NAD M23 Master Series Stereo Amplifier Test Report

The NAD M23 amplifier utilizes the PuriFi Class D topology which is a high efficient (95% at full load), state of the art design. NAD informed me that they manufacture the Class D module and don’t simply just buy it directly from PuriFi. This allows them some leeway with regards to matching with their custom designed input stage. NAD also designed their own SMPS power supply. This is quite a lot of effort on NAD’s part to build a reference amplifier from the ground up based on a SOTA Class D topology. Much could go wrong if care isn’t taken to ensure a clean layout and implementation. Knowing this, I was even more excited to test the M23 amplifier to see if NAD was able to eek out every bit of performance this design offers. Will the attractive package that the M23 dons befit its performance?

NAD M23 Design Overview

The Purifi Eigentakt circuit shares with Bruno Putzeys’ previous designs (UcD and Ncore) the use of full global feedback. This means all of the loop gain is used for the entire circuit, including the output filter.

The big step forward for Eigentakt is that Bruno developed a much more thorough understanding of the large signal behavior of self-oscillating modulators which has allowed him to push loop gain up to nearly 80 dB across the audio band, a 25 dB improvement over his previous work. This has allowed Bruno to relax the switch timing and still get a large net improvement in performance.

The switching frequency of the Purifi 1ET400A design is a fairly standard 550kHz and the output filter physically cuts off at 35kHz, which is actually lower than many competing products. But a 10uH output coil with a loop gain of 10,000 around it means that the output impedance is more like 1nH. That’s less than the inductance of the binding posts!

The high switiching frequency and excellent Low Pass Filter design allows for this unit to be tested without a precondition filter on my Audio Precision that lesser designs would often need for accurate test results. In fact, the PuriFI Class D design is theoretically capable of lower distortion than my Audio Analyzer is capable of measuring. The data sheet shows a THD+N distortion spec of .00017% or 115dB at 100 watt drive and full bandwidth! As a result, there are some measurements that may not realize the true performance of this product and it will be noted within this test report. All measurements were conducted using our Audio Precision APx585 8 Channel HDMI Audio Analyzer.

NAD M23 Testing - I had to remove the alligator clips and measure at the amp side for more precision.

Editorial Note About M23 Testing:

Because my APx585 Audio Analyzer actually has less performance than the Purifi 1ET400A amp module, I do my best to ensure my test gear is NOT the limiting factor in the accuracy of sensitive distortion tests and note it within this test report where applicable.  In fact, during the initial testing procedure, it was determined that the long leads and magnetic alligator clips on my probes were causing higher distortion than NAD specified with this amplifier. When I jerry-rigged new test leads, I somehow blew up the first M23 test sample and sent it back to NAD to investigate the cause of failure. The second M23 test unit performed flawlessly and the measurements in this report have come from that unit but I also attained similar power and noise performance on my first unit prior.  

For more information about how we measure power amplifiers, please see:

Basic Amplifier Measurement Techniques

NAD M23 Amplifier Top View

The NAD M23 is a thing of beauty both inside and out. You can see great care was taken in the layout of this product to ensure minimal noise coupling. They didn’t use bus bars to route power and instead used heavy gauge twisted pair wire. This is much preferred for better signal integrity and reduction of crosstalk coupling. The amp modules are collocated in their own area with short connection paths to the output binding posts to minimize losses. The SMPS power supply is located on it’s own board with the capacitors isolated from heat sensitive areas.

NAD M23 Gain Structure

The NAD M23 has 3 different gain settings which are specified as follows:

• Stereo Mode Gain:  19dB (low), 23.9dB (mid), 29.2 (high)
• Bridge Mode Gain : 25.1dB (low), 30dB (mid), 35.2 (high)

My measurements verified the 19dB for low gain, 23dB for mid gain (vs 23.9dB specified), 29dB for high setting. Oddly, the gain was the same for unbalanced or XLR. Usually in home theater systems (mandatory in THX), unbalanced would have 6dB less gain on the amp side since the balanced outputs of a preamp are usually 6dB hot. This wasn’t the case here but it’s understandable since the M23 provides 3 different gain settings to suite your needs. I measured 30dB of gain in bridged mode for the mid setting which is what NAD specifies.

