ProSoundWeb Community

Sound Reinforcement - Forums for Live Sound Professionals - Your Displayed Name Must Be Your Real Full Name To Post In The Live Sound Forums => Audio Measurement and Testing => Topic started by: John Neil on May 13, 2012, 10:48:16 am

Title: Side Chain Limiting
Post by: John Neil on May 13, 2012, 10:48:16 am

X-Frame 88 Preset schematic:

(http://soundscapesweb.com/files/PSW/MeasureForum/X-Frame88MW12.png)

Langston,

Pardon my n00b showing, but why the sidechained limiter?
Title: Re: Side Chain Limiting
Post by: Langston Holland on May 13, 2012, 09:04:38 pm
Hi John:

I split your post off from that other thread and renamed it as seemed appropriate...

Your question is anything but newbie level (I hope) because I've put many hours into finding answers to it so far.

The reason I bought a free wire processor was to experiment with some ideas I've had for a while concerning side chain limiting. Initially, I was simply interested in playing around with a variant of feed forward limiting commonly called look ahead. After spending some time on that, a flood of other ideas came that have completely changed the way I control level dependent loudspeaker behavior. There are at least a dozen significant advantages to side chain limiting that have convinced me that the way limiting is currently employed in even the best live sound processors is a huge compromise. All of this applies to the more gentle limiters that we call compressors as well, but that's another topic.

I'm just going to point out one advantage in this post because it would take 50+ pages to explain everything I've stumbled upon to date. Charlie Hughes is responsible for striking the match on this with one of his (typical) extremely well written articles and his (typical) amazingly generous personal help.

Using Limiters to Help Protect Loudspeakers (http://www.excelsior-audio.com/Publications/Using_Limiters_to_Help_Protect_Loudspeakers.pdf)

BTW, I reject the "limiter as protection mechanism only" philosophy. I'm absolutely convinced that limiting can be a very large part of why a loudspeaker system does or does not sound good at high outputs and low outputs where the limiter is not even engaged! How can this be true? It's part of the one advantage that will be explained below. :)

I will use a Danley TH118 sub driven by a large amp with 32dB gain to demonstrate. The stimulus used in these measurements is a sine wave burst of 4 wavelengths at 1/24th octave intervals. This sub with this amp can handle a maximum of +11dBu RMS at the output of the processor in the middle of its passband before it starts to sound ugly. It can handle +2dBu at 20Hz before woofer slap (aka ugly). The following plot shows how standard limiting behaves once the threshold is reached and exceeded to the point of clipping the input of the processor. A 30Hz 4th order BW high pass is required in combination with the limiting to keep it sounding good at max output:

TH118 with Standard Limiting:

(http://soundscapesweb.com/files/PSW/MeasureForum/TH118_Limiting_Std.png)

Now I'm going to do something stupid and change the high pass filter to 1st order and move it one octave lower. And experience zero woofer slap or other ugliness. And get a stunning increase in VLF output at levels prior to limiter engagement where the subs are typically run (enough rig for the gig). And get significantly better phase behavior, thus a tighter impulse, thus better impact. And get better protection of the sub. All due to the power of side chain limiting.

OK, I lied. I've mentioned more than one advantage, but there are many others for each end of each passband.

TH118 with Side Chain Limiting:

(http://soundscapesweb.com/files/PSW/MeasureForum/TH118_Limiting_v2.png)


TH118 Composite Standard Limiting vs. Side Chain Limiting:

(http://soundscapesweb.com/files/PSW/MeasureForum/TH118_Limiting_Std_vs_v2.png)
Title: Re: Side Chain Limiting
Post by: Ivan Beaver on May 13, 2012, 09:43:14 pm
This is yet another example of why it is so hard to answer a complicated question with a "simple" answer.

That simple answer will be wrong.

The more you dig into a subject-the mroe complicated it gets.

There is A LOT mroe than just "wattage" that goes into limiter settings.  The freq is a huge part of it.  Overexcursion (as in this example) can be a real problem.

At higher power levels-the response (and output) does not stay linear.  This has to be accounted for when setting limiters.

Of course it gets a bit harder-because limiters with sidechains are not very common.  Thus complicating the whole issue and what you can do about it.

And if youa re talking about a full range product-what the woofer can take (sine wave) and what the HF/horn can take (sine wave) are very different.  Feedback really doesn't care-and can occur at most any freq.  So how do you set a limiter that can protect against feedback on either the lower or higher freq?

Again a sidechain limiter is needed.

Of course just having the proper control/tools/limiter is only the BEGINNING of the procedure of how to actually set it. 

NEXT you avhe to know what the limits are-for the lower and higher freq.  Getting this data from the manufacturers is not easy-and must be done yourself.

Of course the attack times will vary for the lower and higher freq-along with some other variables (such as loss through the crossover-due to the actual rolloff action of the crossover itself).

Why does it just seem to get more complicated the more we learn?



Title: Re: Side Chain Limiting
Post by: Langston Holland on May 14, 2012, 10:52:38 am
Another advantage, this time looked at from the other end of the sub's passband:

A Beautiful Woman that can Cook (http://soundforums.net/junior-varsity/4081-limiters-good-enough-post28256.html)

Black and blue traces use an 80Hz 4th order LR low pass filter. Blue traces use a standard limiting keyed off the post-crossover audio signal, black traces use a side chain limiter keyed from the pre-crossover audio combined with a 2nd order LR low pass filter one octave above crossover.

