Any hints on limiting or compressing sub bass?

[FETlife]

Hello, have been unsuccesfully trying to make a limiter or compressor for a subwoofer (<80hz, strong 30-50hz) for pre recorded music. The main problem is gross distortion/clipping sound on strong peaks (kick drum).  Would ideally like it to respond to peaks, to stop amplifier clipping.

So far I've tried this circuit, which has been the most suitable http://www.thatcorp.com/datashts/dn129.pdf, but I found I needed painfully slow attack and release times, so not really practical. This circuit, http://sound.westhost.com/project67.htm, which had much the same problem and I've tried also been working on something similar to this http://www.diystompboxes.com/smfforum/index.php?topic=89110.0 but no luck.

I'm not overally fussed about modest levels of distortion (around 1-2% sort of thing) but this 'clipping' noise I've been getting is very intrustive.

Has anybody got any pointers please?

 

[XAXAU]

From experiments making bass heavy electronic music I've noticed that distortion comes from the release. I use 25ms release when brickwalling and no matter how much bass or kick drums I throw at the limiter it won't distort. The attack is "instant"!

 

[weiss]

Xaxau is right! It is the release Time that has major influence on distortion. Try to set it as long as possible.

 

[FETlife]

This is the problem I think, I've tried longer release times longer than 1 second without success!

Think the issue is more with the attack of the circuits, slowing the attack right down stops the distortion but lets the peaks though, so a bit pointless. Can't get a fast attack without horrible noises

 

[emrr]

THAT combined with a diode clipper?  Or run it low enough below distortion that overshoot does not clip? 
Or move to DSP. 
Or add another sub/amp system to raise SPL and run both lower.  Maybe combined with one of the above. 

 

[JohnRoberts]

You can make the compressor/limiter arbitrarily fast (make cap smaller) and that will generally just distort the first cycle of attack. This one cycle of distortion is not very audible.

A too fast release. will introduce distortion to every cycle as the gain tries to release during each cycle.

A side chain hold circuit for longer than one typical wavelength can clean up release related distortion while still allowing pretty fast release for real decay of sound level.

There are even more sophisticated approaches but I doubt that is your problem.

KISS, start with something simple like that JFET shunt limiter, or reduce the value of the cap in the THAT RMS detector.  You should be able to tolerate several percent of THD in low bass with  minimal audible sound quality deterioration.

JR

PS If you decide to use a diode clipper (extreme) I got a patent years ago for a bass clipper that was integrated into a Baxandall tone control so just the bass signal path gets clipped. The fact that the HF path remained unmolested, pretty much masked a lot of the negative audible aspects from hard clippers .

 

[ricardo]

Try JR's clipper around a Baxandall control or have a look at my circuit on http://www.proaudiodesignforum.com/forum/php/viewtopic.php?f=6&t=735

 

[gyraf]

NTP (a local Danish company, quite underrated) solved the low-frequency-distotion-when-compressing by weighting very-low vs. higher signal content,

then when low is found dominant they go into a mode where they apply control-voltage change (i.e. compression differences) only when the signal crosses zero

because zero-crossing is the only time you can apply gain changes without distorting

This sets the apparent attack time to be one half-wave of lowest frequency content - much faster than what can be set (without distortion) in any other way..

Jakob E.

 

[JohnRoberts]

NTP (a local Danish company, quite underrated) solved the low-frequency-distotion-when-compressing by weighting very-low vs. higher signal content,

then when low is found dominant they go into a mode where they apply control-voltage change (i.e. compression differences) only when the signal crosses zero

because zero-crossing is the only time you can apply gain changes without distorting

This sets the apparent attack time to be one half-wave of lowest frequency content - much faster than what can be set (without distortion) in any other way..

Jakob E.

Not to quibble (too much), but the dominant audible distortion associated with compression-limiting with LF signals is related to release time. Attack time distortion is generally one and done so a transient artifact is very hard to distinguish from the actual signal attack which is occurring simultaneously.  In fact, waiting until the next zero crossing to apply any gain reduction will lead to even more overshoot than conventional approaches. If this limiter is used to prevent overload, the extra lag could be undesirable.