Frequency Response

NAD M23 Frequency Response under Various Load Conditions

I measured the NAD M23 frequency response in 3 load conditions (open circuit, 8 ohms and 4 ohms). Lesser designed Class D amplifiers will alter the amplitude response significantly at high frequencies but we don’t see that here. Thanks to the global feedback strategy mentioned earlier and the very low output impedance that this produces, we see NO appreciable interaction here. This means, the NAD M23 is a load invariant amp as good as some of the best linear amp designs in this regard. Under all 3 load conditions, I measured the -3dB point of 65kHz which is what NAD specs for this amplifier. Simply excellent measurement!

FFT Distortion Analysis

NAD M23 1 kHz FFT @ 2.83Vrms (1 watt, 8 ohms)

The NAD M23 produced the absolute best FFT distortion measurements I’ve seen in an amplifier. You can see when measured at 1 watt (2.83V, 8 ohms), there were no measurable harmonics at least within the limits of my Audio Precision test gear. When I ran both channels, I found a slight 3^rd order harmonic but quickly realized it was my other resistor load causing it. Measurements of this sensitivity can vary greatly just due to minor test set up differences. Note there is no ground hum in this measurement either. Simply amazing!

 

NAD M23 1 kHz FFT @ (200 watts, 8 ohms)

At rated power (200 watts/ch, both channels driven, 8 ohms), we see an incredibly clean FFT with the 3^rd harmonic being -113dB below the fundamental. The amp is rated it’s cleanest according to the data sheet at 150wpc, but at 200 watts, this is better than virtually any other amplifier I’ve measured to date. Bravo!

 

NAD M23 Frequency Response vs Distortion @ 1 watt and Full Rated Power (200 watts, 8 ohms)

With the AES17 (20kHz) filter employed, I measured the NAD M23 frequency response vs distortion at 1 watt and full rated power (200 watts) into an 8 ohm load. What is remarkable about this measurement is how closely distortion tracks regardless of power level. We see even at full rated power and at 10kHz, the distortion peaks at about 0.005% or 106dB SINAD. Simply incredible!

I also tested with a wider bandwidth filter to measure out to 20kHz but the harmonics beyond that show excessively high distortion since it’s out of the optimized bandwidth of the amplifier feedback loop. Since this is irrelevant to the perceptual audio performance, I decided NOT to publish this test data. I did spot check the amplifier using a two-tone test at 18.5kHz and 19.5kHz and saw similarly low distortion products. I noticed PuriFi published distortion specs on this very test which I show below for completion. The results are as expected, negligible distortion products below the two-tone test frequencies (18.5kHz and 19.5kHz).

SNR & Crosstalk

 

NAD M23 SNR @ 1 Watt (a-wt)

I always measure amplifiers at 1 watt so that apples to apples comparisons can be made between different products that have different maximum output capabilities.  If you want to know the SNR at rated power, then you simply take the 1 watt rating and add 20* log (V*R)^1/2 / 2.83) where V = Vrms and R is the load. In this case, that would be: 102.7dB + 20*log(40/2.83) = 99dB + 23dB = 125.7 dB (a-wt).  I also measured the amp at rated power and it produced 126dB (a-wt) as expected.

NAD specifies the M23 at 101.7dB so I actually measured 1dB better than their spec!

 

NAD M23 CH-CH Crosstalk (1CH, Undriven) @ 1 watt & Full Power

The channel-channel crosstalk was excellent for a stereo amplifier whether driven at 1 watt or full power. I measured > 84dB of channel to channel separation at 20kHz at full rated power and below 100dB at 1 watt!

Power Measurements

Using our Audio Precision APx585 8-channel HDMI analyzer, we conducted a full barrage of multi-channel amplifier tests on NAD M23 per our Amplifier Measurement Protocol. We tested power using three methods all of which were taken at < 0.1% THD + N:

• Continuous Full Power Bandwidth (CFP-BW) from 20Hz to 20kHz into 8 and 4-ohm loads (up to two-channels)
• 1kHz Power Sweep vs. Distortion (1kHz PSweep) - popularized by the print magazines, this is an instantaneous power vs. distortion test at 1kHz. The problem with this test is it often masks slew-related and/or frequency response problems some amplifiers exhibit at the frequency extremes, and thus inflates the measured power results. It does provide an instant gratification number for consumers to argue over on the forums, so we are now incorporating this test to please the masses.
• Dynamic PWR - 1kHz CEA-2006 Burst Method testing. This is a dynamic power measurement adopted from the car industry similar to IHF method only a bit more difficult for an amplifier and more representative of real musical content.