There's an exciting world in limiting that is largely unexplored. :)
Title: Re: Side Chain Limiting
Post by: Arthur Skudra on May 14, 2012, 11:12:04 am
Hey Langston!  Charlie's article on this is fantastic as always.  What are you using/doing to determine the sidechain filter curve?  Let us in on your secret sauce!  ;)
Title: Re: Side Chain Limiting
Post by: John Roberts {JR} on May 14, 2012, 11:32:33 am


There's an exciting world in limiting that is largely unexplored. :)

Unexplored by end users but not exactly virgin territory for captive systems where the informed speaker design engineers get to twist the knobs to protect their own babies..

My world view is a future populated with "Steven Segal" (character) powered speakers that are truly "Hard to Kill".

I played around with designing a stand alone speaker protection black box and decided against it because the logistics of connecting the all the individual speaker specific data dots for end users was too much work for a small company, and such a product could be obsoleted by manufacturers getting their act together.

My preferred platform is active powered cabinets with DSP inside, but the driver protection scheme that chops the power feed (ala PWM) at each speaker, while relatively more expensive is promising too.

JR

PS: I vaguely recall a thread several years ago here that delved into the math for driver excursion protection, while that was more of a theory discussion than a how to do this with practical available gear.
Title: Side Chain Limiting
Post by: Langston Holland on May 14, 2012, 02:30:40 pm
Hi Arthur:

If I were into secrets, I wouldn't be postin' all this stuff. :)

Plus it would be evil - everything I know is a gift from someone else.

Like you, I have a subscription to the AES E-Library and have read everything I could find on loudspeaker protection. There's really two ways to approach this, one is over my head/budget/interest that uses destructive thermal and mechanical measurement of drivers and systems to find limiting thresholds, the other can be done in a field with cows (FWC).

Folks with true mastery of limiters like NEXO probably use both methods, i.e. extensive laboratory measurement and FWC (tuning by ear).

Charlie's article takes the logical approach of keeping the driver within Xmax, which is actually pretty easy and affordable to measure these days with either CLIO or ARTA and a laser attachment. But, the entirely objective procedure of not exceeding the rule of thumb excursion threshold (assume Xmax +10%) does not fully address the issue. This will keep the system within a reasonable distortion range, i.e. probably reasonably good sound as well as protecting the driver.

Why not just cut to the chase and use Don Keele's marvelous tone bursts to find where the driver in a specific system with a specific amp REALLY starts to sound bad? In my experience, this threshold is usually achieved at levels somewhat lower than Xmax, thus driver protection comfortably exceeds the rule of thumb formula.

I've found that CLIO's longest gated tone burst of 4 wavelengths under 150Hz or so starts to sound bad at the same point as the Keele tone bursts. This makes the process much easier and quicker, though the end result is identical. CLIO v10.31 now offers Keele's tone bursts in a full CEA-2010 compliant module for max SPL testing. Incredible.

Bottom line: set the side chain curve such that the sub isn't allowed to go where it sounds bad. Result: the best possible VLF from a given sub plus fame, fortune and beautiful bovines. What else is there?

(this is a Tapatalk test post)
Title: Side Chain Limiting
Post by: Langston Holland on May 14, 2012, 02:33:41 pm
Good points by JR as usual - the all-in-one loudspeaker has to be the future and it surprises me that more have not followed Meyer's lead on that. Based on the access level I've been granted to date, I'm not at all convinced that the practical limits (pun) of side chain limiting have been addressed by any of the mfg's or any available research. They certainly don't appear to have left a paper trail (an ancient form of encoding that used processed trees as a substrate upon which various pigments were applied that one would physically disseminate for decoding by the recipient).

As an example; the 1st order high pass filter I used for the TH118 was simply due to the fact that the 10 year old processor I used as a learning tool could not fully implement the reciprocal response I wanted in the side chain. A newer free wire processor very probably can do this, which means NO HIGH PASS filter at all is needed. That will improve time domain response a bit more and is a far better solution than the dynamic EQ techniques (that screw up phase) currently employed by many mfg's to achieve variable LF extension.

Show me proof that someone has actually applied this and I'll quit bragging about my bovine aided research (BAR)!

Maybe I'll demonstrate how I managed to produce a look ahead limiter without inserting delay into the audio path. It was an accident that occurred at the BAR lab, rest assured that no animals were hurt in the process, though with time they will be eaten.

(2nd Tapatalk test)
Title: Re: Side Chain Limiting
Post by: John Roberts {JR} on May 14, 2012, 03:55:42 pm
Good points by JR as usual - the all-in-one loudspeaker has to be the future and it surprises me that more have not followed Meyer's lead on that. Based on the access level I've been granted to date, I'm not at all convinced that the practical limits (pun) of side chain limiting have been addressed by any of the mfg's or any available research. They certainly don't appear to have left a paper trail (an ancient form of encoding that used processed trees as a substrate upon which various pigments were applied that one would physically disseminate for decoding by the recipient).
paper... I remember that
Quote
As an example; the 1st order high pass filter I used for the TH118 was simply due to the fact that the 10 year old processor I used as a learning tool could not fully implement the reciprocal response I wanted in the side chain. A newer free wire processor very probably can do this, which means NO HIGH PASS filter at all is needed. That will improve time domain response a bit more and is a far better solution than the dynamic EQ techniques (that screw up phase) currently employed by many mfg's to achieve variable LF extension.
Indeed,,, more brain cells were sacrificed to making speakers sound good, than keeping them working, but then was then and now is now... I expect DSP to get so cheap it is silly not to use it in powered cabinets for intelligent protection.
Quote
Show me proof that someone has actually applied this and I'll quit bragging about my bovine aided research (BAR)!
I've been out of the trenches for 10 years, and I wasn't a speaker guy back then either...(as is probably obvious). 
Quote
Maybe I'll demonstrate how I managed to produce a look ahead limiter without inserting delay into the audio path. It was an accident that occurred at the BAR lab, rest assured that no animals were hurt in the process, though with time they will be eaten.

