This is a follow up to my first test of the Cambridge Audio CXA81. Results there showed fair bit of noise which was uncharacteristic of other products I have tested from the company. There were a lot of speculation as to the source of the problem but no data to back any. Fortunately company CEO was kind enough to contact me about a week later expressing surprised at my results. I asked them if they have measurements that are better. After some investigation, they said that they could replicate my results at their China manufacturing facility but not their R&D lab. I expressed that I no longer had the unit so there was not any work I could do to get to the bottom of this. I was offered a new unit and offer for payment to test again. I accepted the second unit but turn down the offer of payment.
In the communication, I was asked (as some members had also suggested) whether I lifted the USB ground using the switch in the back. I had not and ran those tests on the second unit. This is with the unit grounded:
The low frequency noise was much lower in this sample but you can still see the channel in red showing a wide spectrum of high frequency noise. This reduced its sinad fair bit compared to the "good" channel. Lifting the ground made that better but created a new problem:
At first it appears that there is more harmonic distortion given the spikes at 2, 3,4, 5, etc. KHz. But that is not so. USB micro packets are 1 Kbytes each and frame boundaries are showing up there as leakage from digital to analog circuits. So lifting the ground is really not a good solution here although I appreciate it being there as it may help some systems.
Post this we had a conference call with the designers of the product. Issue was brought up that this is an instrumentation thing that the Audio Precision analyzer is creating a loop and ground noise that would not be there. While this is true, I explained that I have tested hundreds of DACs that don't have this problem. I have once in a while seen this but it is exceptionally rare. Since overall results were more positive than the first sample, I offered to re-run all the measurements and post them.
A problem which we did NOT have a disagreement was, was the output limit of the pre-out. Like many AVRs, this is lower than I like to see at 1.5 volt. Above that it will saturate and distortion will increase.
With this intro, let's get into the measurements. Company has seen all the measurements and agrees with them.
CXA81 DAC Measurements (Sample 2)
I wanted to break the electrical link between the analyzer and CXA81 so started my DAC testing using Toslink optical. Here is our dashboard:
Note how I have picked the max output before output saturating. If we sweep the input level we see the full range of output relative to distortion:
As you see at 2 volts, we get down to SINAD of 75 dB which is what I measured with the first sample at that output.
I next switched to USB input:
We see noise intrusion reducing SINAD by a few dBs relative to TOSLINK optical. Here is a key discovery though: this is NOT due to USB input but mere connection of USB cable. You can see that noise even with Toslink when I have USB plugged in (but not used as an input):
Turns out this noise issue will proceed to corrupt the measurements of all other subsystems and was the reason why the amplifier for example did not perform well either in the first round. The impact with this second unit is lower than first sample though. This could be due to noise conditions changing on my test system (nature of computers and grounding) or something is different in the second sample.
To eliminate USB noise as a factor I continued the rest of DAC tests using Toslink so let's go through them quickly:
Linearity test was impacted by another rare issue which is power supply noise very close to the test tone of 200 Hz I use here. There is a strong bandpass filter here but not enough to get rid of that:
We see clear jitter sources but they are benign audibly:
IMD test shows more noise than I like to see but is very good at the limit, beating my reference DAC in blue:
Here is our reconstruction filter:
And THD+N vs frequency:
Pre-amp Measurements
Here again we see the impact of connecting the USB cable or not while testing analog in/analog out:
Power Amplifier Measurements
Let's start with our usual dashboard:
This is above average for all amplifiers tested to date and far better than noise-induced measurements of the first sample. Distortion is at -100 dB which is very good.
Power sweeps are much improved due to lower noise:
We see that in various frequencies as well:
The lower amount of noise allows distinction between the test tones.
Conclusions
The second sample combined with not connecting USB cable shows CXA81 in much better light. Performance is competent now for the class and transparent for most uses. Where would one put the blame for noise being induced using USB? Me? Instrumentation? Vagueness of computer audio/grounding? Some or all of the above?
When testing AVRs, I have started testing with Toslink to remove issues related to HDMI. I should have opted to do the same here but did not since I consider the unit in a different class. Still, I take some blame and should have done this before returning the unit. As I expressed to the company though, this is not a problem that other devices have so I implore them to investigate how to use this variability. After all, your system may be much worse than my configuration as your downstream devices and upstream computer may present even a worse situation. One likes to take comfort in thinking "one digital input is the same as another."
As it is, you have the tough situation that you can't leave USB cable connected lest you degrade the performance of the rest of the system. If you don't have an audible issue, this is not a concern but if it is, experiment with removing USB cable and see if fixes the problem.
For me personally, USB connection is critical and about the only input I would use on such a combo device. So the above is not something that works for me. But may be fine for you.
I also like to see output levels of 2 volts for unbalanced without saturating. We are seeing AVR/AVP companies going this way. No reason to do different in a combo device like this. Higher output level gives better freedom to use external amplifiers.