Keep in mind most review publications don't do continuous power measurements and they usually publish power measurements into clipping at 1% THD + N. Our measurements are very conservative as we use a dedicated 20A line with no Variac to regulate line voltage.  We constantly monitor the line to ensure it never drops more than 2Vrms from nominal, which in our case was 120Vrms. 

For more info on amplifier measurements, see:  The All Channels Driven (ACD) Test

 

NAD M23 Frequency Response @ Full Rated Power (2CH driven, 8 ohms)

When driven at 200 watts/ch (8 ohms, 2CH driven), the NAD M23 produced ruler flat frequency response with -3dB pt at 65kHz, same as it was at 1 watt driven. Bandwidth uniformity is excellent and distortion was under 0.01% THD+N for the entire sweep.

 

NAD M23 1kHz PSweep (2CH) - 8 ohms

The NAD M23 produced 241 watts/ch at 0.1% THD+N and 259 watts/ch at 1% with both channels driven into 8 ohms. The distortion was exceedingly low hovering around .0003% THD+N or 110dB for SINAD nerds below 150 watts. NAD rates the M23 to .00069% THD+N (103dB SINAD) at rated power which is about what I’m seeing here as well. Again, this is the lowest distortion I’ve measured from an amplifier at rated power. Excellent!

 

NAD M23 1kHz PSweep (2CH) - 4 ohms

When driving 4 ohm loads, the NAD M23 was able to produce 444 watts/ch at 0.1% THD+N and 529 watts/ch at 1% THD+N with both channels driven.  This exceeds NAD’s power rating of 380 watts/ch into 4 ohm loads with both channels driven. Distortion was again exceedingly low at 0.0003% THD+N at 250wpc and an impressively low 0.0005% THD+N at 380wpc.

 

NAD M23 1kHz PSweep (Bridged) - 8 ohms

The NAD M23 is a powerhouse when bridged driving an 8 ohm load. It was able to muster 898 watts at 0.1% THD+N and 1038 watts at 1% THD+N. NAD rates the M23 at 700 watts when bridged into 8 ohms, so this rating is exceeded by quite a margin in this test scenario. It's rare when I gotta break out my Big Bertha power resistor to test these power levels, but it was an awesome site to see this amount of clean power produced by an amplifier that wasn’t even warm to the touch delivering it. Distortion performance isn’t as stellar when bridging this amplifier as you can see the SINAD drop from 100dB below 100 watts to about 85dB at max power. This is still a good figure, however. I checked the frequency response bandwidth in bridged mode and it remained similar to regular 2CH operation.

NAD M23 Frequency Response - Bridged Mode, 8 ohm load

 

NAD M23 1kHz PSweep (Bridged) - 4 ohms

CAUTION: NAD Clearly states the M23 should be used with 8 ohm nominal speakers when bridging. I still couldn’t resist testing it into a 4 ohm load to see what it could produce. Be warned that when bridging an amp in a Bridge Tied Load configuration like this, each amplifier sees ½ the load impedance. Thus, each of the two amplifiers in the M23 are really driving a 2 ohm load which as you can see in this test scenario is a bit too much. I had to decrease the sweep time to avoid tripping the protection circuits which still kicked in at the very end of the power sweep. Still, I was able to measure 1,182 watts at 0.1% THD+N and 1,399 watts at 1% THD+N into a 4 ohm load. This is one of the most powerful amplifiers I’ve tested to date! However, note the SINAD drop from 94dB below 100 watts to 70dB at max power. In all honestly, I do NOT recommend bridging this amplifier IF your speakers impedance profile dips below 5 ohms. Using this amp in stereo configuration will provide better distortion performance and increase longevity of the product for difficult speaker loads. IF for some reason you need more power, go with dedicated high power monoblocs instead.

Editorial Note About Amplifier Bridging

Examining the data sheet for 1ET400A Purifi Amp module, it seems they are current limited to about 24 amps which is a limiting factor when bridging two of these amplifiers together like what is done in the M23. Since each module essentially sees half the load impedance, it’s being asked to deliver 26 amps into a 4 ohm load to produce 1400 watts that I measured which exceeds the safety limits of the module.

