High-pass digital filter with knee at 30Hz will have delays substantially longer than 1ms.

 

I can show the change in source sound image quality by only changing speaker delays on my cars 5.1 system. (poorly because of the limited range and channels of the video)
...

I played with multitrack home studio recording for a lot of years and am pretty familiar with audio delays and their effects, from short ER delays and phase effects etc through to echo type delays.

High-pass digital filter with knee at 30Hz will have delays substantially longer than 1ms.

Why? If an analogue filter can do it in real time with RC why can't that be modelled in real time (sample by sample) with DSP?

 

Why? If an analogue filter can do it in real time with RC why can't that be modelled in real time (sample by sample) with DSP?

I'm not a DSP expert but I think it's mainly because analog signals are a superposition of vector/phasor energy em states with infinite levels and the RC filter is a energy storage (so it's not realtime either) device of infinite vector levels.

If the DSP could capture the vector state of the energy directly instead of a scalar quantization reading of the instantaneous voltage value it could directly model the RC filter with a arbitrary waveform instead of gaining the required vector group information by time sampling the waveform.

 

Why? If an analogue filter can do it in real time with RC why can't that be modelled in real time (sample by sample) with DSP?

Analog filter cannot do it real time neither. When you put a signal through filter, you get phase shift. You can look at the phase shift as a delay. 90 degree phase shift is 8ms at 30Hz. This is aready way longer than 1ms.

FIR filter can emulate this behaviour, but high quality FIR filter tries to be more accurate, so it has to wait until it gets enough samples to do its calculations, which produces delays longer than necesssary minimum.

I think the fiters they have now are much more advanced than simple FIR, but I would guess the design critera would be to maximize quality, not to minimize delays. So, we should expect that these filters should produce more delays than primitive FIR filters.

 

If the DSP could capture the vector state of the energy directly instead of a scalar quantization reading of the instantaneous voltage value it could directly model the RC filter with a arbitrary waveform instead of gaining the required vector group information by time sampling the waveform.

Yeah that's pretty much how I saw it. A simple algortihm only needs the previous sample and the current sample (involves zero delay) to get the rate of change or "energy" then modify the output rate of change accordingly. Letting high rate of change through untouched while reducing low rate of change to zero (or close to zero) would model a high pass filter and do so with no delay at all (per sample). You would probably also need to add a small centreing bias too, to keep the output waveform centred after reducing or removing the LF component.

Anyway it could be done simply, with no "delay" caused by the digital processing over what would normally happen in the RC analogue filter.

Analog filter cannot do it real time neither. When you put a signal through filter, you get phase shift. You can look at the phase shift as a delay. 90 degree phase shift is 8ms at 30Hz. This is aready way longer than 1ms.
...

Thanks for clarifying, and yes I already got that.

But the point was "delays caused by digital", and if the digital filter behaves in the same way as the analogue filter and they both give the same phase shift etc then the digital system is not introducing any delay larger than the analogue system. So there's no digital delay, which i thought justified saying the digital filter was operating in "real time" like the analogue one..