Speakerplans.com Homepage
Forum Home Forum Home > General > General Forum
  New Posts New Posts RSS Feed - Peak limiting for driver protection
  FAQ FAQ  Forum Search   Events   Register Register  Login Login

Peak limiting for driver protection

 Post Reply Post Reply Page  123 6>
Author
Message
Augusts View Drop Down
Registered User
Registered User
Avatar

Joined: 07 January 2017
Status: Offline
Points: 150
Post Options Post Options   Thanks (0) Thanks(0)   Quote Augusts Quote  Post ReplyReply Direct Link To This Post Topic: Peak limiting for driver protection
    Posted: 26 December 2022 at 11:39am
Hey!

I have been doing a lot of digging and there are a lot of approaches and its a very complicated process alltogether but to simplify

- If a speaker reaches its Xmax at 1000w AES, than, using the XTA dsp with the look-ahead limiter, should I set the peak limiter +6db or + 3db above the RMS limiter. 
I am asking this because I have almost destroyed a voice coil by hitting it against the backplate,  with using just the rms limiter. As I understand RMS limits the speaker at the given rms value and then it lets the music still be dynamic by letting throught what is called the program material. and then there are peaks that are not important in the subwoofer Xmax context because peaks are so fast that the cone does not travel the distance of what 4000w peak power would do if it was not a peak.

Do I understand correctly or not?
The thermal power rating I understand its aes+3db program plus 3db peak. but I am curious about how these two XTA limiters work together in the coil travel context

Thanks
A
Back to Top
Augusts View Drop Down
Registered User
Registered User
Avatar

Joined: 07 January 2017
Status: Offline
Points: 150
Post Options Post Options   Thanks (0) Thanks(0)   Quote Augusts Quote  Post ReplyReply Direct Link To This Post Posted: 26 December 2022 at 11:43am
If I have an amp with a peak limiter than should I set it plus 3db above DSP RMS limiter or can I set it plus 6db, thus also raising the amps rms power by 3db. The DSP limiter would still be at 1000w but would not that exceed the Xmax if I raise the amps power by 3db

Edited by Augusts - 26 December 2022 at 11:44am
Back to Top
toastyghost View Drop Down
The 10,000 Points Club
The 10,000 Points Club
Avatar

Joined: 09 January 2007
Location: Manchester
Status: Offline
Points: 10861
Post Options Post Options   Thanks (0) Thanks(0)   Quote toastyghost Quote  Post ReplyReply Direct Link To This Post Posted: 26 December 2022 at 5:11pm
Originally posted by Augusts Augusts wrote:

The thermal power rating I understand its aes+3db program plus 3db peak. but I am curious about how these two XTA limiters work together in the coil travel context

Thanks
A


I’m afraid it's more complicated than that.

The AES power handling value on a driver's spec sheet is for the transducer suspended in free air with a particular filtered pink noise test stimulus.

The wattage shown on the spec sheet is a calculated value, based on the RMS voltage and the single-value nominal impedance of the driver, again in free air.

The more in-depth Klippel Large Signal Identification testing system reports the measured Real Power applied to the driver over a given test period, but this is not done by all manufacturers or for all transducers. If it is, the values are not often publicly available.

They can be very revealing, however; as an example, a top-tier best-in-class 18" driver that is very popular across the world has an AES power rating of 1700 Watts, but the plot of the rise in voice coil temperature, and the measured Real Power dissipated in the driver over a 1-hour period shows a different story:


This chart shows that the thermal rise of this particular driver stabilises after 30 minutes of a continuous signal. We see a 125-degrees C increase in the voice coil temperature during the test, which didn't increase further so long as the Real Power applied did not exceed 300 Watts.

During that test, the voltage logged at the driver terminals was 63 Vrms and 201 Vpk, and a little shy of 4 dB power compression was observed in the output SPL.

The DC resistance (Re) of the driver had changed from 5.4 to 8.4 Ohm after the test was completed, too.

Please understand this does not mean the AES rating is a lie. You just need to understand the conditions of the test and how it applies to real-world usage.

These values are unlikely to remain the same when the driver is mounted in your cabinet, as the box loading will not only modify the electrical impedance curve of the driver but also changes the amount of airflow or cooling to various parts of the hardware.

All of this means that finding the best thermal, RMS voltage, peak voltage and over-excursion limiter thresholds not only requires an understanding of how the driver behaves in the cabinet but also the types & durations of signal that it is likely to be driven with.

Spanking a subwoofer with heavy rolling electronic bass lines for 16 hours, every day for half a week at a festival is going to require a more strict threshold than the same box installed in a church with the odd Christian rock band.

For the example driver shown, I'd be comfortable with a thermal limiter threshold of 500 Watts, but know that this will be giving up some headroom for increased safety.

