Saturday, 27 November 2021

MEASUREMENTS: S.M.S.L. A6 as amplifier [ICEpower 50ASX2(SE) Class D module]. (And DAC output - AKM AK4452 AMPT.)

Over the years, there have been devices I've been meaning to measure but one thing or another got in the way. This SMSL A6 amplifier was something I discussed here on the blog back in 2017 (a subjective review and measurements of DAC output). Over the years, this device has been happily amplifying signals for my dad's Klipsch Forté which is quite a sensitive speaker of 96dB SPL/m/2.83V; certainly by now it is very much "burned in". ;-)

I wanted to measure this device as an amplifier out of interest, knowing that inside this is the Bang & Olufsen ICEpower 50ASX2(SE) Class D module. While the datasheet for this amplifier module is already very detailed, there's nothing like measuring oneself!

This ICEpower module is rated at 2x47W into 4Ω, both channels driven to 1% THD+N. The power rating halves to 25W/channel into 8Ω. Within the ICEpower ASX series, there is also the 250ASX2 that will supply up to 250W/channel into 4Ω, and 125W/channel to 8Ω if you need more power.

The nice thing about this Class D module is that it's inexpensive and easily available for hobbyists such as through PartsExpress (current price US$118), sometimes through Amazon as well. You could buy one of these and get a simple case such as this Ghent Audio one and assemble it for yourself quickly.

As a reminder, the product shot above is how the SMSL looks like up front and behind. Build quality is pretty good with strong metal case, relatively sturdy binding posts, coaxial + TosLink S/PDIF inputs, and USB 2.0 port for digital in. There's a matte finish so it's not too shiny but can pick up, dust, fingerprints and oil pretty easily. Analogue inputs include phono (MM) and "CD" ins. There is the "DAC" RCA output as well which we looked at previously conversion duties provided by the AKM AK4452 chip.

The left knob is for input selection and menu changes. The right knob is for volume. Both knobs "click" when selecting up or down. The volume knob is a little "mushy" but works fine. Remote control provided, not shown.

Here's a peek at the inside with another product shot:
You can see the ICEpower 50ASX2 module towards the bottom of the picture.

In the very top picture, we this amplifier on my testbench ready to go. It's connected to the Topping D10 DAC for signal generation duties (high enough resolution). As usual, I have the Linear Audio AutoRanger (2V nominal) as well as the RME ADI-2 Pro FS for measurement of output signals.

I. Voltage Gain, Damping Factor, Frequency Response...

Starting with voltage gain, I measured +30dB with the volume max at "40".

Damping factor into my 4Ω power resistor looked like this:

Thar's very good with an average of 105x as per my standard measurement points across the audible spectrum. Notice that output impedance increases (hence dropping damping factor) as frequencies increase above 5kHz with a steep drop into 20kHz. While not more flat as with the Hypex nCore NC252MP, this is likely fine because the impedance fluctuations for most speakers occur mainly in the lower frequencies. This amplifier should easily exert good control over the woofers in your speakers.

As expected, the amplifier handled the Sony SS-H1600 (rated 6Ω) bookshelf speaker with good control (<0.1dB variation below 10kHz). Notice that the frequency response with this amplifier varied a bit above 10kHz depending on the load. In the bass frequencies, we see about -1dB down at 20Hz compared to 1kHz; this slight low-bass roll-off should not be a problem even if not ideal.

As expected with most amplifiers, phase is essentially flat across the audible spectrum.

II. Single-Tone Harmonic Distortion and Noise

As we go into the distortion measurements, let's start with looking at distortion across the audible frequencies at various output levels:

Because the volume control is not continuous, it's not possible to be more precise with the output power to achieve nice "round numbers".

The distortion amounts listed below each graph are taken at the 1kHz frequency (where the cursor is). The first thing to notice is that the noise floor (brown) is generally quite flat but does have various noise spikes at 100Hz and multiples. The other thing is the "scalloping" THD curve with a relative depression between 100Hz to 2kHz especially noticeable at the higher 5.1V and 9.6V output levels rather than a flat THD across the frequencies.

