AVS Forum banner

1 - 20 of 26 Posts

·
Registered
Joined
·
367 Posts
Discussion Starter #1 (Edited)
This is what my front speakers look like after room correction. They are effectively flat down to 40 Hz.


My chain is Bass Management > Dirac > Amplification. The bass management looks like this.

Step 1: Make a copy of the unmodified channels (to be low-passed for the subwoofer later).
Step 2 to 4: LR-4 high-pass the channels at 80 Hz for the speakers.
Step 5 to 7: LR-4 low-pass the unmodified copy at 80 Hz.
Step 8: Mix the low-passed copy with the LFE channel at the correct levels for the subwoofer.

A source LFE channel gain of -8.2 dB is required for a 5.1 channel system to prevent digital clipping after bass management. The low-passed channels must thus be mixed in at -18.2 dB (-10 dB relative to the LFE channel). I rounded the values to -8.5 dB and -18.5 dB because my amp only has increments of 0.5 dB. My subwoofer is amplified +18.5 dB more than the speakers so the LFE component has a +10 dB gain during playback.

When applying an LR-4 high-pass filter cornered at 80 Hz, the speakers will be -6 dB at 80 Hz and -24.6 dB at 40 Hz. When summed with the corresponding low-passed signal from the subwoofer, the net response will be 0 dB. However, this is only true if the (corrected) speakers are actually flat down to 40 Hz. If the speakers roll off above 40 Hz, the high-pass filter will produce a different slope and the sum with the subwoofer will not be 0 dB. So with a simple LR-4 crossover, you must "waste" an entire octave above the (already high) roll-off point.

The ideal solution would be to design a custom high-pass filter based on the roll-off slope of the (corrected) speakers, such that the combined slope is the same as an LR-4 filter. This would sum correctly with the subwoofer. Unfortunately I don't know the math required to do this.

Instead, I wonder if I can tweak the target curve in Dirac to form a 12 dB / octave slope from 40 Hz to 80 Hz. I can then apply a 12 db / octave Butterworth high-pass filter to the channels, which would sum with the (corrected) roll-off to form an LR-4 filter. The uncorrected signal below 40 Hz would not sum correctly, but I'm assuming that -24.6 dB and below is too soft to make an audible difference. Would this work correctly with Dirac's mixed phase filters? Would Dirac even generate a "smooth" slope like a real high-pass filter that is suitable for summing with another filter?

Most AVRs incorrectly apply a 12 dB / octave slope to the speakers under the false assumption that they are THX speakers which already have a 12 dB / octave roll-off below 80 Hz. If you have an AVR, your crossover is probably wrong.
 

·
Registered
Joined
·
90 Posts
A source LFE channel gain of -8.2 dB is required for a 5.1 channel system to prevent digital clipping after bass management.
I don't get how you come to this statement.

The LFE channel on disc is mastered at -10dB to create more headroom in the mix (to be able to hit 115dB when listening at reference level).

I was under the impression that you have to boost the LFE +10dB, then merge in the low passed content from the other channels. This means that you need to reduce the total gain of the system by a lot to avoid digital clipping. If you pipe in 40Hz at max amplitude into all channels, you would end up with 8 times the original amplitude on the lfe input to Dirac. To be certain that there will be no clipping, ALL channels should be reduced by 20 * log10(8) = 18.062dB.

So IMO the correct bass management would be:
1. Reduce all full range channels by 18.062 dB. and the LFE channel by 8.062 dB.
2. Copy all full range channels.
3. Low-pass the copies of the channels with LR-2 (12dB/oct).
4. High-pass the originals with LR-2.
5. Sum low-passed signals into LFE signal.

Then send the output to Dirac live for room correction. Dirac will apply the equalization gains and delays to the individual channels to make sure distances are compensated for.

My 0.02 cents.
 

·
Banned
Joined
·
15,491 Posts
How low your mains go is of far less significance than how loud they'll go in the lows, compared to how loud the subs will go. If your mains are flat to 40Hz but they run out of excursion long before the subs do at 40Hz then you're crossing too low.
 