I'd also like to do that using a dedicated thermal limiter which tracks the change in current versus the electrical impedance as the driver & coil temperature goes up or down - I'd use separate RMS and Peak voltage limiters, and base those on when the box starts to distort heavily or suffers from other non-linear behaviour.


There's a test process described in the AES2012-2 standard to determine the Maximum Input Voltage a given speaker in a given cabinet can handle, before power compression starts to sap the output level down and thermal runaway begins. It's commonly called the "Toaster Test" because you deliberately heat the thing up, but it's non-destructive when done properly.

In simple terms, you bypass all DSP processing and measure the frequency response of the output while driving the box with an increasing level signal, using pink noise that has been filtered in a suitable way to prevent applying too much signal way outside of the intended frequency range of the speaker. That’s typically using the curves supplied in either IEC 60268-1 or CTA-2034 for subs & full-range speakers, or a simple second-order Butterworth HPF for HF drivers.

Once the level has dropped by 3 dB in comparison to the sensitivity measurement, the input voltage is logged. That goes into GLL or CLF data, and can be used to choose a suitably-sized amplifier to drive the speaker without clipping.

The process has been updated a couple of times over the years, but these two links cover the common approach:
https://www.prosoundtraining.com/2012/09/20/loudspeaker-maximum-input-voltage-test-results/
https://www.prosoundtraining.com/2013/11/22/the-loudspeaker-toaster-test-revisited/

These days, anyone can do this test at home with:
- a powerful amplifier
- a cheap two-channel audio interface
- dual-channel FFT software
- a 1/3 octave band equaliser
- a True RMS DMM/voltmeter
- a microphone ()ideally a measurement one, but a good condenser or even SM58 will do in a pinch we’re only interested in the relative change in level when compared to an initial measurement)
- 30 minutes or so of free time
- friendly neighbours (or a large bribe for them to go somewhere else for the afternoon)

The IEC 60268-1 curve can be approximated on a graphic equaliser with the following adjustment:


An example of traces taken from the EASE SpeakerLab manual’s guide to the test process:


In general, the MIV obtained in this process might seem to be lower than it you’d expect. That’s because it is an approximate test - to do better, you’d need a lot more kit to monitor the temperature of various bits, the impedance curve and so on. Which is why a Klippel Analyser with the relevant modules costs thousands of dollars.


You don’t need those to get in the ballpark though. I could perhaps be persuaded to find time to make an updated step-by-step guide or video on how to do this using the freeware OpenSoundMeter or the free version of ARTA in the new year, if people are keen.



The Peak voltage limit threshold is then typically set somewhere between 6 dB and 12 dB above the RMS voltage threshold, but it again depends on the likelihood that the signal applied to the box has a low crest factor or not.

Another, and in my opinion much better, method to find the peak voltage threshold is to conduct a distortion-limited stepped level test using a dense multitone stimulus. See Production Partner test for an example. The distortion threshold varies, but 10% for subs and 3% for full-range boxes is common.

That way, you ensure the speaker sounds good which is kind of an important goal, although it seems many don’t agree with me from listening to a lot of rigs…

Same for the attack and release thresholds; a limiter that engages before completing a single cycle of the lowest frequency the speaker is designed to play will sound bad.

For subs, the lower Peak overshoot value might be advised - so long as the system operator will pay attention and back the drive level off to maintain headroom and clarity when the limiters are engaging often.

For a 6 dB increase in output you need four times the power, or two times the voltage:
http://www.sengpielaudio.com/dB-chart.htm

Here are some example limiter threshold values for a ported 18” subwoofer cabinet with a processed response of 30 Hz to 100 Hz. It uses 4 Ohm nominal drivers, each rated at 1700 Watt AES, and the magnets have a clear air path to the outside world via a large opening:
RMS: 80 Volts (0.8 seconds attack, 1 second release)
Peak: 160 Volts (32 ms Attack, 256 ms release)
Thermal: 40 Volts (3.0 Second Attack, 3 Second Release)
XMax: 60 Volts at 25 Hz

That last one is often overlooked, but it dictates what kind of filtering the box needs to support the peak voltage limiter threshold.
60/160 = 0.375

A voltage ratio of 0.375 is a decrease of ~8.5 dB, so the electrical HPF used on this speaker must produce an acoustic level drop of 10 dB or more by 25 Hz to prevent over-excursion

For the sake of head maths, let's assume that the box's unprocessed output frequency response is ruler flat with a -3 dB point of 30 Hz and a natural second-order roll off below there (as for a ported cabinet).
In that case, considering port compression and such that makes the large signal frequency response deviate from flat, we may need to add a fourth-order (24 dB per octave) Butterworth HPF at 35 Hz to meet the above condition at all drive levels.


