Print

Please login or register.
1 Hour 1 Day 1 Week 1 Month Forever
Login with username, password and session length

[Mark Wilkinson]

Not really.  You're already hearing sources from the stage that are at least 8ms behind the PA.  Figure out what your crossover/processing "through" latency is.  It's your baseline for "how far upstage is the PA in TIME."  If you're a Mixerperson who likes to delay the FF and PA "to the back line", you are probably going to add additional delay on top of your 8ms DSP/xover latency.  You might even end up delaying the subs...

The times i've tried FIR live, 15ms total latency seemed to work fine.  Between stage depth, mains out to front a bit, and folded horn subs, I think usually I probably could have gone a little higher.

Hey, when I'm just eyeballing distances, trying to get a sense of the "upstage PA TIME" as you called it,  I've been looking to the drum set distance vs the PA distance (usually to FOH desk), thinking the sharpest transients would be the most audible time differences.  Does this make sense to you?

[Mark Wilkinson]

New info! Cool. I will have to look into that more. Granted 500hz is not of real interest for our most common needs, it would seem then that given a Xilica system ( I.E. not common to the average guy ) it would still require about 8ms to deal with common crossover points. While 8ms isn't very long it would be an audible amount of delay. Fine for playback, but a tough sell for small scale live audio.

My take is FIR works great for any crossover once above sub to mains. 
And IF and only IF you have co-located subs, either ground stacked or flown, it can help with the sub to main crossover. (and you can bear the latency)
In the ground stacked case, I've used it with 15ms delay successfully with no audible defect, at least as far as nobody has ever noticed/said anything other than sounds great

But the more I've played with FIR, I believe it's predominantly a tool for the speaker designer...and an incredible tool at that. Maybe for advanced install too.
For live, other than beam steering, venue matching stuff like MLA or Anya etc, I just don't see alot of applicability.

And honestly, even for live and normal alignment, the whole sub to mains alignment stuff is most often bunk...unless of course they are truly co-located.

The more measurements i take, the more convinced i get that for live, it's very seldom about optimization, and far more likely about identifying problems, or simply directing sound and evening out coverage and delays.

Anyway, sorry to swerve and rant a little ...

Back to FIR....from all i see, the biggest use of FIR in speaker design is about making EQ adjustments that are just too difficult other wise.
EQ adjustments that also fix driver phase.
And they don't take much latency.  Below is an example of a filter used by a major player that has just 384 taps.
You can see the degree of equalization possible, and how it is finer, sharper, higher in frequency and gets smoother in effect down low...as per the tap chart would indicate.
Also note the FIR delay is only 0.708ms, as opposed to 4ms if the filter were being used in straight linear phase mode.
I think this kind of tuning....IIR EQ's embedded in the FIR file, is by far and away the major rational behind the FIR craze with speakers right now.

[Frank Koenig]

Back to FIR....from all i see, the biggest use of FIR in speaker design is about making EQ adjustments that are just too difficult other wise.
EQ adjustments that also fix driver phase.
And they don't take much latency.  Below is an example of a filter used by a major player that has just 384 taps.
You can see the degree of equalization possible, and how it is finer, sharper, higher in frequency and gets smoother in effect down low...as per the tap chart would indicate.

This is my observation as well. Furthermore, correcting just the minimum-phase part with a causal FIR filter, which has essentially zero processing delay, gets you most, if not all, of the way there in terms of sound quality on music. This is very useful for flattening monitor speakers. (I haven't needed to "ring out" a monitor in a long time.) My current view is that correcting excess phase (the non-minimum-phase part) is most useful for getting disparate systems to play better together, but has limited value when working with a single species of speaker.

--Frank

[Luke Geis]

I used to do the whole Align backline to the mains bit, but have stopped as it just seemed like a futile task. Now let me say however that I am a sneaky bastard and have used the delay to align things in order to get a particular result. Say a bassist that is too loud and has a huge 80hz EQ bump from either room modes, or just plain being the guy that thinks everything on the block should rattle when he plays. I have used delay on that channel until I have found the point where it begins to reduce at the frequency of interest, or after inverting his phase and delaying until I get a good reduction. My main reason for not delaying the backline to the mains is that it would only ever be right at a very few numbers of locations. While perhaps it would be more correct elsewhere, it was just a bit much more work to pull off unless there was a clear and evident need for it. Bass guitar, as I mentioned, is one such reason.

Given my understanding of the way FIR used to be, it seemed to me that the use of them was limited to a 2-way system or simply beam steering in an array. The latency that I have heard of and in some instances can actually quantify, was to me, the reason why it was not commonplace in more speakers yet. It sounds however that there are new ways of implementing it and reducing latency to negligible amounts making it more viable for the masses. I do find and see its value as a common tool, I guess it just needed some more time to mature. I quit following it more deeply about 5 years ago as it seemed that at the time it was limited. Xilica was leading the pace then as a hardware tool and it also seems that around 2015 is when the technology took real flight. So go figure, I missed the boat on that one. The Eclipse Audio link I provided seems to show more of what an FIR filter would need to be in order to wholly replace an IIR filter. This particular article has been around for more than it suggests, it says copyright 2019 on the PDF, but I remember seeing this several years ago. In either case, this Eclipse Article was/is the most comprehensive one I had seen for the interests in FIR that I have and it does show significantly longer latency than Xilica seems to have figured out. I am not a mathematician, so I can only regurgitate what I have read and understand.