Like with the frequency response, although this is not ideal, at least distortion is lower in the frequencies most sensitive to human hearing. However, this also tells us that there's some frequency-dependence on what "notes" are being reproduced, the total and relative amounts of distortion (ie. which harmonics predominate) will shift. Ideally, I think it's nicer to see a more consistent distortion level with harmonic levels maintained in a consistent relative order across the frequencies such that there is little potential timbral shift. In any event, we are looking at low harmonic amounts so actual audibility needs to be assessed with controlled listening tests.

Okay, let's now have a peek at some FFTs themselves - 1kHz across some output levels into 4Ω.

and similarly, we can have a look at a few FFTs at various voltage levels into 8Ω.

As a generalization, below about 4V, 2nd order harmonic distortion appears to predominate, above that, the 3rd harmonic becomes strongest.

So if we plot that Distortion (THD+N) vs. Output Voltage graph for the 4Ω load, it looks like this:

We see a bit more noise and distortion with more than 0.1% THD+N below ~75mW. Otherwise, the distortion stays below -60dB/0.1% from 100mW to 42W which I think is a pretty good range.

A quick crosstalk estimate:

Averaging the amount of leakage from left to right and right to left at 300Hz and 4kHz, we get -80dB. This is good, similar to the Hypex nCore NC252MP. I do not anticipate any issue with good channel separation when listening.

III. Multi-Tone Testing: Intermodulation Distortion, TIM, and the Triple-Tone TD+N

Moving on beyond just a simple tone like 1kHz. Here's a quick look at the various IMD test signals measured at 2V into 4Ω (~1W power):

Intermodulation distortion looks quite clean at the 2V/1W level into 4Ω. Cleaner in fact than the Yamaha RX-V781 measured last week. The only one of these tests which scored lower than the Yamaha is the CCIF/ITU-R on account of the higher "d2H" component which is ultrasonic up at 39kHz so effectively the ICEpower has better IMD.

Here's the TIM ("Transient InterModulation") test signal:

TIM simply is not much of an issue these days with solid state amplifiers, whether the Class AB or Class D devices that I've tested.

And finally, here is our Triple-Tone TD+N based on the trio of non-harmonic 48/960/5472Hz frequencies, at 2Vrms into 4Ω:

Average of Triple-Tone TD+N at ~1W for the two channels is -58dB; we'll use this value in the composite summary score as the resolution measurement below.

IV. Square Wave and Wideband Noise

As a Class D device, we can anticipate this amplifier will stay cool without active cooling thanks to high efficiency. When turned on and idle, it uses 11W as measured with my Kill-A-Watt meter. When playing a 1kHz 2V tone into 4Ω (1W), the power draw increases to 12.5W. And if I pushed it further to 10W (6.3V), total power draw increases to 29W.

We have seen with other Class D amplifiers like the nCore NC252MP and S.M.S.L. SA300 that the Class D switching results in very high frequency noise. Indeed we can see the same effect on the ICEpower amp here - check out the 1kHz "square" signal (24/384, bandlimited to remove pre and post-ringing) on an oscilloscope: 

Notice all that high frequency "fuzz" at the peaks and troughs in the waveform. On average we see about 0.01V difference between channels which is very good channel balance.

To better define that noise, here it is on my Rigol DS1104Z's FFT embedded with 5kHz and 93kHz signals:

As you can see, the switching noise for this amp is up around 525kHz. Let's define that better and have a look at the noise without any signal playing:

559mV noise at 526kHz. Looking at the datasheet, the idle switching frequency can vary between 500-600kHz depending on the signal. Otherwise, it's looks reasonably clean into the extended ultrasonic frequencies up to 1.2MHz.

While I'm not showing the results here, like with other Class D amplifiers, a bit of load inductance will roll off the ultrasonic tone.

