AnsweredAssumed Answered

FIFO glitch? (first result off)

Question asked by enpa on Nov 7, 2017
Latest reply on Nov 20, 2017 by enpa

I often get a "glitch" in the data-FIFO out of the AFE. The first entry is off (and the DFT results are questionable).


Here is the content of the results I get out of adi_AFE_RunSequence(hAfe, seq005, (uint16_t*)rxBuffer, 8);

This sequence uses the DFT as data source. As you can see, after the first entry, the following results are pairs of similar values (I and Q). The first entry is 14013 and is way off if compared to the rest.

These are the signals during the sequence that generated this measurement:

C2 is the excitation voltage (amplitude = 8 mV), C4 is the signal on the pin TIA_O (i.e. what enters the ADC channel). The trace on the bottom is a zoom-in of the first 5 ms of the acquisition time for C4. The signal amplitude is 475 mV (perfectly in range according to page 394 of UG-587 rev D), there is no saturation and the start transient has died out before acquisition. It looks like a perfectly fine signal to me.

The calculated magnitude is sqrt(59361^2 + 63259^2) = 86750


Then I increase the excitation amplitude and get following results:

The pattern is similar (a point off, at a similar value: 14078 and thereafter value pairs).

The signals look now like this:

The excitation amplitude is almost doubled: 15 mV. Accordingly the TIA output is now 865 mV. The calculated magnitude is here sqrt(54433^2+61404^2) = 82060, which is lower than the previous measurement!


In another sequence, I measure a signal that on the scope has an amplitude of 1.43 V (still in range). Here I deactivated the DFT block and read out the raw ADC samples. This is the result:

It's a sinus and the amplitude is (52235 - 12847) / 2^16 * 3.6 / 1.5 = 1.44 V (close enough).

Everything perfect: the ADC seems to work and probably delivers correct samples to the DFT.

But, still, the first value is always at 14000, the second (correctly) at 51301.


Because of the attachments limit, I'll update the sequences used in a separate reply to this thread.


The question is:

- why is the first result in the FIFO off?

- why is the calculated magnitude from the I/Q-values decreasing when the signal amplitude increases?


Thank you in advance and kind regards, Enpa