MEASUREMENTS: S.M.S.L. SA300 - Infineon MERUS-based Class D desktop amplifier. (Screwed up New York Times - Wirecutter/Butterworth measurements/review?! And Klipsch on TIM.)

"BAS" on screen indicates I'm using the bass-boost EQ here.

Today, let's have a look at the little SMSL SA300 desktop amplifier I'm showing above sitting beside the Topping DX3 Pro V2 DAC (previously measured and reviewed here).

As mentioned previously, I found myself in the position of needing to update my computer workstation desktop speaker system. I figured, instead of staying with powered/active speakers, since I do have a few bookshelf speakers around the home, why not try some passives on the desktop as well?

While active speakers are great in that the built-in amplifiers and transducers can be well-matched and optimized, the ability to "mix-n-match" passive speakers while opening up the potential to upgrade amplifiers I think is fun for the hobbyist. As such, I found myself drawn to getting a small low-power Class D amplifier like this.

Since the amplifier will likely be left on 24/7, I wanted something that's highly efficient but provides adequate power. This SMSL device internally is powered by the Infineon Technologies MERUS MA12070 Class D amplifier which uses their "multi-level" modulation such that the switching output can have half-voltage levels which adds an extra level of control and power savings. The device is rated to provide up to 2x30W continuous into 8Ω or 2x80W into 4Ω but realize this is with 10% THD+N. What will be more interesting to me is how much power is available into something like 4Ω with ≤0.1% THD+N, and the noise this switching device produces. Let's see if this amp lives up to hopes of "high-fidelity" playback...

I. The amplifier itself...

To start, let's have a look at the SMSL device and a few of the physical features:

As you can see, it comes in a typical box. They have different color devices to suite your taste. The blue one I got looks nice and adds a little highlight to my tabletop without drawing too much attention to itself. Almost everything these days can have that Japan Audio Society "Hi-Res Audio" logo which originated back in 2014. Have a look at this quick slide set to see the meaning of that logo. Depending on the device, the "rules for approval" are variable and we have little assurance that a company conforms to a strict standard. For example, as an amplifier, the device has to have "amplification performance of 40kHz or above" - not exactly a difficult hoop to jump through ;-).  For those interested, have a look at Mark Waldrep's discussions and critique.

Comparing the amp with the size of its power supply and cables, this is a small box! About 7cm/2.8" wide x 7cm/2.8" tall x 15.5cm/6.1" deep box (excluding protrusion of antenna and volume knob). The power supply is rated as 24V 6.75A. Workmanship on that CNC metal shell is nice with clean edges, solid little metal buttons for power ON/OFF, and source select (line in, Bluetooth 5.0 with aptX, and USB input). There's a larger turn/button press knob for menu settings which allows selection of EQ options, treble/bass tone +/-10dB, and color of the screen text. The LCD screen is always on; it's not bright or distracting on a computer desktop in use. I would have liked an option to dim the screen nonetheless.

The plastic remote is the typical light, inexpensive variety, 2xAAA batteries needed. It's convenient and replicates the front panel controls with an added mute button.

Here's a look at the rear:

As expected we have the analogue RCA ins, pretty good speaker binding posts, a microUSB input for digital direct input and Bluetooth antenna. I appreciate having that subwoofer RCA out. I'm unlikely to be using an active subwoofer on my computer desktop system any time soon but it's nice to have the "2.1" playback option.

I had no problems with Bluetooth pairing or playback with my phone and as usual, aptX sounded good. Since I have a good DAC with headphone out (the Topping DX3 Pro), I'm generally not using the USB input which allows up to 32/384 input for desktop listening. Having said this, for convenience, you'll see that most of the measurements I make in this post will be with the SMSL's USB digital input. I suspect many/most users will be taking advantage of this feature. USB connection with the various Windows computers and Linux-based Raspberry Pi audio streamer ran stably without special drivers.

Even after hours of music playback, this amp stays cool - even cooler running than the Topping DAC!

