Question regarding DSP usage

[feffa86]

I stumbled over something I thought was weird. I am running my Helix with a Kemper. I use Path 1 as "before amp" with mono effects and path 2 as "after amp" with stereo effects. I've been messing around with a w/d/w setup just for fun, and so I ended up hitting the DSP roof on path 2 a lot. One thing I stumbled over was that I am running a parametric EQ right after the Kemper. For high and low cuts if nothing else. I tried replacing it with the simple "High/Low Cut" effect, and it turns out I couldn't. The simple high/low cut block uses more DSP power than the parametric equalizer block, even though the latter includes a high and low cut as well as many other features. That seemed to make little sense to me, and led me to two questions:

 

1) has anyone done any research and writeups about how much DSP power the different blocks use, or which ones use the most or least, or does anyone have any tricks or weird things they stumbled across regarding DSP power, like the high/low cut vs. parametric equalizer thing?

2)  would it somehow be "better", "more accurate" or in any way preferable to use the high/low cut block over the high and low cut in the parametric eq block?

 

If you can bare with me, lol

[willjrock]

It seems normal that the simple EQ would use more DSP than hi/low. More filters so logic tells me it should just for that reason alone. Also "simple" uses shelving filters on the hi and low side so that could more require computing possibly. 

 

Boy, if you are not using amps or preamps in the Helix (assuming youre doing that part in the Kemper) then running out of DSP in any case should require a bit of effort.

 

1)Yes. Ive seen various mentions on the web. Some comparisons made. Wish i could point you somewhere, but i know theres some useable stuff out there.

 

2) Only for DSP benefits.

[feffa86]

Oh, I actually mean it's the other way around to how you put it. The simple EQ, which has many features, uses less DSP than the simpler high/low cut, which has fewer features. I take that as meaning, the way that the high/low cut block does the high/low cut is somehow a larger operation for the processor. I then take that as meaning "more calculations involved", which leads me to "closer to real life results" Like how a 320kbps MP3 has more information about the sound it contains than a 192kbps MP3. Inaudible, possibly, but "more advanced" at the core of it nonetheless. Another possible scenario would be "bad coding", so I guess that's why I'm getting hung up on it. If it's the latter, it's something to look into, or if it's supposed to be this way, I'm still curious :P

 

I do have some DSP power left in my patches, but on path 1 which sums to mono when it exits into the Kemper and back to path 2. That makes path 1 "unusable" for stereo effects if I want them to actually be in stereo. So I have a lot of mono effects and probably lots of free DSP on path 1, then usually have to make choices on path 2. Again, to keep it stereo, every EQ, compressor, etc. that I put after the Kemper (which I have to, because that's what colors the sound) have to be in stereo. I seem to max out with only around 4-5 effects. Reverb, delay, looper, Double Take and a chorus are usually standard for all my patches. That leaves me one more effect. I mean this is all for living room jamming so none of this really matters anyway, lol, but there seems to be something weird going on with the high/low cut block and I would love to understand what's up and/or find more effective ways to build patches :D

 

[DunedinDragon]

I think it would be a huge leap to attribute it to bad coding since bad coding in the modeling world is more about introducing latency as it is with DSP usage.  My guess is the complexity lies in the algorithm used to model the slope of the hi and low cuts since they aren't typically linear and are generally parametric in the real world which is what they're modeling.  That would mean more advanced math than simple scalar functions.

 

It doesn't surprise me at all that you're running short on DSP in path2 given that's where you have some of the most DSP intensive effects...particularly the reverb and delay.  That's exactly why I reserve path 2 for such things so I have more room to play in path 1.

[feffa86]

Let it be said that this is not an important issue to me. It's, however, significantly a "small" issue that it's likely to never be adressed. I get that. That's particularly why I'm making a thread about it. It's so random that I even noticed.

 

If it's what you say about the calculations of the slope, it still doesn't make any sense. Listen:

 

1) parametric equalizer with five controllers (3 Qs and two slopes for high/low)

2 ) high cut equalizer with two controlles (0 Qs, two slopes for high/low)

 

They must both have a curve calculation for the high and low cut, right? They are both digital and must both be coded. Even still, the one with only two calculations uses more computing power than the one with five calculations. Even though the two calculations must be the same or very similar in both effects, and the other has three extra calculations on top of that.

 

Listen, it's not important to me to have this "fixed" or anything, but it is weird unless I am missing something obvious. At this point, all ll I want is for someone else to see it and agree that it's weird, lol. It's not important. But it also makes no sense unless it's the fact that it's a more "precise" slope than the equalizer has. That then opens a new set of questions, like what the actual difference is. If one leaves in more "upper lows" than the other, that's the one I'd prefer.

[reginaldStjohn]

I agree it seems counter-intuitive.

[phil_m]

According to this, the Parametric EQ and High/Low Cut models have the same DSP load. These values are only estimates, though, and the DSP use of blocks might change slightly over time as Line 6 is always optimizing things.

 

If you're hitting the DSP limit by adding an EQ block, you must be very close to begin with, as EQ uses the least amount of DSP of any of the effects, except the Volume and Gain blocks. There are different way to write DSP for filters, so it could just be the blocks are using slightly different methods. Or it could just be something related to how the Helix is rounding off the DSP values for different things.

[jbuhajla]

Stereo effect blocks use much more DSP than mono versions.