Selah Audio Purezza Bookshelf Speaker Review

  • Sunday, Sep 6, 2020

YouTube Video Review version:


This speaker comes to us from the mind of fellow DIY’r and speaker builder, Rick Craig of Selah Audio. You can find the manufacturer’s information here but I have provided some snippets below:

  • Design: 6.5” 2-way
  • Drivers: Purifi woofer and BZ Labs CQ76B ribbon/planar tweeter.
  • Frequency Response: (+1/-3dB) 40hz-20K
  • Bass Extension: -10dB @ 33hz
  • Impedance: 5 ohms/nominal, 4.2 ohms/minimum
  • Sensitivity: 84.5dB /1W/1M
  • Suggested Power: 30-100 watts/channel RMS
  • Dimensions: 15.5”(H) x 8.5”(W) x 11”(D)
  • Weight: 25 lbs.
  • Standard Finishes: Walnut,Cherry,Maple, and Oak (other veneers available for quote upon request)
  • Pricing: $2,895/pair (includes free shipping to the mainland USA)

It is worth mentioning that this speaker can also be purchased in kit form (without cabinets) from Selah for $1495/pair.

Rick emailed me some additional information he wanted to note:

  1. There are no grills but the front does have embedded magnets for them. I always supply fabric grills so you can mention that in your review. I omitted them with this pair since I mainly built them to be reviewed and for local demos. The grills of course could be added later on if I sell them.
  2. This pair has an oil finish that’s very matte and more in vogue these days with the modern rustic interiors. The surface isn’t ultra smooth but can be buffed out more if needed - again, more in part to the rustic look. The front baffle is Sycamore veneer and sides reconstituted Wenge veneer. The cost for this pair is $2,895 and this includes shipping to the mainland USA. The cost range is $2,695 - $2,895 depending on the veneer used. We can also finish them with a medium gloss conversion varnish if someone wants more sheen than this pair exhibits.
  3. The design axis is the center of the tweeter. A typical stand height is 24”-28”.

As an aside, I have also tested the Purifi Drive Unit from this speaker. The Purifi 6.5-inch woofer is an extremely low distortion midwoofer, possibly the best available in its class - even besting 8-inch woofers I have used/tested. You can find that full review here:

Objective Data

Unless otherwise noted, all the data below was captured using Klippel Distortion Analyzer 2 and Klippel modules (TRF, DIS, LPM, ISC to name a few). Most of the data was exported to a text file and then graphed using my own MATLAB scripts in order to present the data in a specific way I prefer. However, some is given using Klippel’s graphing.

Foreword: Subjective Analysis vs Objective Data (click for more)
If you have seen my past reviews, you know that I am of the mindset that objective data is at least as important as someone's subjective evaluation of a speaker. If not more. Why? Because every room is different. Every listener is different. Some know what to listen for. Others know what they want to hear based on their own preference (i.e., some prefer extended bass, some prefer more midbass punch between 120-150Hz, some prefer a response with a dip around 4kHz, etc., etc.). What one person wants or expects may be opposite of another. Additionally, the room will impact the performance and therefore what the listener hears. This means when you read another's subjective-only review you are left to resolve those variables on your own. That's not likely to happen. Unless there is objective data you can use to get an idea of the performance. With objective data you can begin to understand why the subjective review turned out the way it did. Notably, objective data keeps reviewers honest. It's hard for a reviewer to legitimately bash one product but elevate another when the two measure practically the same in every regard. Not to say two similarly measuring speakers cannot subjectively sound different. Though, odds are if they do measure the same then a huge subjective difference is most likely attributed to other issues such as setup conditions, bias, etc. Objective data is the key to accountability. Simply put: if the measurements are taken with care and you understand them, you can rely on data to help paint a more accurate description of performance than a few adjectives from your favorite reviewer (myself included). As you will see, objective measurements can provide a lot of insight in to how the speaker will perform in your room.

However, when possible, it is always best to demo speakers in your own room. Not simply because of subjective performance. But because of other factors such as aesthetics, pride of ownership, etc. In my experience, all these factors play in to how the listener “connects” with the system. A good shop or manufacturer understands this and allows buyers to try items out at home before purchasing or they offer a reasonable return policy. If you question the performance and can demo the speakers in your own home, I suggest you take advantage of the opportunity.

