qSpins: AudioEngine A2 & Edifier S2000 Mk III 2-Way Bookshelf Speakers

Hi everyone, thought I'd just publish a couple of "qSpins" today as I might be away for a little bit to attend to other duties in the next couple weeks.

In the last while, I've been looking at getting another pair of speakers on my computer workstation desktop. Here's basically what my desktop speaker system has looked like for years:

I. AudioEngine A2

What you see on the table above are my old AudioEngine A2 speakers, the predecessor of the AudioEngine A2+ that came out in 2013 with similar appearance, construction and speaker arrangement (3/4" silk dome tweeter, 2.75" Kevlar woofer, small front slit-shaped port below woofer). While I suspect there have been small sonic tweaks, the A2+ brought to the end user mainly new features: a USB port (no need for separate DAC), subwoofer outs, and wireless Bluetooth capability (aptX in the recent version). These are great features for modern desktop speakers. Due to their diminutive size, notice in the picture that I also have the AudioEngine DS1 plastic/silicone angled "stands"/risers to help direct them towards the seating position.

The A2 are powered speakers, said to be Class A/B, 15Wpc RMS. The left speaker is the "master" where analogue input is attached and contains the amplifier circuitry. When I got these something like 9 years ago, the cost was less than US$200 (A2+ these days go for ~US$270).

Using the "qSpin" CTA-2034A-inspired procedure as documented previously with the KEF LS50 measurements, here are the AudioEngine A2 curves:

You can compare these results with what Stereophile found in 2007. John Atkinson also published a followup with the A2+ in 2015 and described it as sounding "identical". Also, check out NO Audiophile's teardown and measurements of the A2+.

For standardization, I will plot my qSpin speaker measurements with only a 30dB range and 25dB/decade aspect ratio as you see above (and previously with the KEF LS50). Compared to the frequency response graphs in Stereophile which are typically plotted with a 50dB range, my results might look "worse" but allows us to view the graphs with a more critical eye. While similar to Stereophile's frequency response curve, the main deviation I see is that my speakers were able to reproduced frequencies above 10kHz with flatter response.

Notice the upper bass accentuation around 190Hz. You can't fight physics and with a tiny 2.75" woofer, one obviously cannot expect to achieve a great low-frequency response. There's a dip into 1kHz and then the midrange strengthens followed by a rise through the upper-mids and above in the "Listening Window" which probably would be most representative of the nearfield sound dominated by direct energy. There's some irregularity in the directivity indices we can see in the polar maps:

Note that I took some shortcuts and captured the vertical "spin" +/-60° and plotted within a range from +/-90° to show more detail. We can see the 5-10kHz dip in the DI correlates with a bit of "blooming" in the frequency response from about 30-80° off-axis horizontally. Ideally, it's best to listen to these speakers +/-40° horizontally and about +/-15° on the vertical plane. I don't know the exact crossover frequency but we can speculate it's somewhere around that 3-4kHz vertical suckout. Considering that these are desktop speakers, there should be no problem if you need to toe-in or sit at the appropriate level in relation to the tweeter height.

If we look into the time domain and waterfall CSD plots with 25dB dynamic range over the first 3ms measured along the tweeter axis:

We see both the tweeter and woofer are hooked up in  positive polarity, I detected a wee bit of ringing at around 9kHz as per the cursor on the CSD graph. Below 3kHz, there's more energy than we saw with the KEF LS50. Otherwise, it looks quite clean.

I was curious how it handled dynamic bursts; whether it was able to maintain a linear increase in amplitude for brief 200ms durations from 70 to 100dB SPL at 1m:

That looks good. Generally <1dB difference across the 30dB range.

And let's have a peek at the harmonic distortion levels using REW's frequency step across the audible spectrum at around the 85dB SPL level, at 1m measured every 1/3 octave:

As usual, I don't think there's a need to be too hung up about total amount of distortion due to limited audibility (plus I'm not measuring this in an anechoic chamber and the miniDSP UMIK-1 has its limits). We see that the A2's results are hanging around the 1% distortion level. Considering difference in price and likely amplifier quality, it's no surprise that distortion is higher than what was measured for the KEF LS50 previously. Given the poor low-frequency extension, one also expects high distortion below 200Hz. 

