This is a review and detailed measurements of NAD M51 DAC and digital preamplifier. It is on kind loan from a local member. Even though it is still listed on NAD website, other retailers indicate it is discontinued. Looking around, it seems the cost is USD $2000.
The unit has somewhat bland industrial design for my taste:
What distinguishes the NAD 51 from many other DACs are two things 1) custom upsampling DAC and 2) HDMI inputs.
On the DAC portion, this is how NAD describes it:
In simpler terms, the sample rate of content is increased to 844 kHz and then what resembles a little class-d amplifier converts it to analog. The higher sample rate reduces the needed bit depth/accuracy in the PWM portion of the DAC. Whatever DSP they use, must have a 35-bit word length and hence the reason that is the accuracy of the volume control.
The second feature, HDMI is something that is very rare in DACs. In HDMI audio is slaved to video so you need a video subsystem to extract the audio. In my past testing of older Audio/Video Receivers, I found that their HDMI performance was much worse than their S/PDIF (sans a very high-end Mark Levinson processor). The NAD M51 gives us another chance to make this comparison and see if there continues to be a reason to avoid HDMI. If not, this will make it convenient to extract audio out of your blu-ray music discs and have high fidelity playback of them.
Let's get into measurements and see if this wizardry translates into better performance and functionality.
USB Measurements
Starting with our typical dashboard in balanced output, we are greeted with competent results:
This was not meant to be at first. I always test at 0 dB and performance drops like a rock with that setting. Looking around, it seems NAD found the errors of their ways and in an upgraded version of the firmware, reduced max volume by 1 dB. Of course audiophiles being audiophiles, there is talk online about how much better the old firmware was with respect to bass. JA in stereophile confirms the same:
Of course higher volume increase perception of bass so that is understandable. But preferring tons more distortion is not it. NAD was correct in fixing this. You can always turn up your amplifier by 1 dB and get the same effect, without all the distortions.
Anyway, back to our dashboard, we have very good SINAD (signal over noise and distortion) of 108 dB. THD+N beats NAD's spec of 0.0005% in both channels which is comforting. This puts M51 solidly in our tier 2 performance:
RCA output is nearly as good:
And now matches the spec.
Dynamic range lands in the same space with balanced being exceptional:
Good news continues in linearity test where the M51 nails the test:
Usually my audio analyzer struggles to get reliable readings between -110 and -120 dB due to existence of noise. Not here. The measured values were dead on with almost no variation from expected results. This shows very good attention to design. Few DACs do this well.
Jitter is not as good but no audible concern:
Will the good news continue into intermodulation distortion? I am afraid not:
In the past we have seen many instances of this in ESS DACs which I have coined to have "ESS hump." It is a hump because it starts at mid-levels but then goes away as levels increase. Here, it is persistent to almost the end. In that regard this is worse than ESS DACs. I will investigate this further but note the jump in green when I run the test at 0 dB volume. We have a classic, amplifier like clipping occuring at max volume which should be avoided.
Since this IMD distortion is quite pronounced and there have been many questions regarding its audibility, I thought I dig in with more tests.
First, let's look at the spectrum of our IMD distortion at two levels, one where IMD is not dominant and one where it is:
IMD test is dual tone: a 60 Hz and 7 kHz tone. These are the two tall spikes. An ideal DAC would reproduce those two tones and nothing else. When the DAC is not linear, it will then produce sideband distortion products that show up symmetrically around our main tones.
The blue spectrum is the distortion with digital samples level at -45 dB before our IMD distortion start to rise up. We see our noise floor in blue and maybe a few spikes hugging our main tones closely. When I increase the digital value up to -25 dB, the sidebands (in red) multiply and substantially rise in level. We also get a mirrored cluster of them around 14 kHz. the IMD graph above simply sums these and hence the reason distortion rises substantially. There is simply a lot more junk there.
Taking our 7 kHz tone, its level is about -32 dB. The tallest distortions are -108 dB. The difference is 76 dB. Put another way, our distortion is 76 dB below our music level. That makes it very difficult to hear.
Let's look at the same problem, this time with a 1 kHz tone:
This time the graph in red is at full digital amplitude. We see a spray of red spikes indicating harmonic distortion. We also see a rise in noise floor which is typical of noise-shaping which would be mandatory as the bit depth is decreased. What is not expected is continued distortion spikes all the way up to 90 kHz. These should be well below noise floor as it increases.
The related problem to IMD is seen when I *decrease* the signal level to -25 dBFS. Ordinarily decreasing the source level should result in less distortion. Not here. We actually have higher harmonic distortions than at full scale (in blue)! The differential is similar to what we computed in IMD test. The main difference being that perceptual masking would not hide them as harmonics continue at full level until nearly 8 kHz or so.
Again, this lands in difficult to hear category but something is seriously broken. WIth music that is recorded well below max 0 dBFS, you are liable to increase the volume to compensate and then, you will have a better chance of hearing those artifacts.
