Search FAQs on the left to see if your question has been answered. Click on the dropdown to view all of the documents associated with the product. If you can't find your question, click on Ask a Question

AD7401: SINC3 filtering

We're planning to use the AD7401 with an FPGA for current measurement in a
motor control application. The questions concern the post-processing of the
modulator output from the AD7401. I've done some reference reading and also
implemented the sinc3 filter that is proposed from AD in a computation
environment to explore the area, but I don't get the results I expect. My
current suspicion is that far more filtering than the sinc3 is required as post
processing to receive as outputs of the ADC chain (AD7401+FPGA), sampled data
values of the input signal of the isolated modulator. Further, I don't fully
trust the verilog sinc3 implementation presented in the data sheet, since the
accumulator registers seem to grow beyond their 24 bit numerical range when
supplied with an arbitrary pulse train, causing discontinuities to cope with
for the following filter stage. (I ran on floating point, so without numeric
limitations the output is reasonable, but the information in the signal was
nevertheless not what I had expected) Please provide comment on the above.


The code given in the datasheet is what is used in the final test solution for
testing the AD7401. They use the Sinc3 filter and this is suitable for
guaranteeing datasheet specifications.  Higher order filters do not
significantly improve performance.

The filter is not a 24 bit precision filter. It is only designed to give 16 bit
performance. The filter code has internal filter digital buses that are 24 bits
and the data out is truncated to 16 bits, all filter coefficients are set to 
You need this number of bits to guarantee the 16 bit precision, you must drop
the last 8 bits.  Other filter arrangements are possible and some systems may
experience better performance by increasing some of the feedback delays. 

The filter accuracy is determined by the decimation rate, thus a decimation
rate of 256 is required for the 16 bit resolution. If you increase the
decimation rate you will have better performance from the AD7401 but this will
reduce the throughput of the device and the usable BW. Figure 29 in the
datasheet illustrates how increasing the decimation rate effects performance.

With the SINC3 filter detailed on the datasheet you should be able to achieve
16 bit performance without any los of data. It is important to remember the
filter is for 16 bit resolution not 24 bit resolution. Do you require 24 bit
resolution? The accumulator will grow beyond the 24 bits but should not cause
any errors, the size of the registers (24-bits) makes sure that the relevant
data will not be lost.