For all the reasons listed above: What I provide here is objective-heavy analysis. I still provide my own subjective experience but with in-room measurements at my listening position(s) so we can understand why I heard what I heard. Is it the room, the actual speaker itself, my brain, or a combination?

Now that you understand my motives, let’s get started with the review.

Impedance Phase and Magnitude:

Impedance measurements are provided both at 0.10 volts RMS and 2.83 volts RMS. The low-level voltage version is standard because it ensures the speaker/driver is in linear operating range. The higher voltage is to see what happens when the output voltage is increased to the 2.83vRMS speaker sensitivity test.


Impedance vs FR

From the above data we can see the following:

  • The tuning frequency of the enclosure is approximately 35Hz.
  • The minimum impedance dips to about 4.7 Ohms around 5.5-7kHz with an overall nominal impedance at 5 about Ohms. A separate amplifier from the typical AVR would be needed to drive this speaker to typical playback volumes.
  • There are no signs of resonance (typically visible as bumps/wiggles in the impedance that deviate from the curve).

Frequency Response:

Notes about measurements (click for info)

Frequency response data (horizontal, vertical, “Spinorama”, polar, spectrograms, etc.) are all based on a 2.83 volts RMS logarithmic sweep at 1 meter to meet the standard sensitivity measurement spec. These data are captured using Klippel’s TRF module and a mixture of ground-plane measurement and 4-pi free-field measurement. Klippel’s In-Situ Room Compensation (ISC) module is then used with the ground plane measurement to provide a ‘reference’ curve to the 4-pi measurement which then corrects for the room’s influence and allows me to generate a reflection-free far-field response from an indoors measurement. Note: This is not a standard merge of nearfield and farfield nor a merge of ground-plane and farfield. Typical merged responses still suffer low resolution in the midrange where the response is merged due to the necessity of windowing the impulse response to remove reflections. One major downside to “gating” or “windowing” the impulse response is this low-resolution does not show resonance in the midrange. For example, most free-field measurements are only reflection-free until approximately 3 milliseconds, or about 300Hz. That means a data point every 300Hz. If you have a high-Q resonance at 450Hz the 300Hz resolution data will not show this resonance because the frequency resolution only has a data point at 300Hz and 600Hz; skipping right over the 450Hz. You would need a resolution of at least a half the width of the Q-factor; generally, 20Hz is adequate. However, 20Hz resolution is roughly 50ms of window-free response. The only way to achieve this is in a large parking lot or open field. Ground plane measurements are perfect for this but are subject to aiming/ground absorption (grass) and related issues above 400Hz. The ISC module permits results with as-close-to-anechoic as one can achieve without being anechoic. Thanks to the ISC module, the data I am providing here is higher resolution (~30Hz resolution) than an average person can provide without access to an anechoic chamber or the like.

The measurement below provides the frequency response at the reference measurement axis - also known as the 0-degree axis or “on axis” plane - in this measurement condition was situated at the ribbon.

fr horz

The mean SPL, approximately 85dB at 2.83v/1m, is calculated over the frequency range of 300Hz to 3,000Hz.

The blue shaded area represents the ±3dB response window from my calculated mean SPL value. As you can see in the blue window above, the Selah Audio Purezza has a ±3dB response from 65Hz - 20kHz. Even better, a tighter window of linearity is provided in gray as ±1.5dB from the mean SPL and this speaker stays within this window from about 200Hz all the way to 20kHz (ignoring the 0.5dB around 3.7kHz). That is impressive!

The speaker’s F3 point (the frequency at which the response has fallen 3dB relative to the mean SPL) is 65Hz and the F10 (the frequency at which the response has fallen by 10dB relative to the mean SPL) is 34Hz. This speaker has nice low frequency extension.

You will most likely be drawn the low-frequency trim below ~300Hz, showing about 2.5dB drop from 300Hz to 100Hz. I asked Rick about this and this was his reply:

The amount of baffle step compensation was determined from how I thought most users would place the speaker (in close proximity to the rear wall). I could easily change the BSC if someone places it far out into the room and / or wants it voiced differently. There’s always the tradeoff of output capability / sensitivity in the midband and I wanted to keep the sensitivity more reasonable. The tuning is also slightly overdamped which contributes to what you measured as well.

