Mark Levinson No.36 D/A converter Measurements

Sidebar 3: Measurements

Robert Harley generated the test-bench measurements and provided me with the data after I finished my listening tests. The Mark Levinson No.36 has an output of 1.85V (left) and 1.84V (right), unbalanced (3.69V, L&R, balanced) when decoding a 1kHz, 0dBFS (full-scale) sinewave. Its output impedance measured 7 ohms unbalanced, 13.4 ohms balanced—this processor should be able to drive any commercial preamplifier and cable with no adverse effects. The No.36 was non-inverting in its unbalanced mode, with pin 2 positive in the balanced mode. Its DC offset was 0.2mV in the left channel, 0mv in the right, unbalanced (balanced: 0.2mV left, 0.1mV right).

The remaining measurements were taken from the balanced outputs, unless otherwise noted:

The No.36's frequency response was very flat (fig.1). The top curves indicate the normal frequency response, the bottom the de-emphasis error. The latter would only matter on only a small percentage of discs; most don't use pre-emphasis. Nevertheless, the No.36 will play such pre-emphasized discs with vanishingly small response deviation. The Levinson's crosstalk (not shown) is outstandingly low—so low, in fact, that at better than –120dB across the band, it is of little more than academic interest.


Fig.1 Mark Levinson No.36, frequency response (top) and de-emphasis error (bottom) (right channel dashed, 0.5dB/vertical div.).

Fig.2 shows a decoded –90dB, 1kHz dithered sinewave signal. The only artifacts worth noting here are at 120Hz and 240Hz—clearly power-supply noise. But they are below –120dB! Using the same type of spectral analysis,


Fig.2 Mark Levinson No.36, spectrum of dithered 1kHz tone at –90.31dBFS, with noise and spuriae (20-bit data, 1/3-octave analysis, right channel dashed).

Fig.3 shows the result of the Mark Levinson decoding a track of "digital silence" (all data words zero) out to 200kHz, with first-rate results. Again, the very-low-level power-supply noise is visible—and little else.


Fig.3 Mark Levinson No.36, spectrum of digital silence (20-bit data, 1/3-octave analysis, right channel dashed).

Fig.4 shows the fade-to-noise with dither, indicating superb linearity down to the measurement limit at –120dB. In fact, while we've measured D/A converters with slightly less linearity error at –80 and –100dB (the No.36 is off by an inconsequential 0.5dB or so at those points), I have not seen another D/A converter with so little error at –120dB.


Fig.4 Mark Levinson No.36, departure from linearity (right channels dashed, 2dB/vertical div.).

The plots in figs.5 and 6 show the results of the Mark Levinson decoding a 1kHz, undithered sinewave at –90dB, with 16-bit and 20-bit resolution, respectively. Both results are superb in both waveshape and absence of noise, with the classic step-appearance of the 16-bit result clearly evident, as it should be.


Fig.5 Mark Levinson No.36, waveform of undithered 1kHz sinewave at –90.31dBFS (16-bit data).


Fig.6 Mark Levinson No.36, waveform of undithered 1kHz sinewave at –90.31dBFS (20-bit data).

Noise modulation as a function of signal level vs frequency is plotted in fig.7. The results shown here are for six different, progressively lower signal levels decreasing from –50dB to –100dB. In theory, the more tightly clustered the results, the better. The clustering here is relatively poor as high-end D/A converters go; yet the overall levels indicated are generally 10dB or more lower than those in the converters we've tested that have performed best on this measurement. Should the No.36 be criticized because it demonstrates dramatically lower noise, at the expense of the best clustering on this test? I think not. In any event, I tend to be conservative in attempting to relate the result of this test to a converter's audible performance. Below a certain threshold—a threshold I suspect is well above the result here—that relationship may well disappear.


Fig.7 Mark Levinson No.36, noise modulation, –60 to –100dBFS (10dB/vertical div.).

Feeding a full-scale combined 19kHz+20kHz signal into the No.36 and performing an FFT analysis of the output results in the plot in fig.8: the artifacts are very low, comparable to the best we've measured.


Fig.8 Mark Levinson No.36, HF intermodulation spectrum, DC–22kHz, 19+20kHz at 0dBFS (linear frequency scale, 20dB/vertical div.).

The FFT of the Meitner LIM Detector output, an assessment of the No.36's jitter performance, is shown in fig.9. I've shown only the result with a 1kHz, 0dBFS sinewave input; the results with a –90dB, 1kHz sinewave, and with an all-zero input, were essentially identical, except that the spikes at 16kHz and 18kHz disappear into the noise. This is a very clean result. The major jitter component is just under 6kHz. I suspect that the audible result of this—and of the other isolated artifacts indicated—is nil. At least, I'd be unwilling to draw any conclusions unless I could hear an otherwise identical D/A converter exhibiting no artifacts at all—an unlikely circumstance. The RMS jitter of the No.36 measured 50 picoseconds and was independent of signal level. While this is higher than the specified 20ps, it is nevertheless one of the best results we've measured.


