Sigma Filter Foibles - ADAU1701 (and beyond?)

This relates to the development board for the ADAU1701 running under SigmaStudio 4.1. The first observation was related in another recent thread, but because that thread does not have 'filter' in the title, that information is presented again here.

This first problem has to do with the general first-order filter, found near the bottom of the Filters menu tree:

When used as shown, all is well. But this is a stereo project, so for the other channel we need to "Add Algorithm" thusly:

For some reason, SigmaStudio 4.1 does not like this companion channel and puts up that warning. When you try to compile anyway, this is what happens:

...and of course the circuit doesn't work. The fix is either to use a second, single-channel filter, or make another choice.

Back up the tree under Second Order Filters, there's the ol' standby Second Order/Single Precision/2 Ch/
General:

Once that's up on the page, you can select a First Order function. Do note, however, that this filter choice actually provides two cascaded first-order sections. To start, they default to a high-pass followed by a low-pass, as shown above. But you can turn off (deactivate) either of the sections if all you need is a simple first-order filter.

For the most part, this filter choice behaves nicely, not like the troublesome First Order/General discussed earlier. But all that glitters is not gold, which prompts the second observation.

Note that used as a low-pass filter, the two cascaded sections operate as expected. As the following two snapshots show, you can deactivate one section, and the magnitude and phase response act as one would expect. Two sections have a steeper slope and a more convoluted phase response than a single section:

Do note that within the passband this low-pass function maintains signal phase; that is, zero degrees at the bottom-end, whether it's one section or two.

But change the filter to a high-pass function, and look what happens as you deactivate the second section:

Both sections together maintain signal phase at the top of the passband, but deactivating one section gives a 180-degree phase reversal! Not a big problem, as there's a switch on the filter to invert the phase of the signal going through it, be it one section or two. But in the intended application in this particular case, that unexpected phase reversal yielded strange circuit behavior and wasted half a day figuring out what the heck was going on.

Just a 'heads-up.'

Audio Signal Processors

SigmaDSP Audio Processors

adau1701