So, Rick can adjust this amount of trim or you can use a DSP shelf-filter to boost this back up. Personally, in my initial listening tests I actually really, really liked the bass. However, I ultimately used a low-shelf filter to boost the lower frequencies. I will discuss this more in the subjective evaluation below and give some of my thoughts on this.

Another thing that stood out to me was the lack of midrange resonance like I measured in Buchardt’s S400 speaker which seemed to be attributed to the passive radiator breakup and that I believe I have seen in other speakers using a passive radiator. There is no midrange resonance here.

Below are both the horizontal and vertical response over a limited window (90° horizontal, ±40° vertical). I have provided a “normalized” set of data as well. The normalization simply means that I took the difference of the on-axis response and compared the other axes’ measurements to the on-axis response which gives the viewer a good idea of the speaker performance, relative to the on-axis response, as you move off-axis.

fr horz

fr horz norm

fr vert

fr vert norm

As I said above, the provided frequency response graphs were given with a limited set of data. I measured the response of the speaker’s vertical and horizontal axis in 10-degree steps over 360-degrees. Nearly 70 measurements in total are represented in my data. As you can imagine, providing all those data points in a single FR-type graphic below is a bit overwhelming and confusing for the viewer. A spectrogram is an alternate way to view this full set of data. This takes a 360-degree set of data and “collapses” it down to a rectangular representation of the various angles’ SPL. I have provided two sets of data: one set for horizontal and one for vertical. Each set consists of 2 graphics:

  1. Full response (20Hz - 20kHz with the angles from 0° to ±180°) with absolute SPL values
  2. Full, “normalized” response (20Hz - 20kHz with the angles from 0° to ±180°) with SPL values relative to the 0-degree axis

Normalized plots make it easier to compare how the speaker’s off-axis response behaves relative to the on-axis response curve.

spec horz

spec horz norm

spec vert

spec vert norm

The above spectrograms are the standard way of providing directivity graphics by most reviewers. Some prefer not to normalize the data. Some prefer to normalize the data. Either way, it’s a useful visual to get an idea of the directivity characteristics of a speaker or driver.

However, these “collapsed” representations of the sound field are not very intuitively viewed. At least not to me. So, I came up with a different way to view the speaker’s horizontal and vertical sound field by providing it across a 360° range in a globe plot below. I have provided both an absolute SPL version as well as a normalized version of both the horizontal and vertical sound fields.

Note the legend provided in the top left of each image which helps you understand speaker orientation provided in my global plots below.

360 horz polar

360 horz polar norm

360 vert polar

360 vert polar norm

CEA-2034 (aka: Spinorama):

The following set of data is populated via 360-degree, 10° stepped, “spins” from vertical and horizontal planes resulting in 70 unique measurements. Thus, this is sometimes referred to as “Spinorama” data. Audioholics has a great writeup on what these data mean (link here) and there is no sense in me trying to re-invent the wheel so I will reference you to them for further discussion. However, I will explain these curves lightly and provide my own spin on what they mean (pun totally intended). Sausalito Audio also has a good write-up on these curves here. Furthermore, you can find discussion in Dr. Floyd Toole’s book “Sound Reproduction”. Here’s my Amazon affiliate link if you want to purchase it and help me earn about 2% of the price. And, finally, here is a great video of Dr. Toole discussing the use of measurements to quantify in-room performance.

In short, the CEA-2034 graphic below takes all the response measurements (horizontal and vertical) and applies weighting and averaging to sub-sets and can help provide an (accurate) prediction of the response in a typical room. If there is a single set of data to use in your purchase decision, this is probably it.