Fig.9 Mark Levinson No.36, word-clock jitter spectrum, DC–20kHz, when processing 1kHz sinewave at 0dBFS; PS Audio Lambda transport (linear frequency scale, 10dB/vertical div., 0dB=1ns).

The measured performance of the No.36 is superb, consistent in every way with the listening tests.—Thomas J. Norton

Harman Luxury Audio Group
8500 Balboa Boulevard
Northridge, CA 91329
(888) 691-4171

jimtavegia's picture

It is treated like a criminal being interrogated for 24 hours straight until finally breaks; "he talks" and then has to admit that no matter what I do or how good I get I will ever be the equal of my perfect brother, "the vinyl LP" who never gets put on the test bench and had the bright light shined in his face. Mom always liked him best. lol I heard that somewhere.

Glotz's picture

"Yer no good, Digital!"

Axiom05's picture

I had one of these together with a No. 37 transport. This was before jitter became the focus of measurements. Eventually replaced this combo with a Levinison No. 390 CD Processor.

Robin Landseadel's picture

One has to wonder how this DAC compares to Topping's D90.

Bogolu Haranath's picture

What the world needs is a good $5 double cheese burger :-) ......

Bogolu Haranath's picture

.....and/or, a good $5 Margarita :-) ......

rschryer's picture make yourself for much cheaper.

Save your money for audio. :-)

Bogolu Haranath's picture

I don't have a pool table at home ..... Also, I have too many bar buddies to fit into my house :-) .......

Bogolu Haranath's picture

Okto dac8 Stereo, which costs under $2k may be a good value for the money now ..... Stereophile review is forth coming :-) ......

Ortofan's picture

... the RME ADI-2 DAC FS for only $1,149.

Bogolu Haranath's picture

Okto dac8 Stereo under $2k price includes Wi-Fi connection capabilities ...... Also, it is not clear whether RME provides all the 7 selectable digital filters Okto provides ..... ASR doesn't say :-) ........

Bogolu Haranath's picture

RME however, offers several built-in DSP capabilities including, 5-band parametric EQ, bass and treble controls, loudness compensation adjustments and, for headphones cross-feed, M/S capabilities etc ..... Details available in RME user manual :-) ........

Bogolu Haranath's picture

Seems like RME uses some type of minimum phase filter ..... Doesn't look like it uses the standard linear phase filter :-) ......

ASR doesn't provide impulse response measurements for DACs ..... at least not so far :-) ......

Ortofan's picture

... waste $4,000 on this external DAC, but instead bought the $3,000 Sony CDP-XA7ES CD player. It had the sort of sound quality with which TJN "could live happily for a very long time."

Bogolu Haranath's picture

Sony CDP-XA7ES did not have HDCD decoding capabilities :-) ......

jimtavegia's picture

I am glad that I don't have to worry about having to have the best of everything or worry about the few discs that ARE HDCD encoded or MQA or the rest of it. There are so many excellent dacs out their and now the Weiss 502 would seem to be a target for everyone to hit and there are many that cost much more than the 502. I will bet those owner are not in deep despair and are truly enjoying what they own.

I am keeping in mind that ever mastering engineer is probably using a different ADDA in their chain and they are deciding what you are going to hear. They are basing all of that on what they hear on THEIR speakers that I don't own which will correlate to little for me.

Then we praise turntables that can't even spin at the right speed, have to deal with lps that aren't flat or with a center hole not quite in the center, and phono stages that cost more than the Weiss 502 and are still not the best, and we haven't even spent over $10K on a cartridge or know if it is mounted right or not. So we start with a master tape on a mechanical machine that is close to proper speed, but not perfect, the cut on a lathe that spins not at a perfect speed...close but no cigar, and a cutting engineer doing his absolute best to deal with the right groove spacing based on the musical dynamic range, then off to the platers and the pressing plant with all our fingers crossed and the best engineering minds in full play. I respect everyone of them as they are pure music lovers and fully care about what they do.

And even in the pressing of CDs there can be issues as we read about the Steely Dan issue with 2 VS. Nature that was caught in time by the late, great Roger Nichols.

I know of very few industries, other than medical, that goes trough the pains of the record and cd industry and still the hair splitting continues over the gear. I love the reviews and I like reading about the companies and engineers who are working so hard to make superb products to help us enjoy the MUSIC we love, as that is the point.

I will remain content with my newly purchased Project S2 dac AT $299 that has upped my listening in a most affordable way and know that there are million who would think that $299 is crazy money to spend on just a DAC. At nearly 73 I doubt that I could hear all that the Weiss 502 does, but I know to trust JA1, his measurements and his ears.

If I could have afforded it back in the day, I would have bought the Sony as well. It would still be one heck of a transport anyway today.