·
Registered
Joined
·
367 Posts
Discussion Starter #5
How low your mains go is of far less significance than how loud they'll go in the lows, compared to how loud the subs will go. If your mains are flat to 40Hz but they run out of excursion long before the subs do at 40Hz then you're crossing too low.
How much headroom above the roll off point is typically required for speakers like mine (Wharfedale Diamond 155 - Specifications)? I rarely listen above 85 dB SPL.
 

·
Registered
Joined
·
12,738 Posts
+1 to Bill.

Check their distortion with a 40 Hz output signal at 95 to 105 dB and you'll probably want to keep the crossover at 80 Hz. Most speakers distort heavily when presented with large LF signal, and LF signals tend to be by far the largest, so you are giving up a lot of headroom in the speakers (and amplifiers driving them) and picking up a lot of distortion crossing them too low. Lookup Fletcher=Munson loudness curves on Wikipedia or whatever and note that LF signals need to be 20 dB or more higher than midrange signals to sound as loud so LF's tend to be much larger even if they don't sound as loud. In power, 20 dB is a factor of 100.

Your speakers are not real sensitive and do not have real high power handling so you'll end up with better sound overall crossing them higher. You'll find the midrange cleans up when they aren't driven so hard.

My rule of thumb has been to cross over an octave above the speaker's -3 dB point if possible and reasonable; certainly at least a half-octave.

FWIWFM - Don
 

·
Registered
Joined
·
367 Posts
Discussion Starter #7 (Edited)
I was under the impression that you have to boost the LFE +10dB, then merge in the low passed content from the other channels. This means that you need to reduce the total gain of the system by a lot to avoid digital clipping. If you pipe in 40Hz at max amplitude into all channels, you would end up with 8 times the original amplitude on the lfe input to Dirac. To be certain that there will be no clipping, ALL channels should be reduced by 20 * log10(8) = 18.062dB.
The gain structure should be optimized by adding the +10 dB in the subwoofer amplifier, not by attenuating the main source channels.

Subwoofer = -8.5 dB (LFE) + -18.5 dB (L) + -18.5 dB (R) + -18.5 dB (C) + -18.5 dB (SL) + -18.5 dB (SR)
This sum has a maximum possible amplitude of less than 0 dBFS. It cannot clip even if all channels (including LFE) are full scale.
If the subwoofer amplifier is set to +18.5 dB, the low-passed content will be amplified back to 0 dB (the same as the speakers) and the LFE component will be +10 dB.
The main channels' gain is never altered and is always 0 dB from source to playback.

The commonly used -5 dB (LFE) + -15 dB (Mains) does clip if all the channels are full scale. It clips by +3.2 dB for 5.1 channels and even more for 7.1 channels.

When introducing Dirac into the chain, calibration should be performed with the 5.1 System preset and the subwoofer amplifier at +10 dB, or with the Custom System preset and the subwoofer amplifier at 0 dB. This will level match the speakers and subwoofer accordingly. The subwoofer amplifier gain should be manually adjusted to +18.5 dB after calibration is complete.
 

·
Registered
Joined
·
367 Posts
Discussion Starter #8
Nearly forgot: I would also add a 120Hz LR-4 low pass on the final LFE signal.
What is the point in doing this? Why remove content from the LFE channel? The worst that could happen is the subwoofer would produce a localizable sound that is actually present in the content.
 

·
Registered
Joined
·
12,738 Posts
The original LFE bandwidth spec was 120 Hz with a sharp roll-off above that. I have been told that has been raised but I couldn't find the reference after a quick search. The 120 Hz LPF is to suppress images that come about from sampling the signal (see sampling theorem and Nyquist sampling criteria).
 
  • Like
Reactions: ChromeJob

·
Registered
Joined
·
939 Posts

·
Registered
Joined
·
90 Posts
The gain structure should be optimized by adding the +10 dB in the subwoofer amplifier, not by attenuating the main source channels.

