Letting amps clip or running a limiter?

Member
Joined 2007
Paid Member
In a situation where your speakers would have a peak limiter set higher than the maximum peak output voltage of their connected amplifier would it be better to set a zero attack limiter with a short release time (4mS) just below the clip limit of the amplifier or to let the amplifier clip? if so how far below the maximum output voltage of the amplifier would you set the limiter?
 
There is no point setting a limiter above the voltage output of the connected amplifier.. why would you do that? Clipping amplifiers sound bad and if there are high frequency drivers attached they can be overpowered with this clipped/compressed output, so I always set limiters just below amp clipping when that is what you have to work with.
 
Member
Joined 2007
Paid Member
Perhaps I was unclear. What I'm proposing is setting the limiter just below the amplifiers clipping point. This will have the effect of slightly reducing the maximum output power but more predictable behaviour. Also the limiter I have available has zero attack but a minimum release time of 4mS, or it can have longer release. The particular design of my speakers is such that there is no danger of overheating the voice coils using this amplifier, its not required for protection.
 
You correct.. I misunderstood your first post. In that case then yes do as you describe and set the limiter release time based on the frequency range it is covering, longer for low frequencies and shorter for highs and strive for a balance between producing audible pumping and totally crushing the dynamics.
 
Here are the recommendations from Powersoft for K series amps, other processors will behave differently so the numbers can and should be altered to taste.

 
Last edited:
would it be better to choose a zero attack time over the recommended attack release ratio?
Many fear "losing power", worry about "the numbers", and so both set threshold at 99.99% of output power and attack time to ZERO.

Nice on a sheet of paper, a mess in the real world, in exchange of nothing.

1) PA amps are not driven by sinewaves but by Music Program, which has at least some dynamic range.
Which means short time peaks will trigger limiter, and average output will be significantly below that, by definition.
Worrying about how close narrow peaks get to rails is then not too significant, and it´s fine, we are amplifying Music here.

2) ZERO attack time means that waveform tops will be attenuated so flattened as soon as signal reaches threshold voltage, very similar to what plain clipping does, so what´s the point?

3) ultra short release time combined with (2) means lower frequencies (within that band) will heavily modulate higher ones.
Maybe you are (sometimes) avoiding flatb tops, but definitely creating tons of intermodulation, the second distortion mchanism.

In a nutshell, use limiters in the tried and true way arrived at by years of experience, tying to squeeze down to the last Watt may be a self defeating goal.
 
Member
Joined 2007
Paid Member
At this party I'm doing DJ playback and live sound. My processor is a Symetrix Symnet 8x8 DSP. This is a free form DSP has a choice of 'compressor' or limiter blocks. The compressor has attack, release, soft to hard knee and ratio with the minimum attack been very short but not zero, it does not cause any processing delay. The limiter has zero attack time but causes a small processing delay (not enough to be a problem as long as every output is delayed by the same amount). The limiter only has choice of threshold and release time.

In terms of frequency bands I have:
500 - 20 kHz (Apart audio PA2120 120W, ~100dB/1W, top mid/high)
120 - 500 Hz (Apart audio PA2120.120W, ~100dB/1W, top midbass)
30 - 120Hz (nu6000 + P7000s, 4 subs that should do ~128dB/1m/2pi over that range)

I was trying to get a crown CTS4200 up and running for the tops which would give me in excess of 200W per channel but I'm uncertain it will be possible in time. The desk is probably an X32.
 
In my experience, it's fine to let amplifiers clip if:
- They don't sound bad while doing it
- The average power level is still kept under check

If your amp goes "crunch" when clipping, then obviously it sounds bad and that should be avoided.

Remember - loudspeakers don't care what the signal they're being fed looks like. If you push an amplifier 3dB into clipping, it's not the additional harmonics that kill the driver. It's the increase in the average power being delivered by pushing the fader by another 3dB.

A good engineer that's paying attention will be able to spot if/when the PA system is struggling, and act accordingly.

