TLV320ADC6120: Device gain mismatch

Hi Jeff-san,

Thank you for your reply and support.

There are waveforms in which THD+N gradually deteriorates and waveforms in which THD+N rapidly deteriorates.
Is the difference between these two THD+N transitions within the range of device variations?

Best Regards,
Ishiwata

Hi Ishiwata-san,

If the THD+N is deteriorating at or near the full scale range of ADC, then yes these two differences are within range of device variations.

Best regards,
Jeff McPherson

Hi Jeff-san,

Thank you for your support and answer.

Customers are evaluating many devices, but the THD+N curve appears to be polarizing.

Customers think that there will be no polarization and that THD+N curves will appear in many variations.

Do you know the reason for the polarization? Do polarized characteristics appear due to some influence on the device?

Best Regards,
Ishiwata

Hi Ishiwata-san,

Can you elaborate on what you mean by polarizing?

Do you mean that over many samples, the THD+N curves are fitting into only two curves like the ones you shared before?

Thank you,
Jeff McPherson

Hi Jeff-san,

Thank you for your support and answer.

Polarization refers to two characteristics: when THD+N deteriorates gradually and when it deteriorates sharply.
The following is the result with the horizontal axis as the output standard.

There are two types of characteristics that deteriorate. Are there any possible reasons for this?
Also, how much variation is there in distortion between devices near full scale?

Best Regards,
Ishiwata

Hi Ishiwata-san,

This kind of behavior is not tested for our specified so we can't offer a variation. However I can reach out to our design team to see if there's an answer as to what may be causing this. Can you share the data again with the x axis as dBFS? dBu isn't as helpful since that's an analog measurement and this is digital data.

Thank you,
Jeff McPherson

Hi Jeff-san,

Thank you for your support and answer.

Customer plan to measure it in dBFS in the second half of next week in preparation.
please wait a moment.

Best Regards,
Ishiwata

Hi Ishiwata-san,

Thank you. Please reply when you have the data. I'll let you know when the design team responds; it's a holiday right now in India so it will take at least another day.

Best regards,
Jeff McPherson

Hi Ishiwata-san,

The design team responded to me. They said that the difference between the THDN behavior is likely due to clipping occurring at different stages in the signal chain. While the TI datasheet specifies the full scale voltage that creates clipping at a system level, the tolerances between blocks in the signal chain may cause different blocks to clip first between devices.

For example, say the signal chain contains block A and B. Block A has a rapid THDN deterioration when clipping occurs, Block B has a sloped THDN deterioration when clipping occurs. Block A may clip before Block B in some devices and vice versa in other devices while in all devices the clipping voltage is the same (the typical full scale value in the datasheet.)

This behavior isn't tested or designed for since it falls outside of the recommended use case.

Best regards,
Jeff McPherson

Hi Jeff-san,

Thank you for your support and answer.

Thank you for the response from the design team.
Are there only two types of this clipping? I have block A and block B, are there any other blocks?

Best Regards,
Ishiwata

Hi Ishiwata-san,

I only used A and B as a theoretical example. The reality may include more blocks but given your data set, I'd expect only those two patterns of clipping.

Best regards,
Jeff McPherson

Hi Jeff-san,

Thank you for your support and answer.

Customer measured horizontal axis dBFS.
Sudden deterioration and gradual deterioration were confirmed.

I understand that the cause is a difference in clipping.
I would like to understand about clipping.
Will there be steep deterioration if both blocks A and B clip?
Or will the deterioration be steep only in block A? For example, block A will experience steep deterioration, while block B will experience gradual deterioration.

Best Regards,
Ishiwata

Hi Ishiwata-san,

If both blocks clip, then the steep deterioration will dominate in the output. You can see how the two curves approach each other, which reflects that when the input is significantly in the clipping region and both blocks clip, the behavior becomes more uniform.

Best regards,
Jeff McPherson

Hi Jeff-san,

Thank you for your support and answer.

I understand.
I have an additional question: Can you answer how much distortion variation there is between devices near full scale?

Best Regards,
Ishiwata

Hi Ishiwata-san,

I'm not sure what you mean by "distortion variation" but the full scale value isn't given a minimum or maximum. Based on your data showing the worst case clipping occurring around -0.3dB, you could expect the full scale voltage to be at it's lowest 0.967V in single ended mode, 1.934V differentially. Since you already have taken 16 samples, if you take 14 more to give a set of 30, that's what we consider to be a statically significant set. That will give you a good picture of the distortion variations.

Best regards,
Jeff McPherson

Hi Jeff-san,

Sorry for the late reply.
I explained this to the customer. The customer was convinced.
Thank you for your support.

Best Regards,
Ishiwata