Print

Please login or register.
1 Hour 1 Day 1 Week 1 Month Forever
Login with username, password and session length

[mojave]

I've been working on modeling some subwoofers. I've never designed or built one, but I've been using WinISD Pro to model drivers in both sealed and ported configurations. Each driver has an X-Max or amount of excursion it has available. For subwoofers this is in the range of 10-40 mm one way. If given the same amount of power, the excursion increases as the frequency decreases. A driver that moves 2 mm at 50 Hz might move 18 mm at 10 Hz with the same amount of power.

In order to keep from damaging your subwoofer at low frequencies either a high pass filter is used to filter out lower frequencies, or the available power is kept to a level that it won't push the driver to much. For example, I can design a sealed subwoofer with a 15" driver that has 23 mm of X-Max. To reach 23 mm of X-Max at 5 Hz, it takes 800 watts. However, to reach X-Max at 20 Hz it takes 1200 watts (which is the maximum power the driver can handle). The higher the frequency, the more power the driver can take before reaching X-Max. There is usually a trade off between dynamic range in the mid-bass vs allowing the driver to reach maximum excursion in the low bass.

I was wondering if it was possible to create a Limiter DSP that allows the driver to always maximize the power available at any frequency. I don't know for sure of anyone that has one, but I think some of the OEM subwoofer amps with DSP have limiting capability (Seaton Sound, SVS). There has also been a request for this with the MiniDSP. If MC could provide the DIY subwoofer community with some more advanced tools, I think that would be great.

Here is what I think might work:

1.  Measure distortion with REW (or other measuring software) at increasing volumes at set frequencies to determine the maximum output you can have at each frequency.

2.  Once the maximum SPL is recorded for the frequencies you could plug these numbers into the limiter DSP. For example 79 dB at 5 Hz, 93 dB at 10 Hz, etc at 11, 12.5, 14, 16, 18 Hz. The limiter DSP could use these to create a limiting curve.

3.  You would need to play a test tone at a certain frequency and record its dB in the Limiter DSP so MC would have a reference between its volume control and your output (unless there is a better method).

4.  The Limiter DSP could then make sure the maximum output was never exceeded for each frequency range.

Is this something that is even possible? Is there a way to do it based on the electronic signal?

What this does is give a lot more power available to your mid-bass while protecting the driver from over-excursion. In my 15" driver example, if you wanted it to play to 10 Hz, you would need to limit your power to about 770 watts. This only gives you output at 30-100 Hz of 105 dB. The LFE channel can have output up to 115 dB and higher when combined with redirected bass. Therefore, this sub isn't capable of reference levels. However, if I can give it the 1200 watts the driver can handle, then I can get output levels of 113 dB or greater over 30 Hz. Now the sub can almost reach reference levels for most material while still providing a lot of low output. In other words, you are giving yourself the maximum headroom that your driver will allow. This seems to be a much cheaper method than adding more subs. It is also of greater benefit when using a less expensive driver with less X-Max.

[mojave]

Here is a cone excursion chart for three different designs. The Limiter DSP would basically have a reverse curve to allow the driver to receive its maximum power.

[Matt]

I hope jmone doesn't hear that you're thinking of building your own subwoofer

[Matt]

In response to the topic, I think it's a great idea.  I also think you're a year or two ahead of us.

For v16, my signal processing goals are a nice analyzer and a parametric equalizer.

[mojave]

I also think you're a year or two ahead of us.

I'm also probably a year or two ahead of myself.    I may never even use such a feature.

It took almost 8 years for the audio to by synced with the video. 

From February 10, 2003:

You can also stack the filter graphs to include a reclock graph which assists in establishing a perfect sync between the audio and video decoders, this eliminates those micro stutters which have plagued various decoder filters for years due to the compromise that one clock has to be the master clock (audio or video) and breaking an audio stream makes a crack or pop and is very perceptable...

[mojave]

In response to the topic, I think it's a great idea.  I also think you're a year or two ahead of us.