My experience with speakers using FIR is that they tended to be more linear and neutral sounding, were less prone to feedback and more correct sounding at any given listening angle

[Tim McCulloch]

I used to do the whole Align backline to the mains bit, but have stopped as it just seemed like a futile task. Now let me say however that I am a sneaky bastard and have used the delay to align things in order to get a particular result. Say a bassist that is too loud and has a huge 80hz EQ bump from either room modes, or just plain being the guy that thinks everything on the block should rattle when he plays. I have used delay on that channel until I have found the point where it begins to reduce at the frequency of interest, or after inverting his phase and delaying until I get a good reduction. My main reason for not delaying the backline to the mains is that it would only ever be right at a very few numbers of locations. While perhaps it would be more correct elsewhere, it was just a bit much more work to pull off unless there was a clear and evident need for it. Bass guitar, as I mentioned, is one such reason.

Given my understanding of the way FIR used to be, it seemed to me that the use of them was limited to a 2-way system or simply beam steering in an array. The latency that I have heard of and in some instances can actually quantify, was to me, the reason why it was not commonplace in more speakers yet. It sounds however that there are new ways of implementing it and reducing latency to negligible amounts making it more viable for the masses. I do find and see its value as a common tool, I guess it just needed some more time to mature. I quit following it more deeply about 5 years ago as it seemed that at the time it was limited. Xilica was leading the pace then as a hardware tool and it also seems that around 2015 is when the technology took real flight. So go figure, I missed the boat on that one. The Eclipse Audio link I provided seems to show more of what an FIR filter would need to be in order to wholly replace an IIR filter. This particular article has been around for more than it suggests, it says copyright 2019 on the PDF, but I remember seeing this several years ago. In either case, this Eclipse Article was/is the most comprehensive one I had seen for the interests in FIR that I have and it does show significantly longer latency than Xilica seems to have figured out. I am not a mathematician, so I can only regurgitate what I have read and understand.

My experience with speakers using FIR is that they tended to be more linear and neutral sounding, were less prone to feedback and more correct sounding at any given listening angle

You can't delay the backline to the mains, the back line is already upstage of the PA 99% of the time.  You delay the L/R mains, FF and subs to back line... or is the backline processed through Uli's new 'Ultra-Speeder Upper 2000 Pro"?

[Luke Geis]

I did get that backward, but you know what I meant

[Frank Koenig]

The fundamentals of digital filters have been well understood for at least 70 years. Norbert Weiner published "Extrapolation, Interpolation and Smoothing of Stationary Time Series" in 1942. What has changed is the widespread use of the Fast Fourier Transform algorithm which allows computing the Discrete Fourier Transform in N log N time (Cooley and Tukey, 1965), and the availability of more capable computers, digital signal processors and A-D-A convertors. Wiki says that many of the ideas actually go back to Carl Gauss in about 1805.

Sadly, I think many misconceptions come from the sort of turgid, chatty and logically impoverished articles we see in the trade rags that attempt to inform with lots of loose analogies and without any bad, scary math. Manufacturers' glossy claims further muddy the water: "Now with FIR...". The trouble is that while many of the concepts are not difficult (and, admittedly, many are), some amount effort and mathematical rigor is required to gain an understanding.

I wish I knew of a good, concise reference to point people to that lays out the logical framework of this subject while being easier to crack than Oppenheim and Schafer*, for example. Something written more at the level of "The Radio Amateur's Handbook" or "The Art of Electronics". At the very least such a course of study would give us a precise, common language. A table of contents (which I partially plagiarized from Oppenheim and Schafer) might very roughly look like this:

Signals and systems
Linearity
Time invariance
Causality
Impulse response and convolution
Frequency domain representation
   complex numbers
Sampling and reconstruction
Discrete Fourier Transform
   symmetry relationships
Poles, zeros and frequency response (brushing lightly past the Laplace and Z transforms)
   magnitude, phase, delay
   minimum-phase, Hilbert transform, non-minimum phase, excess phase, all-pass systems
IIR (recursive) filters
   properties
   relation to analog filters
   synthesis (briefly)
FIR (non-recursive) filters
   properties
   window functions
   synthesis and optimization for typical audio applications

This is something AES could take up, too, offering some short courses at their conventions in place of yet another session on how to make recordings in your bedroom. If I were a better, more secure educator I might take this on myself. Anyone want to give it a shot? I'll supply the whiteboard and the beer.

*"Digital Signal Processing", Oppenheim and Schafer, Prentice Hall, 1975

--Frank

[ProSoundWeb Community]