You don't say whether the gain changes are applied as a step function at zero-crossings, or ramped in beginning at zero-crossongs. Either way the gain change related artifacts (distortion) will be much reduced. The reason for this is because the gain change is being multiplied by the signal so zero times anything is zero. Clever, and this does reduce sine wave distortion while the gain manipulations still occur.

Gain release can also be applied coincident with zero crossings. I wouldn't try to implement this without using digital processing, but it could be done (a sample and hold side-chain updated at zero crossings.). 

For analog limiters my best sonic result was from using a rate dependent attack/release that was very slow when the gain was near the target result, and got faster when gain was farther away from the desired level.  Artifacts from fast gain changes will generally be masked by the changing signal and audible distortion on steady state sine waves is mostly an issue for say a dedicated bass guitar limiter. Complex waveforms generally have more going on. . 

JR

 

[gyraf]

Thanks!

As I understand it, gain changes are stepped at zero-crossings when low-freq is dominant. This should also work on release-time-distortion

Jakob E.

 

[abbey road d enfer]

Hello, have been unsuccesfully trying to make a limiter or compressor for a subwoofer (<80hz, strong 30-50hz) for pre recorded music. The main problem is gross distortion/clipping sound on strong peaks (kick drum).  Would ideally like it to respond to peaks, to stop amplifier clipping.

So far I've tried this circuit, which has been the most suitable http://www.thatcorp.com/datashts/dn129.pdf, but I found I needed painfully slow attack and release times, so not really practical. This circuit, http://sound.westhost.com/project67.htm, which had much the same problem and I've tried also been working on something similar to this http://www.diystompboxes.com/smfforum/index.php?topic=89110.0 but no luck.

I'm not overally fussed about modest levels of distortion (around 1-2% sort of thing) but this 'clipping' noise I've been getting is very intrustive.

Has anybody got any pointers please?

Your problem comes from the fact that the three examples you've chosen all lack in performance.
They could all work if they had an adequate side-chain, but they all lack appropriate control of attack/release time.
Also are you sure your implementation is correct?  In particular the x-over LPF should be placed before the limiter.

 

[FETlife]

Thanks all, much appreciated.

Haven't got anywhere with the FETs but think I've found a compromise with the THAT circuit. As I understand it, C2 sets the attack and C4 sets the release, can I then simply multiple their value with R10 to find my time constants? LINK: http://www.thatcorp.com/datashts/dn129.pdf

Abbey road, why do you say they lack appropriate control? Whilst it's not user adjustable, values can be changed which adjusts things (with the exception of the opto circuit). Limiter certainly is after the crossover.

Probably should've mentioned, sometimes the sub will be playing by itself so needs to be reasonably clean as there won't be any masking effect from the mids and highs, which I think rules out any clipping circuits!

 

[LeeYoo]

Sounds like you have bought a sub that's too small for your needs.
There is no free lunch there.
If you want clean bass,  you need amps with enough headroom, and woofers that can move air.
Why use compressors that cripple music even more than most music already is.
Bass compression reminds me of an expensive set of Bose speakers I once installed.
Beautiful background bass, but nothing left at party level.
Why not post the size of the woofer, the volume of the box, the amp power, and the room you're in.
Leo..

 

[JohnRoberts]

Thanks all, much appreciated.

Haven't got anywhere with the FETs but think I've found a compromise with the THAT circuit. As I understand it, C2 sets the attack and C4 sets the release, can I then simply multiple their value with R10 to find my time constants? LINK: http://www.thatcorp.com/datashts/dn129.pdf

Sort of but not quite that simple.

C4  with R10 is the slow time constant and C4 with R8 is the fast time constant,

The extra op amp sets a  threshold by the DC current flowing done R9 that the op amp opposes through D4. When the difference between the "slow gain control" current and the DC threshold current from R9 is overwhelmed by the signal level current from R8, diode D4 becomes unclamped and the output of UC2 starts ramping up at a rate limited by C2. So while C2 slew limits the side chain the attack time is still set by C4.

