This is not technically a bug, but the max frequency on the slope being limited to 10khz removes a lot of the treble and affects the sound a lot.
Whenever I use a slope I need to automate the on/off as to not remove everything above 10khz when the filter is not doing anything.
Could there be a better way to do this? Maybe have the frequency knob go all the way to 25khz or something? Maybe use an exponential curve to compensate for the bigger range? When turning the filter on/off the difference between 10khz can sometimes be quite noticable.
Comments (28)
This has bugged me often. I think that the high-pass mode should not touch anything above the cut-off frequency, like a real filter would do. To me it does seem like a bug.
That's true @jordynth
I forgot to specify I meant the low pass in particular :)
Bumping this. High-passing tracks intended to be mixed together is often an important step to get a clean and defined bottom end and give the track enough loudness. A simple pedal like the Slope would be ideal for this task, keeping the signal chain efficient, more than having to use a GraphicalEQ or a Curve. It would be great news if this was addressed. Additionally, I also support this in relation to the Slope: https://www.audiotool.com/board/feature_requests-eGRGOKR3F/slope_pedal_steepness
There might be technical reasons why this isn't (easily) possible. Depending on the filter method (IIR or FIR) higher frequencies impose numeric instabilities or alias artifacts. 25 kHz would be above the Nyquist frequency when sampling with less than 50 kHz. This is impossible or hard or slow to implement.
I think it's possible (by upsampling → filtering → downsampling) for frequencies up to 20 kHz (or at least 17 kHz), but I currently cannot guess the impact on performance and latency.
It might be a workaround to automate the mix-parameter to "soft-off" the effect.
I really don't have the mathematical knowledge to discuss this, but since all synths were redesigned in Next to have band-limited oscillators and sampled audio is already naturally band-limited, wouldn't this remove the need to deal with anything above 22,05 kHz (or 24kHz eventually)?
Valid question and in theory (continuous wave) you're right. But sampled (discrete) audio introduces some non-trivial problems when getting closer to the sampling rate :/
I'll try to explain this with as few math as possible:
The higher the filter's frequency, the less samples have to be taken into account. Fewer samples mean coarser time-resolution. Coarser time-resolution mean you'll have to tell the filter what's "between" the samples. "between" the samples is a matter of interpolation. The zero-cost method (nearest neighbor) will create a discontinuous representation (has hard value jumps in it, like a rectangle wave). Discontinuities cause an infinite series of overtones (of the sampling rate). Due to a so-called alias-effect (https://en.wikipedia.org/wiki/Nyquist%E2%80%93Shannon_sampling_theorem#Aliasing) they get mirrored back into the audible range.
Using a linear interpolation is better but imposes some computation costs. The artifacts are continuous but its first derivative is not (like a triangle wave, which has less overtones than a rectangle wave).
Better interpolations (sub-sampling) have fewer artifacts, but are computationally intensive.
There are lots of other views on the same problem, but I hope this explanation is easier to grasp than the math behind it.
Thanks, I really appreciate your effort in putting this in simple terms and I think I get it. It's really an implementation design problem. In any case, since the Slope pedal is from the first generation and since then much more sophisticated tools have been developed with properly working filters like Curve and Quantum (and even older tools like the Machiniste or Autofilter include proper high-pass filtering), I thought that to fix the Slope's HP filter mode now would be easy.
Depending on how this filter has been designed in the first place, it might be easy or a major rework at the risk of changing the sound of old tracks.
Only the devs know :)
bump again, i actually reported this a while back, not sure what happened, i remember Sandy chiming in too saying he agreed. Note, this effects all of the filters on Slope, not just lowpass, i.e the hipass only goes up to 10kHz as well, which in my opinion isnt a very high hipass haha
Glad that you bumped this too. Yes, for me the main problem has always been the high pass filter. I remember when you reported this too. The more I learn about mixing and mastering, the more I realise the importance of high pass filtering sounds. The slope is the ideal tool for it because it's simple and fast. I propose a Slope Party the day this gets fixed! :)
What about fav'ing this request? It'll be easier to see which request has the most supporters.
my only problem is with lowpass
Right now, you can use the low pass filter of Curve. It goes up to 20 kHz and also has adjustable slope and Q.
Valid workaround but kind of a big hammer for a small nail :)
Slope has less impact on the performance.
Is it still heavy with all other filters disabled and spectrum animation disabled? Definitely takes up a lot of screen space though.
I always wanted to look into this and guess what? You're right :)
I assumed Curve would use another method (FIR) than slope but they're both identical (IIR). Curve has a little more overhead but not in orders of magnitude as I expected.
Another observation make be believe the range for slope cannot be fixed easily (would break existing tracks).
I'll have to get the approval from the elders of audiotool but I believe this will be a "won't fix" with a recommendation to use Curve as a replacement when you wand to use the extended frequency range.
As another option, the Slope steepness can be increased by stacking them in series and syncing the changes with automations.
About the compatibility problem of fixing the Slope, the correct behaviour could be implemented like it has been already in the Pulverisateur (band-limited oscillators) and the Reverb pedal (better Freeverb algorithm): with a context-menu option to use the old behaviour. If that solution has worked in those devices, it would equally work in this one.
I've been recently using the BandSplitter pedal to apply simple filtering to sounds and avoid the Slope problem. You can't have resonance, but it's very easy to apply quick high or low, even mid, pass filtering to sounds without having to use a whole Curve eq.
i love you
Introducing new effect "Onepass"
If the slope gets a good overhaul, it would be the perfect low-CPU low pass / high pass filter.
Curve internally uses the same algorithm with fewer restrictions. AFAIK it's as efficient as Slope when disabling all other filters.
But yes, it's bulkier.
Huh. So using curves with only one filter activated and no spectrogram is just as efficient? Cool
According to the implementation: yes - I didn't benchmark it, though.
Bump.
I’m guessing this was rejected because the Curve is now a thing
I still wish they'd fix it.