Subwoofer = -8.5 dB (LFE) + -18.5 dB (L) + -18.5 dB (R) + -18.5 dB (C) + -18.5 dB (SL) + -18.5 dB (SR)
This sum has a maximum possible amplitude of less than 0 dBFS. It cannot clip even if all channels (including LFE) are full scale.
If the subwoofer amplifier is set to +18.5 dB, the low-passed content will be amplified back to 0 dB (the same as the speakers) and the LFE component will be +10 dB.
The main channels' gain is never altered and is always 0 dB from source to playback.

The commonly used -5 dB (LFE) + -15 dB (Mains) does clip if all the channels are full scale. It clips by +3.2 dB for 5.1 channels and even more for 7.1 channels.

When introducing Dirac into the chain, calibration should be performed with the 5.1 System preset and the subwoofer amplifier at +10 dB, or with the Custom System preset and the subwoofer amplifier at 0 dB. This will level match the speakers and subwoofer accordingly. The subwoofer amplifier gain should be manually adjusted to +18.5 dB after calibration is complete.
Yes, I can see that it makes a lot of sense to add the +10 dB in the amplification at the end of it all, to preserve gain in the main channels.

What does not make sense to me are the -8.5 and -18.5 numbers. How did you get to these?

This is my understanding of things:
In a 5.1 system, you can merge up to 6 channels together. If we consider all channels to be at full scale (1.0) we can calculate the gain as: 20 * log10(6/1) = 15.563 dB. So for me, the proper setup would be:
1. Split the main signals in low and high frequency.
2. Apply a -5.563 dB gain to the high frequency signals.
3. Apply a -15.563 dB gain to the low frequency signals.
4. Apply a -5.63 dB gain to the LFE channel.
5. Merge low frequency signals in LFE channel.
This will maintain the 10 dB gain difference between the LFE and main channels. (To be corrected later by the amplifier gains.)

For a 7.1 system, the numbers change somewhat as 20 * log10(8/1) = 18.062 dB.

In equalizer APO, you are using float32 calculations, which means you do not loose dynamic range by lowering gain levels. You could simplify your setup to:
1. Apply global -5.563 dB gain.
2. Split the main signals in low and high frequency.
3. Apply -10 dB gain to low frequency channels.
4. Merge low frequency channels with LFE.

Does any of this make sense? Or do I have it completely wrong?
 

·
Registered
Joined
·
90 Posts
When introducing Dirac into the chain, calibration should be performed with the 5.1 System preset and the subwoofer amplifier at +10 dB, or with the Custom System preset and the subwoofer amplifier at 0 dB. This will level match the speakers and subwoofer accordingly. The subwoofer amplifier gain should be manually adjusted to +18.5 dB after calibration is complete.
Are you sure? AFAIK, the 7.1 and 5.1 presets will boost the LFE channel by +10 dB in the Dirac filter, where using the custom preset will not apply this gain. I think the only correct settings are:
1. The Custom System preset, with the sub amp at +10 dB. (i.e. Use the amp to add the +10 dB in the analog domain)
2. The 5.1 or 7.1 preset, with the sub amp at 0 dB. (i.e. Add the +10 dB in the digital domain)
 

·
Registered
Joined
·
367 Posts
Discussion Starter #15 (Edited)
What does not make sense to me are the -8.5 and -18.5 numbers. How did you get to these?Does any of this make sense? Or do I have it completely wrong?
You are attenuating the high passed signals unnecessarily, resulting in a higher noise floor from the speakers during amplification. It's better to apply higher attenuation and higher amplification to the subwoofer signal alone, since noise will not be audible from the subwoofer. EMI is an audible problem for me since my connections are analog.