Alternatively, click this arrow, if you want my quick take on what these curves mean without going to another site.
  • On-Axis is simply the on-axis response. This is the 0-degree response curve.
  • Listening Window is an average of the 0° to ±30° horizontal and 0° to ±10° vertical response curves and is used to understand what listeners typically hear in a home at the sweet spot, or Main Listening Position (MLP). The reason for this extended window of sound is simply because your room makeup might differ from another’s. This curve is an attempt to quantify a speaker’s performance over a smaller window that is often the norm for listening angle differences in various homes. It is important for this curve to very closely mimic the on-axis response. Deviations of the Listening Window curve relative to the on-axis response curve indicates a compromise in the speaker; often caused by directivity changes (as a speaker transitions from one drive-unit to another a la midwoofer to tweeter, or as a tweeter’s response becomes highly directional).
  • Early Reflections is very useful because it helps us determine how the room’s influence will alter (corrupt, most of the time) the direct (on-axis) response. Ideally, the speaker radiates sound uniformly with no aberrations; no resonance, no directivity changes as the speaker transitions from the mid to the tweeter and so on. Because speakers often have these issues, however, what is reflected to us from the walls, ceiling and floor is not the same as what we hear from the on-axis, direct sound. And that’s a problem. Why is that a problem? As stated in Dr. Toole’s book “these are very influential in establishing timbral and spatial qualities”. Large deviations in this relative to the on-axis response also indicate areas where the room is of consequence. Also, it is important to understand the Early Reflections response is made up of rear-firing sounds. A speaker drive-unit is omnidirectional (radiating in all angles evenly) until the half-wavelength equals that of the drive-unit diameter. When the diameter is larger than the wavelength being played, the sound transitions from omnidirectional to directional; also known as “beaming”. Even tweeters beam. For example, a 1-inch dome tweeter will beam at approximately 6750Hz (speed of sound ÷ 2 ÷ diameter). In most speakers you have a single tweeter, firing forward. You can imagine that the high-frequency response in the front of the speaker would therefore be quite different than what is measured behind the speaker. So, being that the Early Reflections curve includes rear-hemisphere measurements you can understand that the high-frequency response would slope downward vs the on-axis response. This is understood and accepted.
  • Predicted In-Room Response curve has the benefit of showing directivity mismatches at the crossover as well as resonances easily by comparing them to the overlaid Target curve (further down).
  • Directivity Index (DI) curves are the difference in the Listening Window and the respective Early Reflections or Sound Power curves. My understanding, currently, is the Sound Power and Sound Power DI aren’t quite as useful for typical homes. However, there is emphasis placed on the Early Reflections DI curve. The right Y-Axis provides a value associated with the DI curves. The higher the number, the more directional the speaker. For example: a “0” DI curve - a curve which is completely flat - would be a speaker that is purely omnidirectional; radiating uniformly in all angles vertically and horizontally. A speaker that increases over frequency means that it is radiating in a tighter window as you increase in frequency. This is typical because, as I discussed above, even tweeters beam… and most speakers have a single tweeter facing the front and therefore, the speaker becomes directional at whatever the tweeter’s beaming frequency is. There isn’t necessarily a one-size-fits-all DI curve value. Though, it seems people (myself included) prefer a speaker with a wider soundstage which is found in lower directivity speakers (because more sound is bouncing off the side walls; which confuses the use of side-wall absorption but that’s for a later debate). However, what is important is that the curve, however tall you may prefer it to be, is smooth; almost linear. Dips and peaks mean that something, not good, is going on. But a linear curve indicates excellent transition through crossover regions, no resonance, etc. Since speakers are not perfect, though, linear DI curves are not the norm. Speakers become directional as they increase in frequency, around strong resonances, and as the sound transitions at the crossover from one drive-unit to another and you wind up with areas with peaks and/or dips whether they’re spread through a wide frequency range (low-Q) or very sharp/drastic (high-Q). But when you’re looking at the Early Reflections DI curve, look for this: smooth.

Below is a breakout of the typical room’s Early Reflections contributors (floor bounce, ceiling, rear wall, front wall and side wall reflections). From this you can determine how much absorption you need and where to place it to help remedy strong dips from the reflection(s). Again, as a pointer to the wide horizontal envelope, notice how the Rear Wall Bounces Curve is relatively high in amplitude (for a front-facing tweeter, at least) until about 10kHz.

early reflections

And below is the Predicted In-Room response compared to a general Target curve equaling -1dB/octave.

predicted vs target

The placement of the target curve’s starting point at 200Hz (which is my standard) combined with the previously discussed low-frequency trim of 2.5dB, makes the difference between the target curve and predicted in-room response more apparent than I suspect would actually be. So, I have also provided another version of this image below, using 300Hz as the starting point.

predicted vs target

You may ask just how useful the above prediction is. Well, I’d be remiss for not delving in to that a little bit here. Please see my Analysis section below for discussion on this. :)

Harmonic Distortion:

Measurements were completed in the nearfield (approximately 0.30 meters with room correction applied via Klippel’s ISC module) using Klippel’s TRF module. Multiple output levels were tested to provide the trend of distortion component profiles and to provide a comparison against other drive units I have tested. The SPL provided is relative to 1 meter distance, averaged in the noted bandpass region.


