TLV320AIC3007: PGA noise floor

Hi

Few things to clarify:

• Is this seen on all units and showing same behavior?
• How's the test conducted? Did they provide an input at a known level or just idle?
• What if they use a known level say -60dB input, are they seeing the same behavior?
• Can they run from the opposite direction from -16dB up to 0dB?
• How did they capture the noise with the same period? Is this frequency domain?

Regards.

Hi Pdjuandi,

Thank you for your prompt response.

Our customer prepared 0.5dB step FW script and recode output level, then convert it to time domain graph.

 It was confusion in my previous message but this is DAC side behavior.

Mixer LOP Volume Control Register (R82, R89) are changed by 0.5dB step, input PCM data is 0.

We will confirm following things and get back to you.

- Population of this problem.

- -60dB input measurement with same environment.

- To see another mixer input node (like Line2) noise floor.

- To see the opposite level direction test.

Regards,

Mochizuki

Hi Mochizuki-san,

Thank you for the clarification. Let us know the result of those inquires.

Best regards,

Jeff

Hello,

We have done additional investigation. Here is our observations.

1. All of TLV320AIC3007 device have this noise floor variation and zipper noise, even on the EVM device. But our customer`s configuration is +9dB gain at LO driver.
2. Input signal level is not parameter of zipper noise level, it is not changed at no signal or -60dB signal input.
3. When change mixer input source from DAC to Line2 input, 0.5dB step noise floor variation is not changed. So this 0.5dB step noise floor issue is come from mixer stage itself.
4. When change mixer gain from opposite side, noise floor variation is not changed.

 From those result, we still have questions.

1. We expect Mixer stage gain control and LO driver gain control are not utilized Zipper noise free algorism. Is that correct?
2. Could you duplicate noise floor variation issue by every 0.5dB of mixer gain?  Is there any workaround to improve this noise floor?

Regards,

Mochizuki

Hi,

Can you send me the configuration/register settings for the test so I know the path your are testing?

Are you saying the issue only happens on DAC to LINE output and not from LINE2 input direct to LINE output?

When you said mixer, which block are you referring to?

Here is a block diagram of AIC3107 which is superset of AIC3007, can you draw the path which has and don't have the issue?

Regards.

Hello Pdjuandi,

Let me clarify our customer's concern.

By our additional investigation, we can simplify the problem which is noise performance of HPL/HPR OUT Volume and LEFT/RIGHT LOP Volume function.

They plan to use this function for Master Volume control in the audio system.

At initial test, they had used DAC input path R47,R64/R82, R92 but even if change an input path to LINE2 input R45,R62/R80, R90, concerned noise phenomena is not changed. So input source is not matter in this case.

There are large Zipper noise and non-monotonicity noise floor by 0.5dB step volume gain change, as we sent captured image in first post.

In the customer application HPL/HPR output performance is higher priority, let's focus to HPL/HPR OUT Volume function. The customer apply +9dB output driver gain R51,R65.

Answer to your comments:

Can you send me the configuration/register settings --> I will ask the customer to share the script.

Are you saying the issue only happens on DAC to LINE output and not from LINE2 input direct to LINE output? --> Sorry, there is some confusion, I have explained this in above  

When you said mixer, which block are you referring to? --> That is HPL/HPR OUT Volume block which has MUX and Mixer function.

Here is a block diagram of AIC3107 which is superset of AIC3007, can you draw the path which has and don't have the issue?--> Actually it is not matter input path of HPL/HPR OUT Volume.

This measurement was done by special test script,our customer prepared FW which has 0.5dB step volume change from 0dB to -18dB with some interval. Then recode output noise level and draw the waveform by PC application SW. 

Customer's requests:

1.We expect HPL/HPR OUT Volume control and Output driver gain control are not utilized Zipper noise free algorism. Is that correct?

How can we use HPL/HPR OUT Volume for on the fly function audio system master volume?

2.Could you duplicate noise floor variation issue by every 0.5dB of mixer gain?  Is there any workaround to improve this noise floor?

You will easily hear and identify the noise floor from HPL/HPR OUT if touch HPL/HPR OUT Volume every 0.5dB steps.

By this characteristics, difficult to specify SNR specification in audio system and unconfutable hearable back-grand noise is created.

What is reason of this noise floor change? Will this characteristics different by device by device?

If needed we can have conference call to make rapid progress.                                                                                                                                          

Regards,

Mochizuki

Hi,

Attached  is verification test script on EVM, the other registers are default vale.

Monitor HPL and HPR output signal, then change gain of HPL/HPR OUT Volume R47 and R64 by 0.5dB step.

You will see Zipper noise and non-monotonicity noise floor change.

Regards,AIC3007script.xlsx

Mochizuki

Hi,

You can control the volume gain by changing the DAC volume control (R43 and R44) instead of the analog volume control (R47 and R64) as another option which is normally used to control the volume. Have you tried with 1KHz tone and do you see the same behavior?

I'll check it on the EVM tomorrow.

Regards.

Hi Pdjusndi,

Thank you for the reply.

 Zipper noise and non-monotonicity noise floor behavior is not changed by input signal level, No-signal, 1KHz, or Music.

One of the concern of DAC volume unitarization is EE system level SNR at lower volume setting because keeping higher analog gain at DAC back end mixer and driver gain.

Ideally DAC operate full resolution without applying digital attenuation then reduce gain at backend analog stage to get better system level SNR.

But if it is not allowed by zipper noise concern this time, we should use DAC digital volume for Master volume.

Still we are curious why noise floor level is changed 4~5dB by 0.5dB mixer gain change. What mechanism is in behind?

Regards,

Mochizuki  

Hi,

Thanks for the confirmation.

I have checked on the EVM and I'm seeing similar trend except it's higher at -2.5dB, -6dB etc.

I'm checking internally on the reason behind this and will let you know, but as suggested previously DAC volume might be the option here.

Regards.