NAD M2 Direct Digital Amplifier—A New Take on Class-D
Robert Harley -- Tue, 10/20/2009 - 14:00
Check out my blog on the new NAD M2 Direct Digital Amplifier that uses an innovative new Class-D output stage as well as other interesting technologies. http://www.avguide.com/blog/technology-preview-nad-m2-direct-digital-amplifier-the-future-audio-amplification
All content, design, and layout are Copyright © 1999 - 2011 NextScreen. All Rights Reserved.
Reproduction in whole or part in any form or medium without specific written permission is prohibited.
Reproduction in whole or part in any form or medium without specific written permission is prohibited.
Fascinating new technology! How does it sound?
A good sense of humor makes it ALL sound better!
I posted this in the amplifier & preamplifier forum not noticing there was a forum for class D amps...
Dear RH, What you did was such a tease... not even a hint on the M2's sound quality?
I have a few comments:
- You mentioned that the M2 does not use a digital filter. I think it does or has to since the device uses upsampling and noise shaping just like a regular DAC before the last step of PCM to PWM conversion. I believe it has to use a low pass filter to remove aliasing. In one of the documents published by Zetex, it stresses the importance of the digital filter in a digital amplifier. Zetex has its own proprietary upsampling filter algorithm called the ZTA-filter algorithm. NAD does not mention the type of filter they use. This is one of the key things that I am wondering about. With all the recent developments in the digital filtering technology, is their digital filter up to the current state of the art?
- Regarding the power supply, one of the claims Zetex makes about its invention is that it eases the constraints on power supply design compared to an open loop switching amplifier. As you mentioned, the feedback scheme is meant to compensate for power supply noise as well. The fact the M2 has beefier power supply than other digital amplifiers you have seen is interesting.
- Regarding whether this is the future or not, Mark Levison's latest flagship amplifier is a switching design ( as I am sure you are aware). They claim it was the best in blind tests. I have read their published patent on a switiching amplifier (filed under Harman) which was very sketchy but it uses feedback as well without saying how it is used. The real difference in their implementation is that they interleave the operation of the switching transistors (described in a different patent) to reach a combined switching rate of 2Mhz. For comparison, the M2's switching frequency is about 844khz and that of Tact/Lygndorf is a multiple of the input signal which is less than 400khz.
Now when will the Dec 09 issue of TAS become available on line? Do we get to get to sneak a peek on the sound quality part beforehand?
It is always great to read your reviews.
See my reply under "Preamplifiers." —Thanks.
Thanks RH. I did. I am a subscriber in the softcopy edition available through NextNewsStand.com. Looking forward to read your review, which I think will be the first for this amplifier to appear anywhere.
Hello RH,
I found the review to be focused, to the point and with a very clear explanation of the strengths and weaknesses of the product. It also sent me to read your X100.5 review as well to be able to know more about your reference. The NAD M2 first got my attention when the bold statements from the company seemed too strong to write-off as typical marketing hyperbole. I had been wondering why the idea of a fully digital amplifier was not pushed by many designers further. I only knew of Tact/Lyngdorf. When I read Zetex's patents and documents, their idea seemed to me quite elegant and the error noise shaping seemed to be inspired by the sigma delta modulation used in DAC design. NAD probably spent a lot of the design time just getting the idea to work well. After reading your review, I look forward to see how this design will perform after a couple of years of fine tuning. I have a couple questions and a comment:
- Purely on the basis of emotional impact and enjoyment, how far was this system from your reference system? Does it qualify as a system you can live with or just not yet?
- Did the treble settings help improve the performance of the top frequencies or was performance constant with the settings?