Assuming all source channels are 0dBFS, the standard bass management formula is:
-5 dB (LFE) + -15 dB (5x Channels) = +3.2 dBFS (Clipped)

Applying an additional gain of -3.2 dB to the source channels results in:
-8.2 dB (LFE) + -18.2 dB (5x Channels) = 0 dBFS

Are you sure? AFAIK, the 7.1 and 5.1 presets will boost the LFE channel by +10 dB in the Dirac filter, where using the custom preset will not apply this gain. I think the only correct settings are:
1. The Custom System preset, with the sub amp at +10 dB. (i.e. Use the amp to add the +10 dB in the analog domain)
2. The 5.1 or 7.1 preset, with the sub amp at 0 dB. (i.e. Add the +10 dB in the digital domain)
You've got it backwards. Dirac does not arbitrarily add 10 dB to the LFE channel with the 5.1 System preset. It ensures that your system is already +10 dB LFE. Similarly with the Custom System preset, it ensures that the LFE is already at the same level as the other channels. If the LFE is too loud, it will attenuate it. If the LFE is too soft, it will attenuate all the other channels. You should pre-adjust your amplifier so that Dirac does not do anything to correct the LFE level, to minimize digital attenuation.
 

·
Registered
Joined
·
367 Posts
Discussion Starter #16 (Edited)
In equalizer APO, you are using float32 calculations, which means you do not loose dynamic range by lowering gain levels. You could simplify your setup to:
You don't lose dynamic range in the digital domain since it's 32-bit, but the final analog signal is lower from the sound card resulting in audible noise. Even when using a digital connection like HDMI, the signal is softer and will require more amplification. Increasing the bit-depth increases the resolution, but the maximum volume remains the same.

I believe my setup is simpler than what you proposed.
 

·
Registered
Joined
·
90 Posts
Will look into the Dirac thing a bit more...

Assuming all source channels are 0dBFS, the standard bass management formula is:
-5 dB (LFE) + -15 dB (5x Channels) = +3.2 dBFS (Clipped)

Applying an additional gain of -3.2 dB to the source channels results in:
-8.2 dB (LFE) + -18.2 dB (5x Channels) = 0 dBFS
Maybe I am particularly dense, but I don't get where these numbers come from. What standard are you referring to?
 

·
Registered
Joined
·
367 Posts
Discussion Starter #18 (Edited)
Will look into the Dirac thing a bit more...
If you choose the 5.1 System preset, your subwoofer should already be +10 dB so that Dirac will not do anything.
If you choose the Custom System preset, your subwoofer should already be 0 dB so that Dirac will not do anything.

After calibration, you must manually turn it up to whatever is required to bring the low-passed components to the same level as the speakers and the LFE component to +10 dB, which is entirely dependent on the bass management method.


Maybe I am particularly dense, but I don't get where these numbers come from. What standard are you referring to?
The built-in bass management on most AVRs / HTIBs / PC speaker systems use -5 dB (LFE) + -15 dB (LPF(Mains)), and amplify the subwoofer by +15 dB. This does clip in in theory, but in practice, even during intense scenes, the surround channels will usually be much softer so the mixed signal will not clip.

There is no need to touch the high-passed signal like you are doing.
 

·
Registered
Joined
·
367 Posts
Discussion Starter #19 (Edited)
The practice of attenuating the high-passed channels was popularized by jRiver, because it was the easiest thing to do in the context of their software, without having people manually mess with the subwoofer level before and after calibration etc. But it's a destructive step for bass management in general.
 

·
Registered
Joined
·
367 Posts
Discussion Starter #20 (Edited)
What does not make sense to me are the -8.5 and -18.5 numbers. How did you get to these?
Try to solve this; it will help me verify my numbers too.

  • You are not allowed to modify the source channels or the high-passed signals.
  • You can only attenuate the low-passed signals and the LFE channel and mix them.
  • All source channels are 0 dBFS, and the subwoofer mix must not clip above 0 dBFS.
  1. What is the attenuation on the low-passed signals?
  2. What is the attenuation on the LFE channel?
  3. What is the gain on the subwoofer amplifier?
 
1 - 20 of 26 Posts
Top