Since these are active speakers, there's no point talking about impedance or sensitivity as you will not be pairing then with amplifiers.

Subjectively, given that these have been on my desk for 9 years, I think it's fair to say that I'm intimately familiar with the sound. Basically, while these are small speakers and grossly lacking in the bass department, I've appreciated the relatively clean treble. They are clearly a huge improvement over most of the inexpensive consumer computer speakers and miles beyond the tinny speakers built into most monitors.

That upper bass accentuation which gives the impression of more bass than they're capable of around 200Hz will not be to the liking of everyone. I think it's no surprise that these speakers will likely not satisfy those who enjoy more natural, acoustic albums with bass content like orchestral music. Male voices at times will lack "fullness" and this can sound unnatural also. Intimate recordings like string quartets that won't reach down into the lower octaves are fine, likewise classic jazz sounded OK. Lack of bass will take away from the joys of stuff like bass drops in electronica. For pop and rock, my usual desktop-listening genres, they're again OK and certainly great for casual computer-listening (low-bitrate Internet radio, casual gaming, podcasts, news...).

So, let's try something new on the desktop ;-).

II. Edifier S2000 Mk III

After all the years with the AudioEngine A2 on my desk, I figure it was time to upgrade as I was starting to experience occasional dropouts of the right (slave) speaker - likely an issue with the amplifier.

I purchased the Edifier S2000 Mk III (US$400) online, a recently released (late 2019 - early 2020) active desktop speaker from a well-known computer audio brand. I saw good comments about this speaker on one of the Asian tech forums. Let's have a look (here's the ad page with specs and pics):

I've got to hand it to Edifier for the amazing TRIPLE-boxed almost 50lb package. Each speaker weighs something like 20lbs, the main right speaker slightly heavier than the left. Here they are unpacked on my couch:

Very nice workmanship; solid speakers with a satisfying dense dampened "thud" when you tap on the sides. The wood-grain veneer looks flawless. That long cable on the left connects the 2 speakers together and is thick. Golden Ear audiophiles who like their cables robust will be very pleased. It's also quite long so you can place the speakers up to 16' apart! On the right, you can see the power cable coming out of the "master" speaker, this power cord is only 5.5' so it seems a little short compared to the inter-speaker cable.

It comes with a remote control (quite nice, batteries included), and 3 cables - stereo RCA-RCA, phono-RCA, and a nice set of TosLink cables.

By the way, the amplifier inside is described as Class D, provides 12Wrms for the tweeters, and 50W for the mid/woofers each. They call this "tri-amped" so I'm not sure how the amps are arranged; maybe 2 bass amps and single amp for the tweeter?

I plugged the power into my Kill-A-Watt meter and it registered ~7W when playing music at a normal listening level (~20-30% volume) on the desktop so these appear to be very power efficient.

Here's the rear showing the large, flared, oblong port and front speakers with grille off:

Notice the S/PDIF TosLink and Coaxial digital inputs (no USB); the "Hi-Res Audio" logo refers to the fact that the internal DACs can handle up to 24/192 input. Bluetooth 5.0 wireless audio works well and accepts up to aptX HD (no LDAC). Bluetooth is based on the Qualcomm QCC3031. There are 2 analogue RCA line-ins. Volume control knob which digitally tells the speaker to increase/decrease volume (this is not an analogue pot with 0 and 100% limits). This knob can be pushed in to switch inputs as well. There's a physical power switch but no physical "sleep" switch - use the remote button to go to sleep (<1W power utilization). Under the volume control are bass and treble knobs for +/-6dB tuning.