For v16, my signal processing goals are a nice analyzer and a parametric equalizer.

It looks like you are catching up! 

[Matt]

This is funny.

I told you that you were a year or two ahead of me!

[mojave]

5. NEW: Added 'Subwoofer' mode to Parametric Equalizer > Limiter (an adaptive limiter well suited for subwoofer use, good after a Linkwitz, etc.).

This looks like a great addition. Can you explain what the subwoofer mode does?

Recently I was experimenting with limiting by splitting the subwoofer into User1 (above 20 Hz) and User2 (below 20 Hz) and applying limiting to User1 User2. Then I would combine them back again and route to Subwoofer. This worked well and the signal would stay the same until high output and then just below 20 Hz would be limited.

[Matt]

Because of the Linkwitz in my home system, I sometimes get distortion on the subwoofer if the volume is high.  This is because at close to full volume, there's no extra headroom from Internal Volume.

This new limiter is just an easy way to add an adaptive limiter that works well for a subwoofer.  I wrote the adaptive limiter, and still think understanding the parameters is a little tough.

The subwoofer limiter hard-clamps at full volume, only holds a little, and recovers lost volume quickly.  This way, a huge bass hit that exceeds the limit will still sound reasonably good (just not as loud), but the next bass hit a couple seconds later that's not quite as loud will be reproduced as intended.

In initial testing, it works well.  I should know even more after living with it a few months and watching a variety of movies.

[mojave]

I was backwards in what I stated I was limiting in my previous post. I have fixed it.

I currently have 3 different DIY subs and 4 different amps I could use with them. Depending on sub/amp combination, the limit can be reached much sooner than full volume. Maybe you should allow for a user selectable dB setting. I can currently do this manually by increasing user2 by 5 or 10 dB, applying the limiter, decreasing user2 by the same dB amount, and then adding user2 and user1 back to subwoofer. By increasing the volume, I'm forcing to limiter to act when I want it to and by splitting the frequencies I'm allowing the limiter to only affect the bass below a certain frequency. I can see both of these being beneficial to you.

[Matt]

Are you sure it makes sense to split the signal based on frequency, limit part of it, and add it back together?  Is this common practice?

I'm trying to understand if this is better than just turning all frequencies down equally.  This approach has the advantage that it limits what actually goes to the speaker (voltage) where the other method still adds multiple signals and coherence / incoherence will cause you trouble.

[Matt]

It's worth mentioning that an adaptive limiter will limit low frequencies more than high frequencies in practice.  This is because low frequencies are the only frequencies that use a voltage / speaker excursion large enough to hit the limiter.

There is an issue of what happens if there's a 20 Hz hit and 50 Hz it at the same time.  Each approach would sound different, but I'm not sure which approach would sound better.

[mojave]

Are you sure it makes sense to split the signal based on frequency, limit part of it, and add it back together?  Is this common practice?

It's worth mentioning that an adaptive limiter will limit low frequencies more than high frequencies in practice.  This is because low frequencies are the only frequencies that use a voltage / speaker excursion large enough to hit the limiter.

There is a recent thread at AVS that gets into a discussion about limiting and might be worth reading. Mark Seaton says the following in that thread: "While I know the upper octave of the subwoofer range are fairly easy for most woofers to produce, a 60Hz high level signal requires just as much Voltage as a 16Hz signal, and if you produce equal signals at the same time, that requires 2x the peak Voltage. This reality with program material is not well demonstrated with sweeps which only sweep a single frequency, but can be hinted at in the maximum capabilities, and is also where the compromise of using EQ to achieve a response vs. picking a driver with deeper natural response plays out, particularly if real use will tickle the limits of the subwoofer system."

Mark Seaton uses DSP amplifiers from Speaker Power. Danley Sound Labs also use these DSP amplifiers. I think Mark mentioned once that he splits the bass signal so he can specify where the limiting occurs. I also think I've read of others doing the same. If you look at the testing of subs at data-bass.com you can see that some of the consumer subwoofers with DSP amps have limiting that occurs below a certain frequency when a high level is reached.