It appears that the increased distortion in bridged mode is due to the fact that each module is scrupulously controlling the voltage that appears between its own output terminals. Afterall, this is where the speaker is normally connected. In bridged mode however, the speaker gets connected between two modules’ positive terminals. Any potential difference between the two modules’ negative (ground) terminals thus gets added to the voltage seen by the speaker. Older generations of class D amplifiers had high enough distortion for this problem to go unnoticed, but the native performance of the 1ET400A is so good that any distortion caused by wiring sticks out like a sore thumb.

It turns out that the modules actually have a provision to mitigate this. The feedback connection can be detached from the module output and reconnected elsewhere. In the case of bridging, distortion can be reduced by tying the cold terminals of the feedback lines to some common ground point. That way, the sum of the voltages seen by both feedback loop corresponds to the voltage across the two positive speaker terminals. Unfortunately, it doesn’t appear that NAD has used this trick in the M23. Simply trying to minimise the resistance between the grounds of the module isn’t good enough if you want to maintain SOTA performance in bridged mode.

A closer look at the Purifi 1ET400A Spec sheet confirmed my concerns about Bridging:

Similarly, bridging two 1ET400A Designs will result in performance degradation as the circuit is not configured to sense the voltage differential that exists between the (now unused) OUT- terminals of the two 1ET400A Designs. Bridging is therefore NOT recommended. All operation and performance specs are void in this configuration.

NAD Response about Bridging the M23 Amplifier

With regards to the trick of lowering the distortion by re-wiring of the feedback signal for bridge mode; we’re well aware of this possibility but decided against this. We decided to optimise the M23 for normal stereo use. With distortion levels as low as they are, just adding switching for re-routing the feedback signal will affect distortion in normal operation mode.

We channel-match the Left and Right modules to be as close as possible; this helps reduce distortion in bridge mode also. Indeed, if bridge operation would be permanent and thus not switchable, it’s logical to apply re-routing the feedback loop.

As performance in bridge mode still is as good as some of the best amps out there, we decided this was an acceptable compromise.

 

NAD M23 CEA 2006 Dynamic Power – 4 ohms

The CEA-2006 burst tests simulate musical program material to illustrate dynamic capabilities of the amplifier.  The NAD M23 was able to produce 248 watts/ch at 1% THD+N into 8 ohms and 460 watts/ch at 1% THD+N into 4 ohm loads with both channels driven. It’s a bit odd to see lower power figures in the dynamic power test than the 1kHz power sweep tests but I can only guess that has something to do with the SMPS power supply limitations or because it's regulated unlike a linear power supply. Still, these are good power figures.

# of CH	Test Type	Power	Load	THD + N
2	CFP-BW	200 watts	8-ohms	0.1%
2	CFP-BW	385 watts	8 ohms	0.1%
2	1kHz Psweep	259 watts	8-ohms	1%
2	1kHz Psweep	243 watts	8-ohms	0.1%
2	1kHz Psweep	529 watts	4-ohms	1%
2	1kHz Psweep	444 watts	4-ohms	0.1%
Bridged	1kHz Psweep	1038 watts	8-ohms	1%
Bridged	1kHz Psweep	898 watts	8-ohms	0.1%
*Bridged	1kHz Psweep	1,399 watts	4-ohms	1%
*Bridged	1kHz Psweep	1,182 watts	4-ohms	0.1%
2	Dynamic PWR	248 watts	8-ohms	1%
2	Dynamic PWR	460 watts	4-ohms	1%

NAD M23 Power Measurement Table

*NAD does not recommend bridging the M23 when driving 4 ohm speakers. The amp will current limit and potentially shut down. I had to reduce sweep times to even test the amp in bridged mode when driving a 4 ohm load so the power figured I measured herein are not sustainable.

 Conclusion

I have to pick some nits about NAD’s design decisions with the M23 before I can gush over its performance which I must reiterate is absolutely stellar in every way possible.

