eliya wrote:FWIW, most if not all DAWs have delay compensation, so there's no delay if you're not running the signal out of the computer and back in there should be no delay.Yeah, that works if you just run the signal through an equalizer. Here you're feeding the equalizer as a parallel channel, then feeding that return back through the equalizer. There's no avoiding the processing time for the equalizer algorithm when the signal cycles through it multiple times as feedback.Say the latency of the equalizer at a certain setting is N samples. If you apply the equalizer to a channel, that channel will be offset N samples to compensate for the latency. If you're feeding it from a bus, then you can't offset the channel feeding the bus because it isn't a through-put, and that would put the channel out of sync with the rest of the session by N samples. So the return has a delay of N samples.You then feed that return back through the same bus, so the feedback has an additional N samples of delay, and additional delay on the feedback of that signal coming back into the parallel channel and so forth, regenerating the delay more and more with more feedback.If there were a way to offset the entire session to compensate for the latency of the equalizer, allowing you to offset the send by N samples, the feedback would still incorporate the processing delay, because it can't be offset -- you can't move forward in time a signal that hasn't yet been generated -- so the parallel equalized channel could conceivably be latency compensated, but the feedback cannot.Yes, I have thought about this a lot.

steve wrote:Exactly. If you regenerate the audio through the EQ you increase this effect beyond the design of the hardware unit, tightening the Q to be precisely tuneable and verging on oscillation, and that's the auxiliary signal you mix with the "normal" bass drum. The reason you do it in parallel is to leave the original bass drum track intact, while adding a small amount of the ringing resonance. If the audio is run through the equalizer with high feedback near the point of oscillation, the effect will be much more drastic than the same signal being mixed with the unmolested bass drum to taste.Sorry, I was confusing in what I'd meant by insert. In software conventions inserts now commonly have parallel/blend options built in, so you'd have the choice of keeping it all to a single channel. Of course in the case you described there we would be straying from the original intent.projectMalamute wrote:What Steve is describing involves the signal going from the output and back to the input a theoretically infinite number of times instantly. You can't do this in a DAW. What you can do is calculate what's happening and implement it another way, but you can't just patch the output to the input.Of course you can, take this psuedo-code for example:Output X = (whatever the calculation to perform the EQ).Input = N + (X*0.8 [or whatever the feedback gain value would be to get the desired result]It doesn't actively have to compute the feedback, it just occurs. Now it will take a couple of samples to begin oscillation but that rise time would also be something you'd see in the analogue equivalent. X and N would both have to be zero'd before the calculation could begin, but asides from that there's very little to go wrong. It's not instant in the speed of light sense of electrons, but neither are the time constants involved in the equivalent analogue circuit. Perceptually, I'd find it hard to argue the difference in the effect. Now there may be other non-linearities at play in the analogue system that might augment this in some way what I described would lack, but I'm not in a position to argue them. Outside of doing it within the plugin itself, a lot of DAWs also allow feedback loops now, though you usually have to tick a special box in the preferences. I do it in Reaper sometimes for certain things like inserting effects in the feedback loop of a delay.

The Q of most parametric EQ is set largely through an internal feedback loop. An IIR filter design done in the digital domain should be capable of the same effect, it just needs the Q to be brought to the point of near self oscillation. The delayed samples are fed back to the input through a particular delay network to arrive at the correct level of gain and the correct center frequency, it's entirely the same as an electronic circuit of the same concept but without some of the non-linear aspects. Generally it can be run in parallel, though there's no real need as it could be done as an insert in most DAWs and automated in. Most contemporary DAWs - Ableton, Reaper, Logic by default delay all the channels until the one with the longest latency is ready. PT's usen't to, it may now that it has gone native, I haven't used it in years. I think it's a part of the ASIO standard, it's certainly how the whole concept of buffering is supposed to work.For those about to whip out their Fab Filter 9000 Awesome EQ for this: "Linear Phase" designs where the time domain information is smeared by use of both a feed-forward and feedback network will not do the same thing at high Q's, it will just destroy the sound of the d. As an interesting aside, you should basically always avoid using them for low frequency adjustments because of the massive amounts of pre-ring it would introduce. The feed-forward system needs to be short in order for it to be transparent. I just avoid FIR filters, and increasingly audio designers are moving that way too.

I am actually unclear on this, as something I just read suggested that audio plugins manipulate the audio that is in the buffer, and so they shouldn't in themselves add any additional delay.But in this case I'm unclear on what the need is for delay compensation.

steve wrote:tmoneygetpaid wrote:Worth noting that HD Pro Tools has plugins that have zero samples delay, so you could use a TDM/ AAX DSP plugin for this purpose.This is of course impossible, so the language "zero samples delay" means they are compensating for the algorithmic processing time on the effected signals or the session as a whole.Which is what I was talking about.All audio is delayed by the buffer amount.Plugins access the audio stored in the buffer and manipulate it before it is "queued" in the digital audio stream.So yes, all the audio is delayed in some sense because of the existence of a buffer, but not necessarily with respect to other audio.

So are you acknowledging now that a plugin that doesn't create latency is possible?Bussing does not cause latency/ delay either, so where would this delay be coming from?

Couple questions, and pardon my ignorance:1) This could be done with digital as long as you were able to send to, say a little mixer with busses + some kind of sweepable EQ, then return back to the box? As long as the bussing, EQ and feedback are done in the analog domain, that's the critical element?2) Seems like outboard EQ is used in the examples I see. I assume the reason for that is even if your board has sweepable EQ, having control over the bandwidth is important?Thanks

You're not wrong. I do want to pipe in and say that noisicians have been doing stuff like this for a while, using matrix mixers and regular mixer effect loops. In those cases it wasn't just to make a kick sound more resonant, but that does happen when you run a drum machine or samples through it. Feedback looping mixers are some people's main instrument.

Exhales in a weary groan before we get into yet another a v d debate...For reals though: Yes, the two differ and not every technique can be transferred from one to the other. I really like that this thread has so far gone into the technical part of things and would be very glad if it kept going that way.

all of this is why i'm always left wondering - what is so inherently superior about digital systems?and, yes, i can hear the groans of the a v. d weary. my experience with just about everything digital, is " well, can i do this thing that i do already?" "of course! easily! no problem!, it will simply involve this simple work around, and a little tweaking of your system preferences, and while we're waiting for the system to reboot, after freezing, i'll explain that the additional steps are actually time saving and that the inadequacies of the outcome are superior."