Finally, let me express my sincere and deep appreciate for Cambridge Audio and company CEO for accepting the challenging results of the first test and working with me on better clarity and a constructive path forward. In that regard, they earn highest levels of respect from me. I look forward to quality products from them that come across my bench.
In the communication, I was asked (as some members had also suggested) whether I lifted the USB ground using the switch in the back. I had not and ran those tests on the second unit. This is with the unit grounded:
The low frequency noise was much lower in this sample but you can still see the channel in red showing a wide spectrum of high frequency noise. This reduced its sinad fair bit compared to the "good" channel. Lifting the ground made that better but created a new problem:
At first it appears that there is more harmonic distortion given the spikes at 2, 3,4, 5, etc. KHz. But that is not so. USB micro packets are 1 Kbytes each and frame boundaries are showing up there as leakage from digital to analog circuits. So lifting the ground is really not a good solution here although I appreciate it being there as it may help some systems.
Post this we had a conference call with the designers of the product. Issue was brought up that this is an instrumentation thing that the Audio Precision analyzer is creating a loop and ground noise that would not be there. While this is true, I explained that I have tested hundreds of DACs that don't have this problem. I have once in a while seen this but it is exceptionally rare. Since overall results were more positive than the first sample, I offered to re-run all the measurements and post them.
A problem which we did NOT have a disagreement was, was the output limit of the pre-out. Like many AVRs, this is lower than I like to see at 1.5 volt. Above that it will saturate and distortion will increase.
With this intro, let's get into the measurements. Company has seen all the measurements and agrees with them.
CXA81 DAC Measurements (Sample 2)
I wanted to break the electrical link between the analyzer and CXA81 so started my DAC testing using Toslink optical. Here is our dashboard:
Note how I have picked the max output before output saturating. If we sweep the input level we see the full range of output relative to distortion:
As you see at 2 volts, we get down to SINAD of 75 dB which is what I measured with the first sample at that output.
I next switched to USB input:
We see noise intrusion reducing SINAD by a few dBs relative to TOSLINK optical. Here is a key discovery though: this is NOT due to USB input but mere connection of USB cable. You can see that noise even with Toslink when I have USB plugged in (but not used as an input):
Turns out this noise issue will proceed to corrupt the measurements of all other subsystems and was the reason why the amplifier for example did not perform well either in the first round. The impact with this second unit is lower than first sample though. This could be due to noise conditions changing on my test system (nature of computers and grounding) or something is different in the second sample.
To eliminate USB noise as a factor I continued the rest of DAC tests using Toslink so let's go through them quickly:
Linearity test was impacted by another rare issue which is power supply noise very close to the test tone of 200 Hz I use here. There is a strong bandpass filter here but not enough to get rid of that:
We see clear jitter sources but they are benign audibly:
IMD test shows more noise than I like to see but is very good at the limit, beating my reference DAC in blue:
Here is our reconstruction filter:
And THD+N vs frequency:
Pre-amp Measurements
Here again we see the impact of connecting the USB cable or not while testing analog in/analog out:
Power Amplifier Measurements
Let's start with our usual dashboard:
This is above average for all amplifiers tested to date and far better than noise-induced measurements of the first sample. Distortion is at -100 dB which is very good.
Power sweeps are much improved due to lower noise:
We see that in various frequencies as well:
The lower amount of noise allows distinction between the test tones.
Conclusions
The second sample combined with not connecting USB cable shows CXA81 in much better light. Performance is competent now for the class and transparent for most uses. Where would one put the blame for noise being induced using USB? Me? Instrumentation? Vagueness of computer audio/grounding? Some or all of the above?
When testing AVRs, I have started testing with Toslink to remove issues related to HDMI. I should have opted to do the same here but did not since I consider the unit in a different class. Still, I take some blame and should have done this before returning the unit. As I expressed to the company though, this is not a problem that other devices have so I implore them to investigate how to use this variability. After all, your system may be much worse than my configuration as your downstream devices and upstream computer may present even a worse situation. One likes to take comfort in thinking "one digital input is the same as another."
As it is, you have the tough situation that you can't leave USB cable connected lest you degrade the performance of the rest of the system. If you don't have an audible issue, this is not a concern but if it is, experiment with removing USB cable and see if fixes the problem.
For me personally, USB connection is critical and about the only input I would use on such a combo device. So the above is not something that works for me. But may be fine for you.
I also like to see output levels of 2 volts for unbalanced without saturating. We are seeing AVR/AVP companies going this way. No reason to do different in a combo device like this. Higher output level gives better freedom to use external amplifiers.
Finally, let me express my sincere and deep appreciate for Cambridge Audio and company CEO for accepting the challenging results of the first test and working with me on better clarity and a constructive path forward. In that regard, they earn highest levels of respect from me. I look forward to quality products from them that come across my bench.
