Programmable Devices
CPLDs, FPGAs, SoC FPGAs, Configuration, and Transceivers
20693 Discussions

Playing Vectors for testing

Altera_Forum
Honored Contributor II
892 Views

Vector continuity for signal testing(especially in real time) posed problems in some posts. Vectors may be generated in models from various points of a DSP chain. Additionally you may want to test for bit true accuracy every time you play your vector. I would like to share with you this info: 

 

When a vector is played repeatedly then any discontinuity of signal phase at end/start will have effects similar to data errors (though false).  

 

In a DSP chain model that generates a vector to be played at various testpoints, there are at least two sources of continuity trouble; filters and any sine wave(nco) added. 

 

The nco issue is avoided by having a number of samples equal to integer number of expected sine cycles. 

 

The filter effect is a bit more tricky. You will need to get rid of initial filtered outputs(step response) or the tail response as follows: At the input to the filters copy enough data samples from start of vector and append it to the end. At the filtered output then discard the initial response picking up the required vector length. But you will need to take into account any upsampling or downsampling factor done in the chain. 

 

If your DSP chain also involves block processing e.g. fft/ifft then you are forced to have integer number of frames. In this case you need to generate enough frames to yield your required samples or more then chop off resulting vector as required. 

 

In all cases, it is wise to check continuity of your final vector by simply adding few end samples to the start and view its plot. 

 

When a vector is played as input to realtime nco then you will need to make sure that the realtime nco starts phase at same point if you want your testing to be bit true. 

 

Hope this helps.
0 Kudos
0 Replies
Reply