II. Basic Characteristics

As usual, let's start the amplifier measurements with a look at some basic and "microscopic" characteristics. I will be connecting the USB input to a Raspberry Pi "Touch" streamer where I have the test signals stored.

This is a Class D amplifier. As I have perhaps hinted at in the past, even though I enjoy listening to them, I don't enjoy measuring Class D devices because they inherently have much more ultrasonic noise than a standard Class AB design. As a result, one has to be very careful that noise doesn't result in distorted measurements and depending in the set-up, it's quite easy to exacerbate feedback. I have already blown up an old Class D amp last year when I was starting out with doing measurements on my home rig!

The Infineon chip amp we have here is unfiltered and depends on being connect to an output device (ie. your speakers) with a certain amount of inductance that will taper the high frequency noise. It is possible for the noise to overwhelm the measurement system. If you look at Infineon Technologies' datasheet, their results are published with 22μH in series with whatever resistive load (ie. 4/8Ω typically). Indeed without this inductive load, what I see on my scope is significant >10V high frequency "spikes" even during times of "silence". See this QuantAsylum article to appreciate the situation, and as per that article and Infineon's datasheet, I constructed some 22μH "extender" cables to introduce inductance into the otherwise resistive load. Doing this will result in some high frequency roll-off and in concert with the RME ADI-2 Pro FS (R BE) ADC's filter (AKM AK5574, Sharp filter), we should be able to get stable readings.

Here are the 22μH extenders, one for each channel I'll be using in series with my 4/8Ω load:

Without the inductor in place, the oscilloscope reading is hard to decipher. But with them in place, here's a more reasonable looking +/-2V 1kHz square wave:

The ultrasonic noise from this device is much more "spread spectrum" than previously seen with the Hypex NC252MP. If I put this on my Rigol oscilloscope and peek into the megahertz range (no inductor to filter out the noise), we see quite a bit of stuff up around 1.2-1.5MHz particularly. Presumably this is related to the clock speed, harmonics, or maybe intermodulation products from the power MOSFET switching frequencies (which varies depending on power level).

Using the RCA analogue inputs, I was able to determine that at maximum volume, this amplifier can provide up to +34dB voltage gain with the signal fed into it.

I was not able to get a good reading for the damping factor using my usual procedure of comparing V(load) vs. V(unloaded). I'm not able to see output when the speaker terminals are open. We can still compare the frequency response between the 4Ω load and my old Sony SS-H1600 speakers to get an idea of how well the amplifier handles a variable impedance load:

Into the 4Ω load, it's quite a flat response across the audible frequencies. For completeness, in dotted pink, we see the effect of that 22μH inductor with a -1.7dB dip into 20kHz (see cursor at 20kHz). This is as expected when putting such an inductor in series with the 4Ω resistance:

Although I could not get a good reading of damping factor across the spectrum as I usually do, we can see that the Sony speaker response varies within <0.4dB range from 20Hz to 20kHz. Not the tightest I've seen unlike the "iron grip" of the Hypex NC252MP, but way better control than the Pass Amp Camp Amp 1.1! Also we're seeing flatter bass response down to 20Hz than the inexpensive TI TPA3116 Class D amp.

III. Single-Tone Harmonic Distortion and Noise

Time to have a peek at the THD(+N) results. Let's have a look at the FFTs across a few voltage levels into 4Ω+22μH. As usual, these measurements were captured with the Linear Audio Autoranger (2V nominal output) and the RME ADI-2 Pro FS ADC (detailed "Measurement Of Amplifiers Rig" [MOAR] procedure and scoring as discussed here). The amplifier was set to "Direct" mode with no EQ/tone control:

Apologies that I had changed the display range for the 12V (36W) reading.

As you can see, we're looking at results from 16mW to 36W to a 4Ω load. Overall, this looks pretty good. For the most part, distortion tends to be second order and starts to transition around 6V and by 12V, the third order predominates. We'll graph the THD(+N) vs. power a little later.