The idea is that you should very rarely see the RMS voltage limit engaged in normal use. A well-chosen peak limiter can flick gently on signal peaks.
If you're seeing either of these limiters engaged often then you probably should have brought more speakers. If it's on a lot, you're may start to be heat up the coil because the signal's peak-to-average ratio goes down - it's getting compressed. That's when a dedicated thermal limiter (with a very slow attack and release in seconds) can save the day, acting as a last line of defence to avoid burnt coils.


There's plenty of reading out there on general practice for limiting. Likewise for simple calculators for various tools. Powersoft provide some useful bits of both, which don't require you to use their amplifiers
https://www.powersoft.com/wp-content/uploads/2019/01/powersoft_TN009_LimiterSetup_en_v1.0.pdf

https://www.powersoft.com/en/ufaq/how-to-calculate-limiter-values/

If you're after a quick answer, I'm afraid there aren't many of those that don't massively over simplify the situation.


Unfortunately, your XTA only features a single RMS limiter, so you're going to have to choose a value that's somewhere between absolute thermal protection and keeping a good amount of headroom in the system.

Edited by toastyghost - 26 December 2022 at 5:47pm
Back to Top
Conanski View Drop Down
Old Croc
Old Croc
Avatar

Joined: 26 January 2006
Location: Canada
Status: Offline
Points: 2315
Post Options Post Options   Thanks (0) Thanks(0)   Quote Conanski Quote  Post ReplyReply Direct Link To This Post Posted: 26 December 2022 at 5:16pm
An RMS limiter offers zero driver excursion protection, it's purpose is to control voice coil heating which occurs over an extended period of time so it completely ignores momentary peaks.

I don't know if this will help or just confuse more but here goes.

As a measurement AES does not equal RMS despite lots of online sources suggesting it does. And if we are being rigid about it RMS is always a voltage measurement, there is no such thing as RMS power, but everybody uses it so we may have to igore that.

The AES test signal has a 6dB crest factor(peak - continuous average ratio), that continous average is what heats a voice coil and if we are talking about power in this case it is 25%(-6db) of the peak value. The RMS value of that same test signal would be 0.707 of the peak value or 50%(-3dB) of power. If you want to thermally protect a driver with an RMS limiter then it has to be set at the AES continuous average, if you set it at the calculated RMS value it offers no protection.  

The peak limiter should be set at the peak voltage that results in driver xmax or maybe xmax+3dB depending how much warmth(distortion) you like to hear. Forget about the published ratings for the drivers, they are calculated(theoretical) and depending on the cabinet design the driver may never be able to handle that much power.

So in your example if a driver hits xmax at 1000w that is what the peak limiter should be set at or perhaps +3db as mentioned above, and the RMS limiter should be set at 1/4 of that or 250w. See the link for some actual driver test results.. PowerPoint Presentation (aesmelbourne.org.au)


Edited by Conanski - 26 December 2022 at 5:30pm
Back to Top
Earplug View Drop Down
Old Croc
Old Croc


Joined: 03 January 2012
Location: Europe
Status: Offline
Points: 6917
Post Options Post Options   Thanks (0) Thanks(0)   Quote Earplug Quote  Post ReplyReply Direct Link To This Post Posted: 26 December 2022 at 5:24pm
In my experience, the maths here are only a rough guide. There really isn't anything that can be cast in stone - like so much when it comes to sound - or engineering in general, when you are balancing several factors that may be in conflict with each other.

It will depend on the processor, amp/s & driver/s.

The best is to get levels set by injecting some noise into the system, listening carefully & maybe monitoring the o/p of the amp with a scope, or multimeter to check on the voltages/clipping.

And if you can also monitor the temperature of the drivers, even better. There is no point trying to squeeze that extra 1/2 dB from the speakers, when you'll just loose 3/4, or more due to overheating. It makes no sense. If you want louder, take more gear!!  Smile



Earplugs Are For Wimps!
Back to Top
Augusts View Drop Down
Registered User
Registered User
Avatar

Joined: 07 January 2017
Status: Offline
Points: 150
Post Options Post Options   Thanks (0) Thanks(0)   Quote Augusts Quote  Post ReplyReply Direct Link To This Post Posted: 26 December 2022 at 7:26pm

Ok, so Powersoft suggests RMS or True Power for a woofer that has a 4" voice coil 250W max to not to overheat.


Then Hornresp says that my simulated speaker RMS is 1000W@4ohms or 63Volts, that hits the Xmax. So I guess than it does not matter how is it called but I know that at that point it hits Xmax.


