Cambridge Audio CXA81 Review (Sample 2)

[laudio]

No.

Well this should get interesting then

Cambridge CEO response was pretty clear and professional, plus sounded good to me from an engineering explanation. But there are so many other posts on this thread that bring up other situations to obfuscate what appears to be a reasonable explanation.

 

[audio_tony]

Well this should get interesting then

Cambridge CEO response was pretty clear and professional, plus sounded good to me from an engineering explanation. But there are so many other posts on this thread that bring up other situations to obfuscate what appears to be a reasonable explanation.

Earth loops between a PC and DAC (with mains ground - not all have a mains ground) are a very real issue.

So often the USB port or PC motherboard is blamed for this noise, when in fact it's simply a ground loop and neither the USB port or motherboard is noisy (in a modern PC, noisy USB ports and boards are very, very rare and usually indicative of a fault elsewhere).

I performed various measurements using a Raspberry Pi, PC and a laptop.

The full tests are on my site here but the relevant spectrum graph is reproduced below - and as can be seen it looks remarkably similar to one of Amir's graphs (a noise curve has formed at -120dB rising in amplitude but then begins to drop off at around 10kHz where the ground loop effect begins to diminish.

 

[pma]

Earth loops between a PC and DAC (with mains ground - not all have a mains ground) are a very real issue.

Exactly. The plot you are showing reflects loop issue. There are many ways how the loop may be created. Every engineer experienced enough in low level analog signal transmission systems can tell, from the plot shape, that the loop is the origin of the issue.
And the issue is serious. I have shown similar example in post #154

Cambridge Audio CXA81 Review (Sample 2)

@amirm, Thanks for this follow-up review and nice to see you unmistakingly nailed the core issue of this product: The device quite likely pollutes its own audio signal from suboptimal internal layout and cabling. Some of the balancing current that enters at USB (and other digital connectors)...

www.audiosciencereview.com

To the unexperienced reader who looks at THD and THD+N (SINAD) numbers only the difference (usb ground loop x isolator) may seem unimportant, but looking at noise we can see it went from 8uV to 23uV (very much higher) and the modulation is in the most audible part of the spectrum, plus possible spectrum spikes. So, this is an important issue, not to be overlooked. And many users have it at home, though they have no idea.

 

[laudio]

Earth loops between a PC and DAC (with mains ground - not all have a mains ground) are a very real issue.

So often the USB port or PC motherboard is blamed for this noise, when in fact it's simply a ground loop and neither the USB port or motherboard is noisy (in a modern PC, noisy USB ports and boards are very, very rare and usually indicative of a fault elsewhere).

Can you draw a picture of your setup? Are 2 USB ports on the same PC being utilized to make a measurement. Most users have just one USB port connected as a playback device.

 

[audio_tony]

Can you draw a picture of your setup? Are 2 USB ports on the same PC being utilized to make a measurement. Most users have just one USB port connected as a playback device.

There is a description of the setup I used on my website (reproduced below).

 

[Harmonic Study Admin]

I am no Electrical Engineer, but surely there must be a way to filter out noise from the USB without breaking the bank? Cheap DACs seem to have no such issues.

 

[laudio]

There is a description of the setup I used on my website (reproduced below).

Not sure how anything you are posting is relevant to the Cambridge discussion. Sure... tests using 2 PCs can cause a lot of noise when measuring a DAC.

Nobody here has refuted Cambridge audios explanation IMO. Waiting for somebody to actual refute what they said. The meaurement technique originally used to discredit was suspect.

 

[Rottmannash]

Not sure how anything you are posting is relevant to the Cambridge discussion. Sure... tests using 2 PCs can cause a lot of noise when measuring a DAC.

Nobody here has refuted Cambridge audios explanation IMO. Waiting for somebody to actual refute what they said. The meaurement technique originally used to discredit was suspect.

Pretty sure their explanation has been refuted-do you work for CA?

 

[KSTR]

The test setup caused the problem, not the DUT.

No.

Let's not make things more complicated than they are.

This is what's going on here and what I had already explained in post #155 :

• There is a voltage differential between two nominal "GND" points of the device, from the connected source devices (USB and balanced analog).
• There are further voltage differentials between these GND points and Mains PE (internal GND may or may not be directly bonded to PE. There might be the typical "isolator" in place, antiparallel diodes shunted with a capacitor -- which is a short for higher frequencies, again).
• The voltage differentials will cause cross currents. These cross currents are the Signal Integrity Aggressors.
• These currents develop voltage drops along their paths which find their way into the analog audio signal. The effects of these voltage drops are the Signal Integrity Victims.

The green current path is probably the main culprit here. I've seen enough devices with the dreaded "horseshoe" PCB ground plane design (in an attempt to create a "star GND") which failed in the cross current test for exactly that reason, conducting all the cross current right across the PCB, along sensitive GND reference nodes. This is also a EMI disaster, violating the #1 rule of EMC.

I'm pretty sure we would see the exact same noise pattern if the GND loop current injection points were different ones (as mentioned), say SPDIF coaxial and one of the unbalanced inputs. The connection and signal transport between top and bottom PCBs appears to be the weak point. Further, I am also confident the issue persists when there is no PE connection anywhere, removing the blue current paths completely.