Back to our suite of measurements, the noise shaping and harmonic distortions in ultrasonic range hurt THD+N versus frequency with 90 kHz bandwidth:
The measurements actually became so noisy that the analyzer gave up a few times (as indicated by "T" symbol and sudden jump). Even without that, we have much elevated THD+N compared to "ordinary" budget DACs that cost less than the shipping for the NAD M51.
Multitone test was not very revealing other than not being able to get a stable reading every time:
HDMI Measurements
I had a hell of a time finding a way to access HDMI audio. My normal monitor is 4K and the NAD being an older unit of course doesn't support that. I reduced resolution to 1080p, 1080i and even 720p and no cigar. I would get no signal from my monitor when I routed the output through the M51. By itself connected to the computer was no problem so it was not a capability issue of my monitor. I dug out another cheap PC monitor I had sitting around and that worked at 720p.
Next problem was how to play audio to the HDMI port. My Nvidia card shows up as a sound device but ASIO4ALL does not see it. WIth no native ASIO interface, I was out of business using the Audio Precision analyzer software to drive the port. Fortunately I could access it using WASAPI exclusive mode using Roon media player. That limits me to a couple of specific tests so let's get into that.
FIrst is the dashboard:
Ah, most excellent! Performance is nearly identical to USB!
Jitter test is where I had seen issues with HDMI but not here:
The graph is a bit more busy than USB but nothing audible.
I did run into a problem here though. Using fixed output I was just getting microvolts output. I switched to variable and all was well. Not sure what is going on there.
Conclusions
I am always concerned about custom DAC implementations as reimplementing the wheel is not easy. The NAD M51 at first showed me they had gotten the basics right with very low noise and good distortion metrics with full scale signal. Alas, that fell apart when we look at distortion and noise at levels below max. There is a sharp increase in distortion, costing us some 30 dB or so in performance. What a shame.
The NAD M51 gives me hope that there are more good HDMI implementations out there. Crossing fingers as I test more units already here for review on that.
Given the mid-level distortion I can't quite recommend the NAD M51. If you need HDMI connectivity though, and can get this unit on the cheap in secondary market, it may work out well.
------------
As always, questions, comments, recommendations, etc. are welcome.
It is snowing here for the first time this year. I need some money to buy snowshoes. Yes, it is only one inch of snow but you don't want me to fall and get hurt, do you? So please consider donating money using:
Patreon: https://www.patreon.com/audiosciencereview), or
upgrading your membership here though Paypal (https://audiosciencereview.com/foru...eview-and-measurements.2164/page-3#post-59054).
The unit has somewhat bland industrial design for my taste:
What distinguishes the NAD 51 from many other DACs are two things 1) custom upsampling DAC and 2) HDMI inputs.
On the DAC portion, this is how NAD describes it:
In simpler terms, the sample rate of content is increased to 844 kHz and then what resembles a little class-d amplifier converts it to analog. The higher sample rate reduces the needed bit depth/accuracy in the PWM portion of the DAC. Whatever DSP they use, must have a 35-bit word length and hence the reason that is the accuracy of the volume control.
The second feature, HDMI is something that is very rare in DACs. In HDMI audio is slaved to video so you need a video subsystem to extract the audio. In my past testing of older Audio/Video Receivers, I found that their HDMI performance was much worse than their S/PDIF (sans a very high-end Mark Levinson processor). The NAD M51 gives us another chance to make this comparison and see if there continues to be a reason to avoid HDMI. If not, this will make it convenient to extract audio out of your blu-ray music discs and have high fidelity playback of them.
Let's get into measurements and see if this wizardry translates into better performance and functionality.
USB Measurements
Starting with our typical dashboard in balanced output, we are greeted with competent results:
This was not meant to be at first. I always test at 0 dB and performance drops like a rock with that setting. Looking around, it seems NAD found the errors of their ways and in an upgraded version of the firmware, reduced max volume by 1 dB. Of course audiophiles being audiophiles, there is talk online about how much better the old firmware was with respect to bass. JA in stereophile confirms the same:
Of course higher volume increase perception of bass so that is understandable. But preferring tons more distortion is not it. NAD was correct in fixing this. You can always turn up your amplifier by 1 dB and get the same effect, without all the distortions.
Anyway, back to our dashboard, we have very good SINAD (signal over noise and distortion) of 108 dB. THD+N beats NAD's spec of 0.0005% in both channels which is comforting. This puts M51 solidly in our tier 2 performance:
RCA output is nearly as good:
And now matches the spec.
Dynamic range lands in the same space with balanced being exceptional:
Good news continues in linearity test where the M51 nails the test:
Usually my audio analyzer struggles to get reliable readings between -110 and -120 dB due to existence of noise. Not here. The measured values were dead on with almost no variation from expected results. This shows very good attention to design. Few DACs do this well.
