AMB Laboratories DIY Audio Forums

[amb]

OK, I'll need to check my γ3 and compare against your results. It might take a few days before I have an opportunity to do that.

[cobretti]

One more thing, not related to my issue.
Previously you mentioned that USB connector shell shouldn't be touching the rear panel. And yes, there is a 1Meg resistor from miniB USB connector shell on ζ1 to GND, which is LVDD GND, also chassis GND.
What it means is, that when USB cable is plugged to laptop or other PC, the USB cable black GND wire gets connected to outer shell through the PC. GND wire is miniB PIN5 that is connected to LVDD GND through the SMD ferrite. The shell is now ground referenced. So the 1Meg resistor doesn't longer apply. The wanted USB shell isolation from the rear panel is void. All USB cables are like this. Even USB cable from Audioquest I have.
I found only one USB cable that doesn't have shell to shell connection.

[amb]

Yes, but the purpose is to prevent multiple ground paths from USB source to the ζ1, or there will be a ground loop.

[cobretti]

I had some time today to dig deeper into the poping noise problem as I described earlier. What I found out is a DC offset problem.
Selected input is input 0 (USB). When music is played through the USB or not, (measurement taken on γ24 board), the DC voltage on pin 54 is 2.422V and pin 55 has exatly the same voltage of 2.422V as expected. The same applies for other channel, pins 64&65. No problems there either.
Now. Music is being played through input 0, and now I swith to input 1 coaxial. I measure the DC offset like this(measurements taken each pin separately against the γ24 board GND):
pin 54: 1.92V
pin 55: 2.92V
pin 64: 2.04V
pin 65: 2.80V

Exactly the same voltages were measured on α24 boards input. Single ended output on α24 produces exactly this voltage differences DC offset!
I got 0.99V on one channel and 0.74V on the other channel DC offset!
I did sanity check and all voltages included the VREF and all points on the boards are exactly spot on. Isolation between GND and IGND is all fine.
As I wrote earlier, the unit performs flawlesly otherwise and has no sonical issues.
So, for some reason the WM8741 outputs go in to DC offset-imbalance. Once the SRC chip sees the signal on input 1, S/PDIF channel gets locked, all is fine. Then I can disconnect the S/PDIF from input 1 and everything is fine. It just needs that one time lock.
I replaced the SRC chip twice, so I am pretty confident it is not the chip itself. I still think somehow that it can perhabs be resolved by firmware because if there were other problems like soldering joint etc. the problem would be persistent and it wouldn't go away. Can you please suggest an idea where else I can check? Thank you.

[amb]

That's definitely strange! The firmware has no control over the WM8741's analog output DC offset. They should be the same regardless of what input is selected on the SRC4392. All signals up to the inside of the WM8741s are digital, and there is certainly nothing in the firmware that could alter the "DC offset" of that! If there was a firmware issue that could influence this, then all γ3s should exhibit this issue, but clearly that is not the case.

I am as baffled as you are on this.

[cobretti]

I was thinking if by any chance when the other input is selected the I2C command gets passed to WM8741 and doesn't do proper acknowledge or something. It would or wouldn't invoke the other channel volume level correctly, that is stored in memory for that channel perhaps? Other observation also is that when music playing through input 0 is louder, then when switched to input 1, the DC offset gets higher. So it is also volume dependent. Could those WM8741 be faulty? But both of them at the same time? There is nothing else besides IC2 that performs those control commands. What else can I check? I don't have any logic analyzer on hand that I can use and capture the I2C data so I can compare them against the firmware if they match.

[cobretti]

That's definitely strange! The firmware has no control over the WM8741's analog output DC offset. They should be the same regardless of what input is selected on the SRC4392. All signals up to the inside of the WM8741s are digital, and there is certainly nothing in the firmware that could alter the "DC offset" of that! If there was a firmware issue that could influence this, then all γ3s should exhibit this issue, but clearly that is not the case.

I am as baffled as you are on this.

I remember :whomper: also once described this issue. If other guys don't have components DC coupled, maybe they don't know they have DC offset on the output.

[amb]

I2C commands are sent to each WM8741 separately, without read-back. The volume settings sent down to the WM8741s are absolute (not relative to any previous step). The latter is handled within the firmware, but the same value is sent to both chips.

How does the volume change with the DC offset? I'm not sure I get what you're describing in terms of cause and effect.

Many people are driving DC coupled amps with the γ3. Whomper and myself are using β24 power amps which are definitely DC coupled. I don't think there were any reports of symptoms like yours (DC offset shift when changing inputs). It's very odd, because the chip that does the input switching is the SRC4392, and I don't see how it could cause a DC offset at the analog output of the WM8741s. Even if one or both of your WM8741s are defective, then the symptom should be constant, not dependent on the selected input.

Are you 100% sure of the correlation between input switching and DC offset? Maybe the problem has nothing to do with the SRC4392. Have you tried reflowing the solder joints of both WM8741s and their surrounding resistors and capacitors? Also, make sure that the 5V analog power supply (individually regulated with ADM7150-5.0s for each WM8741) have constant 5V on them under all circumstances.

[cobretti]

How does the volume change with the DC offset? I'm not sure I get what you're describing in terms of cause and effect.]

The other way around. DC offset changes with volume. The higher the volume level, the bigger the DC voltage differences between pins 54&55 and pins 64&65.

Are you 100% sure of the correlation between input switching and DC offset? Maybe the problem has nothing to do with the SRC4392. Have you tried reflowing the solder joints of both WM8741s and their surrounding resistors and capacitors? Also, make sure that the 5V analog power supply (individually regulated with ADM7150-5.0s for each WM8741) have constant 5V on them under all circumstances.

Yes, I am 100% sure, measurements were taken several times by Agilent U1252B. I am not saying the problem lies in or around SRC4392 chip itself, because I really don't know. I will have to ring out and look around SCL-SDA traces. I2C is basically responsible for everything. Some of the I2C data maybe don't reach the chips all the way.....there definitely is a ghost in the machine.