Maximum Long Term SPL:

The below data provides the metrics for how Maximum Long Term SPL is determined. This measurement follows the IEC 60268-21 Long Term SPL protocol, per Klippel’s template, as such:

  • Rated maximum sound pressure according IEC 60268-21 §18.4
  • Using broadband multi-tone stimulus according §8.4
  • Stimulus time = 60 s Excitation time + Preloops according §18.4.1

Each voltage test is 1 minute long (hence, the “Long Term” nomenclature).

The thresholds to determine the maximum SPL are:

  • -20dB Distortion relative to the fundamental
  • -3dB compression relative to the reference (1V) measurement

When the speaker has reached either or both of the above thresholds, the test is terminated and the SPL of the last test is the maximum SPL. In the below results I provide the summarized table as well as the data showing how/why this SPL was deemed to be the maximum.

This measurement is conducted twice:

  • First with a 20Hz to 20kHz multitone signal
  • Second with a limited 80Hz to 20kHz signal

The reason for the two measurements is because it is unfair to expect a small bookshelf speaker to extend low in frequency. Applying both will provide a good idea of the limitations if you were to want to run a speaker full range vs using one with a typical 80Hz HPF. And you will have a way to compare various speakers’ SPL limitations with each other. However, note: the 80Hz signal is a “brick wall” and does not emulate a typical 80Hz HPF slope of 24dB/octave. But… it’s close enough.

You can watch a demonstration of this testing via my YouTube channel:

Test 1: 20Hz to 20kHz

Multitone compression testing. The red line shows the final measurement where either distortion and/or compression failed. The voltage just before this is used to help determine the maximum SPL.


Multitone distortion testing. The dashed blue line represents the -20dB (10% distortion) threshold for failure. The dashed red line is for reference and shows the 1% distortion mark (but has no bearing on pass/fail). The green line shows the final measurement where either distortion and/or compression failed. The voltage just before this is used to help determine the maximum SPL.


Test 2: 80Hz to 20kHz

Multitone compression testing. The red line shows the final measurement where either distortion and/or compression failed. The voltage just before this is used to help determine the maximum SPL.


Multitone distortion testing. The dashed blue line represents the -20dB (10% distortion) threshold for failure. The dashed red line is for reference and shows the 1% distortion mark (but has no bearing on pass/fail). The green line shows the final measurement where either distortion and/or compression failed. The voltage just before this is used to help determine the maximum SPL.


The above data can be summed up by looking at the tables above but is provided here again:

  • Max SPL for 20Hz to 20kHz is approximately 103dB @ 1 meter. The compression threshold was exceeded above this SPL.
  • Max SPL for 80Hz to 20kHz is approximately 102dB @ 1 meter. The compression threshold was exceeded above this SPL.

Surprisingly, the max SPL test from 20Hz to 20kHz is higher than that of the 80Hz to 20kHz test. However, only 1dB different. A re-test yielded the same results.

Extra Measurements:

Simulated Response Using Low-Shelf Filter at 200Hz.

Given the LF trim of this speaker, I decided to see what a simple low-shelf filter would do in order to correct the response and flatten it. Think of this as a ‘tonal contour’. I applied a single low-frequency shelf at 200Hz with a gain of 3.5dB and a Q of 0.70 to arrive at the below CEA-2034 response curve set.


As you can see, the response is incredibly smooth and flat wtih this single band of EQ applied. I discuss my subjective impressions of this adjustment in the subjective evaluation section below.

Nearfield measurements.

Mic placed about 0.50 inches - relative to the baffle - from each drive unit and port. While I tried to make these as accurate in SPL as I could, I cannot guarantee the relative levels are absolutely correct so I caution you to use this data as a guide but not representative of actual levels (measuring in the nearfield makes this hard as a couple millimeters’ difference between measurements can alter the SPL level).

