MEASUREMENTS: Hypex nCore NC252MP Amplifier (+ Subjective opinions around objective testing, and CES2020)

Alright, you knew this was coming from the post last week with assembly details for this amp.

As discussed, this is an easily assembled single-board OEM Class D stereo amplifier rated at 150Wrms into 8Ω, 250W into 4Ω, and 180W into 2Ω. The enclosure comes from Ghent Audio. The amplifier is based on the Hypex nCore "phase shift controlled self-oscillating loop" design.

Let's run this box through my usual MOAR testbench as previously described and see what kind of data we get. In general for these tests, the signal path for measurements looks like this:

RME ADI-2 Pro FS AKM DAC (signal generator) --> XLR --> Douk/Nobsound NS-05P passive attenuator --> XLR --> Hypex NC252MP amplifier --> Test leads --> BNC unbalanced input of Linear Audio Autoranger MK II --> Balanced TRS/XLR --> RME ADI-2 Pro ADC --> USB --> Surface Pro 3 laptop

I'll also have some oscilloscope measurements where the test leads will be directed towards the oscilloscope instead of Autoranger and RME ADI-2 Pro.

Unless stated otherwise, all tests here were done with both channels driven. To reduce mains hum and lower the noise level, other than the amplifier itself, the rest of the measurement chain listed above was powered with lithium battery packs.

All cables were of the "generic" variety including some Cable Matters XLR cables, a few feet of Amazon Basics 14G speaker zip cords connecting amplifier to load, and 16G IEC AC power cord.

I. Basic Amplifier Characteristics

Amplifier Voltage Gain was measured as +25.84dB for my unit. This is consistent with the datasheet listing voltage gain typically +25.5dB (25-26dB range).

Amplifier Damping into 4Ω load is high:

Impressive! We're looking at an average of 340x, with high damping consistently across the audible spectrum. Above a damping factor of 100, measurement precision becomes very important so I had to perform a few readings and average things out to get that graph above. To have an even more accurate look at the output impedance, check out page 13 of the data sheet. Remember that we don't necessarily need very high values (something like >50x average would be great) but the stronger the damping factor, the more "load invariant" the amplifier becomes and we can show this when we look at the frequency response of an actual reactive speaker load:

As you can see, a speaker with complex impedance curve like the Sony smoothed out completely with this amplifier! Compare this to the equivalent graph from the Yeeco TI TPA3116 amp to see the difference low vs. high damping factor makes.

Note that there is a little bit of high frequency roll-off at 20kHz; I'm seeing about -0.9dB with the 4Ω load. The official NC252MP datasheet is showing around -0.7dB at 20kHz.

Phase Response along with the frequency response into 4Ω resistive load:

-15° at 20kHz; nothing to be concerned about.

Since this is a stereo amplifier, let's check to see if the 2 channels have the same output level (Channel Balance) and frequency response:

Very good channel balance. Essentially perfect overlay of levels where human ears are most sensitive (400Hz - 5kHz) and an insignificant 0.1dB difference by 20kHz. (Note: I made sure to use the same output from the signal generator to ensure the only variable was the amplifier channel being measured.)

II. Single-Tone Harmonic Distortion and Noise

To start, let's look at the Harmonic Distortion vs. Frequency at various output power levels into 4Ω - 2V/1W, 10V/25W, and 20V/100W. These graphs were created with Room EQ Wizard's "Stepped Sine" function:

Very nice and in fact I think rather impressive for <US$750. Notice that at 1W, the grey THD+N measurement is already better than -90dB across the audible spectrum. As you can see, the harmonics and noise remain low and well controlled across all frequencies. THD+N is down to -100dB by 100W output.

The cascade of harmonics at 1W has the 2nd order predominating. By 25W, we're seeing a rise of the 3rd order harmonic to around the same level as the 2nd order and by 100W, the 3rd order harmonic has taken over as the strongest component.