I have tried to read the manuals for the Peavey IPR DSP amp and the Behringer iNuke DSP amp, but they don't say much about how their limiter works. I don't think it is common practice, but I think it is done.

I experimented with a clip from a movie (I think Iron Man's Jericho scene) and unless I split the frequencies, I still got limiting at the higher frequencies even though I didn't need it. I was going to split the frequencies and combine them again without and limiting and see if they were identical, but I never got around to it. In other words I wanted to see if splitting the signal was a detriment in any way.

[Matt]

a 60Hz high level signal requires just as much Voltage as a 16Hz signal

That's true, unless you're adding a Linkwitz.  Then you're boosting the 16Hz signal a lot more than the 60Hz signal.

I experimented with a clip from a movie (I think Iron Man's Jericho scene) and unless I split the frequencies, I still got limiting at the higher frequencies even though I didn't need it.

I'm intrigued by the idea of frequency based limiting.  But I'm not sure how to handle the math.  Imagine:

Split [Input] to [Low] and [High].  Imagine the goal is to limit [Low] + [High] to -1 to 1.

Do you limit both [Low] and [High] to -0.5 to 0.5?  This would mean [Low] couldn't use the value 1, even if [High] was at zero.

Or do you wait for [Low] + [High] to cross the limit, and then turn down _only_ [Low] until you're under the limit?  

The second approach seems reasonable, but what if [High] alone puts you over the limit?

And do you want to take [Low] right down to 0 to preserve [High]?

So maybe you should have a limiting ratio where you limit 80% [Low] and 20% [High]?

The slope of filter used to make [Low] / [High] would naturally allow some [Low] through in [High] and vice-versa, so that also has to be considered.

[Matt]

The general idea is that limiting is applied as a smooth frequency curve that limits lower frequencies more strongly.

One way to do this might be to base the limiter around a high-pass filter (probably at 24 dB/octave, but we would have to test).

The limiting would be accomplished by varying the frequency of the filter.

If no limiting was happening, the frequency would be at 0 or 1 Hz.

If lots of limiting was happening, the frequency might be like 100 Hz.  This would make 50 Hz -24 dB and 25 Hz -48 dB.

The only question I have is if continuously adjusting the frequency of the high-pass would lead to phasing artifacts, or have a delay that causes problems (so a peak would slip through before it could adapt).

[Matt]

I don't think the sliding cross-over frequency idea is a keeper.  When you change a parameter of an IIR filter, you need to reset it which causes it to change behavior for a little bit.  That means it's not a good candidate for being continually adjusted.

Instead, I think a series of a few cross-over filters would be better.  I think you could break the signal into a two or three buckets, then limit each bucket separately.  The limiter would only engage when the sum of all the buckets exceeds the set limit, meaning it would work better than the approach you mentioned using in Parametric Equalizer.  As long as it's designed properly, the sum of all the buckets should make no changes when the limiter is not engaged.

But I need to figure out conceptually the right way to engage the limit.  Should it be willing to bring 20Hz down to zero to preserve 80Hz?  Or is there some relationship, where it's willing to bring 20Hz down 4x as much as 80Hz or similar?

In other words, what should limiting do with the following scenarios to get the total to 100% or less:
20Hz: 200%;  80Hz: 0%
20Hz: 0%;  80Hz: 200%
20Hz: 200%;  80Hz: 100%
20Hz: 100%;  80Hz: 200%

[Matt]

I'm sort of talking to myself here, but I think I've figured out something that would work well.

You take the incoming signal and split it with a low-pass / high-pass combination at like 50 Hz with 12dB/octave filters.  This gives you the low and high components.

Then, when the signal crosses the limiting threshold, you run the limiter in two stages.

First, you bring the low all the way down to zero.

If it still needs limiting, you bring the highs down.

Since you have a 12dB/octave slope on the highs, even when you bring the lows all the way to zero you still have a nice smooth frequency response slope.