(2nd Tapatalk test)


God bless you and your precious family - Langston

Keep on keepin on...

JR

PS: There could be some rather short look ahead capability based on the delay inherent in typical digital filters, but using that gets a little involved since you often want to crunch data after it is filtered, but I am wandering here... never mind. 
Title: Re: Side Chain Limiting
Post by: Langston Holland on May 14, 2012, 08:48:52 pm
Quote from: JR
PS: There could be some rather short look ahead capability based on the delay inherent in typical digital filters, but using that gets a little involved since you often want to crunch data after it is filtered, but I am wandering here... never mind.

Never mind my eye! You don't realize how much of a great idea that is...

With a typical processor, that won't help of course because everything must be converted to digital to be usable. BUT, the future is FIR and FIR involves a second processing stage with significant delay from the perspective of the shorter wavelengths involved with practical latencies in live sound. If one could feed the output limiter's side chain prior to the FIR stage, you'd have more than enough look ahead delay for the mid and high frequencies. This would be huge and I'm trying to talk my favorite processor mfg into implementing it - just been so busy with work and personal stuff lately.

With the shorter wavelength limiters able to look ahead WITHOUT actually adding delay to the audio path due to the FIR stage, how do we obtain look ahead delay for the sub passband WITHOUT adding delay to its audio path? There will be no FIR involved here because the latency penalty is prohibitive, thus the fact that all live sound FIR based processing is hybrid - FIR on top, IIR down low.

For usable look ahead, it's nice to have at least 1/4 wavelength (the first peak of the sine) of the frequency of interest for the limiter to predict the worst of what's coming. Given that the kick is the primary threat to the sub and its fundamental is about 60Hz, a 1/4 wavelength is about 4ms. I found a way to get about 5ms for free - Shazam! :)

How?

The upper orange trace in the first plot in my earlier post shows the sub's high and low pass response ballooning outward under heavy limiting. The high pass balloon endangers the drivers and the low pass balloon creates high level, high distortion mud. This ballooning occurs because the limiter is keying on the post-crossover audio signal. As the filters properly reduce the signal above and below the sub's passband, this signal also reduces the action of the limiter above and below the passband. The ballooning has to happen for the limiter to see enough level to act upon.

To fix this, you could simply feed the limiter's side chain input with the full range audio prior to the crossover, but then you'd have the upper frequencies modulating the sub output - unacceptable.

What to do? Compromise. I've found that a 2nd order LR low pass at twice the sub's actual low pass frequency works very nicely. For a sub with an 80Hz 4th order LR low pass, the low pass to the limiter's side chain would be at 160Hz. There is a great deal less group delay with the filter feeding the limiter's side chain vs. the one passing audio to the sub and this is where the free look ahead delay comes from.

Thus we are able to get usable amounts of look ahead for all passbands with no additional delay inserted into the audio.

Measurement:

(http://soundscapesweb.com/files/PSW/MeasureForum/Group_Delay_Look_Ahead.png)
Title: Re: Side Chain Limiting
Post by: John Neil on May 15, 2012, 02:04:17 am
Langston,

Thanks for the in depth response.  I've been devoting some brain space to limiting lately due to a smoke-letting incident with one of my clients.  (How'd you know I was thinking about subs?)  The real issue there is an inadequacy in the transducer department but the most available tool is limiting.  Sigh.

I am having difficulty wrapping my head around the difference of pure tone bursts and program material...how each would uniquely impact the frequency domain for each of the three apparent tools (limiting compression, side chain limiting, and dynamic EQ). Is the additional VLF extension shown above worth driving the band pass into limit due to content that was previously a non-factor?

BSS also purports a "mid-band" method...
Title: Re: Side Chain Limiting
Post by: Langston Holland on May 15, 2012, 12:20:49 pm
Quote from: John
I've been devoting some brain space to limiting lately due to a smoke-letting incident with one of my clients.  (How'd you know I was thinking about subs?)  The real issue there is an inadequacy in the transducer department but the most available tool is limiting.  Sigh.

A few thoughts:

1. Make sure your limiters are set per mfg recommendation or at the amp's max rated output into that load, whichever is less.
2. If the high pass slope on the sub is less than 4th order (24dB per octave), change it to 4th order BW.
3. Raise the high pass frequency 5Hz and listen to the difference with the type of music they play. Leave it there if it's acceptable.

Quote
I am having difficulty wrapping my head around the difference of pure tone bursts and program material...how each would uniquely impact the frequency domain for each of the three apparent tools (limiting compression, side chain limiting, and dynamic EQ). Is the additional VLF extension shown above worth driving the band pass into limit due to content that was previously a non-factor?

None of this stuff I've talked about is worth it. Monetarily. Even worse is the fact that it's just a partial introduction. I do this for personal entertainment between sessions volunteering at the DMV as a crash dummy and shooting myself with tasers.

Quote
BSS also purports a "mid-band" method...

This is simply an attempt to find more appropriate time constants for attack and release settings. It results in more conservative (shorter) times. Like Ivan would say, that little comment would require quite a bit of time to unpack and explain. In itself, it's more of a tool for the marketing department than the end user.

