TAS5548: Noise persistent on CH1 PWM output

Part Number: TAS5548
Other Parts Discussed in Thread: , NE5532

I am working on a design utilizing the TAS5548. My Dev PCB for software has op-amp based low pass filters at the output of CH1-8 PWM outputs for line level output. The layout and schematic for CH1-8 was performed using Altium's Multichannel design, so i am confident that all channels are identical in layout and schematic. The layout for CH1-2 is literally copied and pasted to equal CH3-4, CH5-6, and CH7-8. With that said I am experiencing a noise that is only persistent on CH1.

This noise sounds like a repeating tearing sound that is down around our noise floor. It is repeating at a rate of around 2Hz.

What is unique about this noise is that it does not track my input signal. If I connect the APx to I2S input of the TAS5548 and turn off the generator the noise is still there. If I then use the input or output mixers of the TAS5548 to route, say input 2 to output 1 the noise is still there. Or conversely if I route input 1 to output 2 the noise is not present.

If I set the shared volume 0xD9 to 0x02 0x45 (minimum) the noise on CH1 is still present.

If I set CH1 volume 0xD1 to 0x02 0x45 (minimum) the noise is still present.

If I enable DAP automute register 0x04 bit D4 = 0, then the noise is muted when signal disappears. However, if i keep the DAP unmuted by sending signal to CH2 and removing it from CH1 the noise persists on CH1.

If I route the output of CH1-2 to the I2S output and feed that back into my APx the noise is not detected.

This leads me to conclude that there is some block adding noise to the output of CH1 between the output mixer and the DAP automute block. Unfortunately I cannot find any information in the data sheet about anything that exists between those blocks.

I have the TAS5548EVM, and am unable to make a measurement of this noise, as the noise floor of this EVM is greater than the noise floor on my PCB. It kinda sounds like it is there, but it is hard to say definitively.

I am using the following register writes to perform a minimal interaction with the TAS5548:

0x03 = 0xA0

0x04 = 0x13

0xD9 = 0x00 0x00 0x00 0x48

My application cannot have an automute function, and I am at a loss as to why this issue is only present on CH1. I have also gone through and verified the power on sequence, and manually toggled all of the control lines with delays of seconds, manually written the I2C registers with an aardvark and still see the issue.

CH1 = Blue

CH2 = Red

Image 1: Noise Floor With Auto-mute Enabled.

Image 2: Noise floor with automute disabled and shared volume = 0dB (0xD9 = 0x00 0x00 0x00 0x48)

Image 3: Noise floor with automute disabled and shared volume 0xD9 = 0x00 0x00 0x02 0x45

Image 4: CH1-2 routed to I2S output. Shared volume = 0dB. 0xC5 =  0x07 0x00 0x01 0x00 0x03

  • Hi Adam,

    Did you check the Valid pin to see if there's any noise on that one? If that pin is clean then we will need to look somewhere else.

    Also did you try a swapping the parts to see if that helped? 

    Best Regards,

    Robert Clifton

  • I have not done a part swap, but the issue is being seen across units from this build, and only on CH1. Regardless I will order a couple ICs up and swap them out.

    The boards were made in the US by an aerospace board house, who ordered direct from TI. So I do not think that it is a counterfeit part/assembly issue, but it could be an assembly or storage issue.

    When probing Valid I do not see it toggling. I do see power supply noise, but only ~2.5mVrms when the bandwidth is set to 20MHz (passive 1x probe). I get about 33mVpp when probing at 200MHz bandwidth with a standard 1x passive probe.

    Let me know if schematics or PCBA samples are needed, and I will ship them out.

    Right now this platform is being targeted at subwoofers, and we chose CH1 to connect our output stage. And we need to know if we should switch to using another PWM output channel. Ideally we would have identical noise across all channels. Additionally we are seeing this issue on our first PCBA with a power stage, but the noise floor on that board is higher.