In the meantime, we can double check that THD remains flat across the audible frequencies - the Harmonic Distortion vs. Frequency graph. Here's the sweep at 1V (0.25W into 4Ω), USB digital input:

Looks good. THD stays below -60dB/0.1% across most of the audible spectrum at this power level. Note that I only plotted down to the 6th harmonic to keep things cleaner although the THD includes the usual 2nd to 9th harmonics. As you can see, with the USB input, most of the distortion at this power level is 2nd harmonic and the 3rd harmonic becomes most prominent above 3kHz.

Let's now have a look at the S.M.S.L. SA300's THD(+N) vs. Voltage/Power curve into 4Ω and 8Ω (with the 22μH inductor in series), both channels driven:

Although I think most of us are aware these days that even relatively high THD is of limited audibility (as per our blind listening test a few months back for example), for the purpose of "high-fidelity", I do like to see my amps being capable of <0.1%THD+N and we do get such low distortions with this little SMSL. Into 4Ω, we get up to about 7V (12W) with <0.1% THD+N, and with an 8Ω load, I'm seeing about 6V (4.5W) of clean sound <0.1% THD+N.

If we relax the distortion limits to <1% THD+N, it can provide up to 14V (49W) into 4Ω and around 15.5V (30W) into 8Ω. Definitely not bad at all! Remember, this is measured with both channels driven by the stock power supply and continuous over a number of seconds.

Since this is a stereo amplifier, let's have a quick peek at the cross talk:

This is my simple procedure where as you can see, the left channel is playing a strong -3dBFS 4kHz signal and the right channel concurrently playing a -3dBFS 300Hz sine wave. Notice that there is minimal "seepage" from one channel to the other - impressive! Also, since we know the frequency response is flat from 300Hz to 4kHz, the very slight difference in dBFS peaks between the 2 channels suggest only a 0.08dB imbalance. On my desktop, music sounds balanced coming out of the speakers with well-anchored center image and a wide soundstage.

IV. Multitone Testing - Intermodulation Distortion and Triple-Tone TD+N

Alright, let's move on with a few more challenging test signals that might be more audibly significant than simple THD. First, the IMD tests at 2V (1W) and 6V (9W) into 4Ω + 22μH:

Not bad - IMD results ranging from -57 to -64.5dB depending on the signal used at 2V output.

Likewise, we see good results at the higher power 6V output ranging from -48 to -59dB intermodulation; all significantly <1% distortion.

Next is the TIM signal (1kHz square, 12kHz sine, 96kHz bandwidth), again, 2V and 6V into 4Ω+22μH load:

No problem. With 0dBFS at 1kHz, we see the 2kHz distortion sidebands around the 12kHz signal is peaking at less than -74dBFS with both 2V and 6V output levels. (**See "Sidebar" discussion below regarding TIM and Paul W. Klipsch.)

Finally, for this section, let's end off with my Triple-Tone Distortion and Noise measurement using the standard 2V (1W) into 4Ω as a simple benchmark of fidelity; as a bonus, I've included the 6V (9W) result also. Again, I'm using the amplifier's USB in from the Raspberry Pi "Touch" as digital source:

The 2V result at almost -62dB is quite clean. Compared to previous amplifiers tested, it's sitting at the "middle of the pack"; the Amp Camp Amp 1.1 got -30dB, the Yeeco TI TPA3116 amp scored -51dB,  Emotiva XPA-1L -81dB, and Hypex nCore NC252MP at -91dB.

V. Subjective Listening

At the time of this review, I've had this amp on my desktop for about 6 weeks and have tried running a number of speakers already including the previously measured KEF LS50. Despite the ultrasonic noise we're seeing on the oscilloscope tracings and the use of the inductor for measurements, when listening with actual speakers, the audible noise floor is very quiet. With the amp turned up to 100%, I literally will have to put my ear against the tweeter to hear a faint hiss. My computer is relatively quiet and the fan noise is much louder than this minuscule hiss.