Traditional time constant calculation:

Attack in msec = (1000/HP Frequency of Passband)/2
Release in msec = 16x attack

Mid band time constant calculation:

Attack in msec = (1000/Midpoint of HP and LP Frequencies of Passband)/2
Release in msec = 16x attack

Package the latter calculation in your processor and send it to marketing for inclusion in the brochure. While you're at it, add an equally simple calculation for notched filters, give it a fancy name, imply complexity that can't be achieved on other processors (false), interview somebody with a professional frown that says it's good (true) and send it to marketing.
Title: Re: Side Chain Limiting
Post by: Brandon Romanowski on May 22, 2012, 10:53:35 am
Would anyone like to explain how to measure excursion with an accelerometer ? Does anyone use this sidechain limiting commercially ?
Title: Re: Side Chain Limiting
Post by: Langston Holland on May 22, 2012, 02:05:53 pm
Quote from: Brandon
Would anyone like to explain how to measure excursion with an accelerometer ? Does anyone use this sidechain limiting commercially ?

Hi Brandon:

I think a laser is a more accurate way of measuring excursion and the nicest software implementation of this happens to be the least expensive as well; the STEPS module contained in ARTA.

See pp 45ff of the STEPS manual: http://www.artalabs.hr/download/STEPS-user-manual.pdf (http://www.artalabs.hr/download/STEPS-user-manual.pdf)

IMO, this kind of measurement is more useful to the loudspeaker designer than to us, but the pursuit of knowledge and toys is always encouraged. :) My take on this subject is that once distortion becomes objectionable, you need to limit the system so that it doesn't go there. This may coincide with Xmax or may be a bit higher or lower depending on the quality of components and skill (and/or luck) of the designer.

The best outline for "objectionable" testing was penned by the great Tom Danley 81 moons ago:

With a number of listeners there is another test one could conduct. If one had a group of listeners, (a good distance away so ears aren’t overloaded from excess loudness), each speaker / system could be cranked up with the same program material until say of the group said “obviously musically compromised” and that peak level at one meter recorded as that boxes “maximum musical output”. Next increase the level until half said it was “unlistenable” and record that level as the “maximum usable output”. While they may go louder, being unlistenable seems like a good place to stop, no need to risk letting the magic smoke out by going higher..

There is magic in those words.

---

As to commercially available processors that employ side chain limiting, there are very few. One example that takes baby steps with the concept of look ahead limiting is available from Four Audio (http://www.fouraudio.com/en/technical-articles/inside-the-hd2.html). It appears to be a world class loudspeaker processor.

The extensive use of side chain limiting as I’ve described is the domain of the higher-end free wire install type processors. There are other things I haven’t mentioned yet, but I’m sure I’ve thrown out enough bait to attract some interest. As mentioned earlier, I don’t see evidence of anyone else taking this seriously – but in the words of an immortal – I could be wrong. :)

---

You are engaging in dangerous thinking. This is your last chance. After this, there is no turning back. You take the blue pill (http://www.ehow.com/how_3391_set-stereo-speakers.html) - the story ends, you wake up in your bed and believe whatever you want to believe. You take the red pill (http://www.linkwitzlab.com/frontiers.htm) - you stay in Wonderland and I show you how deep the rabbit-hole goes.
Title: Re: Side Chain Limiting
Post by: Doug Fowler on May 22, 2012, 04:24:53 pm
Would anyone like to explain how to measure excursion with an accelerometer ? Does anyone use this sidechain limiting commercially ?

Yes, NEXO have implemented this for quite some time. Their term is "VCEQ" or voltage controlled equalization.  The NX242 manual offers a nice explanation of what they are doing.
Title: Re: Side Chain Limiting
Post by: Brandon Romanowski on May 23, 2012, 05:12:40 pm
Langston,
    Thanks for taking the time to reply. Have you ever used a laser to measure excursion ? What's the setup ? specifically ?
Title: Re: Side Chain Limiting
Post by: Arthur Skudra on May 23, 2012, 08:23:27 pm
Langston,
    Thanks for taking the time to reply. Have you ever used a laser to measure excursion ? What's the setup ? specifically ?
Essentially the laser measures distance, and converts it to a voltage, which in turn is measured by software.  I know that Clio has that capability.  Essentially it's a linear scale reading.  The laser I've been looking for is made by Baumer, and it's at least $1,000!!!

Excursion measurements can be done a lot cheaper using an accelerometer, here's a link:
http://www.libinst.com/accel.htm

For a completely unrelated task, I built my own custom accelerometer preamp in a compact, watertight, indestructible Hammond project box that works off phantom power and uses the ACH-01 accelerometer, works great with Smaart for vibration analysis in the Spectrograph.  Haven't figured out a convenient method to calibrate it to translate that signal to excursion though!  You should have seen the look on the sales clerk's face at the doll store when she asked me what I was planning to do with the Quik-Stik wax!  LOL!

There is a DIY project in June 2011 Audio Xpress magazine to build a preamp for the ACH-01 specifically for speaker excursion measurements that I have been wanting to build myself for a while! I just don't have any time to design and etch the printed circuit board (I wish they would have published that in the article).  If there is anyone enterprising enough to make one, I would gladly collaborate with them for a copy of the board for me.  Hint Hint!   ;)
Title: Re: Side Chain Limiting
Post by: Langston Holland on May 23, 2012, 11:41:08 pm
Quote from: Brandon
Have you ever used a laser to measure excursion ? What's the setup ? specifically ?

Hi Brandon:

I haven't found a reason to measure excursion directly given my interest in optimization (vs. loudspeaker design). I've found that indirect measurement via distortion combined with one's ears are ideal for finding the peak usable output of a system. This will not protect the system from overheating and the sonic degradation involved with power compression, but the crest factor of music will keep you away from that problem 99% of the time in my experience. That other 1% can be expensive, thus RMS type limiting is necessary to cover all the bases.