To the right, we can see the speaker front with grilles off. We can see a phase plug and waveguide design to the presumably ~1" tweeter. Woofer is 5.5" aluminum. The grilles are made of a low-density fabric material wrapped around a light plastic frame:

There's a nice little OLED display which shows volume change, DSP setting change and which input is playing:

As far as I can tell there is no dimmer or way to turn the OLED off. It's not too bright to be distracting on the desktop. I know Edifier has suggested that these speakers can be used in the living room and for home theater applications. The little display might be a bit small if you're sitting at a distance and in a dark room, it would have been preferable to be able to turn off or provide a few levels of dimming.

One of the features the company has implemented into these newer Edifier speakers is their DSP. I didn't see any hi-res frequency response charts on what the settings looked like so I created my own ;-). Since I suspect most users would just leave the grilles on, here's what the settings do on-axis with treble & bass at neutral, measured at 1m:

As expected, the "MONITOR" setting has the flattest frequency response (meant as studio monitor of course). "CLASSIC" according to some company literature is meant for listening to classical music; we can see that they've implemented a 1kHz dip and accentuation of the treble. "DYNAMIC" is sort of like a "smile  curve" with accentuation of bass and treble. "VOCAL" adds some midrange lift.

I noticed right at the start of this evaluation that the grille makes a significant difference with these speakers:

Notice that the grille tones down the highs and limits the dip at 1kHz. It's important to think about this because some speakers seem to sound better with grilles off, whereas others like these, appear to be designed to account for the grille diffraction effects.

Despite knowing the above, for consistency, let's measure with grilles off unless specified. Let's also focus on the "MONITOR" DSP setting since it's intended to be the most "accurate". Test signals were sent to the speakers using Line 1 RCA input from my Topping D10 DAC. Without further ado, here are the CTA-2034-like "qSpin" results, grille off:

Frequency response range looks reasonable for small 2-way speakers. The company lists frequency response as "45Hz - 40kHz". While my microphone cannot confirm the ultrasonic frequencies, indeed the 45Hz low-end looks reasonable.

As shown above, I believe these speakers must have been "voiced" with the grilles on. Notice that the moment I take off the grilles for the qSpin (as opposed to the DSP graph above with grille on), frequency response on axis (and subsequently the other curves) become more uneven with the "MONITOR" DSP setting. In particular we see a dip at 1kHz along with 3-3.5 and 10kHz peaks. 

Directivity indices look relatively smooth but notice that they're not flat, but rather sloped up beyond 10 by 20kHz. This suggests that we're looking at progressively narrowing dispersion, a "beaming" effect into the high frequencies. We can see this on the polar maps:

Yup, on the horizontal map we can see the gradually narrowing dispersion as predicted by the DI. Notice that I was lazy again and only measured the vertical spin data +/-60°. The data suggests that it's best to stay +/-15° above and below tweeter axis and +/-30° horizontally. I don't see the crossover frequency listed but it's probably around 2.5kHz based on the dips we're seeing in the vertical spin data which accentuates the asymmetry between the tweeter and mid/woofer due to the vertical orientation of drivers. The literature seems to suggest that the crossover is performed by DSP.

The polar maps above are with grilles off - have a look at what happens with grilles on along the horizontal plane (I only measured to +90° and had the software mirror the other side since we know it's symmetrical):

Clearly the grille has an effect on the speaker directivity. There's an accentuated 6-7kHz narrowing for example. Plus we see an undulating effect likely secondary to diffraction off the grille. I've rarely seen test results with/without the speaker grilles in place so it was interesting to capture this here with the Edifier's plastic-framed protruding grille that probably exacerbates the effect. Audioholics has a nice article to review grille effects.

Let's have a peek at the time-domain...

Looks like a typical crossover design with both the tweeter and mid/woofer arranged in positive acoustic polarity like the AudioEngine A2. And here's the CSD waterfall plot:

Looks good. Steep waterfall over 3ms with a range of 25dB. A few little ripples around 4kHz and 6.5kHz but nothing to be concerned about. Notice how quickly the <2kHz frequencies decay compared to the A2 above. Whether this is meaningful while listening is unclear and might depend on the person and music; it at least suggests that the drivers and box are better dampened than the A2.

So how well does it track in terms of amplitude linearity between 70-100dB SPL at 1m?