As noted above, I've already plugged a few speakers into this amp and the differences I hear I would say are mainly qualities of the speakers themselves rather than any special limitation from this amplifier. Transients sound tight, plenty of dynamics. Even with relatively inefficient speakers - I have a borrowed pair of vintage 1976 Spendor SA1's on my table currently and the amplifier can easily drive these sealed acoustic suspension speakers to good volumes that easily fill the room. I'm listening to Kylie Minogue's new Disco (2020, DR6) album as I type this - cheerful album for these dark pandemic autumn days. For those into more traditional audiophile fare, the new Diana Krall This Dream Of You (2020, DR11, noise floor clearly not "hi-res") is well recorded, sparse, vocal jazz, a bit more subdued than usual but otherwise a typical recording from Ms. Krall. If this were a normal year with audiophile shows, I'm sure some of these standards like "Autumn in New York", "Almost Like Being In Love", or "Just You, Just Me" would be right at home featured as part of showroom demos. ;-)

Channel separation as suggested by the measured result is excellent. The bass sounds well controlled and I certainly did not feel I was lacking anything. Of course, with small desktop speakers, by nature we have limited low bass. It is convenient to use the tone control to add a few dBs for taste and I didn't hear anything unnatural when boosting the bass. I did not have any issues like unexpected shutdowns (suggestive of thermal or excess current protection) at loud playback.

As ridiculous as this might look, here's the S.M.S.L. in my sound room in the company of the DIY Hypex nCore NC252MP, Yamaha RX-V781 receiver (pulled out from rack to switch some cabling), and one of the Emotiva XPA-1L Class AB monoblocks under my Technics SL-1200 turntable (the other XPA-1L is being used for speaker measurements).

With the Paradigm Signature S8 v3 front speakers and SUB1 subwoofer connected through the Emotiva XSP-1 preamp, as you might imagine, this still sounds very decent. While we know that objectively this little amplifier isn't capable of the same performance when it comes to low distortion compared to either the Hypex or Emotiva amps, in a set-up like this, frequency response remains full, mids sound palpable though not as detailed as with the Hypex, and trebles are able to convey the sense of "presence" although transients seem to not be as dynamic. The main speakers "disappear" into a wide and appropriately enveloping soundstage on well recorded tracks like the ones I mentioned in my recent Soundroom Article. I can also hear a bit more hiss and low-level hum connected to the more sensitive larger speakers in the quiet room (to some extent related to the use of unbalanced cables). While I would not recommend an amp like this for one's main "hi-fi" system, in a pinch, this will work. :-)

I see that Zeos/Z Reviews liked this amp on the desktop, sure, it's good but I wouldn't go as far as his extremely positive review. There will certainly more performance in a small package like this to be had in the days ahead!

BTW, I really enjoyed the folk/country/indie album from Canadian Donovan Woods' Both Ways (2018, DR10) listening on the soundroom set-up one evening. Particularly love the sweet sentimentality of "Next Year" which I think has special relevance given the events of this year and all the times of togetherness we have lost with friends and family.

VI. In Summary...

Overall, I'm quite happy with the performance of the S.M.S.L. SA300, an inexpensive (currently ~US$140 street price), small desktop amplifier for use on my computer workstation. As a device which is left on 24/7, I appreciate the high efficiency of the Infineon amplifier. Using my Kill-A-Watt meter, I measured power utilization of <3W when silent and typically 4-5W with normal volume playback. If pushed to extremes, you can see the amp sucking up >25W. The enclosure remains cool even after hours of music listening.

While I did not bother to measure the DSP effects like EQ or bass/treble boost, know that these features are present which adds an extra level of flexibility for the end user. I have pushed up the bass a tad for example when doing some bookshelf speaker evaluations (like with the KEF LS50), and have not noticed any unpleasant distortion.

Subjectively, I have no doubt that this little amplifier has easily enough resolution to allow me to listen to the differences between speakers under evaluation on my desktop. Plus it provides enough power to drive even some lower-sensitivity boxes. Which gets us to the summary objective results graphic:

Due to technical reasons I could not do my usual damping factor measurement across frequencies, but as we can compare the frequency response between the 4Ω (no inductance added) load and Sony SS-H1600 bookshelf, this amplifier is able to control a reactive load reasonably well. If I had to guess, my suspicion is that the damping factor would be around 10X if we compare these results with something like the Onkyo TX-NR1009 (again, much better control than the Pass ACA 1.1).