BTW, you can get a very close approximation of excursion with a direct measurement without buying anything:

http://www.linkwitzlab.com/faq.htm#Q29

Accelerometer measurements interest me for reasons other than excursion, i.e. cabinet and other boundary resonance behavior. Small signal impedance measurement is a pretty good indirect method of spotting cabinet resonances, but an accelerometer would be ideal.
Title: Re: Side Chain Limiting
Post by: Barry Singleton on July 21, 2012, 12:44:26 am
  Hi Langston;

  I am keenly interested in this subject. I read Charlie's paper a couple of years ago and couldn't quite wrap my head around the possibilities this opens up. Reading through this thread has more than kindled my interest.

  I don't think I have a piece of equipment capable of Side Chain Limiting, nor do I know what unit would be a good candidate. Do you have a suggestion?

  I have several measurement platforms and a fair collection of measurement gear including an acceleromoter and a lazer position system. I have never claimed to be a smart guy but maybe I could be of some value if I can find some direction on this.

  With high hopes I am looking forward to hearing from you.

  All the best,
                  Barry.

 
Title: Re: Side Chain Limiting
Post by: Langston Holland on July 22, 2012, 03:16:02 pm
Hi Barry:

If you play with all that stuff you're more than ready to blow things up. :)

Overview:

You have a fair amount of homework to do before you need to think about finding a free-wire processor with side-chain comp/limiters. IOW, what you need to do first is learn how to measure the thresholds of where your subs say uncle with and without the current processing you have. Keep in mind that it might not be your subs that give up at full amplitude - it may be your amp. For example, with a single 4Ω TH118 on one channel of a FP14000 in stereo mode with VPL set at max volts and "hard", the peak output of the amp is about 1dB shy of what the sub can handle. With a pair of TH118's on one channel as I actually use them, the peak output of the amp is 2.5dB shy of what the subs can take. More on the procedure later.

Since it's easy to talk about, I'll start with processors. There are several out there that will do the trick. I have a decade old MediaMatrix X-Frame 88 that Ray Rayburn1 sold me that works very nicely, but I bet the current stuff would make the process easier and possibly allow slightly better results. From the point of view of electrical control of loudspeaker behavior, it appears that the NION and Q-Sys processors are the most powerful available. Interestingly, they are both based on the same code. Few people seem to appreciate that Peavey is responsible for the most powerful audio processing in the world. I do not yet own either of these fancy processors, but I'm leaning toward the Q-Sys (250i) due to the fact that it has predefined FIR crossover filters as well as direct FIR coefficient import. The latter is in my future, the former I can use immediately. The NION's only have direct FIR coefficient import, but are all about Dante and that's valuable. QSC justifies their decision to go their own way with their network protocol with some drivel about better operability with standard Ethernet networks. What idiot is going to do a high dollar install and not run a dedicated network for the audio anyway? All processors in this league can do AES3 I/O, and that's all I need - but I'm going to be buying CL5's if things go well for business in the coming elections and Dante would be convenient. Maybe QSC will see the light (a.k.a market).

Remember that several other mfg's offer side-chain comp/limiting for probably less money and fewer additional features - you'll have to pursue that (and report!) yourself, because I want a Ferrari even if I never go over 70mph. Hopefully I'll learn how to get comfortable at much higher speeds one of these days. :)

Homework:

I use CLIO, a scope and my ears to do this, but you really can do this with just your ears and a voltmeter. You will connect the amplifier to the subwoofer without a processor. You will measure the gain of your amplifier, I suggest setting it to 32dB and verifying without the sub connected with a very small sine wave and a voltmeter. You will verify, preferably with a scope, that the setup you use to drive the amplifier input will cleanly pass a 10Hz, 20Hz, 30Hz, 40Hz, 50Hz and 63Hz continuous sines at +20dBu at the same level. Do not trust anything prior to measurement. Treat audio equipment like riding a motorcycle on the street; pretend that the mfg's are trying to hit you, but do not give them the opportunity. I've also heard that you may find user error situations that you've lived with through months of shows without knowing it.

Do this outdoors so the neighbors can hear and so you can hear the onset of distortion more clearly. If possible, remove whatever's needed to be able to see the woofer(s) movement. Remove the grill and gaff the wheels or anything else that can rattle - you want to hear the drivers, not the other stuff - fix the rattlers later if needed. Connect a scope across the balanced output of the source (or the passthrough output on the amp) and monitor the stimulus peak levels. If you only have a voltmeter, that'll work - just make sure you use that same meter when you're tuning your processor's side-chain limiters. The downside to using anything other than a true peak reading meter (scope is best) is that your readings won't be comparable with anyone else's unless they have that same meter. The methods that average or RMS reading meters use to do their thing varies. I can't imagine anyone going to the trouble of this kind of tuning without owning a scope.

Now feed the 10Hz Keele tone burst to the sub at a very, Very, VERY low level. Turn it up slowly while observing your meter and listening and wondering why on earth your going to all this trouble when it won't net you a dime more in rental fees. Once you find the threshold where you hear the onset of objectionable distortion (woofer slap, etc), record the meter reading. Turn it down and back up again to verify you got it right. Rinse, repeat with the other tone bursts. It gets very easy once you do it a bit.

These are the results on a pair of TH118's. I've included 1 meter peak SPL of the pair of subs for entertainment. Not bad for single 18's.

