (prior to this version, only VST2 was supported), Reaper DAW, and Audirvana can handle VST3 in Windows (use HLC's AU plugin for Mac). I see Chris Connaker's article on HLC also discussed applying convolution to DLNA streams; alas I have not used DLNA in years here at home, but this is definitely significant.

For my listening, I used the current JRiver 28 with the Surface Pro laptop, accessing test music either on the internal SSD or through the WiFi network. Notice that I just have the USB 2.0 cable connected to the RME ADI-2 Pro FS R BE DAC (16' Amazon Basics cable, no worries about sound quality - it's digital bits), and the DAC's XLR output to my Emotiva XSP-1 preamp:

And here's what the JRiver 28 screen looks like with the VST3 plugin activated, loaded with some DSP filters to try out as shown on the right window:

HLC loaded filters can be any samplerate up to 384kHz and 32-bits. ZIP files with multiple samplerate filters supported. Efficient processing with <10% CPU utilization even with the Intel i5-4300U-based Surface Pro from 2013.

I've got a couple of albums showing in the screenshot. Ēriks Ešenvalds' There Will Come Soft Rains (2020, DR13), a wonderful collection of choral works beautifully conducted by Richard Nance and performed by the Pacific Lutheran University Choir of the West (Tacoma, Washington). This album has a complex overlay of voices, wide and deep soundstage, expansive reverb. And Olivia Rodrigo's recent album SOUR (2021, DR7) which gives us a taste of highly produced modern pop.

The beauty of HLConvolver is the combination of instantaneous filter switching and the flexible volume control facilities including an "autogain" function. You can use the included pink noise tone (file in the HLC folder) to fine calibrate the filter amplitudes with the "bypass" level. Then with a single click, you can switch between each of six "filterbanks" or bypass the DSP instantaneously. The "Master Trim" control allows you to add some attenuation if you see the output level meters clip. There's simply no easier or better way to accurately compare the differences between DSP room correction filters than this! The instantaneous switching not only allows you to appreciate the frequency-domain changes as I did with the Spendor SA1 curve (and variants), but also the effect of time-domain correction. There's also a convenient Load / Save feature to quickly recall all the settings and filterbanks.

For anyone interested in DSP room correction, this is the "next level" in filter evaluation.

How does it sound with those Spendor-like filters? Obviously it's impossible to fully describe in words. The differences are not hard to hear compared to a typical "flattish" room target curve.

The dip going from 1-3kHz adds a "polite" factor to the sound that's less fatiguing. Some describe it as adding a subjective impression of depth as well, perhaps related to Linkwitz's comments here (see "Psycho-acoustic 3kHz dip") discussing the difference in diffuse field frequency response of microphones vs. human ears. In combination, I like the "punchy" sound in the "Spendor SA1 -3dB Tilt & +3dB Harman Bass Bump" version for rock music with that extra low-bass kick.

Here's what the impulse response looks like in order to "correct" for my left speaker + room (focusing on -50ms to +200ms of the 65k-taps FIR filter) with both frequency and time-domain correction. Notice the complexity. Whatever you think about MQA these days, their simple filters would obviously perform little if any actual correction for "time smearing" in playback systems! To this day, we still have little idea of what they're trying to "de-blur".

Time-domain changes are a bit harder to put the finger on but can be readily appreciated using HLC's quick A/B switching. I created versions of the Spendor SA1 filter with and without Acourate doing excessphase correction. In general, I hear a better "focus" with phase correction on. Voices like the choral parts and solo for example in "Amazing Grace" on There Will Come Soft Rains "snapped" into place better in the soundstage. Even the reverb of the "space" around the voices sounds better controlled and not as broadly diffused, as if localized to a smaller "volume" in the soundscape (obviously better appreciated with actual listening than words!).

Even with adult-contemporary pop as I revisited Peter Cetera's World Falling Down (1992) the other night, it wasn't hard to notice that his voice had extra depth and became more front-focused with time-domain correction when flipping back and forth instantaneously. This is a good example of the kind of change that I would not be able to tease out listening to an unfamiliar sound system or room such as at an audio show or the local hi-fi dealership.

