ADAU 1452 FIR length Sigma Studio

Reini Question

Nov 29, 2019 00:00

Hi all,

Am I right assuming that, if I want to implement a FIR filter in Sigma Studio for a ADAU1452 of 2400 taps, I have to set the length in SigmaStudio to 2401? I've read somewhere that the last coefficient of the filter is used as a bypass function. If the last coefficient is 1.0 then the FIR is bypassed, if 0.0 the filter is active.

For debugging purposses I implemented a 10 tap FIR filter and saw that the last coefficient toggles between 1.0 and 0.0 when I toggle the bypass control on the component in SigmaStudio :-)

[update]

I feel a bit stupid, now that I found that the last coefficient sent, is actually the first. I now think the "bypass" button on the FIR component does nothing more than creating a kind of unity fir filter when bypassed is chosen and uses the coefficients from the table when the FIR is not bypassed. Is that assumption correct?

Note: I installed SigmaStudio 4.5, but the help is empty...

With kind regards,

Reini

Edit Notes
Added [update] section

[edited by: Reini at 12:44 PM (GMT 0) on 29 Nov 2019]

Audio Signal Processors

adau1452

SigmaDSP Audio Processors

ADAU1452 FIR

Report

Reini Reply

Nov 29, 2019 00:00

After some checking, it is now clear to me that the FIR is working flawless. The reason why my FIR's didn't act as expected is that I had added a 2-way mux for bypassing the FIR. These multiplexers expect a floatingpoint number to set the correct input. I used them before and I should know that :-) Of course, I supplied a integer number 1 instead of a fixed point 8.24 number 1.0 and the mux always bypassed the FIR...

So, my issue isn't an issue anymore.

Thanks for watching!

Report

DaveThib Reply

Dec 5, 2019 00:00

Hello Reini,

Yes, the last coefficient is the bypass and when you click on the bypass button all the other coefficients are changed to zero. So when you select a 10 tap filter you use 11 memory locations. However, you do not have to plan for this. If you want a 2400 tap filter you select 2400 taps. The compiler will add the one extra storage location for the bypass.

Now regarding the 2-Way MUX. They do NOT expect a floating point number. They expect an unsigned integer. So if you use a readback cell to look at the data you need to set it to HEX or 32.0 in fixed format. If you use a DC cell to test the Mux you set it the same, for a 32.0 or Hex number. So the 2-way will react to the LSB only. If you give it a number that is higher you will get unpredictable results. The number is used as a pointer to the location of the audio data. If it is pointed to some memory location that it not part of the switch you will get unexpected results.