For the lab evaluation, 24bit/96kHz signals through a coaxial S/PDIF input were used (except where stated otherwise).
The digital interface seems to offer a quite good jitter attenuation with the relevant graph being in the range between -60 and -50dB without any significant deviation.
Jitter susceptibility measurement draws a similar picture with the curves for medium and high-frequency signals below -100dB and the curve for low frequencies at a somewhat higher level (-75dB).

In terms of the jitter generated by the device itself, Femto Clock appears a great solution with calculated values below 10pS (a figure that includes the total jitter and not just the clock jitter), with most findings to reside at very low and low frequency.
Frequency response is, as one would expect, practically flat up to the limit of the measurement with a deviation less than 0.1dB in the band around 40kHz. Channel balance is exemplary.

Out-of-band response behavior was as expected in terms of the selected filter. With the “Fast” setting the transition region is quite narrow and the stop-band very clear.
The “Slow” setting, of course, (not unsurprisingly) changes that picture, the corresponding graph showing a stop-band with significant content, the price to be paid for a filter that potentially offers a better transient behavior.

Level linearity proved to be one of Manhattan's strongest sides. The relative curve maintains the optimum shape up to the -110dBFS point where the first visible deviation occurs but is still acceptable up to -120dBFS. A similarly good behavior was witnessed in the case of 192DSD, sample, therefore the use of the ESS chip and the parallel/balanced topology adopted were good choices, judging by results.
THD and THD+N versus signal level measurements showed a circuit with good behavior both at low and at high signal level. THD+N graph remained below -40dB even for -60dBFS signals, while THD (the harmonic distortion figure without taking the noise into account) is almost 20dBFS lower. For signal levels above -15dBFS, distortion shows a quite mild (but visible) upward trend but, even in the output limit of 0dBFS, it remains below -70dB showing no evidence for an analog stage overload.

Variation of the THD+N as a function of the signal frequency is in practice zero. The curve for -20dBFS signal level remains constant near 0.006% while with signal level near the limit of the output stage (-1dBFS) the graph confirms the slight upward trend witnessed in THD/THD+N Vs Level measurement. The relevant curve is slightly above the 0.02 %, but without any trace of significant variation depending on the signal frequency.
Manhattan appeared to be a particularly noise-free device, which justifies Mytek's choices both in the power supply and the chassis design and construction. The relevant graph remained below -120dBFS in the low/medium frequency range and despite the (normally expected) upward trend, never exceeds -110dBFS.

Broadband analysis reveals that the largest proportion of the noise concentrates towards the lower part of the spectrum and probably comes from the power supply. In the higher part of the spectrum, a barely visible component near 90kHz exists, but at a level beneath -120dBFS probably have no practical significance.
Noise analysis from the power supply includes visible 50/100Hz components (which hardly exceed -120dBV) and some intermodulation products at even lower levels. With 5uVrms of total power-related noise, one can say with certainty that Manhattan's power supply stands to the occasion.

Modulation components spectrum offers minimal information. If there are any traces of modulation, these are located very close to the fundamental (24kHz) and can barely be identified.
Narrowband signal noise modulation analysis highlights small peaks at 100Hz and 12Hz, the first probably due to the power supply and the second due to jitter, however the levels are very low in both cases.

Finally, channel crosstalk remained at very low levels with the relevant curves being overall bellow -90dB, reaching -120dB for mid and medium-high frequency signals. It is interesting to note that the shape of these curves is very similar to that of 192DSD, something that can hardly be a coincidence, but probably is indicative of a manufacturing feature, perhaps related to the physical structure and topology of the analog stage.

Previous | Next | More Reviews