To neutralize the aggressor currents, we can

• either make them flow as far away from the internal ground reference as possible. Normally (said #1 rule) one would connect each and every connector GND & shell directly to the metal chassis and double up that "GND bus bar" with as much copper as you can on the PCB along the board edge. Cambridge is using more than one PCB here and the device is mains earthed (but probably "GND-lifted" as noted) which makes things way harder
• or just live with any developing reference GND differentials and mitigate their effect by clever circuit design (internal balanced signaling at critical points, proper local GND peninsulas, etc. For example, the final filter and subtractor stage of the DAC filter could be placed on the lower PCB which would allow balanced signalling at zero extra cost or effort).

The best devices I've seen actually use both approaches in combination. I could run several hundred mA from one connector GND to another with only very little error voltage created at the output.

I think it is safe to say that this is NOT a measurement setup problem. It can and will happen in any typical user setup, hopefully without clearly audible artifacts in most cases.

 

[pma]

I think it is safe to say that this is NOT a measurement setup problem. It can and will happen in any typical user setup, hopefully without clearly audible artifacts in most cases.

Yes. But I am afraid that in a public forum debate scenario, with experts at one side and unqualified public at the other side, even such explanation will not find the fertile soil. It is time consuming with not much outcome.

 

[KSTR]

Yes. But I am afraid that in a public forum debate scenario, with experts at one side and unqualified public at the other side, even such explanation will not find the fertile soil. It is time consuming with not much outcome.

I'm more optimistic ;-)
It will stick at least for a few people and might help the manufacturers and designers to build better devices

 

[Geert]

Once this is working as intended, you also don't need that "GND lift" on USB input anymore.

For me this was a red flag from the start. First thing I did was to check if my Cambridge Audio Edge A integrated also has this switch. Luckily it doesn't.

 

[KSTR]

For me this was a red flag from the start.

Yep. A giveaway that they found grounding problems early on and tried to mitigate them this way (not very successfully).

 

[Snoopy]

So if I'm not using any of the inputs expect XLR everything is alright?

 

[audio_tony]

I am no Electrical Engineer, but surely there must be a way to filter out noise from the USB without breaking the bank? Cheap DACs seem to have no such issues.

To be clear; it is not noise from the USB. It's noise caused by two (or more) disparate ground points, which results in a loop (hence the term 'ground loop').

In simple terms, not all grounds (within various devices) are at the same potential (voltage levels), and this causes minute currents to flow between the different grounds, resulting in the noise we see in these tests.

In summary, your PC will definitely be connected to mains ground (it should be by default), and then there are HiFi devices like this Cambridge CXA81 which is also connected to mains ground. But due to minor differences in ground connection resistance, minute currents will flow between the ground connections.

Any DAC (or other HiFi device) that does not have a mains ground is unlikely to suffer from this issue (hence your comment about 'cheap DACs'), which typically have a power supply with no earth connection.

 

[BDWoody]

But there are so many other posts on this thread that bring up other situations to obfuscate what appears to be a reasonable explanation.

You seem to be confused as to the source of the obfuscation.

 

[sweetsounds]

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.

thanks @amirm, this is an impressive, investigative piece of work, taking pride in scientific work.

Very insightful, where defects can come from through noise injection.

This amp looks like a competently designed power amp, also in intermodulation.

Yet in general, channel separation or input and output impedances get rarely analyzed or mentioned (except by Atkinson).

The CXA8 has >40k input, and 0.115Ohm output, which is quite good.

Given that some speakers drop below 2Ohm, Isn't there a level, when a high output impedance makes amplifiers distinguishable (e.g. when frequency response starts to vary by >0.3dB)?

I wanted to have an amp 0.1Ohm or less to keep control.
Not a relevant thought?

 

[amirm]

Yet in general, channel separation or input and output impedances get rarely analyzed or mentioned (except by Atkinson).

Audio Precision software/hardware make it a pain to measure impedances. This is why I don't post them unless I have a reason to run them.

 

[amirm]

I wanted to have an amp 0.1Ohm or less to keep control.

Impedance is not one number. It will vary with frequency.

 

[Kosimo]

Dear All,

after all such interesting discussions it would be useful to have some hints on what are scenario works better, replying to some questions and doubts.
I report here for anyone who wants to elaborate on those helping to choose best scenarios.

Here I have sketched my two possible scenario based on my setup:

The solution 1 uses internal DAC of CXA 81 starting from an USB power supply (that has no ground pin) and uses the internal CXA 81 DAC over USB input.
1) In this case I positioned the USB switch on LIFT. It make sense to set USB input as lift even if raspberry pi has no ground pin? I would consider it equivalent since there is no ground pin in the raspberry input.
2) Does the fact that Rpi4 doesn't have a ground pin impact the noise source highlighted in post # 169?

The solution 2 uses my Topping E50 external DAC and hence only XLR inputs are used for CXA81. In this case both Rpi4 and Topping E50 have their own usb power supply and there is no ground pins (and no switch to choose for input).

I tested both these solutions qualitatively from a listening point of view without relevant audible differences. Maybe the second one feels more well-balanced but honestly we are speaking of nuances and I can not say how much these are objective.
Having said that I would like, from a signal integrity point of view, a hint on which solution you think is better and why, just to draw some conclusions on how use this unit at its best.

Thank you!