We can see this change in predominance of the 3rd order harmonic with select THD(+N) graphs of a 1kHz signal from 16mW to 250W output:

You can compare this panel with the Emotiva XPA-1L measurements previously to see how this Class D device stacks up with a more "classic" Class AB amplifier. By the time we hit >20V output, the 3rd order harmonic rises above the 2nd (as you can see in the 28.3V tracing). The Emotiva XPA-1L amp is more powerful which is why by 250W, we can see the Hypex here (rated as 250W into 4Ω) show a bit of strain with increased noise floor, odd-order harmonics starting to rise, and the THD+N rising to -75dB. Consider however that this amplifier is rated to 250W into 4Ω, we're still looking at below 0.02% THD+N - not bad at all!

I know there's quite a gap between 10W and 200W in the graphs above. For brevity sake, let's just say there were no issues and harmonic distortion was very low.

Just a couple more THD+N plots for completeness:

This is the 5W into 4Ω output graph showing an average SINAD of more than 95dB. Notice that my build here did not have as much 60Hz hum as Amir's measurements of the IOM NCore Pro PWR Amp. The right channel had a little more hum than the left, but had lower 2nd order harmonic.

And this is what clipping looks like with this amp when we push it to slightly >280W:

Notice the exacerbation of odd order harmonics as the signal starts to clip. THD+N rises to -47dB (0.45%) by this point.

Since we're measuring a 2-channel stereo device, my simple Crosstalk procedure using 0dBFS 300Hz and 4kHz tones playing simultaneously at 2V output in either right or left channels resulted in an average of -78dB seeping through to the other channel. This is getting close to the limit of the crosstalk through the passive preamp I'm using. The result is significantly better than the Onkyo TX-NR1009 receiver previously measured which scored -69dB. (Again, for some context, good turntable playback with a phono cartridge is something like -30dB, and likely lower than that if one were to use the same procedure I'm doing if such a test LP existed.)

This then brings us to the important THD(+N) vs. Output Voltage curve - left channel measured, both channels driven:

0.1%/-60dB THD+N is hit at just over 32V; 32.7V to be more precise or just above 267W into 4Ω. Impressive continuous output that performs as promised.

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

Alright, so THD+N is looking good with a single tone... Now what about the multitone tests and intermodulation distortions?

As with previous amplifier measurements, here's the panel of SMPTE, CCIF/ITU-R, and "Linkwitz" 2-tone intermodulation signals at 2V and 10V into 4Ω:

A busy set of graphs for those who want to peek into the fine details. The bottom line is that this is looking very good again. The total calculated IMD values are for the most part at least -10dB superior to my Emotiva XPA-1L monoblocks! The only one that did not surpass the Emotiva interestingly was the 2V CCIF/ITU-R where "d2L" and the ultrasonic "d2H" components were just a little higher. (Note that in my CCIF measurements, I've included the ultrasonic "d2H" component at 39kHz into the calculation for completeness even though this would not be audible.)

Next, we have the synthetic TIM ("Transient InterModulation") signal consisting of a 1kHz square wave with 12kHz sine at 192kHz samplerate (96kHz bandwidth), measured at 2V, 10V, 15V and 20V output:

As you can see, this amplifier was essentially perfect at the 2V and 10V output levels. However, by 15V and 20V we are seeing the development of distortion down below -80dB from the 1kHz peak along with probably some power-related low-level noise. Interesting finding and perhaps a useful data point for future comparisons. The idea is that this test signal might be an indicator of slew-rate limitation when the amplifier is reproducing rapid transients. Considering that we're still seeing just distortion below -80dB from the peak with a 20Vrms continuous output of this very challenging synthetic signal, I find it hard to believe one should have any worries with real musical transients.