V. Impressions and Conclusions

As noted at the top, I've been wanting to have a better look at the S.M.S.L. A6 as an amplifier for quite awhile. The machine sounds quite good as I had described in the subjective review in 2017. Furthermore, it has been in daily use in my dad's system for the last 4 years or so without any problems as DAC and amplifier feeding his Klipsch Forté I speakers. As we saw in 2019, these Klipsch speakers drop down to 5Ω in the bass and also have a ~100Ω peak up at the 2.15kHz crossover region. It's good I think to have an amplifier like the A6 with high damping factor to keep things controlled.

I know my dad flips back and forth between this Class D amp, the Melody Onix tube amp and I think an old rebuilt Dynaco which I'll have to have a better look at another time. He does not have any strong preferences between them and he doesn't need a lot of watts. I concur that the various amplifiers sound quite good with the speakers. Regardless of sound quality, he has told me more than once that the SMSL A6 is the most convenient and runs the coolest on a hot summer day (no surprise).

Here's my summary multi-component AMOAR score into a 4Ω load, measured with both channels driven:

Overall, this SMSL A6, based on the ICEpower 50ASX2 Class D amplifier module did reasonably well in many ways. The damping factor is good although higher up in the spectrum (see the 20Hz-20kHz graph in Section I), it drops off substantially beyond 10kHz. We can see that this results in a nice and tight Sony SS-H1600 bookshelf frequency response. I usually don't bother with the 8Ω frequency response on the summary graphs like this because for most amps, it overlays on the 4Ω. However for this device, it's interesting to see that the 8Ω curve does deviate above 10kHz showing a flatter frequency response with higher impedance (the Sony speaker has >10Ω impedance above 10kHz so it shows an even flatter response). The variation is small though with around 0.3dB at 20kHz between all the curves.

[One can easily appreciate how the high damping factor in the low frequencies would make inexpensive, efficient, yet powerful Class D amps such as these ICEpower models good as subwoofer amplifiers.]

On the low end of frequency response, we see a -0.5dB dip by 35Hz and -1dB down to 20Hz. I trust this is not likely an issue unless you're running quick A/B comparisons. Ideally, I'd like to see flat down to 20Hz, but this is nothing to be worried about.

Speaking of power, that ability to deliver a continuous 13V into 4Ω (42W) while maintaining <0.1% THD+N should be enough for most home users with reasonably efficient speakers and small to medium sized rooms.

Finally, if we talk about fidelity/resolution, the -58dB TD+N on the triple-tone test at 1W is okay though not going to win any hi-fi awards. Assuming no special anomalies (for example if an amplifier physically buzzes or has an unusually high 60Hz hum), for my listening, a result of -70dB TD+N on this test (which typically correlates with -75 to -80dB 1kHz THD+N) is approximately where I would put my "good enough" threshold for high-fidelity playback in my sound room with low ambient noise level in the evenings. Sure, lower TD+N is better objectively, but I would not say that I "need" better than -70dB in order to enjoy the music any more. As usual, I think considering what is "Good Enough" for oneself is a wise exercise for all audiophiles.

Digging into some of the results and graphs, although distortions like IMD are low with this ICEpower amp, it's the noise level that is holding back the overall TD+N rating.

SMSL A6, a bit finnicky USB playback with Raspberry Pi "Touch" with Volumio often not recognizing the device (didn't try other Pi audio software). Computer USB compatibility no problem at all. Pi running off lithium battery pack.

I'm having some renovations done at home so my sound room is in a bit of disarray and I did not get a chance to have a good listen with this amp in there (although I had it hooked up in my previous report). Nonetheless, I connected this to my secondary system in the living room as per the picture above. Yeah, it sounds pretty good through my older Tannoy mX2 bookshelf speakers. Although I don't think the ICEpower amp sounds harsh at all, the pairing with the Tannoys and their smooth-sounding soft dome tweeters was a nice match I think.

Though obviously not perfect, this inexpensive amplifier sounds less "colored" than something like the much more expensive Pass SIT-2. Another interesting comparison I think is with the even smaller, inexpensive SMSL SA300 which is based on the Class D Infineon amplifier (measured here, I'm still using it on my desktop computer). While this ICEpower 50ASX2 device can provide more power and better damping with similar audio resolution, the SA300 has a flatter frequency response. I would stick with the SA300 as a desktop amp while this A6 would be appropriate for a modest secondary system.

