sinc interpolation in fpga

[lgeorge123]

For high sampling rate in Gs/s of scope passing data to fpga ,interpolation will be used , for a 8 bit ADC , multiple and sum in sinc interpolation will result in 16 bit data . 255 pixel of lcd is representing 8 bit data,but after interpolation 16 bit is comes out , how to put this 16 bit into 8 bit "heights" ?

 

[kripacharya]

Throw away the lowest 8 bits, just use the higher 8 bits

 

[lgeorge123]

Throw away the lowest 8 bits, just use the higher 8 bits

but the number represent +32768 to -32768 , 8 bit number is +128 to -127 , just use the higher 8 bits is an easy job, how to understand its meaning ? can kripacharya give me an example ?

 

[kripacharya]

Exactly.

You have to "scale" the +-32768 down to +-128, since you only have 255 pixels.
This 'scaling' is done by dividing the 16 bit number by 256.

In binary, doing this division is the same as a right-shift by 8 bits.

Which is the same as throwing away the least sig 8 bits and just directly using the higher 8 bits.

 

[lgeorge123]

Thanks a lot , kripacharya !

 

[lgeorge123]

Other than sinc interpolation I know that there is a filter called cic comb filter for upsampling , is it suitable for interpolation of dso ?

 

[FvM]

Did you check the effect of sinc interpolation on real oscilloscope signals? In my view it's rather a theorectical concept than a practical method for presentation of time domain data. I remember an early digital oscilloscope from HP that actually performed sinc interpolation of display sampled data. As a result, you got funny oscillations with monotonous steps. The presentation of the new instrument caused big laughter in the audience.

Sinc interpolation corresponds to a low-pass filter with rectangular shape. This "ideal" filter may be appropriate for small band signals, but it's rarely useable for wide band signals analyzed in time domain. A CIC filter in contrast has a wide transition band and may introduce an unwanted bandwidth reduction. What you probably want is something similar to a bessel or butterworth filter with no or little overshooth in step response. I wonder if recent oscillocopes are using interpolation filters or e.g. spline interpolation of the waveforms?