(2) TH118 w/ FP14000 at 32dB gain:

1. 63Hz 150.57dB SPL at +14dBu
2. 50Hz 150.78dB SPL at +14dBu
3. 40Hz 149.40dB SPL at +14dBu
4. 30Hz 139.05dB SPL at +14dBu
5. 20Hz 118.63dB SPL at  +2dBu
6. 10Hz  94.79dB SPL at   0dBu


Study this: Keele Shaped Tone Burst AES Paper (http://www.xlrtechs.com/dbkeele.com/PDF/Keele%20(2001-09%20AES%20%20Preprint)%20-%20Dev.%20Test%20Signals%20EIA-426-B.PDF)

Download this: Keele Shaped Tone Bursts and CBT Presentation (http://www.xlrtechs.com/dbkeele.com/Don's%20CBT%20Goodies%20Zip%20Download.html)

That download is huge and you only need a small portion of it, but Don2 bundled all this stuff for a reason and I don't want to interfere with that. The places his brain goes are well worth one's time to follow and it's just amazing to me that he puts his work into the public domain for free. He's one of the great ones.

Extra Credit:

Now add your sub processor to the chain. Put your meter across the output of the processor or input of the amp and repeat the tests. Keep your ears open. What maximum dBu levels are you now reading? Is your current processing REALLY protecting your loudspeakers? Ignorance is bliss until the brother-in-law of the lead singer shows up.

Notes:

1. Name drop! You impressed?
2. How about now? :)
Title: Re: Side Chain Limiting
Post by: Mark Wilkinson on November 12, 2019, 12:37:24 pm
Hi Barry:

If you play with all that stuff you're more than ready to blow things up. :)

Overview:

You have a fair amount of homework to do before you need to think about finding a free-wire processor with side-chain comp/limiters. IOW, what you need to do first is learn how to measure the thresholds of where your subs say uncle with and without the current processing you have. Keep in mind that it might not be your subs that give up at full amplitude - it may be your amp. For example, with a single 4Ω TH118 on one channel of a FP14000 in stereo mode with VPL set at max volts and "hard", the peak output of the amp is about 1dB shy of what the sub can handle. With a pair of TH118's on one channel as I actually use them, the peak output of the amp is 2.5dB shy of what the subs can take. More on the procedure later.

Since it's easy to talk about, I'll start with processors. There are several out there that will do the trick. I have a decade old MediaMatrix X-Frame 88 that Ray Rayburn1 sold me that works very nicely, but I bet the current stuff would make the process easier and possibly allow slightly better results. From the point of view of electrical control of loudspeaker behavior, it appears that the NION and Q-Sys processors are the most powerful available. Interestingly, they are both based on the same code. Few people seem to appreciate that Peavey is responsible for the most powerful audio processing in the world. I do not yet own either of these fancy processors, but I'm leaning toward the Q-Sys (250i) due to the fact that it has predefined FIR crossover filters as well as direct FIR coefficient import. The latter is in my future, the former I can use immediately. The NION's only have direct FIR coefficient import, but are all about Dante and that's valuable. QSC justifies their decision to go their own way with their network protocol with some drivel about better operability with standard Ethernet networks. What idiot is going to do a high dollar install and not run a dedicated network for the audio anyway? All processors in this league can do AES3 I/O, and that's all I need - but I'm going to be buying CL5's if things go well for business in the coming elections and Dante would be convenient. Maybe QSC will see the light (a.k.a market).

Remember that several other mfg's offer side-chain comp/limiting for probably less money and fewer additional features - you'll have to pursue that (and report!) yourself, because I want a Ferrari even if I never go over 70mph. Hopefully I'll learn how to get comfortable at much higher speeds one of these days. :)

Homework:

I use CLIO, a scope and my ears to do this, but you really can do this with just your ears and a voltmeter. You will connect the amplifier to the subwoofer without a processor. You will measure the gain of your amplifier, I suggest setting it to 32dB and verifying without the sub connected with a very small sine wave and a voltmeter. You will verify, preferably with a scope, that the setup you use to drive the amplifier input will cleanly pass a 10Hz, 20Hz, 30Hz, 40Hz, 50Hz and 63Hz continuous sines at +20dBu at the same level. Do not trust anything prior to measurement. Treat audio equipment like riding a motorcycle on the street; pretend that the mfg's are trying to hit you, but do not give them the opportunity. I've also heard that you may find user error situations that you've lived with through months of shows without knowing it.

Do this outdoors so the neighbors can hear and so you can hear the onset of distortion more clearly. If possible, remove whatever's needed to be able to see the woofer(s) movement. Remove the grill and gaff the wheels or anything else that can rattle - you want to hear the drivers, not the other stuff - fix the rattlers later if needed. Connect a scope across the balanced output of the source (or the passthrough output on the amp) and monitor the stimulus peak levels. If you only have a voltmeter, that'll work - just make sure you use that same meter when you're tuning your processor's side-chain limiters. The downside to using anything other than a true peak reading meter (scope is best) is that your readings won't be comparable with anyone else's unless they have that same meter. The methods that average or RMS reading meters use to do their thing varies. I can't imagine anyone going to the trouble of this kind of tuning without owning a scope.

Now feed the 10Hz Keele tone burst to the sub at a very, Very, VERY low level. Turn it up slowly while observing your meter and listening and wondering why on earth your going to all this trouble when it won't net you a dime more in rental fees. Once you find the threshold where you hear the onset of objectionable distortion (woofer slap, etc), record the meter reading. Turn it down and back up again to verify you got it right. Rinse, repeat with the other tone bursts. It gets very easy once you do it a bit.

These are the results on a pair of TH118's. I've included 1 meter peak SPL of the pair of subs for entertainment. Not bad for single 18's.