Let's now have a look at the Triple-Tone Distortion and Noise measurement used in my AMOAR rating system, here's the TD+N at 2V into 4Ω for both channels:

An identical TD+N reading of -91.9dB (let's just round it down to -91dB so as not to overstate anything) for both channels at 2V is excellent - again, both channels are driven simultaneously. Remember that I use this value as a "summary" distortion score. This result is better than the Emotiva XPA-1L's -81dB and Onkyo TX-NR1009's -73dB; blowing away the inexpensive Yeeco amp with the TI TPA3116 Class D chip's result of -52dB.

For completeness, here's the left channel triple-tone TD+N at 10V where the distortion is still very low and with low noise:

IV. Square Wave and Wideband Noise

Up to this point, we're overall looking at very impressive results! However, remember that there is something about Class D "switching" amplifiers that we need to keep in mind. Because of the oscillating architecture, without a filter, one can detect potentially large amounts of "switching noise". This is why measurements of Class D amps are recommended to be done with a strong low-pass filter in place. The Audio Precision measurements Hypex publishes in their data sheet for the nCore NC252MP are done with the AES17 20kHz low-pass filter in place which is a "brick wall" design providing about -60dB attenuation by 24kHz! BTW, it's because of the use of this kind of steep filtering, instead of 19 & 20kHz CCIF/ITU-R intermodulation tones, they use 18.5 & 19.5kHz.

As you can see, I am using no such thing other than the RME ADC's filter which is certainly good enough to demonstrate excellent performance already. However, once we connect this amplifier to an oscilloscope where the bandwidth is much higher than the ADC, we start seeing the very high frequency switching noise show up:

Wow. Will ya look at those 1kHz "square" waves? ;-)

The digital scope used above is just an inexpensive one with 100kHz bandwidth at best. It's showing us that without any form of filtering, there are significant amounts of high frequency "stuff" in the output... Let's now put the amplifier signal to my Rigol DS1104Z with an even higher bandwidth of up to 100MHz, stop the square wave playback, and let's see what the switching noise looks like in the background:

"Zoomed" in, what we see is a 340-350mV, 410-430kHz oscillation coming out of the two channels (with no signal being amplified). 400+kHz is obviously ultrasonic so even if reproduced to a high level, we wouldn't be able to hear it unless it resulted in distortions in the audible frequencies; obviously no loudspeaker is designed to reproduce 400+kHz!

We can see from the FFT above that indeed, the noise is confined to ~420kHz while playing the dual 5kHz/93kHz tones. This is similar to Audio Science Review's measurement although I'm not seeing any significant 840kHz harmonic.

Although 350mV is a pretty high level, remember that because of the high frequency nature of it, it will naturally be filtered out by inductance in one's audio system. To demonstrate this, I looked in my parts box and found a couple tiny 10μH inductors. Placing this in series with my 4Ω power resistor, we create a simple passive first order RL low-pass filter with calculated -3dB point at 63.7kHz. Tapping the signal after the inductor in the left channel only, this is what happens to that 350mV noise:

The high frequency noise level drops by about -19dB to ~40mV (yellow Vrms reading). 10μH is a very small amount of inductance in concert with a low 4Ω resistance. Most tweeter voice coils have inductance and resistance values higher than this (and that's not even mentioning the effect of one's crossover or the contribution from the various feet of speaker cabling in an actual sound system).

With the small 10μH inductors in place on both channels:

Here's a repeat of the "non-aliasing" 24/384 1kHz, 2V square wave with high-frequency noise now significantly attenuated:

Voilà... Much cleaner square waves, more like what your speakers will be trying to reproduce (if you tried to do such a silly thing as play square waves!).

The bottom line is that while it would be preferable to not have the high frequency noise there at all, it's simply a fact of life with Class D amplifiers. However, due to this being way up at 400+kHz, even relatively low levels of inductance in your loudspeakers would effectively filter it out.

V. Impressions and Conclusions

Here's the summary AMOAR Score graphic I usually end off with for each detailed amplifier test done:

IMO, this is not just "good", it's a great amplifier. Other than the fact that my Emotiva XPA-1L monoblocks are more powerful, within its limits, in every one of the major parameters, this little <6-pound Class D amplifier surpasses the Emotiva's performance.