Interesting, beyond 90dB SPL, it's "losing steam" with tracking the intended output amplitude. Should be fine for most purposes especially on the desktop but double check if the power is adequate with loud transients in a larger room. While unlikely to cause damage since it's a powered speaker with matched drivers and amplifier, I didn't push the speaker to the extremes to check the maximum amplitude.

Set to 85dB SPL measured at 1m, here's the THD based on 1/3-octave stepped frequency readings across 20-20kHz:

Hmmmm... While I don't trust the UMIK-1 microphone's results absolutely, this is telling me that the Edifier's distortion level is higher than even the AudioEngine A2. Note the -40dB line representing 1% distortion. There's a substantial amount of >1% THD here. This is unexpected and I wonder if this is a combination of factors like the drivers themselves, the internal Class D amp, or side effect of the DSP processing. I suppose it's possible that distortion might be lower if one used the digital S/PDIF inputs to the speakers. Presumably, the RCA analogue inputs must be redigitized internally for processing by the DSP prior to playback.

As I've mentioned before, I don't often see measurements checking consistency of right/left speakers to make sure they're precisely balanced. Let's have a look at how this Edifier S2000 Mk III pair does (I'll use the "CLASSIC" DSP setting here to switch it up a bit):

As you can see, I'm targeting a relatively loud 90dB SPL output level, measurements shown for both the in-room frequency response and the "gated window" response which excludes the room reflections (gated result separated by -10dB). We see reasonable consistency between the 2 speakers. At the 3.70kHz mark we see about a 2dB spread.

I trust that the distortion levels above and up to 2dB difference between the speakers across the frequency spectrum are an important reminder to audiophiles about just how important speakers are when it comes to achieving "accurate" sound - way easier to find <0.01% THD amps and DACs with <0.01dB interchannel imbalance!

We can compare the waterfall cumulative spectral decay plots for the 2 speakers to see the slight differences within the first 3ms:

Obviously, as mechanical devices, speakers cannot be perfect "copies". We're basically looking for "close enough" for our ears to not hear a difference. This is also a good reminder that when we examine measurements especially for speakers, there will be some variability depending on manufacturing standards (higher quality would result in more consistent products) and some "luck of the draw" whether one receives a "good copy" of the product from the store!

For subjective listening, I set these speakers up on my computer workstation desktop as you can see:

Compared to the diminutive AudioEngine A2, these are large speakers measuring 13.5" high, 7.8" wide, and 10.5" deep; multiple times the volume! Tweeters are at the appropriate ear-height with these and I didn't feel there was a need for stands or angulation other than a little toe-in although the table will add some distortion from diffraction.

Listening almost exclusively in "MONITOR" mode, with grilles in place, I thought these sounded "smooth" and generally very pleasant. Lindsey Stirling's Artemis sounded great with a wide soundstage and nice rendering of her violin even during complex and dense passages, capable of separating out the various instrumental parts. The Weeknd's After Hours (great trilogy of tracks - "Blinding Lights" / "In Your Eyes" / "Save Your Tears") sounded excellent on these speakers, conveying the necessary pop "energy" without sounding muddy or unpleasantly harsh. As 2-way bookshelf speakers, one can't expect good reproduction of the lowest octave from 20-40Hz but the speaker is at least giving 45-50Hz a good effort. Remember that in bass frequencies, room and placement will no doubt make a huge difference. Given the digital inputs (TosLink, coax, and Bluetooth), it would have been nice to see subwoofer output (which the AudioEngine A2+ has, a single RCA channel ".1" would be great).

"Audiophile" type music like female vocals sounded great like Norah Jones' Come Away With Me - very fine, natural sounding reproduction of her voice, detailed guitar-work, quite even bass-line on the title track. Overly dynamically compressed albums like Kanye West's Jesus Is King, as expected sounds wimpy - come on Kanye, you can do better than this kind of distorted sound with only DR5!

