Hello everyone,
I was going through the DSC-260 user manual when I noticed the following (on Page 1, Introduction)
"Output limiters to protect system transducers from overload damage with adjustable thresholds and automatic attack and release settings based on crossover frequency..."
Since the use of different attack/release rates for different frequency bands (bass, mid, treble etc.) has become commonplace, I was wondering if someone knew how to arrive at proper values for a band, based on its frequency content. Besides, it would also be nice to know of any existing methods/tables/standards for obtaining these values from the crossover frequencies.
Please note that the method(s) do not have to be scientifically proven (working methods will do). In case, anyone uses values that they've obtained otherwise (by trial & error etc.), kindly share the same.
Thank you.
(Please note that I am not in possession of the above-mentioned processor.)
I was going through the DSC-260 user manual when I noticed the following (on Page 1, Introduction)
"Output limiters to protect system transducers from overload damage with adjustable thresholds and automatic attack and release settings based on crossover frequency..."
Since the use of different attack/release rates for different frequency bands (bass, mid, treble etc.) has become commonplace, I was wondering if someone knew how to arrive at proper values for a band, based on its frequency content. Besides, it would also be nice to know of any existing methods/tables/standards for obtaining these values from the crossover frequencies.
Please note that the method(s) do not have to be scientifically proven (working methods will do). In case, anyone uses values that they've obtained otherwise (by trial & error etc.), kindly share the same.
Thank you.
(Please note that I am not in possession of the above-mentioned processor.)
Last edited:
Hi,
I think you have to decide to a number of cliped occurence allowed and then deduce attack time from crossover wavelength. For release let's say you take the same...
So if you have a 1khz xover point and allow for 1 cycle clip for low way you take 1ms and 0,1ms for high way ( i assume 10khz as the highest place where a peak could happen).
I would be more concerned about where to put thresholds ( as with such fast limiting it better has to happen only on some peaks and not often, iow only for protection purpose).
I think you have to decide to a number of cliped occurence allowed and then deduce attack time from crossover wavelength. For release let's say you take the same...
So if you have a 1khz xover point and allow for 1 cycle clip for low way you take 1ms and 0,1ms for high way ( i assume 10khz as the highest place where a peak could happen).
I would be more concerned about where to put thresholds ( as with such fast limiting it better has to happen only on some peaks and not often, iow only for protection purpose).
here's a good paper on limiters ( albeit deeper than might be wanted)
https://www.powersoft.com/wp-content/uploads/2019/01/powersoft_TN009_LimiterSetup_en_v1.0.pdf
Pages 8 & 9 are less technical and have charts for suggested attack and release, vs frequency & driver size
https://www.powersoft.com/wp-content/uploads/2019/01/powersoft_TN009_LimiterSetup_en_v1.0.pdf
Pages 8 & 9 are less technical and have charts for suggested attack and release, vs frequency & driver size
krivium said:
So it's then a direct (reciprocal) relationship that the attack/release times have with the highest frequency within each band, as in your two-way 1kHz example. I never thought it to be that simple. Thanks!
krivium said:
For playback of pre-recorded content, it's 0dB(peak) attack and -3dB/-6dB(RMS) release, I think. PA standards could be different, as they tend to use significant dynamic compression. I'm planning to use 0dB/-6dB (peak/RMS) and as you said, it's only for limiting (protection).
mark100 said:
Don't worry, I'm technical person and will definitely go through the article. Thanks.
Last edited:
Cool. What i think you'll find as you get into limiters, is that peak and RMS limiters do not correspond to any kind of 0db, -3dB, -6dB analysis.
Because they serve entirely different purposes. Peak is to protect from over excursion or drivers. RMS is to protect them thermally, and avoid compression.
The voltages and time durations used for peak and RMS are very different, not falling into easy dB or time comparisons.
Ironically ime, PA most often has significantly less dynamic compression than home audio.
It's because in PA, the compression is on the input channels, and not on the output channels driving amps.
Compression on input channels is often simply to avoid clipping them. Limiters are then put on output channels to allow PA's greater dynamic range, without consequences to drivers.