JR

Abbey road, why do you say they lack appropriate control? Whilst it's not user adjustable, values can be changed which adjusts things (with the exception of the opto circuit). Limiter certainly is after the crossover.

Probably should've mentioned, sometimes the sub will be playing by itself so needs to be reasonably clean as there won't be any masking effect from the mids and highs, which I think rules out any clipping circuits!

 

[abbey road d enfer]

Abbey road, why do you say they lack appropriate control? Whilst it's not user adjustable, values can be changed which adjusts things (with the exception of the opto circuit). Limiter certainly is after the crossover.

in order to fine tune the limiter, you need to have a form of adjustment easier than soldering components, something like a potentiometer. Once you're satisfied with the settings you may (or may not) choose to replace with fixed components.

 

[ricardo]

I note there is absolutely no discussion in this thread on what are the 'limits' for the sub except for the link I pointed to and sorta in JR's clipper around the Baxandall.

Speaker overload is much more complex than amps.  At times the amp will overload first and at others, the speaker.  The overload slope is 12dB/8ve so a naive approach will almost certainly be non-optimal.

The 2 approaches I suggest are also the only ones where 'limiting' can be made 'inaudible'.

 

[gyraf]

Aah - I didn't realize that the aim was to prevent power amplifier clipping. Must read closer.

For preventing power amplifier clipping, it is often a good idea to monitor actual amplifier output voltage vs. actual amplifier power-supply lines: This difference function gives one of the best/simplest clipping warning systems. Then use this information to control your limiter - possibly with a hard-clipper for catching the part before your gain-reduction sets in..

Jakob E.

 

[ricardo]

I didn't realize that the aim was to prevent power amplifier clipping. Must read closer.

The aim is to prevent audible overload of speaker/box AND driving amp

A sub is easy cos you can make the amp big enough so it is never the limiting factor.  Then you only have to look at the speaker in its box ... but you MUST look at it.

When a speaker hits the stops, it will produce a certain amount of bass and loadsa distortion.  A good intelligent sub will produce the same amount of bass at the same volume setting but get rid of the distortion.  I show one solution using THAT compressors in my earlier post.

 

[abbey road d enfer]

Speaker overload is much more complex than amps.

Agreed. I would say it is essential to properly understand the excursion behaviour before undertaking such a project. I have found WinISD to be a useful tool in that respect.* I have corroborated the excursion simulation with actual measurements (using accelerometer, laser and ruler).

The overload slope is 12dB/8ve so a naive approach will almost certainly be non-optimal.

This is for constant SPL. For constant voltage, the slope is 6dB/octave with a dip (bass-reflex) or two (dual-chamber) or none (sealed box). The excursion does not asymptotically go infinite (as the integrator would imply), but has a limit when the speaker enters compliance-limited area, which is better modeled by a low-pass filter. I have found that trying to make an higher-order filter model the actual excursion curve is an almost impossible task, leading to errors that leave some areas unprotected and others overprotected.

The 2 approaches I suggest are also the only ones where 'limiting' can be made 'inaudible'.

I've never been satisfied with bass and sub-bass limiting. I use an excursion-driven variable HPF; when "limiting" occurs only the lowest spectrum is turned down, the rest is still there to fool perception.

*It doesn't replace a good theoretical analysis, but it allows building some instinctive sense of what is doing what.

 

[ricardo]

I have found that trying to make an higher-order filter model the actual excursion curve is an almost impossible task, leading to errors that leave some areas unprotected and others overprotected.

It's trivial for closed box.

In my 12may post on http://www.proaudiodesignforum.com/forum/php/viewtopic.php?f=6&t=735, U3b does this.  It is a LP filter which matches the Q & frequency of the sealed box speaker.  The relationship is exact for a closed box.

I use an excursion-driven variable HPF; when "limiting" occurs only the lowest spectrum is turned down, the rest is still there to fool perception.

This is indeed one correct way to do what's required. 

There are even simpler and better (less audible distortion, artifacts, bla bla) methods which are in my patented Powered Integrated Super Sub technology.  8)