In summary, I actually quite enjoyed my time with these speakers but ultimately returned the pair mainly because of the physical characteristics. They look great but were a little large for my set-up. Yeah, I knew going in that the size could be an issue on my computer desktop but was willing to give it a shot. Alas, my wife uses the space behind my monitor to do work in the evenings sometimes and she found it annoying having the large rear ports pumping low frequencies at her. While I am relatively free of the WAF down in my soundroom, this is a significant factor on the computer desktop upstairs! :-)

I think it's a nice feature to have the 4 DSP modes as well as the physical control of bass and treble. Remember that even though the DSP settings can yield subtle differences, those differences are much more significant than the filter options in DACs. Bluetooth playback was reliable with my smartphone (Huawei P30 Pro) and music sounded great over aptX. The S/PDIF inputs are great to have if you're using these speakers in the living room and have a computer with appropriate outputs, streamer or CD player connected. Ideally, a USB input and subwoofer output would have been nice. It would also have been nice to be able to dim/turn off the OLED, and a button to put these speakers to sleep without using the remote control; I consider this last bit to be small issues. The remote control is plastic but had a good feel and all functions were easily accessible although the button labels could be hard to read due to low contrast in dim light.

For these speakers, the measurements suggest that they were "tuned" to be listened to with the grilles in place. If you're listening with "MONITOR" mode and prefer to have the grilles off, I felt that dialing in -3dB on the treble knob toned down a bit of the high-frequency "edge".

I suspect many listeners will very much enjoy the sound. For example, I see this person on YouTube presenting a very positive review although I find it hard to take seriously his subjective comments comparing these to the more expensive Edifier S3000Pro (~US$800) without having both available side-by-side. Yes, these speakers certainly have a good number of features and sound very good but I'm just not sure about using the adjective "incredible". Maybe I'm just not that easily impressed. At around US$400, I think the Edifier S2000 Mk III's build quality is superb and a fair price for the feature set (DSP settings, tone controls, OLED, remote control, S/PDIF, Bluetooth aptX HD) and sound quality.

To end off, let's compare the pSpin "Listening Window" graphs of the KEF LS50, AudioEngine A2, and Edifier S2000 Mk III (let's use 50dB Y-axis and 25dB/decade aspect ratio for comparisons of this sort):

Of the 3, clearly the KEF LS50 achieves the flattest frequency response, has great time-domain step response, and a nice cumulative spectral decay pattern (previously published). If I were in a studio and demanded the most "neutral", predictable sounding speaker, for sure the LS50 would be preferred.

Clearly compared to the others, the small AudioEngine A2 speakers have very limited frequency response and even though the upper frequencies sound great, you're simply not able to hear all that's there.

Despite some limitations with the Edifier including higher harmonic distortion, grille diffraction effects, perhaps limits to dynamic linearity, the Edifier S2000 Mk III does push deeper into the low frequencies than the KEF, the slight bass accentuation probably adds to a greater sense of impact, and I must admit that I enjoyed the Edifier more than the KEF on the desktop with grilles on for longer duration listening. This is an example of where personal subjective preferences play quite a role in sound/music enjoyment. If you're wondering, yes, I had both the KEF and Edifier at home for side-by-side listening for a few days so this comparison is not just by memory separated by days or weeks.

Unfortunately, I did not try the Edifier speakers in the larger listening room so cannot comment on how they would sound in a larger space.

The nice thing with seeing the objective results is learning to correlate one's own subjective impressions with what was found. As usual, measurable differences are not equally audible; for example frequency response anomalies vs. time domain vs. THD differences will be different, at times inaudible, other times subtle, sometimes even obvious perhaps over longer listening. Without a doubt, for me the frequency response is much more important, especially harsh upper-mids put a damper on my ability to enjoy music beyond 60 minutes.

Okay folks, that's all I have for this week's musings ;-).

Stay safe and enjoy the music!

Addendum: December 20, 2020

FIR DSP filter for AudioEngine A2, should work for A2+ model as well:

AudioEngine A2 - Correction Filter (64/96, 64k Taps).zip

Enhanced a bit of the bass but not too much to exacerbate port noise I hope. Strengthen midrange, and temper the upper treble a bit.