This thread has been locked.

If you have a related question, please click the "Ask a related question" button in the top right corner. The newly created question will be automatically linked to this question.

TAS5727: Sound quality change when going from AD to BD modulation

Part Number: TAS5727
Other Parts Discussed in Thread: TAS5717

I am using one channel of a TAS5727 amplifier as a mono output. I designed the filter based on App note SLOA119B and using components that would fit in the space we had on the board. The schematic for the output filter is as follows: 

 The PVDD inputs of the TAS5727 are being supplied by a 17.5V power supply.

 We decided to try switching to BD Modulation to test power consumption and ripple differences between the modulation schemes. To do this, I removed C43. Registers 0x11 through 0x14 were set according to table 16  on page 46 of the datasheet. Register 0x20 was also set accordingly. 

 The result of doing this was that the sound quality at the output is greatly improved with BD modulation. In AD mode, there is a constant white noise component when the amplifier is turned on. In BD mode, this white noise is gone and the audio is better. 

 My understanding is that changing the modulation scheme should not effect the sound quality of the output, but could have other effects on the power supply and emissions. One thing I noticed when looking at the output ripple on the 17.5V power supply is that the ripple increased when using BD modulation. 

 Can anyone provide an explanation as to why changing the modulation scheme would have such a great effect on the audio quality? Are there other register settings that might need to be changed or adjusted to improve audio quality for the AD modulation scheme? 

 Any feedback on this would be appreciated. 

Thank You,

 Ben Missele

  • So to update my own post, I found out that removing C43 eliminates the white noise regardless of modulation. The issue is that it adds an audible high pitched noise. Not sure how to describe it. Sort of a squeal.

    I updated C42 and C45 to 0.68uF capacitors (same as the eval board). In AD mode, the white noise is increased again. In BD mode, the white noise is much reduced, but the high pitched noise is there.

    It seems my sound issues are due to the output filtering.
  • Hi Ben,

    Do you get white noise using the EVM in your test setup? Also what is the resistance of your speakers?

    That is correct, with both AD and BD modulation you should not hear distortion. The final section of that app note you referenced has a nice comparison of THD achieved using AD vs BD modulation.

    Regards,

    Alex
  • Alex,

    Thanks for the reply. I've been trying to figure out what is going on here in our lab and I still do not have an answer.

    We are using an 8 ohm speaker. We do not have an EVM to test against. The design was originally tested with a TAS5717 EVM that I do not have access to.

    I have now tested across several boards and found varying results.  I do know that with AD Modulation, we get white noise across the spectrum.  Using BD modulation, I see varying noise across the spectrum.

    Here's an example of the noise using AD Modulation:

    Using BD modulation I see varying noise across the spectrum, but here is one example: 

    Looking at my power up sequence, we have PVDD set to 17.5V. The 17.5V supply is an input to the 3.3V supply, so looking at the ramp, the 3.3V supply comes up before the 17.5V supply and both Reset and PDN are tied to 3.3V through 10K resistors. The scope plot below shows the 17.5V trace coming up and the 3.3V and reset traces ramping at the same time.

    This one shows the voltage level of the 17.5V supply when 3.3V is at full voltage. When 3.3V is fully on, the 17.5V (PVDD) supply is at about 9V. It then takes an additional 9 ms to get up to the full 17.5V

    Looking at the software side of things, the oscillator trim is done first, then we wait 50us per the datasheet. Volume level is set, and then other register values are set. We are using most of the default settings, so DRC is turned off. 

    I'm just wondering if any of this could be having an effect on the audio output of the amplifier or could contribute to the differences I am seeing in the audio output between different boards and modulation schemes. Any help would be appreciated. 

    Thanks,

     Ben

  • Hi Ben,

    I will take a look into this and get back to you tomorrow.

    Regards,

    Alex
  • Hi Ben,

    Thanks for the additional information. Those noise plots make sense to me. Checking the datasheet, "output integrated noise (rms) A-weighted" is spec'ed at 56 μV, which translates to -85dBV. If you would like to do a one-one comparison though I would recommend taking an A-weighted integrated noise measurement so you can compare the values directly to the datasheet.

    I was under the impression answering your first post that you were having a problem with audible noise. Is the white noise you mentioned in your original post audible? All the noise of ~-85dBV shown in the plots should be inaudible at that low a level.

    I'd also like to clarify that C43 in your design is required for and should only populated for AD modulation. That is why you see more noise when it is removed in AD modulation mode. Likewise, BD modulation needs C43 removed to function properly as I see you have done in your plots.

    Regards,

    Alex
  • Alex,

     Thanks again for the reply. 

      The noise is noticeable. It does not dominate sound coming out of the speaker, but it is noticeable, especially when we have volume turned down to lower levels. 

      I understand that C43 needs to be placed for AD modulation and needs to be removed for BD modulation. 

     With C43 placed and the board in AD modulation, I see the noise spectrum shown with what sounds like barely audible white noise. 

     With C43 removed and the board placed in BD modulation, that white noise is gone, but it replaced by spikes in other areas of the range. On some boards, it falls out of the audible range, on others, it's right around 8khz and is noticeable. 

      My understanding of the modulation scheme is that changing modulation schemes should not change the output to the speaker, yet when I switch between modulation schemes, I see a difference in the audio output spectrum and I am trying to understand why that might be the case. I'm concerned that this could point to something else that may be wrong in the design. 