The <1dB 20kHz frequency response roll-off is not significant IMO. This amp can control speakers with an iron grip thanks to its very low output impedance and high damping factor providing a load-invariant frequency response; not just at lower frequencies but across the whole 20Hz-20kHz audible range. With a triple-tone distortion factor better than -90dB, this amplifier is "clean", producing very low audio frequency noise, harmonic and intermodulation distortions at the important ~1W level (and beyond of course). Furthermore, the Hypex is able to reach the rated 250W into a 4Ω resistive load with <0.02% (-75dB) THD+N with a few watts to spare before hitting my usual "hi-fi" threshold of 0.1%. Unless you have a large room paired with low-sensitivity speakers, I think this amount of clean power will easily suffice.

Remember, compared to Class A or AB devices, these Class D amps are extremely efficient up to 90+%. In my daily use, this amp runs cool and I rarely see it drawing >25W at normal playback levels compared to routinely >100W each when I run my Emotiva monoblocks! There should be some savings on the electric bill if you play lots of music over time :-).

If there is one "anomaly" to be aware of objectively, it would of course be the switching noise up at 400+kHz. It's just what Class D devices do and we see these relatively high levels of ultrasonic noise with higher-end Class D Hypex amps like the NC500 as well without low-pass filtering. In time we'll likely see the switching frequency be pushed even higher up - for example, the Sigma Acoustics Vivace "GaNTube" I mentioned at RMAF 2019 is advertised with a 768kHz switching frequency. While higher frequency noise would be filtered out even more easily, I don't believe there's any evidence to suggest that noise around 400kHz isn't already completely innocuous.

[Have a look at this Exogal ION PowerDAC whitepaper. It's said to operate at 100kHz and supposedly different from standard Class D PWM... Curious whether there is switching noise at 100kHz. If so, it seems that reviewers aren't noticing (here, and here for example) with words like "strong sound quality", "no spurious sounds", "perfectly silent operationally" and even "legendary". The Hypex NC252MP clearly performs objectively superior compared to these ION PowerDAC measurements which showed poor performance IMO - too bad no spectral analysis done out to 100+kHz to show switching noise level.]

In summary then, for less than US$500 (typically <US$800 retail), this simple self-assembled Hypex nCore NC252MP amplifier performed very well and lived up to my expectations both objectively and subjectively. This amplifier has more than enough resolution for hi-res digital playback; the question is whether the rest of your audio chain can maintain the fidelity. For sure, one can spend more money and achieve even lower noise level and lower distortions (yeah, I know, the Benchmark AHB2, Hypex NC400, NC500, and Purifi amps are all awesome - anyone got level-controlled blind test results to show audible differences?!). If 150W into 8Ω, 250W into 4Ω, and even 180W into 2Ω is all the stereo power you need (you might want to see here for a quick check), it's going to be very hard to justify the "need" for something of even higher fidelity.

What I'm excited by is the idea of ongoing refinement, even better power efficiency, smaller physical footprint, and of course even lower prices in the days ahead as technology marches forward and provides value to those seeking high fidelity audio. Progress does not look or sound like 100-pound solid state behemoths or anachronistic tube-jobs costing 4+ figures, IMO. Rather, this amplifier is a nice example of high-fidelity in the real world of the 21st Century - potentially unassuming devices that can provide beautifully clean, "clinical", and some might even call "sterile", empty, sonic canvases upon which the "colors", noise, "grunge", and harmonics of your music arises from with ease.

Let's throw down the gauntlet with this little Hypex amp. For less than US$800, are there any challengers that can surpass this level of objective performance with the same amount of output power?

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

On subjective opinions formed while obtaining objective results...