- I was a bit surprised that the M2 could not resolve the smallest details to the level of the reference system given the great linearity and SNR highlighted by the company. However, I will go out on a limb and a suggest a reason for this as well as an experiment to check the validity of the explanation I give. I think the culprit is the digital volume control but not in the usual way it has been known to deteriorate a signal. By using a whopping 35 bit bandwidth to allow (among other uses and benefits mentioned by NAD) significant headroom for volume control without distortion, I think NAD has spread the available resolving power of the system a bit too thin. Pulses of the switching stage have a fixed number of possible widths (128 in case of the M2) and there is a finite number of all the possible combinations of pulse widths within the duration of any audio frequency cycle. For a system without noise shaping, this finite number will determine the maximum resolving power of the system for a given audio frequency. Conceptually, noise shaping of a given order provides a finite improvement upon that basic resolving power by picking the pulse sequence that will average out closest to the signal from all the possible pulse sequences (higher order noise shaping should give better picks). In the Tact/Lyngdorf approach to volume control (not to imply their product is better or worse), the DC offset representing the volume is applied by the power supply as some voltage level. All the resolving power of the system is dedicated to representing the music signal. The pulses will have a voltage related to the volume setting but their widths are only determined by the music signal. In the NAD, the volume level is embeded in the signal sent to the switching stage. The voltage of the pulses is fixed. The pulse widths have to portray both the music singal and volume level. At any volume setting, many of the possible combinations of pulse widths can not be chosen because they represent the same signal at a higher or lower volume setting. It is only when such a system has a dynamic range corressponding to 35 bits (a very difficult 210db) that digital volume control will not affect the system's resolution of the music signal. As can be expected, this limitation will have the most effect on the signals which need the most resolving power, i.e. the lowest level signals and the highest frequency signals or a combination of both as a worst case scenario.
The experiment to determine whether this theory is correct or not is to take a 16 bit music track and re-encode it as 24 bit but the value of each 16 bit number has to be mapped to an equivalent 24 bit number as if the track was recorded in 24 bits since just placing the 16 bits in a 24 bit package with the other bits being zero will effectively have the same limitation as digital volume control. If the explanation given above is valid, this simple change should result in a decent improvement in low-level resolution and overall sound quality vs the 16 bit playback. This is because there are only 11bits to represent volume when a 24 bit music signal is used (more signal than volume will be modulated so to say) vs.19 bits for volume when a 16 bit signal is used. I wonder if the digital output sound card or some playback software can be set up to do this on the fly.
Best regards
I couldn't help but smirk that a $6k integrated amplifier occupied the "Cutting Edge" section. Thank you Robert Harley- that was a great article. The most exciting thing I've read in TAS in a long time, in fact. I enjoyed all of the technical information you provided in the review and in the sidebars. It obviously stoked my interest in the M2, but I also felt as though I actually learned something.
Thanks so much for the comment, EasyBigFella. The Cutting Edge section isn't just for high-priced gear, but for products that break new ground technically and are worthy of more depth.
Those are interesting observations, HE74. The M2 came closer to the reference system for musical involvement than one would think considering the price difference, but it won't replace the reference system. The treble adjustments did allow me to dial in the treble to some degree, but not completely.
Robert,
You state that this came closer to the reference system than one would think with the mention of the "pricing caveat". Would you say that the M2 is or would be comparable to many 30k systems? It is generally hard to read these reviews, no matter how good some are, and get a real sense of things. People (reviewers and general populace) always state things relative to price. What we/I try to decern is what the comparison is without the price factor. Some things in audio are just over-priced and it is a shame that sound quality is most often delineated on a price scale instead of just "sound".
So, I do wonder how far above it's weight class you feel this amp would punch? A "slam dunk" to equally priced seperates is one thing but how about 15k, 20k, 25k? As you may be able to tell, I am really excited about this piece and especially the technology. Hopefully many other manufacturers will embrace the Zetex module and we will be in a brave new world.
I suppose your question boils down to this: If I could have the M2 or $15k worth of separates, which would I choose? I think that $15k of carefully chosen separates would out-perform the M2, but it is also possible to put together $15k of separates that don't sound as good.