While I believe correction of "time smearing" of the speakers and room is worthwhile and would be my preference, I can appreciate this opinion is also a subjective preference. Although less focused, the spacious subjective "bloom" in my system without DSP can be appealing as well without the time-domain correction. Perhaps we can draw an analogy here, sort of like how vinyl and tube amps have more distortion which to some can sound "fuller" and judged to be "better". Presumably this is why stripping away the time correction in PGGB can sound fine for some users even though I believe this lowers the ultimate rendering quality.

Even though I would not characterize my Spendor SA1-like filters as capturing all that is special about those speakers, with DSP, I believe I'm capturing a significant chunk of the mid/upper midrange tonality especially when listening to vocal music and some of the "presence" those little speakers can convey. Just another playback option to use as the "spirit moves" on those enjoyable evenings when not trying to be analytical about sound!

As I discussed in the past, feel free to design whatever subjective preference or even flight-of-fancy EQ curve you might want to use. Whether it's a flat response, B&K-like, Harman-like, throw in a "BBC Dip" (the Spendor curve would be a variant of this), accentuate the bass, emulate a favourite speaker, etc., by all means audiophiles, tweak to your ears/mind/heart's content and explore for yourself what sounds "good" to you.

Don't forget though that while we can create and explore countless filter design variations, there is still something to be said about "accuracy". This is especially important if one is an artist or in music production since you're creating "the art" itself and presumably you would want this to be experienced as intended with a neutral reference playback system which is really the most logical standard to aim for rather than some arbitrary "colored" sound. Sure, one could be selective and target the music to sound awesome on Apple AirPods but how's that going to come across with a good hi-fi system?!

Compared to the electromechanical transducers (speakers, headphones), the electronics side of audiophile hardware has over time evolved to become extremely "accurate" these days. For example, DACs even at very low price points (like the Topping D10 Balanced, or D10s) are capable of fantastic resolution which will likely never be needed by most playback systems (to a large part dependent on your room's noise floor)! If we combine that resolution potential with the computing resources we have available, we can transform the sound of our audio systems in ways that home hobbyists could not achieve with such ease (and low cost) until recent years. 

[For a fascinating look at an early digital correction device, have a look at the SigTech TF 1120 "Time Field Acoustic Correction System" review recently posted on Stereophile. The cost was about US$11000 in 1996 or ~$19,000 inflation-adjusted today, if you want the "full meal deal" including hardware, software, a computer and omni mic to do what we're talking about here but with lower resolution and less flexibility.]

I would argue that future developments that significantly elevate the audiophile's experience of sound quality and overall "quality of life" has already started shifting from hardware devices (such as streamer, DAC, even amps) towards developments in software (such as playback apps, DSP room correction abilities, virtualization of multichannel streams, metadata availability, UI).

As a music lover, I think the rise of computer audio and streaming allowing us to access the music through all kinds of apps and user interfaces has been the most important change in the last decade plus. Roon for example with its extensive metadata embedding I think has added more to my enjoyment of music than DAC, amplifier or speaker swaps in the last few years. Even though I enjoy testing out technical performance, on the evenings listening to music, it is the interaction with the software and how it allows me to tweak the sound, and learn about the artists and albums, or browse the lyrics easily that has been the most rewarding "upgrade". In terms of sound quality, software has allowed audiophiles to explore with infinitesimal precision things like playback filtering (eg. HQPlayer), providing opportunities to transcend whatever limits of the filter implemented in the hardware. Of all these techniques, DSP room correction is currently the most powerful, allowing the audiophile great latitude; not just targeting for "accuracy", but also achieving one's subjective preferences as discussed above.

Over the years, I have talked about nonsense hardware tweaks. Stuff like the ridiculous Synergistic Research HFT stickies, weird "Vibratrons", or even just expensive cables in general. Then there are folks who think otherwise bitperfect server computers can have a major effect, or fear "noise" from switching power supplies and suggest various typically expensive options with no evidence of anticipated effect. Are we sure these things truly change the sound, or allow the user to have any degree of control?

