It has been a little over 7 years since ESS Technology introduced the revolutionary ES9018 audio D/A converter chip. This converter delivered a major improvement in audio conversion and, for 7 years, it has held its position as the highest performing audio D/A converter chip. But a new D/A chip has now claimed this top position. Curiously the successor did not come from a competing company; it came from ESS. On October 19, 2016, ESS Technology announced the all-new ES9028PRO 32-bit audio D/A converter. In our opinion, ESS is now two steps ahead of the competition!
This application note examines the differences between the ES9018 and the new ES9028PRO. It also compares the Benchmark DAC2 and DAC3 to demonstrate the performance improvements that can be achieved in a commercial product.
For several months, Benchmark Media Systems has been working with some early samples of the new ES9028PRO converter chip. We have conducted numerous measurements and listening tests, and I have to say that we are very impressed! As a result of these tests, we have decided to add two premium converters to our existing line of converters. The new DAC3 HGC is a high-performance version of the popular DAC2 HGC. The DAC3 L is a high-performance version of the DAC2 L (for users who don't need a headphone amplifier).
The Benchmark DAC3 is one of the first products to use the new ES9028PRO converter chip. The DAC3 is available in two models; the DAC3 HGC and the DAC3 L. Both are now in full production. These products began shipping November 14, 2016.
The block diagrams below show the internal workings of the ES9018 and ES9028PRO converter chips.
Clearly the overall architecture is similar, but there are four significant performance improvements that have been added to the ES9028PRO:
Of these, we consider the THD compensation the most clever and unique feature. The improved filters deliver a slightly flatter frequency response. The improved phase locked loop allows virtually instantaneous switching between digital inputs. The improved power supply distribution contributes to the effectiveness of the THD compensation by minimizing THD-inducing interactions between the various subsystems.
This 32-bit digital processing block compensates for the THD produced by the analog outputs on the chip. It can also be adjusted to compensate for the off-chip analog components in the DAC. This system even allows independent nulling of the 2nd harmonic and 3rd harmonic distortion. The result is a conversion system that is virtually perfect in terms of 2nd and 3rd harmonic distortion!
Our tests show a reduction in 2nd harmonic distortion over the entire audio band, with the largest improvements in the midrange where our ears have their highest sensitivity. In the midrange, the compensation circuit reduces the 2nd harmonic distortion by 8 to 12 dB.
The performance of the 2nd harmonic compensation system is demonstrated in the plot below. Both traces show the 2nd harmonic distortion at the XLR outputs of the DAC3. The black trace shows the DAC3 in a special test mode where we have disabled the THD compensation. The red trace shows the normal operation of the DAC3 with the THD compensation turned on. When the THD compensation was turned off, the THD performance of the DAC3 was virtually identical to that of the DAC2. The red trace demonstrates the effectiveness of the THD compensation process. This unique processing function is one of the major performance improvements in the DAC3.
The ES9028PRO has an improved selection of filters. We have chosen the filter which provides the lowest passband ripple. This new filter exceeds the pass-band performance of the best filter available in the older ES9018.
It is important to note that the DAC2 and DAC3 converters frequency-shift the selected built-in filter by driving the converter chip at a fixed 211 kHz sample rate. This frequency shifting places the near-Nyquist region entirely above the audio band. The Benchmark upsampling system completely eliminates the time-domain errors that would have been produced by the built-in filter. The 211 kHz upsampling frequency sets the cutoff frequency of the internal filter at 105 kHz. This places the filter transition region safely above the 96 kHz Nyquist frequency of a 192 kHz input sample rate. This unique Benchmark system prevents the near-Nyquist errors that would normally be produced by the filters in the D/A converter chip. This unique 211 kHz upsampling system allows us to implement a filter that is optimized for minimum passband ripple instead of using the default filter in the ES9028PRO.
The DAC3 and DAC2 both use the Benchmark 211 kHz upsampling system, and both provide outstanding frequency-domain and time-domain performance. But, due to the improvements in the ES9028PRO, the DAC3 has slightly less passband ripple than the DAC2.
The ES9028PRO has an improved digital phase locked loop that allows it to lock more quickly than the ES9018. The new chip can lock with high precision in less than 6 milliseconds. For all practical purposes, this is instantaneous. We have leveraged this performance and have modified the timing of our UltraLock2™ system to create the new UltraLock3™ system. Both versions of Benchmark's UltraLock™ system have near-perfect jitter attenuation and they exceed the performance of the phase locked loop that is located inside the ESS converter chips. In the DAC2, we had to delay our mute circuits by 450 ms to wait for the ES9018 to lock. But, with the ES9028PRO we have been able to reduce the mute interval to less than 6 ms.
Remarkably, the phase of the DAC3 is completely settled at the end of the very short 6 ms mute interval. In contrast, the DAC1 converters unmute in about 50 ms, but the PLL phase takes another 2 seconds to settle. The DAC2 converters mute for 450 ms, but like the DAC3, they are fully settled at the end of the mute.
The following three graphs show the improvements in lock time over the three versions of Benchmark's UltraLock™ system. Please note the time scales when comparing the plots. In each case the time scale is adjusted to show the complete mute interval. The red lines show the approximate envelopes of the mute functions. Please note that the DAC1 did not use an ESS converter chip but the DAC1 mute timing is shown to add some perspective to this discussion.
The ultra-fast phase-accurate settling of the new UltraLock3™ system allows seamless switching between digital inputs. The DAC3 input selector is suitable for conducting A/B comparisons between two digital inputs. The longer 450 ms mute time of the DAC2 and the long 2 second PLL settling time of the DAC1 prevent the use of the input selectors when conducting A/B tests with these products. In contrast, the DAC3 provides seamless 6 ms switching between inputs and it exhibits none of the frequency shifting that occurs when the DAC1 is settling. With the DAC3, it is now possible to make accurate comparisons between two digital sources.This is a capibility that will be appreciated by many audio professionals as well as by many home audiophiles.
The performance and complexity of the support circuitry outside of the ES9028PRO can be configured to match a manufacturer's budget and space constraints. When Benchmark designed the DAC2 we started by using all of the ESS-recommended performance options and then we added our own set of enhancements. This same strategy is carried forward to the DAC3. Our performance-driven strategy greatly increases the number of support components, but these components add important performance improvements. For this reason, it is important to understand that there can be a wide variation in the performance of an ES9028PRO converter solution. The overall performance will be determined by the manufacturer's budget, goals, and skills.
In the DAC3, Benchmark uses the following techniques to maximise the performance of the ES9028PRO:
|SNR - A weighted||116 dB||126 dB||128 dB|
|THD+N, 1 kHz, -1 dBFS, 20 kHz LPF (dB)||-106 dB||-109 dB||-113 dB|
|THD+N, 1 kHz, -1 dBFS, 20 kHz LPF (%)||0.00056 %||0.00035 %||0.00022 %|
|Frequency Response, 20 to 20 kHz||+/- 0.1 dB||+ 0, -0.04 dB||+0, -0.015 dB|
|Passband Ripple||+/- 0.005 dB||+/- 0.003 dB||+/- 0.002 dB|
|Intersample Headroom||0 dB||3.5 dB||3.5 dB|
|PLL Lock Time||60 ms||450 ms||6 ms|
|Jitter Attenuation System||UltraLock™||UltraLock2™||UltraLock3™|