Notice how there is no resonance in the nearfield measurements. It’s not at all unusual to see port resonance or woofer resonance outside the crossover; these issues tend to creep up as distortion (linear and non-linear) in the speaker evaluation. However, there is none of that here.



step response

Subjective Evaluation:

Before I dive in to the subjective feedback let me first give you the layout of my room…

room layout

If the false wall part is odd to you, here's some background. I don't like seeing speakers when watching a movie. So, I built a false wall and used an acoustically transparent screen with speakers behind it. The wall is only 2x4's; no panels of wood or anything. Just a skeleton of a wall to give me something to attach the screen and acoustic treatment to. There is 2-inch wedge foam affixed to the 2x4 studs and between the false wall and back of the room are the front speakers (L/C/R & 18-inch subwoofers).

My demo music:

Title Artist Album
Enjoy The Silence Depeche Mode Best Of Depeche Mode, Vol. 1
Higher Love Steve Winwood Back In The High Life (MFSL UDCD-611)
24K Magic Bruno Mars 24K Magic
Magic The Cars Heartbeat City (MFSL)
Everlasting Love Howard Jones The Best of Howard Jones
Kodachrome Paul Simon There Goes Rhymin’ Simon
Everybody Wants To Rule The World Tears for Fears Songs from the Big Chair (2014 Deluxe Edition - Disc 1)
Know Your Enemy Rage Against The Machine Rage Against The Machine (Hybrid SACD)
Doo Wop (That Thing) Lauryn Hill The Miseducation of Lauryn Hill
Tell Yer Mama Norah Jones The Fall
Don’t Save Me HAIM Days Are Gone
He Mele No Lilo Mark Keali’i Ho’omalu and Kamehameha Schools Children’s Chorus Lilo And Stitch
Wrapped Around Your Finger The Police Synchronicity
Sledgehammer Peter Gabriel So
Feel It Still Portugal. The Man Woodstock
Free Fallin John Mayer Where The Light Is
Whiplash The Swampers Muscle Shoals Has Got The Swampers

Note: I don’t generally audition speakers with the typical “audiophile” music. I have thousands of high-quality albums ranging from pop to metal to jazz and all around. I don’t typically listen to “audiophile” music because I just don’t enjoy it. It is far more important that your evaluation music be something you are familiar with than it is to be esoteric for the sake of being esoteric. You also want to listen to music you enjoy because auditioning a stereo system shouldn’t feel like a chore. Such is the case in my evaluations. Besides, the subjective evaluation is purely to help tie to the objective data and make sense of what I am hearing to help you all get an understanding of how relevant the data is. As you will see below, my music selection did a great job at providing enough range for me to identify the issues that readily appear in the data.

Subjective Analysis Setup:

  • The speaker was aimed on-axis with the vertical listening axis on the tweeter axis.
  • I used Room EQ Wizard (REW) and my calibrated MiniDSP UMIK-1 to get the volume on my AVR relative to what the actual measured SPL was in the MLP (~11 feet from the speakers). I varied it between 85-90dB, occasionally going up to the mid 90’s to see what the output capability was. In a poll I found most listen to music in this range.
  • All speakers are provided a relatively high level of Pseudo Pink-Noise for a day or two - with breaks in between - in order to calm any “break-in” concerns.
  • I demoed these speakers without a crossover. However, based on the simulations discussed previously, I demoed these both with and without EQ. The EQ applied was done so via my MiniDSP using a simple low-shelf filter with Q = 0.70, Gain = +3.5dB and Fo = 200Hz.
  • Components: Oppo BDP-103 playing music off my thumb drive feeding signal via HDMI to a Denon AVR-X4000 which then feeds in to a refurbished Adcom GFA-545 for power.

I initially listened to these speakers and made my subjective notes before I started measuring objectively. I did not want my knowledge of the measurements to influence my subjective opinion. This is important because I want to try to correlate the objective data with what I hear in my listening space in order to determine the validity of the measurement process. I try to do a few listening sessions over a couple days so I can give my ears a break and come back “fresh”.

Upon measuring and finding the low frequency trim, I applied the EQ above and did another round of listening.

Below is the measured in-room response at the main listening position (MLP) in blue, across the front row of seats (red) both compared against the predicted in-room response (black). You can see the prediction follows pretty reasonably to what is measured in the seated position, aside from the major modal dip at ~125Hz and the dip between 2-3kHz.