I don’t recommend using this amplifier in bridged mode as previously stated in the report UNLESS you are assured your speakers don’t dip below 5 or 6 ohms. Using two of these as monoblocs isn’t very cost effective and you lose some of the low noise/low distortion benefits of the design when you switch to bridged mode. So, what’s the point?

Ergonomically it’s a challenge to connect any type of fancy speaker cable to the bridged +- terminals that are a whopping 8+ inches apart. I don’t think the engineers at NAD considered this but if you’re rolling some nice Kimber Kable 8TC in your system, good luck reaching the speaker terminals in bridged mode. It would also be nice if NAD would have marked the left input terminal as the “bridged” input on the unit silk screen. There was  no way of determining which input to use in bridged mode without consulting the user manual.

NAD M23 Back Panel View

Editorial Note About Bridging Connector Spacing:

Ergonomically it’s a challenge to connect any type of fancy speaker cable to the bridged +- terminals that are a whopping 8+ inches apart. I don’t think the engineers at NAD considered this but if you’re rolling some nice Kimber Kable in your system, good luck reaching the speaker terminals in bridged mode. It would also be nice if NAD would have marked the left input terminal as the “bridged” input on the unit silk screen. There was  no way of determining which input to use in bridged mode without consulting the user manual.

NAD claims that the wide spacing of the speaker terminals is a consequence of the lay-out chosen by going for a dual-mono type configuration of the input stage and keeping the modules away from the input section, yet close to the back panel. As it so happens, they made a last-minute change to the back panel by replacing the input/output terminal section, made out of steel, with an anodised aluminium strip. This change, with aluminum being non-ferrous unlike steel, reduced interference between high power speaker signals with inputs to improve distortion figures yet again, albeit beyond the 0.000x% mark.

I like that NAD gives you 3 different gain settings on this amplifier. If you’re connecting this unit to an AV receiver, I recommend using the high gain setting. Use the medium or low gain setting IF you connect to a dedicated preamp or IF you decide to bridge this amplifier. The lowest noise performance was achieved for the low gain setting (+6dB for low vs high setting in bridged mode) but the noise was so low in any of these settings that  you’ll be fine to choose any as long as your preamp has enough voltage drive to reach full rated power (ie. 4Vrms for low gain, 2.5Vrms for mid gain).

 

NAD M23 Front Panel View

The capacitive touch power button at the top of the chassis is cool but the novelty wears off quickly each time you accidentally touch the case in its vicinity and the M23 powers on or off when you don’t want it too. I would much prefer a push button power on the front panel NAD logo LED.

 

NAD M23 Magnetic Foot Coasters

The magnetic foot coasters have me scratching my head. They seemingly look neat when placed under the pointy feet but a bit tricky to get lined up properly and area easy to misalign if you move the amp around in your rack. The discs serve not other purpose but to protect the surface on which the pointy-coned feet (for acoustic isolation) would otherwise leave scratches or dents on the furniture’s surface. This feature was introduced for all full-width Masters Series in this design back in 2014. The NAD isolation feet is a somewhat hip and trendy feature some audiophiles may appreciate.

Aside from my nit picking, the NAD M23 is a gorgeous looking amplifier. From its high quality brushed aluminum casing, to its radiused edges and slick looking vents on the top and sides of the chassis, it’s really a product to behold that makes you just want to sit still and gawk over. I highly recommend leaving this unit out for display instead of stuffed into a rack in a closet if you can.

I love the fact that you get a high power two-channel, audiophile grade power amplifier that would spank most Class A amplifiers in power and performance, but it’s so light weight at a little over 21 lbs that you can carry it with one arm.

Linear Amps Dead. Long Live Class D!

Linear Amps Are Dead, Long Live Class D!  - YouTube Discussion

NAD managed to get most of the ultra-low distortion performance out of the Purifi amplifier module making it a truly SOTA design that has pushed the envelope of performance and what we can expect from Class D amplification. NAD has proven to me that linear amps are becoming a relic of the past and it’s time to embrace the future of Class D amplification that offers both superior efficiency and performance! In almost every category of performance, the M23 produced some of the best measurements I’ve seen to date regardless of amplifier topology or price.

Stay tuned for my formal review with listening tests to follow. Are you considering purchasing this amp or do you already own one? Share your thoughts in the related forum thread below.

Unless otherwise indicated, this is a preview article for the featured product. A formal review may or may not follow in the future.