I think this whole thing results in a reasonably simple design that would nicely accomplish the goal of limiting the lowest frequencies more than higher frequencies.

[JustinChase]

I'm sort of talking to myself here

sort of?? 

[Matt]

An improved subwoofer limiter is in tonight's build.

Thanks to mojave for the idea (almost two years ago).

The actual implementation tuned some of the parameters I listed above, but the general idea is the same.

(p.s. if you had tried the subwoofer limiter in past builds, please remove it and add one of the new style; the old style was removed)

[mojave]

I just tried this out and can't notice any difference whether it is on/off or varying dB level settings. I downloaded and played Jurassic Lunch which has content down to 10 Hz. I played back multiple times and watched it in GlissEQ. I had all my amps off and the volume set to 100%

[Matt]

I tested that same THX sample.

First, I added +6dB so it would actually clip.

It looks like this at 45 seconds:


Then I added a subwoofer limiter to the left and right channels (there's no subwoofer channel in this clip).  Then it looks like this at 45 seconds:


See how the energy slope is much steeper at the low frequencies with the limiter engaged?  A regular limiter wouldn't change the slope at all.

So it seems to be working for me.

[mojave]

I had a chance yesterday to get back to this. I wasn't seeing anything happen because I didn't add any gain in.

I played around yesterday with pink noise to try to figure out what was going on. I'm trying to understand the relationship between the internal volume control, the level field in the limiter, and adding gain before the limiter.

Example A
+10 dB (Left, Right)
Subwoofer Limiter (0,Left)
Subwoofer Limiter (0,Right)

With the internal volume at -10 dB I get the first image. With the internal volume at 0 dB, I get the second image. Is the Analyzer representing the actual db levels? If so, the sub is quieter with the internal volume at 0 dB than it is at -10 dB. The low frequencies should be staying the same with the upper frequencies still increasing in gain. If the Analyzer just shows relative levels, then it makes more sense.

What exactly does the level field do? I've tried it with both positive and negative numbers with different results and don't really know what I'm supposed to enter.

[Matt]

Is the Analyzer representing the actual db levels?

The 'Options' button at the top right of Analyzer lets you pick if it should process independently of Internal Volume (-10 dB in your case).  I think the default is to process independent of internal volume, or else the graph is almost invisible if you turn down the volume.

What exactly does the level field do? I've tried it with both positive and negative numbers with different results and don't really know what I'm supposed to enter.

It's dB.

0.0 is the normal full level (100%).

If you enter something like -6.0, it will limit before it gets to full range, so the peak level will only go to like 50%.

[mojave]

If you enter something like -6.0, it will limit before it gets to full range, so the peak level will only go to like 50%.

The field isn't intuitive that it should be a negative number. I had been putting in positive numbers and just this morning realized it should probably be a negative number. I was thinking "I want to limit by 6 dB so I should enter 6 in the field."

You take the incoming signal and split it with a low-pass / high-pass combination at like 50 Hz with 12dB/octave filters.  This gives you the low and high components.

Then, when the signal crosses the limiting threshold, you run the limiter in two stages.

Is this how you ended up doing it? If so, can the split frequency be made variable? As I mentioned earlier, since I started this thread I have made 3 different sub systems. I would use 50 Hz on one, 30 Hz on one, and 20 Hz on the last. Above those frequencies I am amp limited so don't need much limiting. However, below those frequencies I am driver limited and might need limiting.

I'm sort of talking to myself here, but Matt did a good job on this. 

[hulkss]

What would be great is if your clever limiter does not introduce group delay distortion.

For an ideal filter, the phase will be linear and the group delay would be constant. Group delay, whose unit of measure is time in seconds, can also be thought of as the propagation time delay of the envelope of an amplitude modulated signal as it passes through a digital filter. Group delay distortion occurs when signals at different frequencies take different amounts of time to pass through a filter.

Any typical high pass subwoofer protection filter has a problem with group delay.