Home audio is effectively compressed by the average-to-peak ratio of the playback track, whereas PA can be whatever the operator sets up..
Because they serve entirely different purposes. Peak is to protect from over excursion or drivers. RMS is to protect them thermally, and avoid compression.
The voltages and time durations used for peak and RMS are very different, not falling into easy dB or time comparisons.
Ironically ime, PA most often has significantly less dynamic compression than home audio.
It's because in PA, the compression is on the input channels, and not on the output channels driving amps.
Compression on input channels is often simply to avoid clipping them. Limiters are then put on output channels to allow PA's greater dynamic range, without consequences to drivers.
Home audio is effectively compressed by the average-to-peak ratio of the playback track, whereas PA can be whatever the operator sets up..
Mark thank you for the article.
I agree with Mark about protection in pa, there is no real defined standard as it'll depend from soundsytem, amps, kind of music,...
The protection threshold have to be defined in situ. Even when there is a whole digital front end (digital mic to daw with remote controller to amp dsp) you still have to keep headroom.
For playback to be honest i don't see the point: the recordings are theorically already maxedout ( normalized) at -0,1dbfs ( on the loudest peak of the loudest track on the album), so you shouldn't need protection ( if your dsp don't bring cliping through it's own internal processing which i suppose should be the case with 'serious' units).
I even see a case where you'll run into trouble: first version of Metallica's 'death magnetic'.
With 3db dynamic range (!) and a multiband limiter set to protect at 3db peaks... you'll be limiting 80% of time ( in that case there was so much distortion -from the artistic choices of the band at first mastering- you won't hear your limiter own distortion anyway...).
About simplicity, well yes it could be simple or very complicated: given you allow for a bit of delay and you have a strategy ( allow low and sub ways to clip from time to time) you could even make some 'predictive' protection ( you allow 1 or 2 ms of delay for the limiter out while driving the sidechain realtime for mid and high) and nobody will hear anything clip ever. If the musicians are not bothered by 3ms delay of course ( i've worked with a guitarist which was but... statistically he is a weirdos! Usually we start to be bothered around 5ms).
On my own dsp i skipped the limiters totally ( home/studio use). For pa i would use it though.
I agree with Mark about protection in pa, there is no real defined standard as it'll depend from soundsytem, amps, kind of music,...
The protection threshold have to be defined in situ. Even when there is a whole digital front end (digital mic to daw with remote controller to amp dsp) you still have to keep headroom.
For playback to be honest i don't see the point: the recordings are theorically already maxedout ( normalized) at -0,1dbfs ( on the loudest peak of the loudest track on the album), so you shouldn't need protection ( if your dsp don't bring cliping through it's own internal processing which i suppose should be the case with 'serious' units).
I even see a case where you'll run into trouble: first version of Metallica's 'death magnetic'.
With 3db dynamic range (!) and a multiband limiter set to protect at 3db peaks... you'll be limiting 80% of time ( in that case there was so much distortion -from the artistic choices of the band at first mastering- you won't hear your limiter own distortion anyway...).
About simplicity, well yes it could be simple or very complicated: given you allow for a bit of delay and you have a strategy ( allow low and sub ways to clip from time to time) you could even make some 'predictive' protection ( you allow 1 or 2 ms of delay for the limiter out while driving the sidechain realtime for mid and high) and nobody will hear anything clip ever. If the musicians are not bothered by 3ms delay of course ( i've worked with a guitarist which was but... statistically he is a weirdos! Usually we start to be bothered around 5ms).
On my own dsp i skipped the limiters totally ( home/studio use). For pa i would use it though.
Well, it's nice to know that PA people compress at the mixer and not the amplifier.
My understanding is that the 0dBFS attack could be effectively used for anti-clipping action after which the release can happen when the RMS has dropped to -6dBFS. So, it's 0dB peak attack and -6dB RMS release. Correct me if that's likely to be an issue.
Table 7 in the Powersoft paper seems to recommend timings based on lowest frequency (vs. highest/short times suggested by krivium) in the low and mid bands with comparable attack/release times for HF (>1kHz). That's a great start.