I just wanted to add a comment at the end here about the process of making measurements. I don't know if this is the experience of others out there, but what I have noticed is that when measuring high quality gear, the subjective "feel" of the measuring process itself is better. I notice less unanticipated challenges and one is less likely to be surprised by what one finds or need to double check the set-up because of odd results. I think this is especially the case when I'm doing this with hobbyist components and software rather than any kind of automated professional measurement gear ;-).

For example, when I was testing the inexpensive Yeeco amp a few months back, I found that I had to fiddle around with placement of the device because it was picking up hum from a nearby lamp. Little things like the volume knob didn't smoothly control output and it was difficult at times to hit output levels precisely. When the device got warmer, sometimes little momentary anomalies will show up on the scope but unreproducible and hard to comment on in the write-up.

These things that I experience in the measurement process beyond what I hear I think provides a deeper appreciation of the engineered quality of the devices being tested. It's an experience which I think purely subjective reviewers would unfortunately not be able to speak of. A subjective reviewer would also not experience the same type of surprise when expectations are high but quickly dashed when objective results clearly show subpar performance especially when at best the sonic effect is subtle and one has to specifically listen for them once tipped off by the measurements. I felt this disappointment a few years ago when testing the well built and expensive Vitus Audio SS-010 yet it performed worse than my much less expensive Emotiva amp (the superiority of the Emotiva was audible). Likewise, more recently on the DAC side, the Dragonfly Cobalt was surprisingly uninspiring with sonic anomalies subtle and requiring careful listening comparisons with the cleaner sounding Dragonfly Red.

I guess this is all to say that I would encourage audiophiles to consider trying their hand at measurements if possible. The process itself can be enlightening and will allow for exploration of your own perceptual limits. The integration of subjective experience and objective knowledge I believe allows one to be more insightful when there is a need to address inconsistencies between build quality, actual sonic fidelity, and the hype that accompanies products.

[I would say in general, integrating both subjective experiences and objective knowledge is crucial for gaining wisdom through all of life - and doing so for the audiophile hobby is but a logical application of this principle...]

Having said this, the little Hypex amplifier in this enclosure was actually a joy to measure. Results came smoothly, and I was quite positively impressed by the objective performance. I was already expecting the switching noise. There were no weird hums, no strange distortions, no heat issues, no safety shutdowns.

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

This first week of January historically has been somewhat interesting for audiophiles due to CES announcements. Based on what I see, CES 2020 looks very tepid for audio gear. I wonder if the audiophile press even showed up in Las Vegas for a report of what was announced (other than echoing a few company press releases). The most interesting audio product seems to be the latest Nuraloop Bluetooth headphone that measures one's otoacoustic emissions (OAE) for sound optimization (I assume they're using some kind of transient-evoked OAE response). Hmmm, interesting idea and worth seeing if this function catches on over time!

Also, lossy codec development continues with Bluetooth LE Audio where they've improved perceptual quality with lower bitrates and power output. Must be approaching modern MP3 and AAC levels of quality at the same bitrate. Plus the multi-stream sharing feature should be great when you're on a flight and want to share audio with family members (watching a movie on your tablet for example). Also there's the introduction of the LC3plus protocol for "beyond perceptual transparency" at 500kbps; a lossy "high resolution" audio codec at bitrates about 1.5x typical 320kbps MP3 these days. This will be good for wireless sound quality though for "perfectionist audiophiles" this is obviously not on the level of true lossless high resolution.

In the next week, there might be more interesting stuff from the NAMM 2020 tradeshow where we should see some good pro-level audio gear announcements.

As I put the finishing touches on this post, I'm listening to the excellent Alexandre Desplat Little Women (2019, DR10) soundtrack. Looking forward to catching the movie with my wife and daughter soon.

Gonna be busy with some work, a Windows 10 reinstall here to freshen things up, a ton of E-mails to respond to and some non-audio projects I want to try over the next couple weeks.

Hope you're all enjoying the music...

[PDF Version for download / printing.]