While the -62dB Triple-Tone Total Distortion + Noise result at 2V into 4Ω isn't going to win awards for lowest distortion, this is still <0.1% and as we've seen with recent posts, this level of distortion is tiny compared to the limitations of loudspeakers! With 7V into 4Ω and <0.1% THD+N "Power Factor", this tells us that there should be enough juice here to provide a clean 12W into low-impedance transducers for desktop nearfield listening - I certainly have not had a problem with a number of small speakers.

To end off, let's talk about one more thing. Remember that the measurements above were done with the USB digital input to the amplifier. Suppose we have a better DAC lying around, could we improve the sound? Let's have a look with the Topping DX3 Pro RCA into the SMSL:

While the differences are subtle, notice that with the DX3 Pro DAC, the noise level is a little lower than with the USB input (Section IV above). This is also reflected in the TD+N score which has improved by about 3dB to almost -65dB. I would have no problem just using the SMSL's USB input; but since I have the Topping DX3 for headphone listening on the desktop, it's nice to see that a better DAC with analogue out to this amp can improve total distortion and noise a bit.

If the results above suggest that this amplifier fits your needs, go for it. ;-)

A look at the setup menu with Input (USB/Line In/Bluetooth), EQ selections (like "Super Bass", "Rock", etc...), Treble/Bass control, screen color (blue/yellow/cyan/orange/purple), and "BT Clear" to unpair from devices if you scroll down. 

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

As of this article's publication, I know of one other review with measurements of the SMSL SA300 amplifier. The results were published in this Wirecutter column "The Best Mini Stereo Amplifier" from the New York Times by Brent Butterworth comparing a number of inexpensive desktop amplifiers.

For the sake of illustration/convenience, here is the distortion vs. output graph from that article:

The "S.M.S.L. A300" line is in fact the S.M.S.L. SA300 I'm reviewing here based on the Infineon amplification chip. Clearly ladies and gents, looking at the graph above, we have a problem! ;-)

First, does it make sense that at 0.1W (into 8Ω) basically all these amplifiers have ≥0.5% THD at 1kHz (not even THD+N)? Specifically, the SMSL looks terrible on this graph with supposedly >5% THD at 0.1W and 1% at 1W, barely dropping below 0.5% across it's useful power range! Really? This is the kind of THD result one might see from loudspeakers, not amplifiers!

I was shocked when I saw these results while researching for this article because they certainly did not jive with what I was hearing on my desktop or subsequently seeing once I put the device on the measurement test bench.

Like I said at the beginning of this post... Measuring Class D amplifiers can a bit of a pain due to the ultrasonic noise inherent in the high-speed switching technology. Since Mr. Butterworth did not specify how he measured the amplifiers, we don't know what equipment was used, whether he added some inductance to the load, whether an appropriately sharp filter was applied to the ADC, or what bandwidth measured.

It's not just with the distortion results, I'm also perplexed as to why his S.M.S.L. SA300 measurement was unable to show that this device in fact is capable of producing significantly more than 10W at less than 1% THD! Furthermore, it's not just the SMSL. While I didn't find the Yeeco TI TPA3116 amp all that powerful or distortion-free in my measurements, this is the same amplifier as the "Facmogu F900" on that graph. I'm pretty sure the Yeeco can provide more than just 2 or 3W at <1% THD into 8Ω with just the stock power supply!

Even though Butterworth's graph shows the Yeeco/Facmogu F900 having much better distortion than the SMSL SA300 up to 2W, without a doubt guys and gals, the SMSL SA300 has far superior sound compared to the Yeeco/Facmogu using the stock 12V power supply. Here's his assessment of the SA300:

"In our tests, however, it sounded slightly thin, and both our ears and our measurements indicated that it was noisier and had higher distortion at low levels than our picks."