My understanding is that the 0dBFS attack could be effectively used for anti-clipping action after which the release can happen when the RMS has dropped to -6dBFS. So, it's 0dB peak attack and -6dB RMS release. Correct me if that's likely to be an issue.
Table 7 in the Powersoft paper seems to recommend timings based on lowest frequency (vs. highest/short times suggested by krivium) in the low and mid bands with comparable attack/release times for HF (>1kHz). That's a great start.
Last edited:
Glad you found the article worthwhile Krivium.
I strongly agree with you about the general lack of need of limiters in home audio.
Since with home audio it is rarer to have amps with significantly greater power than drivers can handle.
I do use limiters at home, but that's just because i run proamps on everything.
Heck, I'll test a single TC9 with 1500W @ 8 ohms
Haven't smoked one yet Lol
I strongly agree with you about the general lack of need of limiters in home audio.
Since with home audio it is rarer to have amps with significantly greater power than drivers can handle.
I do use limiters at home, but that's just because i run proamps on everything.
Heck, I'll test a single TC9 with 1500W @ 8 ohms
Haven't smoked one yet LolMy take newvirus2008,
has come to view RMS and peak limiting as completely independent functions.
Because the time structures are so different.
That said, i often find the voltage values i use are on the order of 0dB peak and -6dB RMS, as you suggest.
I think if there is variance from that rule of thumb, it's that peak can be +9dB or more to RMS, provided it's quick enough acting.
I would run +6 dB peak and 0 dB rms with 10 ms attack and 100 ms release times. If you only have one type, use peak when running the levels close to max power handling of cabs (small drivers working hard) and rms with lower levels ie. acoustic music with more system headroom.
Smaller VC HF drivers crossed low are the most vulnerable and need the most consideration, especially with the potential of feedback. Alot of people who dont know what they're doing tend to use the main limiters as their master compression, which is silly and abusive, both to the equipment and audience. I would run frequency dynamic compression at a minimum to protect people's hearing and for less fatiquing sound.
Smaller VC HF drivers crossed low are the most vulnerable and need the most consideration, especially with the potential of feedback. Alot of people who dont know what they're doing tend to use the main limiters as their master compression, which is silly and abusive, both to the equipment and audience. I would run frequency dynamic compression at a minimum to protect people's hearing and for less fatiquing sound.
Like almost always, I think we have a bunch of variables floating around.
When talking multi-band, any attack and release times need to be given for a specific band.
(and of course specify whether for peak or RMS)
Biggest variable, and the unspoken one so far...is relative processor dB levels are completely meaningless, when it come to setting limiters.
Final amp output voltages are what matters....and must be measured at amp outputs. Processors must be set to limit measured voltages.
We've been talking relationships between what those amp voltage ratios, peak to RMS, might be after going though limiters.
But that's just a ballpark relationship type overview....
..and of this i'm certain, limiting amp voltage and for how long, is all that matters... other than knowing what type of limiting you need to do.
And it's why if you don't have two limiters per multi-band channel, you have to choose.
On lower freq bands, i choose RMS.
On higher freq bands, i choose peak. With the caveat of looking at impedance curves to make sure RMS isn't more important.
When talking multi-band, any attack and release times need to be given for a specific band.
(and of course specify whether for peak or RMS)
Biggest variable, and the unspoken one so far...is relative processor dB levels are completely meaningless, when it come to setting limiters.
Final amp output voltages are what matters....and must be measured at amp outputs. Processors must be set to limit measured voltages.
We've been talking relationships between what those amp voltage ratios, peak to RMS, might be after going though limiters.
But that's just a ballpark relationship type overview....
..and of this i'm certain, limiting amp voltage and for how long, is all that matters... other than knowing what type of limiting you need to do.
And it's why if you don't have two limiters per multi-band channel, you have to choose.
On lower freq bands, i choose RMS.
On higher freq bands, i choose peak. With the caveat of looking at impedance curves to make sure RMS isn't more important.
Last edited:
Straight to the point Mark.
This is how it's done in the Lake ( more or less: it doesn't measure but the dsp needs to be setup to the parameters of your amp and then it does it's thing).
