- Written by Chris Heinonen
- Published on 24 February 2014
The Yamaha CX-A5000 SSP On the Bench
First I take a look at the changes offered by the YPAO room correction. For each channel it seems to offer 7 bands of EQ with frequency, adjustment, and Q. Compared to other room correction systems this is a much smaller number of adjustments than they can offer.
Testing with my speakers, Definitive Technology Mythos STS towers, running full range I see the adjustments that it is making are minor in size. The lowest frequency being adjusted is 78.7 Hz which means most bass issues are not being addressed. It also touches on the response at 16 kHz that is both unnecessary and possibly impossible to hear. Many room correction systems start to roll-off their adjustments past 500 Hz, or at least 5 kHz, as they aren't really making a noticeable change. The table below shows the adjustments made to my left speaker by YPAO.
|Frequency||78.7 Hz||250 Hz||500 Hz||1 kHz||8 kHz||8 kHz||16 kHz|
|Gain||+1.0 dB||+1.0 dB||-1.0 dB||+1.0 dB||+1.5 dB||+1.0 dB||+3.0 dB|
As the STS are bringing run full-range with no subwoofer I would expect a larger adjustment below 80 Hz but YPAO seems to not correct the bass on the main channels. Looking at this data it is easy to see why I did not hear a large difference with YPAO engaged. The changes it makes are small in nature and not in the bass frequencies that would benefit most.
I also ran the CX-A5000 through the Audio Precision APx 585 and APx 582 test instruments. Use of these was generously provided by Audio Precision and allows us to really investigate how well the ESS SABRE performs and determine what other design decisions might be holding back its performance. This unit uses the same low cost AVR LSI chip used throughout the Yamaha multichannel line, the Renesas R2A15220FP. For more detail I highly recommend looking at the article that Dr. David Rich put together on SSP and receiver design.
To avoid graph overload, I will summarize some of the general measurements first. Tested over HDMI, THD+N on a 0 dBFS 1 kHz signal at unity gain measures at 0.0045%. The volume control was set for an output of 2 VRMS at 0 dBFS. The THD component for that measurement is 0.0042%. Both of these are filtered to only include 20 Hz to 20 kHz data. Signal to Noise Ratio, also filtered, checks in at 111 dB while Crosstalk measures -75 dB at 20 kHz and -100 dB at 1 kHz. The power supply noise spurs are below -110 dB as well.
Moving onto the HDMI graphs, with THD+N vs. Frequency at 2 VRMS we see that it remains below 0.005% up to 5 kHz. There we see a slight rise out to 20 kHz where THD+N approaches 0.008%.
THD+N vs Level with a 1 kHz signal shows that the level falls until 1 VRMS, at which point it begins to rise again. Measurements here start at 5 µVRMS and so that data is dominated by noise.
The 1 kHz FFT shows a 2 VRMS signal and a 2nd harmonic that comes in at -92 dBV relative to it. The even-order harmonics are more prevalent than the odd-order ones.
Our 19 kHz + 20 kHz IMD test shows a B-A peak that is over -110 dB below the fundamental frequencies. Sidebands are also below this level with out-of-band 2nd order harmonics close to -100dB below the fundamental tones.
Testing the linearity of the DAC, it is good down to -100 dBFS, at which point an error term becomes visible. By -112 dBFS it is 3 dB off the baseline value. The SABRE DAC is capable of better performance and it is likely the analog section which is introducing the noise here.
Looking at the time domain data for a -90 dBFS, 1 kHz, 16-bit undithered sine wave we see good, fast response and well-defined 1-bit levels. For comparison, I performed the same test on my Marantz AV7005, which uses the AKM AK4358 DAC, and the result is below the Yamaha CX-A5000 result here.
Looking at a -90 dBFS, 24-bit undithered sine wave, we again see very good resolution thanks to the SABRE DAC. For comparison I've again included the same test on my Marantz AV7005 to help illustrate the difference.
The spectra for a -60 dBFS tone at 1 kHz shows a noise level down below -140 dBV for the FFT bin width used for all graphs in this review.
The spectra for the same test done at -90 dBFS also shows a low noise level other than a power supply spur at just below -110 dBV.
Moving onto analog testing, Dr. Rich examined the service manual to determine how the CX-A5000 is designed. The left and right channels of the multichannel input are wired internally as the two-channel RCA inputs are. They go through the same switch and so have higher distortion associated with them. The other channels on the multichannel input do not pass through this switch and so distortion is lower on them compared to the left and right channels. This is fairly unique to see.
Using Pure Direct mode on an RCA Input, we see 0.0175% THD+N and 0.0174% THD at 2V RMS. We also see an 115dB Signal to Noise Ratio with the same 2V RMS signal.
Using the multi-channel input we see 0.0173% THD+N and 0.0173% THD. This corresponds to the non-multichannel input, just as we expected it to. The Signal-to-Noise ratio checks in at 113dB on the multichannel input as well.
Frequency response is flat from 20 Hz to 40 kHz, and only down by 0.1dB at 100kHz. The power supply noise is below -125dB with spurs visible at 60 Hz and 180 Hz only. Crosstalk measures at -88dB at 1kHz. It is only -120dB at 20Hz and reaches -60dB at 50kHz.
The analog power supply noise and crosstalk results are excellent. The crosstalk shows only capacitive coupling between the channels. Both results reflect careful design of the voltage regulators and PC board wiring one expects in a unit at this price point.
Unfortunately, the THD and SNR performance is similar to products at much lower price points.
THD+N vs Level with a 1 kHz signal over RCA shows that distortion starts to rise at 250 mVRMS, while over HDMI it didn't rise until 1V. There is a hard knee at 3.2 VRMS as well.
Looking at the same test with a 10 kHz test tone as see almost identical results. THD+N bottoms out at 0.0015% at 300 mV and then rises up until a knee at 3.2 V.
THD+N vs. Frequency over RCA with 2 VRMS in shows a THD+N below 0.02% that extends out to 10 kHz, with a slight rise up to 0.025% by 20 kHz.
A close look at the spectra from 0 Hz to 1 kHz shows those power supply noise peaks at 60 Hz and 180 Hz. Otherwise the noise floor is very low at -140 dBV.
Looking at our same 19 kHz + 20 kHz IMD test over RCA we see much higher sidebands than on the HDMI version, and secondary harmonics that are at -80 dBV instead of -100 dBV.