Premier Rock Forum

projectMalamute wrote:People didn't seem to be picking up on it.Story of my fucking life. Say some simple thing and 28 jagoffs go in circles guessing you're full of shit until they step on the rake you were telling them about.I kid.

Stinky Pete wrote:Okay, so the issue is the feedback's latency?Yes. The feedback being the only significant feature of this technique, which would otherwise be "boosting the bass a little."

tmoneygetpaid wrote:steve wrote:Stinky Pete wrote:Okay, so the issue is the feedback's latency?Yes. The feedback being the only significant feature of this technique, which would otherwise be "boosting the bass a little."Right, but they are discussing whether each feedback needs to have a 1-sample delay, and it seems the consensus is there will need to be. I said N samples, but theoretically there might be a way to do the math fast enough that N=1.Stinky Pete says the time constants of the electronic components and slew rates and what not of the analog components in the EQ and console mean that there is a non-zero delay in the analog version of this trick, as well.Right. I mean, I said it first but... as I mentioned, those time constants are on the order of a small phase shift, not an audible delay.

projectMalamute wrote:The way to do this on a digital system is with an EQ with some sort of controllable resonance built in that is simulating the same process.Can I point out that I said this at the outset? ctrl-F for "moog filter" if you like.

projectMalamute wrote:People didn't seem to be picking up on it.Story of my fucking life. Say some simple thing and 28 jagoffs go in circles guessing you're full of shit until they step on the rake you were telling them about.I kid.

projectMalamute wrote:Your code is just running it through the EQ one time and then scaling it. There is no feedback going on.No, because once you assign a portion of the last sample output by the program to the input you will have a feedback loop. You simply scale it down to less than the full output to stop it being a complete mess. Assigning a unique ID to every output sample and then adding that back for all time would definitely be the hard way.Think of a more conceptually simple plugin like a delay. How do you add feedback? You take the output of the buffer and you feed it back to the input of the buffer, and the percentage of feedback is determined by the factor you multiply it by. The data is stored in a circular buffer and you just read and write around it. Why, on a conceptual level at least, would it be different here?I've my Audio DSP book at the University. I'll grab it and stick up the C++ for a working BiQuad with feedback if you're interested.

projectMalamute wrote:If you add it to the next sample you've made yourself a one pole filter. Which is exactly the kind of delay based artifact that started this whole conversation. Look at the code for the biquad you offered to post. You will find no zero delay feedback, what you'll find is that the previous output sample (z^-1) and the one before that (z^-2) are what are being fed back in to the filter.Okay, so the issue is the feedback's latency? Earlier you were saying that there was no feedback, which was my confusion. Yeah that's an unfortunate design limitation that the output has to be calculated first. It definitely does effect filter design in certain situations, and correcting for it can be tough. Stuff like shelving filters behaving as LP as they approach the Nyquist and loss of "bell" symmetry in boosts near the Nyquist are tricky problems to overcome.But would it effect the resulting output of this filter? I don't think so. Perhaps it would, perhaps some oversampling would be required to minimize the artifacts enough to render them inaudible. The delay artifact I believe Steve was referring to would have been the bussed signal and the direct channel, which we already said can be synchronized but I think there's several "Resonator" plugins on market that could already be done.

steve wrote:Stinky Pete wrote:Okay, so the issue is the feedback's latency?Yes. The feedback being the only significant feature of this technique, which would otherwise be "boosting the bass a little."Right, but they are discussing whether each feedback needs to have a 1-sample delay, and it seems the consensus is there will need to be. At question is whether this 1-sample will actually impact the effectiveness of this trick, as Stinky Pete says the time constants of the electronic components and slew rates and what not of the analog components in the EQ and console mean that there is a non-zero delay in the analog version of this trick, as well.Does anyone know the delay time that routing a track in the analog domain in this way would create? And how would that compare to 1 sample delay at 48k or 96k?As an aside, I did try this trick in a DAW last night amidst some mixing. The results were... a bit confounding.First, to check how the DAW was behaving with audio bussed back in on itself in this feedback loop scenario, I setup a track with a single kick hit. Also setup a click track set to 120. I bussed this to a bus at unity gain, and created an aux with input set to this bus, output to stereo. And on the aux I also created a send back to itself at unity gain. On this aux I put a quarter-note delay via the good old Pro Tools Extra Long Delay. My thinking is that if the latency is zero and delay compensation and such are working correctly, the delay would stay perfectly in sync with the click.When I used the "native" version, it did not track to the click, it phased in and out of time.When I used the "DSP" version, it tracked to the click.The native version did add considerable delay for processing- about 1000 samples or so.Of course, with a delay you can just add feedback to the feedback control, and it tracks perfectly to the tempo. So it occurred to me that you could use the Soundtoys rack with a single instance of their Sie-Q plugin to do this trick. The rack itself has a feedback control, so that with whatever set of effects you setup in the rack you can create a controlled feedback to the effects. I didn't try it with the Soundtoys effect rack, but I did try it in the DAW with the DSP version of the basic Pro Tools plugin and A/B'ed it against a simple boost. Once level matched I was surprised by the difference in sound.I had the original track's output to the EQ bus set to unity, and the feedback send set to -10dB. I think I boosted around 10dB at the EQ. I don't really know if this compares to how you would set it up on a desk.I'm happy to upload the level-matched results if anyone is curious.

Do you mean to say that the one-sample delay actually disallows the feedback? Or that the one-sample delay is compounded?Also, I'm unclear on how a resonance control would simulate this. Doesn't a resonance just narrow the Q of a filter/ EQ band? Does it actually commonly use some feedback, as well?To the point of "this is an analog trick..." Even if I don't use this trick, the bus feedback concept has other uses and implications. I will somewhat regularly create a delay that feeds back on itself this way with a filter on it so that successive iterations are increasingly filtered, or I will do the same with a distortion. Understanding this, I know that if I want the delays to stick to the song's tempo, feeding the bus back in on itself is limited in viability.

Thanks for your replies. It's good to hear from someone who understands this stuff, even in broad strokes, at the level of the underlying math and code. Same to you, StinkyPete.How did you get into this stuff? Are there resources you'd recommend?