(2) TH118 w/ FP14000 at 32dB gain:

1. 63Hz 150.57dB SPL at +14dBu
2. 50Hz 150.78dB SPL at +14dBu
3. 40Hz 149.40dB SPL at +14dBu
4. 30Hz 139.05dB SPL at +14dBu
5. 20Hz 118.63dB SPL at  +2dBu
6. 10Hz  94.79dB SPL at   0dBu


Study this: Keele Shaped Tone Burst AES Paper (http://www.xlrtechs.com/dbkeele.com/PDF/Keele%20(2001-09%20AES%20%20Preprint)%20-%20Dev.%20Test%20Signals%20EIA-426-B.PDF)

Download this: Keele Shaped Tone Bursts and CBT Presentation (http://www.xlrtechs.com/dbkeele.com/Don's%20CBT%20Goodies%20Zip%20Download.html)

That download is huge and you only need a small portion of it, but Don2 bundled all this stuff for a reason and I don't want to interfere with that. The places his brain goes are well worth one's time to follow and it's just amazing to me that he puts his work into the public domain for free. He's one of the great ones.

Extra Credit:

Now add your sub processor to the chain. Put your meter across the output of the processor or input of the amp and repeat the tests. Keep your ears open. What maximum dBu levels are you now reading? Is your current processing REALLY protecting your loudspeakers? Ignorance is bliss until the brother-in-law of the lead singer shows up.

Notes:

1. Name drop! You impressed?
2. How about now? :)

Hi Langston, I've become interested in sidechain limiting and found this really cool older post of yours.
Have you done more with this?

It looks like a brilliant way of handling over-excursion. 
Programs like hornresp seem to be doing a pretty job of predicting excursion, so I'm thinking I should give this a try.
That said, i get that burst testing is the only way to know for sure...what excursion actually is, and does the sidechain limiter work.
Anyway, if you did take this further...would love to hear about it..
Title: Re: Side Chain Limiting
Post by: Marcel de Graaf on November 15, 2019, 10:59:31 am
Hi Mark (and of course Langston),

Thnx for putting this up again, I really have missed this wonderfull post at that moment.

I have a few questions:

- The subject is in main based on the excursion of the driver. What is a good and pratical way to measure this?
- I don`t come across loudspeaker DSP with a side chain input. What is a good way to do this after an analogue mixing table?

gr. Marcel
Title: Re: Side Chain Limiting
Post by: Frank Koenig on November 15, 2019, 07:55:03 pm
- I don`t come across loudspeaker DSP with a side chain input.

The newer Powersofts (after the k series) allow you to insert filters in the limiter side chain. I assume some of the other new DSP equipped amps do too. I think the question here is what to do with those filters to make the peak voltage limiter a better proxy for an excursion limiter. Interesting subject.  --Frank

(http://SideChainFilterScreen.JPG)
Title: Re: Side Chain Limiting
Post by: Chris Grimshaw on November 16, 2019, 08:42:07 am
I wasn't sure about how useful that'd be - Armonia says that'll onset whole-band limiting, when really I just want to stop 25Hz getting too loud.
I don't want 25Hz to pull 60Hz down with it.

I have a load of Powersoft T-series here, so I'll see about rigging something up and doing a test.

Something along the lines of:
- Apply some guesswork settings on the screen Frank has shown above
- Set limiting threshold nice and low so I don't melt the speaker
- Mic in front of speaker
- 25Hz and 60Hz tones into the mixing desk, so I've got control over relative (and overall) signal levels. Feed to amplifier.
- Feed the mic into an RTA.
- Watch the 60Hz level on the RTA as I push the 25Hz fader.

My suspicion is that, with the Powersoft limiter, 25Hz will pull 60Hz down. I suppose this test will find out.

Chris

PS - This'd be much easier with a spectrum analyser, but I don't have one that'll cope with the output voltages these amps are capable of, even with the limiters clamped right down.
Title: Re: Side Chain Limiting
Post by: Mark Wilkinson on November 16, 2019, 10:10:38 am
Hi Mark (and of course Langston),

Thnx for putting this up again, I really have missed this wonderfull post at that moment.

I have a few questions:

- The subject is in main based on the excursion of the driver. What is a good and pratical way to measure this?
- I don`t come across loudspeaker DSP with a side chain input. What is a good way to do this after an analogue mixing table?

gr. Marcel

Hi Marcel, hi guys,

Yeah, if anybody knows an accurate cheap way to measure excursion, I'd love to hear it. 
Apart from listening to tone bursts as Langston suggests, I just make a jig that holds something in front of the cone where I can make visual guesstimates.
But I suspect nothing can replace the tone burst listening.

I guess a good DSP after an analog mixer is required...I've no clue how this could be done without dsp...barely a clue with it haha
Although I think I've made some progress using q-sys and a Core110f...

First  response below is limiting via sidechain.  No hpf, xover, eq, etc. 
It's for getting the most out of a lightweight sub I mentioned in another thread.  Xmax is 9mm.
The thing is, excursion exceeds 9mm above tuning at 41Hz, ....at tuning excursion is minimal.
I've found a regular full-band limiter completely kills 40 Hz response, which needs a lot of power, but is excursion safe. (Thermal is real issue.)
So hopefully the sidechain limiter is gonna help here.

Second set of responses are curves from top down, -24dBFS RMS, -21, and -18, using pink noise.
The idea is I wanted full sidechain  limiting at -18dBFS, and wanted to see how limiting fell off at lower signals.
I can see the crest factor of the pink giving a bit of limiting at -21, and limiting looks gone at -24dBFS, so i think I'm ok.
Oh, by limiting is gone, I mean for the response notch countering over-excursion, 50-80Hz.
There are two stages of sidechain, a second one for the replacement of the hpf, which kicks in at about -36dBFS.