And what i've seen from some 'top' (french) PA engineer: limiter to protect from clip, rms for thermal management.
But the requirements in PA are not the same that at home ( or even in studio): who need to be out of thermal compression? ( if you are then maybe we should speak LOUDER as you may here more Tinitus than my voice... )
Newvirus: about cinema and movies i disagree on your conclusions. In fact i even think the situation is easier than for music thanks to one parameter: there is a reference spl level, a standard ( in France until 2017 levels were setup for 85dbspl (a) with 20db headroom allowed, since it is 82dbspl (a) with allowed peaks of 20db for 15mn. For bass it is now 118db (C) over 15mn too).
This will let you manage same thing at home ( given you'll very probably need less than 85dbspl in the end: domestic rooms are smaller and less absorbent than a theatre and this will make you feel unconfortable at this kind of average spl).
With music we don't have reference so most bet are off about dynamic.
Except if you choose to do the same as in a movie: choose a reference level* with a given dynamic range allowed (20db is a worst case scenario so it even gives you the max level you'll ever see to size your system) and from then compensate the level of the different materials for them to match between each others. ( eg: a classical uncompressed record could need to be played at 0db out while if you listen to the Metallica album i talked about your volume will need to be setup at -17db: both will be played at same perceived loudness.
In other words apply a K monitor set up. K-system - Wikipedia
But well,... i understand that some won't set up a mastering facility at home! That said the approach could be a good source of inspiration though imho.
About the limiters and the multibands it seems to me that some point of views mix differents approach: technical ( protection) and esthetics.
Please guys don't fall into the rabbit hole to try to perform esthetic tasks ( mastering) with technical tasks ( protections of the sound system).
Yes tools used could be exchanged** but the goals and targets are not the same so better use dedicated mastering tools for 'remastering' than rely on tools used to be used occasionaly for protection duties (limiters which can lack finesse in this kind of situation imho).
*
82/85dbspl are convenient because our auditory system is already starting to linearize itself -take a look at Fletcher and Munson curves- and it allow 8hours exposure/day with total recovery after one night rest.
**
I use the Lake's eq to perform mastering duties (they are very transparent and powerful) but on independant slots - in/out trough Aes lines from and to pc- than the loudspeaker management process.
This is how it's done in the Lake ( more or less: it doesn't measure but the dsp needs to be setup to the parameters of your amp and then it does it's thing).
And what i've seen from some 'top' (french) PA engineer: limiter to protect from clip, rms for thermal management.
But the requirements in PA are not the same that at home ( or even in studio): who need to be out of thermal compression? ( if you are then maybe we should speak LOUDER as you may here more Tinitus than my voice...
)Newvirus: about cinema and movies i disagree on your conclusions. In fact i even think the situation is easier than for music thanks to one parameter: there is a reference spl level, a standard ( in France until 2017 levels were setup for 85dbspl (a) with 20db headroom allowed, since it is 82dbspl (a) with allowed peaks of 20db for 15mn. For bass it is now 118db (C) over 15mn too).
This will let you manage same thing at home ( given you'll very probably need less than 85dbspl in the end: domestic rooms are smaller and less absorbent than a theatre and this will make you feel unconfortable at this kind of average spl).
With music we don't have reference so most bet are off about dynamic.
Except if you choose to do the same as in a movie: choose a reference level* with a given dynamic range allowed (20db is a worst case scenario so it even gives you the max level you'll ever see to size your system) and from then compensate the level of the different materials for them to match between each others. ( eg: a classical uncompressed record could need to be played at 0db out while if you listen to the Metallica album i talked about your volume will need to be setup at -17db: both will be played at same perceived loudness.
In other words apply a K monitor set up. K-system - Wikipedia
But well,... i understand that some won't set up a mastering facility at home!
That said the approach could be a good source of inspiration though imho.About the limiters and the multibands it seems to me that some point of views mix differents approach: technical ( protection) and esthetics.
Please guys don't fall into the rabbit hole to try to perform esthetic tasks ( mastering) with technical tasks ( protections of the sound system).
Yes tools used could be exchanged** but the goals and targets are not the same so better use dedicated mastering tools for 'remastering' than rely on tools used to be used occasionaly for protection duties (limiters which can lack finesse in this kind of situation imho).
