TLV320AIC3254: TLV320AIC3254

Hi Don,

would you mind sharing your schematic with me? if you would like yo avoid showing the schematic in public, you can send it in a private message.

best regards,
-Steve Wilson

Hi Steve, Thanks for picking this up.  Tony Janke-TI has the schematic and code.  He will forward to you.  

Thanks, Don

Don, I've received the schematics.

With the common mode set to 0.9V the full scale input level is 0.5vRMS. with the common mode set to 0.75 it would be 0.375Vrms.
So that is where the ADC will start to clip. Where are you seeing the input start to clip? With the voltage divider you have your mic signal would need to be huge to clip the input.

best regards,
-Steve Wilson

Hi Steve, I took some scope shots this AM.  What I see is not clipping, but distortion? Tony is forwarding these now since I don't see a way to upload here.

JPGs sent with brief explanation embedded in filename.

The audio path is 'AUDIO_IN', through U13 (analog switch), R-divider (R64/R225), IN1_L.  IE the U20 Mic pre-amp is bypassed. the audio  source is a PC audio  signal generator.. Audio freq is 1.0 kHz.

The linear case:

1. P1:  100 mV nominal Vin, BEFORE R-divider.  (I have verified the R-divider output; its clean but difficult to grab on scope since there ismodulated RF nearby.

2. P2:  This is the CODEC output, (1 channel), unfiltered.  There is some RF on the waveform. Not a problem, its just my probe.

But note there is gain produced by the CODEC.  Could it be in the digital filters involved?

3. P3: I have an active op-amp filter that accept P2 as an input.  The op-amp output (scope shot) goes to the RF modulator.  Obviously, we are clean and linear at this point.

A non-distorted, max input case:

P4: Into R-divider.

P5: Corresponding unfiltered CODEC output.

A distorted Case to illustrate CODEC output:

P6: high level input-linear (I verified linear to actual CODEC in)

P7:  CODEC out-distortion.

………..  So I don't really know what is happening inside the part.. If you look at present gain setting, I can try various things you suggest. But I'm seeing the part (or my filters) going non-linear somewhere around 35 mVrms.

If the filters are a concern to you: The filters coeff were developed by someone I no longer have access to.  I'm happy to share them with you.  Basically they take the mono input signal and form a quadrature I-Q  pair over the freq range (very accurately). The I signal is -45 deg shifted.  The Q signal is +45 deg shifted; ie 90 deg total.

Steve, I should state my goals here. This radio is in production and I'm not looking to do anything with the R-divider. I put it in place when I could not resolve a clipping issue 18 months ago.

The scope shots sent are using an auxiliary audio input to the radio. What I am hoping to do is use the CODEC to raise gain on the mic input (user control). I have users who have various mics in use and are having problems getting enough RF output. Even though there is a gain leveling pre-amp, having some control with gain using the CODEC (6 more dB) would take care of that issue.

It appears I do have some gain headroom. But I thought it would be good to understand what is causing the limited input levels first.
Thanks, Don

Don, I've taken a look at your configuration and schematic but don't see where the input would be clipping. There is no gain on the PGA... Are you using the MiniDSP? if so can you share your process flow?
-Steve

Steve, I just forward the process flow to Tony.  He will forward to you.  I have not run in 3 years since we moved to production. You will likely see both mini-DSPs in use.  The mono input is split into 2 outputs (-45 deg and + 45deg) to form quadrature signals.  The filters work precisely from 500-3000 Hz as intended. 

It seems there is process gain judging by the input and output levels.  Please advise, Don