I find this suspect and seems to be more informed by the graphs than actual listening. The SMSL has better bass response, sounds quieter and certainly not as "thin" as the Yeeco/Facmogu to my ears. BTW, where are the frequency response results that he measured? That would be just as important if not more than the distortion/power graph.

The bottom line is that I suspect the comparison graph above likely has a number of gross inaccuracies. It's likely that this is the result of measurement error although we can't rule out that the SA300 he tested might have been broken.

It's good to see these days more audio reviewers include objective measurements especially for the mainstream like this Wirecutter article. I hope the importance of objective analysis continues to grow because, IMO, the typical poorly controlled, purely subjective impressions/"reviews" by untrained listeners are frankly useless and often worse than useless when they includes all kinds of misinformation as is often the case with audiophile articles.

Having said this, remember though to also evaluate objective results with some critical thinking as suggested by my concerns above! Not only do we need to ask ourselves: "Are these measurements audible?", but also "Do these measurements look right?" given what we know about performance of the class of device being evaluated and related information like the datasheet for the core components (eg. the Infineon amplifier). Measurements/objective tests are essentially science experiments done independently to replicate performance claims. The results of these "experiments" are only as good as the techniques used are appropriate and the write-ups complete enough that the reader understands the context. The details need to be adequate for the reader to also replicate the results if he/she desires. 

Now if one were to use calibrated Audio Precision gear, that will provide a level of accuracy and standardization to testing, but at a significant cost as one would expect for professional-level devices. For what I'm doing here as a hobbyist (and presumably for Butterworth's article), it is essential that I'm transparent with telling you about my test gear and techniques. I can't tell from the Wirecutter article what Mr. Butterworth did to get those results beyond his generic link to the TI measurement guideline. As a reader, one should be mindful of this not just with audiophile articles, but also more broadly, employing critical thinking when we come across "news" items of all sorts.

To me, since measurements are so much more important and have greater impact for serious hobbyists than mere testimony, I think Mr. Butterworth needs to have a closer look at his published results again and consider if he needs to either update them or retract these if indeed inaccurate. Objective measurements demand more discipline than subjective reviewing and things have to be done right especially given the long memory of the Internet. Grossly inaccurate results are like "fake news" items that create confusion rather than appropriately educating the consumer. In the audiophile world where there have been so many unsubstantiated "items of faith" promoted over years, there is a need to try and rationalize the "communications ecosystem" around this fractured and perhaps understandably marginalized audiophile hobby.

Certainly over the years I have made errors here as well. I have had to update various articles, include some addenda and over time upgraded my equipment and evolved different ways of measuring more accurately - all of this part of an empirical process to seek objective truth with the hopes of achieving high-fidelity reproduction of sound.

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

Happy November friends. With Remembrance Day / Veterans Day just passed, we're on to the Christmas stretch (Thanksgiving already passed here in Canada) - still my favourite time of the year despite pandemic restrictions.

Stay safe and I hope you're all enjoying the music!

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

** Sidebar:

Here's an interesting letter written by Paul W. Klipsch from 1984 which I believe has been passed around the Klipsch discussion forums over the years. I saw this recently on the Steve Hoffman Forum posted by Dennis0675 (click on image to enlarge):

Klipsch shared his thoughts on solid state vs. tube amps. He talks about the "controversy" around "digital disks versus analog recording".

What caught my eye most was his comment and focus on TIM (so-called "transient intermodulation distortion") of -70dB as an "approach to perfection" and this being presumably one of his criteria for "now having solid-state at home" with the vintage Crown D-60 and BGW Model 100 (misprinted as "MGW-100") amplifiers of the day.

The Crown D-60 is a 10lb stereo 35Wpc (8Ω) amp, and the BGW Model 100 is 18lbs and provides 30Wpc (8Ω) stereo. Both were designed and produced back in the 1970's. Performance wise, those amps still sound great I'm sure (remember the recent 1978 Pioneer receiver measurements). It's interesting however to see how technology has moved on over the years with a little Class D amp these days providing similar amounts of power weighing mere ounces with digital input and wireless communications capabilities!