Chris
 
The limiter ... causes a small processing delay

That's so it can anticipate the attack and soften the application of limiting to prevent what would otherwise be a clipped signal.

Ive got my PA management unit setup with a 40:1 compressor on the inputs which ultimately stops anyone (DJ) from overloading the PA. It's set to compress just a tiny bit before the sources master output level shows it touching the reds. So there is plenty of gain and a bit of red light flashing to keep DJs happy. (Certainly on the pioneer mixers with a bit of red flashing there is still some headroom under the hood, because they know DJs love smashing the reds). When going further into the red the compressor then starts crushing the life out of the music leaving it with no dynamics and unpleasant to listen to, although that doesnt always stop DJs from doing it. At least that cant overdrive the system.
 
At this party I'm doing DJ playback and live sound. My processor is a Symetrix Symnet 8x8 DSP. This is a free form DSP has a choice of 'compressor' or limiter blocks. The compressor has attack, release, soft to hard knee and ratio with the minimum attack been very short but not zero, it does not cause any processing delay. The limiter has zero attack time but causes a small processing delay (not enough to be a problem as long as every output is delayed by the same amount). The limiter only has choice of threshold and release time.

In terms of frequency bands I have:
500 - 20 kHz (Apart audio PA2120 120W, ~100dB/1W, top mid/high)
120 - 500 Hz (Apart audio PA2120.120W, ~100dB/1W, top midbass)
30 - 120Hz (nu6000 + P7000s, 4 subs that should do ~128dB/1m/2pi over that range)

I was trying to get a crown CTS4200 up and running for the tops which would give me in excess of 200W per channel but I'm uncertain it will be possible in time. The desk is probably an X32.

Ime, limiting has to be split into two functions, with separate limiters for each.
RMS, or thermal limiting, like given in the powersoft chart earlier.
And peak, or excursion limiting, with near instantaneous action....

I know you know all this, just repeating for any who don't..

Since the Symetrix only appears to allow one of those limiter types at a time, I'd use it for Peak, and the compressors in the X-32 for RMS.

I didn't see what drivers your using...do you have the voltages in mind, RMS and peak, for each section?
 
Member
Joined 2007
Paid Member
The DSP is free form so I can have unlimited (within the processing power available) of each type of limiter as any point in the signal path. Tops are three way with a passive mid high crossover:
15CM V2
4*3FE22
P-audio BM-D446
There is over 10dB of padding in the mid high crossover performing equalization functions and bringing the sensitivity of the mid/high down to 100dB/1w
 
The DSP is free form so I can have unlimited (within the processing power available) of each type of limiter as any point in the signal path. Tops are three way with a passive mid high crossover:
15CM V2
4*3FE22
P-audio BM-D446
There is over 10dB of padding in the mid high crossover performing equalization functions and bringing the sensitivity of the mid/high down to 100dB/1w

Good, i mistook with Symetrix "a choice of compression or limiter block" to mean you had to choose one.
Which doesn't make any sense, because who would design a free form processor without both.

Anyway, a quick look at your drivers...and I would go in order:
40Vrms, 105Vpk
20Vrms, 50Vpk
15Vrms, 35-40Vpk

where pk is true pk (1.4 x rms)
Oh, and 20Vrms, 50Vpk assumes the 4 drivers are 2 sets of parallel in series.

I've found setting predictive limiting to be a bit of a pain, needing to measure actual amp outputs. But well worth it for peace of mind with live.
RMS and peak limiting are one piece of dsp i like in the amp, better than in the processor.
 
Member
Joined 2007
Paid Member
Thanks for those numbers. I have already measured the gain and maximum output voltage of the amplifiers (2* Apart Audio PA2120) used for the above drivers . It can do 120Vpp so 42.4Vrms, 60Vpk (clip point). It also has a slow acting limiter built in that reduces the level to below the clip point if that happens repeatedly. The mids are wired series parallel:
BM-D446 on PH-4220
considering that the crossover is padding the mids down at least 5dB it would seem that the only driver that could be vulnerable to damage on these amps would be the tweeter if there was excessive >10kHz content.