On the implicit Low-Cut Filter of the CABINET Slot

  • Short version

    If you plan to profile (bass) amps or want to use (bass) cab IRs: Be aware of the implicit low-cut filter at 70Hz of the cabinet slot.

    Humble suggestion to the Kemper team: How about having frequency and slope of the low-cut filter as (internal) free parameters?

    Long version

    I've tried to profile a bass amp through the DI out but completely failed at first. I've then tracked down the cause and was really surprised by the findings. So why not share them?

    Already a coarse comparison of the frequency responses of the amp and the profile using white noise showed major differences: There was a steep low-cut at about 70Hz present in the profile and a ripple of alternating bumps and dips up to about 2kHz.

    But I've seen many bass amp profiles with deep lows. So how do they differ? It turned out that all such profiles were direct profiles, i.e. without a cab. In contrast, I naively hadn't engage the No Cabinet option for profiling the amp because the highcut filter of the amp was engaged which provides a gentle high-end roll-off like cabs do.

    Now having a clear suspect I imported an IR with a totally flat frequency response and analysed the resulting CABINET preset: A steep low-cut at about 70Hz, a light bump at about 100Hz, flat above; see attached plot frequencyResponse-KPA_CABINET-flat.png.

    Out of curiosity I've then tried to get a truly flat response by inverting the "flat" CABINET response to compensate for the unwanted "add-ons". The resulting frequency response was then transformed to an IR and imported via RM: Still a steep low-cut, above prominent ripples with decreasing intensity; see attached plot frequencyResponse-KPA_CABINET-compensated.png. Apparently, the algorithm is trying hard to match the desired (low-end) response but needs to introduce over- and undershoots in order to minimise the overall deviation. Actually, this is a common problem for filters with a limited number of uniform filter stages/taps: Large gradients in the low-end response are just hard to reproduce.

    Possibly said effects (low-cut, ripples) are the reason for some people finding it hard to get the low-end/low-mids right when profiling?

    KAOS version used: 8.5.8. Infos on the methods used to do the analysis: here.

  • There is even more to learn from the above analysis:

    1. There is also an implicit high-cut present at about 21.83kHz; see attached plot frequencyResponse-KPA_CABINET-flat-highCut.png.
    2. The ripples seem to be evenly spaced with a raster of slightly over 86/172Hz. Dividing 22.05kHz by this results in approx. 256/128 support points of the CABINET filter.

    Another humble suggestion to the Kemper team: If it is possible to individually define the center frequencies (like for a cascade of IIRs), how about a log(f) spacing? See attached plot filterSupportPoints.png.