Hey there,
Just wondering if there was an inherent limit to the number of biquad filters you can use in one design, beyond just system memory and number of instructions.
Cheers
David
David,
Not that I am aware of. why the question? Is there anything that you have in mind that you think might not work?
Susan
Susan,
Just an incongruity I was coming up against yesterday, it seems to have gone away. I asked because we had some strange results with the TAS3108, where once a certain number of biquads was reached, the dsp would refuse to output anything, and also exceeding 50% of the chip capacity in code words caused strange phase distortion.
Thankfully it seems to be ok again on this chip.
Thanks for the reply.
~David
An update: Using the TAS3204, when I do a build that has over 2030 DSP Code(words) (65.65%) according to the resources window, I get no output.
I did a check with this, when I remove a single component and bring it down to 2027, I get throughput. That includes just running input to output, as soon as it hits over 65.5% or so, it fails.
Is there an error in the resources reporting? Does the chip simply fail to output if you go over the maximum code words? All the other resources are under 60%, except for micro program but that's always around 89%.
Sorry for the delayed response. The message hit my mailbox when I was on vacation, but when I went through my mail when I came back, I missed this one.
Yes, that is somehow a known issue. When the cycle count >2000, then the DSP overran. It was a surprise when it appeared to me the first time. I think the reason is that since TAS3204 has analog in/out and it was oversampled when converting to digital, the up/down sampling took some cycles on the DSP, but when they do the resource calculation, they forgot to take that into account, only calculated what's in the process flow.
I hope this answered your doubt.