    Thanks, 

     Ben

     

  • Hi Ben,

    I see. Less noise with BD modulation is expected though I am surprised that you can hear distortion with AD modulation. If you want to get to the bottom of this I would take an integrated noise measurement with an A-weighted filter. That way we can tell if the device is operating within spec. If it is out of spec then there is certainly something going wrong. I can also take a look on the EVM and do a sweep to see if it correlates with your measurements, but in the end we will want to see how the performance performs compared to the datasheet.

    Regards,

    Alex
  • Alex,

    I think we have come to terms with the audio output. I do have a question about the register values, though. IT seems that several of the register values do not match the datasheet default values. This is consistent from board to board, so maybe the values in the datasheet are just outdated? Can you take a look and verify that all of the register values we are seeing are okay?

    reg=00 dat=6c
    reg=01 dat=c1
    reg=02 dat=00
    reg=03 dat=a0
    reg=04 dat=05
    reg=0e dat=f0
    reg=10 dat=01
    reg=11 dat=ac
    reg=12 dat=54
    reg=13 dat=ac
    reg=14 dat=54
    reg=19 dat=35
    reg=1a dat=68
    reg=1b dat=c0
    reg=1c dat=57
    reg=20 dat=00 msb
    dat=01
    dat=77
    dat=72
    reg=21 dat=00 e=00 msb
    dat=00 e=00
    dat=43 e=43
    dat=03 e=03
    reg=25 dat=01 msb
    dat=02
    dat=13
    dat=45
    reg=46 dat=00 e=00 msb
    dat=02 e=00
    dat=00 e=00
    dat=00 e=00
    reg=4f dat=00 e=00 msb
    dat=00 e=00
    dat=00 e=00
    dat=08 e=06
    reg=50 dat=0f e=00 e!= msb
    dat=70 e=00
    dat=80 e=00
    dat=00 e=00

    Thanks,
    Ben
  • Hi Ben,

    Just to double check, are these the default values that you read from the device or the values you are writing to the registers? Can you confirm that nothing was written to the devices before reading these values? For example I noticed register 0x03 has default value 0x08 and you have value 0xA0 listed. This means that the default setting is "Soft unmute on recovery from clock error" whereas you have hard unmute set on your device. If you wrote this to the device intentionally this makes sense, but if you read 0xA0 as the default value without changing register 0x03 then that would be something I could look into.

    Also if you could please specify which specific registers you are seeing the values not matching that would help me know which ones to look into. It is possible that the datasheet has the wrong default value.

    Finally, when you write "reg=21 dat=00 e=00" I was interpreting it as register 21 has value dat=00, but I was not sure what the e=00 means. Could you please clarify this for me?

    Regards,

    Alex
  • Alex,

    Thanks for the response.

    The registers we write are writing the following registers and they appear to be written to the value we expect:

    0x1b
    0x4f
    0x72
    0x73
    0x76
    0x77
    0x19
    0x20
    0x04
    0x07
    0x05

    The full readout is below. e=xx is what we expect the register to read. xx is the value shown in the datasheet. e! means the register does not match the default value in the datasheet including the reserved bits. em! means it does not match, but does not include the reserved bits. So to use register 0x03 per your statement above, we expect it should read 90 based on the datasheet, but it reads a0 and we have not written it to that value. Please let me know if you need further clarification.

    reg=00 dat=6c
    reg=01 dat=c1
    reg=02 dat=00
    reg=03 dat=a0 e=90 em!= e!=
    reg=04 dat=05
    reg=0e dat=f0 e=f0
    reg=10 dat=01 e=02 em!= e!=
    reg=11 dat=ac
    reg=12 dat=54
    reg=13 dat=ac
    reg=14 dat=54
    reg=19 dat=35
    reg=1a dat=68 e=0f em!= e!=
    reg=1b dat=c0
    reg=1c dat=57 e=02 em!= e!=
    reg=20 dat=00 msb
    dat=01
    dat=77
    dat=72
    reg=21 dat=00 e=00 msb
    dat=00 e=00
    dat=43 e=43
    dat=03 e=03
    reg=25 dat=01 msb
    dat=02
    dat=13
    dat=45
    reg=46 dat=00 e=00 msb
    dat=02 e=00 e!=
    dat=00 e=00
    dat=00 e=00
    reg=4f dat=00 e=00 msb
    dat=00 e=00
    dat=00 e=00
    dat=08 e=06 em!= e!=
    reg=50 dat=0f e=00 e!= msb
    dat=70 e=00 e!=
    dat=80 e=00 e!=
    dat=00 e=00

    Thanks,
    Ben
  • Hi Ben,

    Thanks for the clarification. I will look into this and report back.

    Regards,

    Alex
  • Hi Ben,

    Just wanted to give you an update. I need to verify this on the EVM but did not have one on hand so I am currently in the process of obtaining one. I should be able to get back to you in a few days.

    Regards,

    Alex
  • Hi Ben,

    Sorry for the delay on this. Are you still working on this? I had checked the registers and I got different values for the register 0x03 value: 80 (same as datasheet). However, for registers 0x02, 0x1c, 0x46, 0x50 I got values differing from the datasheet defaults. Checking the values you provided it looks like these are the only ones you were having problems with. I wanted to double check that this is correct or were there others as well?

    Best,

    Alex
  • Hello Ben! Per Alex' message is your question answered here? Thanks, Jeff