XTA manual calls the limiter a Program limiter BUT points out that I should be looking for the speaker RMS power rating, and calls the peak limiter - a limiter to protect the speaker from over excursion.


So concluding from all this, and what I read from your posts, the XTA peak limiter should be set to 63Volts and the Program limiter should be set to 1/4 of that, being 250w for a 4" voice coil and plus 6db for the peak limiter.


And if I would like to be very precise with the power rating I should do the test that you suggested and take that as the limit.


I guess my delusion was because in my mind I had calculated that a conventional limiters' treshold is meant to limit the musical material to at say a 1000W and it lets trough only peaks that are shorter in time than for example a 100hz for a subwoofers LP filter that is 10ms. That would mean that the coil would not move further than the treshold set to a 1000W if the attack time is 8ms for example.

Back to Top
Augusts View Drop Down
Registered User
Registered User
Avatar

Joined: 07 January 2017
Status: Offline
Points: 150
Post Options Post Options   Thanks (0) Thanks(0)   Quote Augusts Quote  Post ReplyReply Direct Link To This Post Posted: 26 December 2022 at 7:37pm
But, in that case, I don't understand how all the manufacturers except for D&B Audiotechnik perhaps, provide insane AES and RMS, Program, Peak Power ratings for their speakers. I cant imagine an 18" subwoofer with a 4" voice coil that the specs would say 250rms (or AES) and 500w Program. Maybe 10 years ago.
Back to Top
Augusts View Drop Down
Registered User
Registered User
Avatar

Joined: 07 January 2017
Status: Offline
Points: 150
Post Options Post Options   Thanks (0) Thanks(0)   Quote Augusts Quote  Post ReplyReply Direct Link To This Post Posted: 26 December 2022 at 7:44pm
I have done a lot of speaker simulations and in no case it would be possible to have a 4000W RMS speaker that plays 40hz at that power.
If we assume that that would be the peak power value that a sim like Hornresp provides.
I think its not.


Edited by Augusts - 26 December 2022 at 7:46pm
Back to Top
Augusts View Drop Down
Registered User
Registered User
Avatar

Joined: 07 January 2017
Status: Offline
Points: 150
Post Options Post Options   Thanks (0) Thanks(0)   Quote Augusts Quote  Post ReplyReply Direct Link To This Post Posted: 26 December 2022 at 7:57pm
Ok I get it I think. The simulation simulates 1000W RMS amplifier power. And an amps Vpeak would be 3db more.
Back to Top
Augusts View Drop Down
Registered User
Registered User
Avatar

Joined: 07 January 2017
Status: Offline
Points: 150
Post Options Post Options   Thanks (0) Thanks(0)   Quote Augusts Quote  Post ReplyReply Direct Link To This Post Posted: 26 December 2022 at 8:00pm
But I still don't get where the big guys get 1500W AES and 6000W peak from a speaker enclosure.
Back to Top
Conanski View Drop Down
Old Croc
Old Croc
Avatar

Joined: 26 January 2006
Location: Canada
Status: Offline
Points: 2315
Post Options Post Options   Thanks (0) Thanks(0)   Quote Conanski Quote  Post ReplyReply Direct Link To This Post Posted: 26 December 2022 at 10:01pm
Originally posted by Augusts Augusts wrote:

But I still don't get where the big guys get 1500W AES and 6000W peak from a speaker enclosure.

It's theoretically possible with program material that has 12db crest factor or more, that would result in 375w continuous with 6000w peaks.
Back to Top
toastyghost View Drop Down
The 10,000 Points Club
The 10,000 Points Club
Avatar

Joined: 09 January 2007
Location: Manchester
Status: Offline
Points: 10861
Post Options Post Options   Thanks (0) Thanks(0)   Quote toastyghost Quote  Post ReplyReply Direct Link To This Post Posted: 27 December 2022 at 5:11am
Edit: double post due to terrible signal halfway up a mountain

To make it not be a complete waste of space…

Like most software modelling tools (and real world constant voltage amplifiers and loudspeakers) Hornresp applies a drive voltage to the modelled design, using the value set in the Eg field when defining the model.

It calculates the current demand versus frequency based on the predicted electrical impedance curve.

The power values shown are then calculated from that. This produces Apparent Power, not Real Power.

There are scaling tools on the various plots of data it produces, and some of these allow a wattage value to be entered for convenience. However, this is back-calculated to determine a new drive voltage, before the model’s output values are recalculated.

Edited by toastyghost - 27 December 2022 at 5:19am
Back to Top
 Post Reply Post Reply Page  123 6>

Forum Jump Forum Permissions View Drop Down

Forum Software by Web Wiz Forums® version 12.05
Copyright ©2001-2022 Web Wiz Ltd.

This page was generated in 0.078 seconds.