*
82/85dbspl are convenient because our auditory system is already starting to linearize itself -take a look at Fletcher and Munson curves- and it allow 8hours exposure/day with total recovery after one night rest.
**
I use the Lake's eq to perform mastering duties (they are very transparent and powerful) but on independant slots - in/out trough Aes lines from and to pc- than the loudspeaker management process.
when i was still doing pa (mostly dj soundsystems) we used those limiters in the dsp only for protection, hitting them was too loud. We set the shortest attack and release possible and ratio as brickwall limiter so it would start pumping if the dj hit that limits and he would back off. standard the treshold was set at -3dB VU. But this was for to protect our system, not to sound good. A good dj kept headroom and sounded good, a bad dj hit those and sounded like ****. We did protect that setting with a password off course (because dj's sometimes try to change it to go louder).
profiguy said:
Do you run separate attack/release times for peak and RMS limiting as suggested by mark100?
mark100 said:
Well, I only want clipping protection, which is also quite unlikely to be triggered as correctly mentioned by krivium. However, since we're into the topic, let us discuss RMS limiting and their characteristics as well.
I strongly believe your limiters are using dBu units (1dBu=0.775V), that definitely makes your limiting dB value dependent on absolute amplifier output voltage. And with another setup, the values required to achieve the same function would entirely be different.
I purposely used dBFS units to keep amplifiers (and their gains) out of question and thereby make the discussion easier. For example, your 0dBFS could be +2dBu and mine could be -2dBu, but we could still have this discussion without confusing each other or worrying about system gain, if these values be interpreted as 0dBFS by both you and me, respectively.
As to using different settings for peak / RMS limiting, it makes sense as RMS is about thermal damage, whereas peak limiting is about clipping / excursion damage.
krivium said:
Yes, I agree with that. But what if 105dBspl at the listening position is equal to my 0dBFS (not dBu)? After all, that's the peak attack threshold I suggested for pre-recorded content in the beginning, remember?
BTW, cinema and studio speakers do seem to use limiting (attached). It's just that they don't disclose the settings, or maybe the attack/release times vary with transducer model and type (cone/compression driver etc.).
Last edited:
Hi, I should stop and make sure i understand your goals.
I say that because clip protection and limiting are two different functions in my mind.
So far, all my comments have been about limiting amplifier power for the protection of drivers.
Moving now to clip protection, do you mean protection that keeps the amps from clipping? (For amps that don't have internal clip limiters already built in, and don't have enough power to cause either thermal damage or over excursion.)
If you do mean that, then you still have to take amplifier gain into account. Because it takes a certain amp voltage input, to push the amp into output clipping.
So either a voltage measurement of the amps output, or a very accurate calculation of processor and amp gains, is still necessary.
Along with knowing how dBFS maps into the processor's voltage output.
DBFS has to be converted into voltage, since voltage is what we ultimately want to limit.
It varies a little among pro audio processors/mixers/studio gear, but typically -18 to -20dBFS maps into +4dBu.
(Oh, yep, my limiters are set using the pro/studio standard of +4dBu = 1.23V, or as you gave 0dBu = 0.775V.
Hope this helps and made sense
mark100 said:
Yes, anti-clipping only. However, it's said that peak limiting would increase the RMS voltage, so I thought of a release that would be based on RMS value. I also managed to find a diagram (from a datasheet) that uses my concept of peak attack / RMS release (attached), that might (hopefully) help you understand the philosophy beneath.
mark100 said:
And that's the voltage that I call 0dBFS. It varies from amp to amp but fortunately does not still cause any confusion during discussions, as (unlike dBu) it is not an absolute value, and could mean different numbers to different people.
Well, my 0dB is equal to 1.35V (amp clips beyond this), but I'm 100% sure that's this information is of no use to you (or anyone else) other than myself.
So getting back to the topic, what attack/release settings do you suggest for anti-clipping protection? I was able to gather that the timings are usually much faster than those used for RMS limiters, which was also possibly why krivium recommended calculating them from the uppermost frequency of the band.
