Inconsistency between Coeff. Calc filters and general filters
Hi,
I am using Sigma Studio v4.4 on ADAU1462.
There seems to be an error in the Coeff Calc. filters. When set to parametric mode, the gain measured at the output is only half that which is set in the UI (and shown on the stimulus probe).
See the attached images
The first image shows the Sigmastudio settings. On the left is the 'Coeff Calc' filter, and on the right there is a normal 2nd order filter block. Both are set to Q1.41 and gain of +10dB. The Coeff Calc is set at 500Hz whereas the normal filter block is set to 3kHz.
The second image shows the Stimulus/Probe simulation. All looks good.
The third image shows the measured output. As you can see, the peak at 500Hz is only +5dB instead of +10dB.
Is this a known issue?
Thanks
Audio Signal Processors
SigmaDSP Audio Processors
adau1462
I have also just found the issue is present on shelving filters too
I have noticed that the same issue is raised in the post below
https://ez.analog.com/dsp/sigmadsp/f/q-a/106814/measured-filter-characteristic-is-half/317930#317930
DaveThib's response in March 2019 is that there was no estimated fix date. Is there any update on this? It will be disappointing if such a basic issue has still not been resolved a year after being brought to AD's attention. (The issue is still present in Sigmastudio v4.5)
I seem to have found a further issue with the Coeff Calc blocks when set to high shelf filter. When they are set to anything other than 0dB boost there is a large increase in LF noise. See the below images - first image shows the filter set to 0dB boost, the secodn is with it set to 1dB boost. Again this issue is not present on the noraml 2nd order filter blocks.
Hello eepmc,
Thanks for the continued posts. I looked up the bug report and renewed it so it will get some attention.
The other issue you are seeing with the noise floor has to do with something that was a hold-over on the older filters. When the boost is set to zero it does not use the coefficients for the filter, instead it uses bypass filter coefficients. So when you set it to zero it is bypassed. If you put in a very slight boost it will still use the filter coefficients but with the gain set really small. So it is not a bug but a symptom of what the code is doing if the boost is set exactly to zero. With it set to zero it is using very different coefficients.
Dave T
Please log in to reply.