It's not hard to find various B&O ICEpower amps online as parts, making them popular for DIY builds like this. While I cannot speak to more powerful versions of the amps like the 200AC/ASC, I expect that some of the performance qualities would be shared. For example, if we look at the datasheets, the increase in output impedance >10kHz appears quite typical.


As I discussed in the previous SMSL A6 test using the DAC output, there's the AKM AK4452 DAC chip inside. Well, since I probably won't have this device on my testbench again in the future, let's record my AMPT "Standard Music Track" with this DAC for you to hear/compare:

S.M.S.L. A6 DAC output (AKM AK4452 chip, RCA) AMPT

As usual, it's a hi-res 24/96 recording of 16/44.1 material, volume normalized for comparison with other DACs. I think it sounds great even if using an older AKM.

The qualitative (and quantitative) differences between DAC output these days is subtle for the most part. Speakers, rooms, and to some extent amplifier differences affect your playback quality much more than would a decent DAC IMO.


Well, with American Thanksgiving and "Black Friday" passed, I guess the rush into Christmas has officially begun!

Despite all that's messed up over the last couple of years with travel plans among other restrictions, I still love this time of the year while taking steps to be thankful and being deliberate with the joys of the season. For this year, with news of the COVID-19 Omicron variant, I hope my Hawaii holiday over Christmas with the family will not be disrupted - in the Archimago household, we're already double-vaccinated and I've received my 3rd as well.

I'm hoping you're all enjoying the atmosphere of the season heading into year end as well.

For those of you who use Acourate for DSP room correction, you might have received the news about an upgrade to version 2. I'm still playing with the new version and appreciate the updated interface which runs faster on my machine, better tool tips to remind what parameters do (Target Designer comments on subjective effect and frequency is cute), as well as a few more features here and there like high frequency limit on the inversion step (which will help with performing partial frequency corrections) are welcome.

There's also the new Interchannel Phase Alignment (ICPA) feature. I see that Mitch Barnett's Accurate Sound DSP video from a few weeks back already incorporated discussion of this feature as Room Macro 6 - "Asymmetry Treatment":

Alright, I'll have to play with this and listen to the effect myself the next time I update my room DSP filters, perhaps over the winter!

That's all for now friends, time to bring out some of the Christmas tunes soon including the suggested high dynamic range albums from Allan Folz last year. ;-)

Hope you're enjoying the music!


  1. Hi Archimago,

    Here's an interesting Chinese digital amplifier on TPA3255 for less than $100:

    It would be great to see your review of it :)

    1. Thanks for the idea fgk,
      Will keep an eye out for stuff like this...

  2. A little late to the party, but if one is willing to spend some more money ($349), there's now the Topping PA5 amp which got a glowing review at ASR. With so many excellent DACs and headphone amps around, I've been waiting for a small amp with the same kind of performance. Not that I really need any more stuff :)

    1. Yeah Freddie, the PA5 looks great especially paired with the D10 Balanced DAC for a low-cost but high performance "system".

      IMO, nice how companies like Topping are systematically producing devices at various budget tiers pushing the envelope of hi-fidelity.

  3. I am all in favour of class D amplifiers, and I am sure the better ones represent the future of amplification. However, in the Netherlands the trusted Yamaha AS501 costs about the same, has more features, and quite a bit more power. It is bigger of course, and probably not quite as energy efficient. I bought one for my daughter.

    1. I agree Willem, that Yamaha A-S501 looks amazing! I certainly would not have any concerns with recommending that!

  4. Thanks for doing a performance analysis of this. I had been interested in it a couple of years ago when I was putting together a high performance desktop system. I wound up purchasing a Klipsch PowerGate when it was heavily discounted at $150USD - it was originally ~$500USD.

    While the DAC section of the PowerGate left a lot to be desired, its power amplifier section was acceptable. So, I bypassed the PowerGate's DAC with a $9USD Apple 3+USB-C to 3.5mm dongle for external input and a Chromecast Audio (analog output) for streaming, and am quite happy with the results.