The measurements/prediction below do not have EQ applied.

measured vs predicted

Here’s the rundown of my subjective notes (in quotes) and where it fits with objective:

Before I start, let me talk about Low Frequency Trim: As the speakers are (without low frequency adjustment) the sound is very, very neutral in the midbass region. What stood out to me the most in my initial listen is simply that the typical resonances of my room weren’t “activated” like they normally are. Which made for a pleasant listening session without a lot of overbearing ringing. I found 2000’s rock music to be more listenable as well. For example, I played Andrew W.K.‘s “Party Hard” on a whim. This song has multiple elements/tracks in the mix thrown together and typically sounds very congested/compressed because of the actual limited dynamic range itself… typically at higher volumes it just doesn’t sound as “fun” anymore. But with the Purezza “stock” it was enjoyable at higher output levels. I attribute this to both the low distortion but also the low frequency trim. When adding the low shelf EQ of +2.5dB @ 200Hz, bass quality improved to a more ‘expected’ fuller sound but brought out resonances in the room more easily.

With the above said, the below notes are a mixture of me using the speaker in “stock” form as well as with adding a low shelf.

  • Soundstage Depth: Depth to stage (from my seated position to the front of the soundstage) is 5 feet in front of me. The stage depth itself (front to back of soundstage) is deeeeeep.
  • Soundstage Width: Crazy wide.
  • The Purezzas possess the best soundstage I have heard to date. A very wide, very deep but also very symmetrical soundstage without one side lagging or the other side coming more forward or being more extended. Most people don’t seem to catch on to assymetry in soundstage but once you do, it’s hard to ignore and makes the stereo rendering seem fake (as if the illusion of stereo isn’t already, but I digress). The Purezzas do not have this problem.
  • Higher Love - guitar playing in first verse sounds like it’s 5 feet to the left-front of me. Very deep soundstage. Timbre of midrange is so freaking pure. Shakers/tambourines??? add layer of sound. Width is to wall 3 feet to the left.
  • 24k Magic - Excellent sound of ‘air’ even at moderate volumes (don’t have to turn it up to hear the details). 98dB. Even at 101dB the distortion is incredibly low but I can hear the woofers give (PR on back?). Crazy how loud these get with no hint of distortion.
  • The Cars - Magic - Shaker at R-C just comes out of stage. High Hat on L-C clear as a bell even though it’s somewhat low in the mix. Bass guitar; can hear the mechanical vibration of the string itself (in addition to the obvious amplified string).
  • I wanted more ~ 120-160Hz but room mode there.
  • Tears For Fears - opening synth; can hear higher frequency portion I hadn’t heard before. Resonance in mix doesn’t show up as boomy in reproduction; actually helps give the acoustic space more room. Hadn’t noticed that before. Heard backup singers @ 2:00 I never noticed before.
  • RATM - Know Your Enemy: Enveloping sound due to wide stage left and right
  • Tell Yer Mama - Great sense of space in the recording; not overly bright like sometimes ~ 0:39 + (background sound fading in and out)
  • Wrapped Around Your Finger - Bass line is incredible. Can hear the vibrato of it and it actually has a tangible feel as it energizes the room

Bottom Line

The Selah Audio Purezza is a very accurate and enjoyable speaker. While the low frequency trend native to the speaker may be a problem to some, it can be resolved by Rick easily from what he says. I simply used an electronic low-shelf filter to resolve this and was on my way (and, thus, my comments below are with this adjustment).

The soundstage is incredible. Focus is top notch. There were no gripes of mine once I made the LF trim correction (before this, the top end sounded a bit hot).

The distortion is at dangerously low levels. What does that even mean?… Well, most of the time distortion is an indicator that the volume is too loud. We hear distortion and we turn it down. With these speakers, I was often in the 95-100dB level (at 11 feet, no less) without having a clue it was that loud.

I have yet to hear a speaker of this size with this level of neutrality. The midrange is so distortion-free that it’s like listening to your music collection with a new set of ears/memory. I have to think that this is pushing the envelope, if not exceeding it, for performance in this size class and cost.

I wholeheartedly recommend the Selah Audio Purezza speakers if they are within your budget.


If you like what you see here and want to help me keep it going, please consider donating via the PayPal Contribute button located below. Donations help me pay for new items to test, hardware to build test rigs (some speakers require different test stands), miscellaneous items and costs of the site’s server space and bandwidth. All of which I otherwise pay out of pocket. So, if you can help chip in a few bucks, know that it’s very much appreciated.

You can also join my Facebook and YouTube pages if you’d like to follow along with updates.