ADC6140EVM-PDK: ADC6140EVM-PDK

Paolo,

Thanks for the feedback. We are I interested in how we can improve the performance of the DRE. Unfortunately, the current version of PurePathConsole (PPC) only allows the control of the DRE threshold and gain. However, you can manually write the DRE registers by clicking the I2C Monitor button at the middle bottom of the PPC window. This opens a window where you can manually write any register with the following commands:

Dear Pedro,

please keep this thread alive if possible. Unfortunately, due to the current health situation in the country I work, I have not been able to do more experiments as you suggested, by changing more parameters. And I don't know when I will be able in the next future.

What I can tell you regarding this point:

"I would also like to understand the glockenspiel audible artifacts. Could you please upload an example waveform that generates the problem and the parameters used for the DRE."

I cannot upload an example as it is confidential and proprietary information. We could proceed with further exchanges only with an NDA and by continuing this conversation privately. Please let me know what you think about this. We have of course interest that TI does the best possible job, and we would be happy to help.

In the meantime, I can ask you to perform the test with an audio signal made of two sines summed together: one at 1kHz, and another one about 4 Hz apart. Do to the beat effect, the envelope will oscillate very slowly, and the switching of the PGA will be triggered periodically. You will hear extreme artefacts, especially if you normalize the signal up to an rms close to the -54dB threshold.

Could you give me a feedback on the latter experiment? 

Thank you!

Best regards,

Paolo Castiglione

Paolo,

I will send you a private message where you can share whatever information you feel comfortable to share. The DRE algorithm changes the gain of the system on the fly to boost low volume. Once the threshold is crossed, it will boost the gain until it reaches maximum gain. If the signal goes above the threshold, it will stop boosting the gain. This gain is simultaneously digitally removed to cancel the PGA boost and increasing the overall dynamic range of the system.  Plotting the response of the DRE into the ADC with hdB(n) being the input and ydB(n) the output of the DRE into the ADC, it looks las follows:

Boosting the low-level signals in analog keeps the input to the ADC significantly above its noise floor and thus prevents the ADC performance from being the limiting factor. Once the threshold is crossed, the signal is not boosted. If the threshold is set too high and the signal crosses this threshold too often, the DRE will be constantly releasing and attacking. By configuring the DRE parameters, choosing a lower threshold, increasing hysteresis, or changing the release and attack constant, the artifacts can be minimized. Note that the DRE takes time to respond to varying signals since it is estimating the RMS energy of the signal and deciding how to act. We normally set the attack rate to be fast to prevent the ADC from clipping due to too much gain and have the release be much slower, about 10x. If the signal is crossing this threshold often, increasing the hysteresis and the debounce would be my first recommendation.

As far as a test, what amplitudes for the sine waves should I use, same amplitudes for both? I performed some tests with trumpets, cymbals, and trombones and did not notice any audible artifacts. However, my threshold was much lower, around 70dB.

Best regards,
Pedro

Dear Pedro,

yes please, we can continue sharing files and more details about our application privately. I believe you have my email address.

Regarding the sinus, they shall both have about the same strength to obtain the beat effect. This artificial test is exactly to show what happens when the threshold is crossed, and due to the high tonality of this type of sound, the artifacts are not masked at all. In my experience this happens also with the glockenspiel, which is a very tonal signal. More harmonic and rich signals might mask farily well the artifact.

Yes, I would like to play with the parameters once I can reach our lab. At the moment, I really cannot move forward in this matter.

Tahnk you for your time.

Best regards,

Paolo