Now I need to add regular full-band peak and RMS limiters and do some listening and tests.  I'm encouraged so far.
Title: Re: Side Chain Limiting
Post by: Frank Koenig on November 16, 2019, 12:28:04 pm
My suspicion is that, with the Powersoft limiter, 25Hz will pull 60Hz down. I suppose this test will find out.

My suspicion, too. It's just a limiter, not a sub-band limiter. But please do experiment. And it still may be a step forward from a broad-band voltage limiter. If you're hitting the threshold because of 25 Hz and it's pulling down the 60 Hz, you, the human, has to remove some 25 Hz from the program with EQ or, in other words, rejigger the system high-pass.

A true excursion limiter needs either to measure excursion directly or a have a computational model of the physical system that can estimate excursion based on the electrical input. Does anyone know whether the big boys who make fully integrated speaker systems (Meyer, for example) do either of these? Or something simpler, like some sort of sub-band limiter, that's good enough? Better yet, has anyone reverse-engineered any of these systems?

--Frank
Title: Re: Side Chain Limiting
Post by: Frank Koenig on November 16, 2019, 12:46:48 pm
First  response below is limiting via sidechain.  No hpf, xover, eq, etc. 
It's for getting the most out of a lightweight sub I mentioned in another thread.  Xmax is 9mm.
The thing is, excursion exceeds 9mm above tuning at 41Hz, ....at tuning excursion is minimal.
I've found a regular full-band limiter completely kills 40 Hz response, which needs a lot of power, but is excursion safe. (Thermal is real issue.)
So hopefully the sidechain limiter is gonna help here.

Second set of responses are curves from top down, -24dBFS RMS, -21, and -18, using pink noise.
The idea is I wanted full sidechain  limiting at -18dBFS, and wanted to see how limiting fell off at lower signals.
I can see the crest factor of the pink giving a bit of limiting at -21, and limiting looks gone at -24dBFS, so i think I'm ok.
Oh, by limiting is gone, I mean for the response notch countering over-excursion, 50-80Hz.
There are two stages of sidechain, a second one for the replacement of the hpf, which kicks in at about -36dBFS.

Now I need to add regular full-band peak and RMS limiters and do some listening and tests.  I'm encouraged so far.

Good work, Mark. Keep it coming.

On measuring excursion: If you're downtown get a laser interferometer. Short of that, I wonder if there are any geometrical optical tricks that might be good enough -- we don't need a great deal of precision. Something that involves sticking a small mirror to the speaker and measuring the width of a line on a screen, maybe with a linear photodiode array or video camera, or a ruler for that matter. For detecting whether a certain peak exclusion has been reached it seems a simple, well-damped, spring, that is not too sensitive to the air motion, could be rigged to touch the cone and make noise. Maybe a piece of soft rubber or urethane foam with the right properties? Just free associating here.

--Frank

Scratch the video camera. It's not going to work because of aliasing by the frame rate of the video, at least in its simplest form.

As for transient test signals, how samples of actual kick drum hits, at least as a guide for choosing a tone burst envelope?
Title: Re: Side Chain Limiting
Post by: Chris Grimshaw on November 16, 2019, 02:05:44 pm
I saw a nice way of measuring cone excursion a while ago:
- Get a strip of cardboard, and cut one end at 45 degrees.
- Attach strip to the cone, so the 45 degree end is standing up away from the cone
- Play the test tone
- Hold a ruler to the cardboard strip, and use persistence of vision to make a measurement

Might've been Linkwitz that suggested that - can't remember.

Chris
Title: Re: Side Chain Limiting
Post by: Mark Wilkinson on November 18, 2019, 10:50:18 am
Thx Frank, Chris,

Yeah, I also saw the taped strip to cone thingy of the Linkwitz site somewhere. I've tried that and a few other gizmos that all come down to visual guesstimates.
Wish I was rich, i'd get one of the laser devices like in the ARTA manual.
Must say though, I think listening to the sine burst tests, like CEA-2010, is the real way to judge excursion. Or better yet, just stopping when measurable distortion fails the test.
REW has a CEA-2010 routine that I've toyed with in past and plan to use dialing in the sidechain limiter attempt.

Been having good luck with the sidechain limiter, at least so far with voltage measurements and transfers.
Two really cool things have popped out.
 
If the sidechain limiter follows the normal peak limiter, where the normal peak limiter is set to allow maximum output for frequencies where excursion is no problem, it's very easy to define an almost "parametric" limiting curve at lower output levels for any excursion limited frequency range or ranges.

Second, it's looking like sidechain limiting can totally replace a HPF for subs, in a very superior fashion. 
So far, I'm getting better protection (steeper curve/higher order) with considerably less phase lag / group delay.  I've become more excited over this than the frequency dependent excursion limiting.
Title: Re: Side Chain Limiting
Post by: Mark Wilkinson on December 11, 2019, 11:18:41 am
OK, between wind, rain, and too damn cold just to make another measurement...I haven't been able to measure the sub itself with the sidechain limiter.
But I have refined the limiter a fair amount...

Below is a set of traces that show how the limiter phases in a various drive levels.
The part of the limiter that acts as a high pass filter starts engaging below -40dBFS,
whereas the part that acts a frequency dependent excursion filter starts engaging at about -18dBFS.

It is all fully engaged at -12dBFS, and the peak voltage by frequency set up with a PL340 is shown in the side chart.