Hi,
We are using a 12J4000 ADC in a real-time frequency measurement system.
The ADC works at 4Ghz without background calibration.
One of our tests showed some errors in the frequency measurements in very specific situations.
The test consists on sending a pulsed single-tone signal on the ADC input. ADC raw data are stored on the FPGA and read-back on a host computer where the frequency analysis is done.
The frequency analysis consists on splitting the signal into 256-samples windows and apply an FFT on each window. For each window the frequency corresponding to the maximum bin power is extracted and compared to the input signal frequency.
In most cases it works very well but we identified a specific case where there is always a difference between the measured frequency and the input signal frequency.
This error happens in the case where:
- The frequency of the input signal is a multiple of 250MHz (250MHz, 500MHz, 750MHz …)
- The beginning or the end of the pulse is included in the windows. Ie a part of the buffer contains the signal; the other part contains noise.
In that case we systematically observe an error in the frequency measurement. The bin corresponding to the maximum of energy is not the one corresponding to the input signal frequency but adjacent. Depending on the input frequency we observe an error of + or – one bin.
Please find below a picture of a typical acquisition that causes issues.
We performed tests with several boards, several synthesizers and several FFT algorithms before considering that this problem may be related to the ADC itself.
Have you ever encountered this kind of behavior? Do you think this could be related to the ADC?
Thanks for your help
Florian