  • I will also double check our crystal tuning tomorrow. I previously checked our DSP and it is at -6ppm average with a 10ppm XTAL. The TAS5548 also has a 10ppm XTAL, and I used the same estimated parasitics for calculating it. Based on the performance of the DSP I would assume we are well within spec.

  • XTAL output on MCLKO is sitting at 12288202Hz. So I do not think we have a clock problem.

    I also hacked the APx back into the TAS5548 as the source, and the noise is only present when I enable dither on the APx. If I disable dither the noise is not there. I am also using our newer apx to monitor I2S and Analog out simultaneously. Below is a screen capture showing the noise present with dither enabled, and it not present with dither disabled. One item of note is that the noise caused by the additional dither seems to be present on both CH1 and CH2 when looking at the I2S, but only CH1 when looking at the analog circuit.

    Dither On:

    Dither Off:

    I also tested out disabling the DC blocking filter, thinking there may be a precision issue down there. This had no effect.

    I am still waiting on the chips I directly ordered from TI. However, I am seeing this across PCBAs.

  • Hi Adam,

    I'm taking a look into this to see why there's a difference in the results. 

    Regards,

    Robert Clifton

  • Hi Adam,

    I had the chance to take a look further into this today. First, I'd recommend changing register 0xC5 to 0xC5 =  0x07 0x00 0x01 0x00 0x00. This might help minimize any noise since it will give the ASRC a higher sample rate. 

    Second, I couldn't recreate your issue. Now my setup doesn't have the output op-amps. I suspect there might be something wrong with that layout around channel 1's output. 

    Let me know if changing the sample rate helps. 

    Best Regards,

    Robert Clifton 

  • Robert,

    I have 3 PCBA builds in front of me. Below is a summary of the builds.

    Proto 1: Built in US, has all 8channels broken out to NE5532 low pass filters to provide line level output. Each stereo pair utilizes an identical layout. CH1, CH2 = CH3, CH4 = CH5, CH6 = CH7, CH8. Altium multi channel design was used to copy and paste the layout of a room 4 times over.

    Proto 2 (USA): built in US, has discrete class D output stage.

    Proto 2 (China): built at target contract manufacturer in southern china. Same layout and schematic as US build.

    All three builds have the issue with noise on CH1.

    Proto2 us a mono build. If I remove my series resistors on the PWM output from CH1 and route CH2 into the output stage the noise is not there.

    Ultimately the noise tracks all three builds across 2 unique output stages.

    I would be happy to share the schematics and layout with you over a secure file transfer or private email. I would also be more than happy to ship out some units if you are interested in looking at the target hardware.

    On another note; I tried changing the ASRC per your recommendation to no avail. The noise is still present. Below are plots from the high power class D output stage with either CH1 PWM output connected or CH2 PWM output connected to the output stage.

    PWM = CH1: No input signal.

    PWM = CH1: 1kHz Sine

    PWM = CH2: No Signal

    PWM = CH2: 1kHz Sine

    As a last item of note, I finally RX'd my ICs from Singapore, and will swap them out tomorrow or Wednesday. However, Seeing as this issue is persistent across 3 builds, does not lead me to believe that the issue is the ICs unless there was a quality issue at the factory. So I guess I will sacrifice an EVM's IC to see if that is a possibility.

    Thank you,

    Adam

  • Robert,

    Is it possible that the noise shaper is misbehaving? After posting I am noticing that the harmonic content of CH1 VS CH2 is different. Specifically, harmonics at 8kHz and 16kHz are present on CH1, and not on CH2. These are both harmonics of the switching frequency and source sample rate of 48kHz. Where as the main harmonics on CH2 are products of the 1kHz source signal.

    Just a shot in the dark, but that's all I've got right now.

    PWM = CH1:

    PWM = CH2:

  • Hi Adam,

    I think it would be beneficial to take a look at the schematic and layout offline. Expect to see an email from me shortly.

    Best Regards,

    Robert Clifton 

  • Issue seems to be damaged ICs. Working to track down the root cause of the damage.