Audiophiles, let's make sure to synchronize with what we can accomplish in the 21st Century and start implementing fine-tuning that can be measured, and objectively demonstrated to unquestionably affect the sound. Ultimately, to combine objective techniques of verification and fine-tuning informed by insightful subjective individual preferences is powerful and highly rewarding; arguably I believe the most fulfilling forward direction that makes sense for the "high-fidelity" hobbyist.

For those of you who might not be experienced with DSP and all this stuff above, check out Mitch Barnett's Accurate Sound website and his services. I'm sure he'll be able to provide guidance and ideas depending on your needs.

September 2024: Also have a look at the Dolby Atmos Music Target Curve discussion as an example of a contemporary target standard.

---------------

Pssst... With Mitch's permission to post, here's a sneak peek at what he's working on ahead as he expands HLConvolver:

I have started developing an easy-to-use, headphone calibration application that uses high resolution FIR correction filters. The application is called Kick Back DSP (KB-DSP), and Hang Loose Convolver will host the FIR filters as part of the same application/plugin. To summarize, calibrated binaural microphones would be worn in the ear. Headphones worn over top. Test tones will run in both ears, and within 90 seconds, the headphones will be accurately and precisely matched to the wearer’s ears for perfect sound reproduction. Such an application could become the ultimate high resolution personal EQ system for headphones in that the correction filters are orders of magnitude greater in frequency correction resolution than a handful of PEQ filters. In addition, existing PEQ filters are based on measurements using a “dummy head” with faux ears. The measurement difference between faux ears and your ears is considerable, which renders the dummy head PEQ ineffective. Those who try out the application need to get ready for the most accurate sound reproduction they have ever heard.

Nice! Looking forward to what he can achieve in the headphone DSP correction space (I assume this procedure will be best for larger circumaural headphones). Just like our rooms will treat the sound differently, so too are the nuances of our head/pinnae.

---------------

Patricia Barber's Clique! (2021, DR14) is out now for those who like her style of vocal jazz. The audiophile press highlights music like this because it's well recorded so hardware audiophiles will like this, but to be honest, the audience among music lovers I suspect is limited. I see there's also the SACD listed to be released on September 3rd which might be a better deal for those looking for something to spin since it includes the 5.1 multichannel mix.

The CD sounds great, very clean recording with excellent tonality. Wide dynamic range (DR14 average), fantastic resolution, love the smooth decays on the cymbals. Also love the soundstage and guitar work on "I Could Have Danced All Night". Barber's voice perfectly captured without excess sibilance. Will need to listen to it a few times to get a better sense of the album as a whole.

The other CD I got this week was The Killers' Pressure Machine (2021, DR6). This one will definitely split fans of the band! Still "alt. rock" genre, but the songs are generally a bit slower, more meditative, more melancholic (although "Quiet Town" and "In The Car Outside" are quite sprightly). The standard version has "snippets" of commentary at the start of a number of the songs from rural American folks. There's an Abridged version without the comments (less distracting for streaming and playlists). At a time of diminished attention spans, I think using this system with interspersed comments between songs is an interesting way to have listeners engage with the album as a whole instead of just picking individual songs since the spoken comments do add to the mood and will "prime" the listener to the music coming up.

We've got stories of small-town tragedies like a railroad-related death, injured horse in stampede, themes around evangelical Christianity, opioid addiction, pyromania, hunting, physical abuse, working-class life, etc. The Killers do Americana, channeling their inner Springsteen.

As I said above, I think fans will have mixed reactions to this one but it's good to see the band expand their repertoire with some quieter, stripped down tunes (like "Runaway Horses"). And there's a sweetness and empathy across the album away from the glitz of Las Vegas. My favourite track is "In Another Life". Sound quality's decent, I think better than most of their other albums; admittedly, I have not heard their previous album Imploding The Mirage from 2020 yet.

Stay safe everyone! Hope you're all having a good end of August. I'll be out-of-town visiting family next week, so off on vacation again for a little bit ;-).