Jitter is not as good but no audible concern:
Will the good news continue into intermodulation distortion? I am afraid not:
In the past we have seen many instances of this in ESS DACs which I have coined to have "ESS hump." It is a hump because it starts at mid-levels but then goes away as levels increase. Here, it is persistent to almost the end. In that regard this is worse than ESS DACs. I will investigate this further but note the jump in green when I run the test at 0 dB volume. We have a classic, amplifier like clipping occuring at max volume which should be avoided.
Since this IMD distortion is quite pronounced and there have been many questions regarding its audibility, I thought I dig in with more tests.
First, let's look at the spectrum of our IMD distortion at two levels, one where IMD is not dominant and one where it is:
IMD test is dual tone: a 60 Hz and 7 kHz tone. These are the two tall spikes. An ideal DAC would reproduce those two tones and nothing else. When the DAC is not linear, it will then produce sideband distortion products that show up symmetrically around our main tones.
The blue spectrum is the distortion with digital samples level at -45 dB before our IMD distortion start to rise up. We see our noise floor in blue and maybe a few spikes hugging our main tones closely. When I increase the digital value up to -25 dB, the sidebands (in red) multiply and substantially rise in level. We also get a mirrored cluster of them around 14 kHz. the IMD graph above simply sums these and hence the reason distortion rises substantially. There is simply a lot more junk there.
Taking our 7 kHz tone, its level is about -32 dB. The tallest distortions are -108 dB. The difference is 76 dB. Put another way, our distortion is 76 dB below our music level. That makes it very difficult to hear.
Let's look at the same problem, this time with a 1 kHz tone:
This time the graph in red is at full digital amplitude. We see a spray of red spikes indicating harmonic distortion. We also see a rise in noise floor which is typical of noise-shaping which would be mandatory as the bit depth is decreased. What is not expected is continued distortion spikes all the way up to 90 kHz. These should be well below noise floor as it increases.
The related problem to IMD is seen when I *decrease* the signal level to -25 dBFS. Ordinarily decreasing the source level should result in less distortion. Not here. We actually have higher harmonic distortions than at full scale (in blue)! The differential is similar to what we computed in IMD test. The main difference being that perceptual masking would not hide them as harmonics continue at full level until nearly 8 kHz or so.
Again, this lands in difficult to hear category but something is seriously broken. WIth music that is recorded well below max 0 dBFS, you are liable to increase the volume to compensate and then, you will have a better chance of hearing those artifacts.
Back to our suite of measurements, the noise shaping and harmonic distortions in ultrasonic range hurt THD+N versus frequency with 90 kHz bandwidth:
The measurements actually became so noisy that the analyzer gave up a few times (as indicated by "T" symbol and sudden jump). Even without that, we have much elevated THD+N compared to "ordinary" budget DACs that cost less than the shipping for the NAD M51.
Multitone test was not very revealing other than not being able to get a stable reading every time:
HDMI Measurements
I had a hell of a time finding a way to access HDMI audio. My normal monitor is 4K and the NAD being an older unit of course doesn't support that. I reduced resolution to 1080p, 1080i and even 720p and no cigar. I would get no signal from my monitor when I routed the output through the M51. By itself connected to the computer was no problem so it was not a capability issue of my monitor. I dug out another cheap PC monitor I had sitting around and that worked at 720p.
Next problem was how to play audio to the HDMI port. My Nvidia card shows up as a sound device but ASIO4ALL does not see it. WIth no native ASIO interface, I was out of business using the Audio Precision analyzer software to drive the port. Fortunately I could access it using WASAPI exclusive mode using Roon media player. That limits me to a couple of specific tests so let's get into that.
FIrst is the dashboard:
Ah, most excellent! Performance is nearly identical to USB!
Jitter test is where I had seen issues with HDMI but not here:
The graph is a bit more busy than USB but nothing audible.
I did run into a problem here though. Using fixed output I was just getting microvolts output. I switched to variable and all was well. Not sure what is going on there.
Conclusions
I am always concerned about custom DAC implementations as reimplementing the wheel is not easy. The NAD M51 at first showed me they had gotten the basics right with very low noise and good distortion metrics with full scale signal. Alas, that fell apart when we look at distortion and noise at levels below max. There is a sharp increase in distortion, costing us some 30 dB or so in performance. What a shame.
The NAD M51 gives me hope that there are more good HDMI implementations out there. Crossing fingers as I test more units already here for review on that.
Given the mid-level distortion I can't quite recommend the NAD M51. If you need HDMI connectivity though, and can get this unit on the cheap in secondary market, it may work out well.
------------
As always, questions, comments, recommendations, etc. are welcome.
It is snowing here for the first time this year. I need some money to buy snowshoes. Yes, it is only one inch of snow but you don't want me to fall and get hurt, do you? So please consider donating money using:
Patreon: https://www.patreon.com/audiosciencereview), or
upgrading your membership here though Paypal (https://audiosciencereview.com/foru...eview-and-measurements.2164/page-3#post-59054).
