Bass-management headroom - how to avoid clipping your sub out

[Kvalsvoll]

Clipping of the low frequency signal somewhere in the chain is a real problem, and care must be taken to ensure this does not happen. This is a solved problem, it is known. Still, I can not see that this issue is tested for in the processor/avr reviews here on this site.

Sub out from processor/avr is the input signal to the bass-system, and in a decent cinema system, this system is designed for both full frequency range and necessary output capacity, which is >125dB.

LFE is +10dB which ends up at 115dB for 0dB. This does not mean that the sub out must handle 10db higher voltage, it just means that for a 0dB signal the spl from the bass-system should be 115dB. So the requirement for output voltage is no different from other channels.

The problem is when bass from other channels is routed to the sub channel, and added together. In total, a full worst case signal ends up at 125dB for 7.1. This means there needs to be an additional 10dB headroom in the chain to ensure no clipping occurs. You would think the manufacturers have full control of this, after all, it is not that difficult to get right. On the other hand, in a business where a power cable is claimed to improve sound, well, who knows. And indeed, it turns out this is a problem in MOST processor/avrs's, and for some it is possible to fix it by choosing a proper gain structure,

First possibility for disaster is clipping the signal digitally when summing the bass inside the dsp. Then there is no salvation, the signal is destroyed.

Then there is the possibility of clipping the sub out. This is possible to avoid with proper trim level settings and gain on the bass system.

I use Denon/Marantz in my systems, so I can provide information on how they work and how to avoid clipping on those, perhaps coming later.

 

[Kvalsvoll]

For Denon/Marantz units:

Max output level is 4Vrms, but distortion increases above approx 2Vrms (I believe Amir's reviews confirms this). So:

- Max 4Vrms
- Try to avoid >2Vrms

Clipping is easy to detect by looking at the signal on the sub out. First thing is to verify that it is possible to get a clean non-distorted sum signal of all 7 channels. Send a 0dB signal on LCR+SR(4x)+LFE, check clipping and output voltage. Here I found that the signal is not clipped inside the processor, and the output level is:

4Vrms for MV 0dB and trim level -12dB, with full 7.1 signal in.

This means the trim level must be at its lower limit at -12dB to avoid clipping of the sub out signal. It also means that any combination of trim level + MV must be less than -12dB to ensure a clean signal.

The full 7.1 is an extreme scenario that occurs only in a very few movie scenes in some very special movies. Most bass-systems also have problems reproducing full-scale 125dB clean, the bass level is actually higher than that due to the tilted target frequency response with some lift in the bass, so in practical installations we are asking for around 130-135dB. To dimension for LFE+LCR only, seems reasonable, and that requires 120dB (actual in-room spl 125-130dB). For 120dB, there is 5dB trim level left to play with.

 

[QMuse]

..the bass level is actually higher than that due to the tilted target frequency response with some lift in the bass, so in practical installations we are asking for around 130-135dB.

Room EQ systems never lift anything, all lifting is done on account of reduced headroom and output level never exceeds 0dB. If that wouldn't be the case digital clipping would occur.

With the tilted target curve you're mentioning HF is actually reduced and LF is left as it is to achieve tilted curve.

 

[Kvalsvoll]

Room EQ systems never lift anything, all lifting is done on account of reduced headroom and output level never exceeds 0dB. If that wouldn't be the case digital clipping would occur.

With the tilted target curve you're mentioning HF is actually reduced and LF is left as it is to achieve tilted curve.

I see my description is confusing. The actual spl level in the room has no relevance for sub output clipping, it is mentioned only to put thing into perspective, to explain why it may not be necessary to have full capacity.

All examples above are with no room-correction enabled on the processor/avr. All corrections are handle by separate dsp units anyway, but that is another discussion very far off the scope for this thread.

The point here is to verify and if necessary adjust gain structure to avoid clipping of a bass-managed signal with full level on all channels.

 

[QMuse]

I see my description is confusing. The actual spl level in the room has no relevance for sub output clipping, it is mentioned only to put thing into perspective, to explain why it may not be necessary to have full capacity.

I can see you are in the sub business but the idea of say 40Hz tone played at more than 110dB is out of my music enjoyment perspective.

 

[Kvalsvoll]

I can see you are in the sub business but the idea of say 40Hz tone played at more than 110dB is out of my music enjoyment perspective.

Fortunately you don't need to, because there is a volume control. But that will not prevent distortion if the signal is clipped inside the processor.

Trying to tell the story slightly different; 2 things can go wrong:

1. Bass-managed signal is clipped inside the processor, due to digital clipping or clipping in any analog stage before master volume control. If this happens, the distortion will be present at any volume.
2. Bass-managed signal is clipped at preamplifier output stage, due to too high voltage. This may be possible to prevent by lowering trim level on sub out. If you turn down the volume, say play at -20dB, this will not be a problem.

These are very real and very audible faults, and should be tested for in the reviews.

 

[QMuse]

Fortunately you don't need to, because there is a volume control. But that will not prevent distortion if the signal is clipped inside the processor.

Trying to tell the story slightly different; 2 things can go wrong:

1. Bass-managed signal is clipped inside the processor, due to digital clipping or clipping in any analog stage before master volume control. If this happens, the distortion will be present at any volume.
2. Bass-managed signal is clipped at preamplifier output stage, due to too high voltage. This may be possible to prevent by lowering trim level on sub out. If you turn down the volume, say play at -20dB, this will not be a problem.

These are very real and very audible faults, and should be tested for in the reviews.

That was a joke.

I fully understand the problem you're talking about although I am quite surprised learning that processors are handling bass management so poorly.

 

[Kvalsvoll]

That was a joke.

I fully understand the problem you're talking about although I am quite surprised learning that processors are handling bass management so poorly.

Joke or not, some people never want loud bass, and I respect that. I even recommend bass-systems that can never play very loud and will never shake the house. And good bass does not need to be loud.

Yes, this came as a surprise. Discussion around this must have been on data-bass, this was at least a couple of years ago. Denon-Marantz were among few that did not clip the signal. Among the crowd with multiple 18" driver systems or large horns, this obviously did not fare well.

 

[QMuse]

Joke or not, some people never want loud bass, and I respect that. I even recommend bass-systems that can never play very loud and will never shake the house. And good bass does not need to be loud.

Personally I believe I'm with the "typical" crowd as 10dB tilted down curve works for me. My speakers can reproduce 40Hz at 105dB level at my LP (app 3.5m from the speakers) with <2% THD and that is enough for me as I usualy listen at average level of 75-80dB.

 

[Martin_320]

Trying to tell the story slightly different; 2 things can go wrong:

1. Bass-managed signal is clipped inside the processor, due to digital clipping or clipping in any analog stage before master volume control. If this happens, the distortion will be present at any volume.

Hello,
I have a Marantz pre-pro (AV7705) and sometimes I dis-engage my large sub (in the 5.1 speaker config settings) and set the front L/Rs to "Large" and all the others to "Small".
This way the bass-management routes the bass below 80Hz to the mains -- ie on top of the bass that they would already normally handle.
Doing so I can definitely hear more bass coming out of the front L/R speakers. And they seem to be fine (no distortion).
Would you happen to know whether in its digital DSP Marantz has extra bits reserved specifically as headroom for bass-management operations?
For example, say a Flac with a 0dB peak in all channels simultaneously, pre-DSP, would be a 24bit PCM word (for each channel) composed entirely of ones, ie -- 111111111111111111111111
Use-case / Question: Would the Marantz's Bass Management DSP scale this 24bit word (the source) into a 32bit (or 64bit) word, and do so in floating-point math, and without using-up all of the 32 or 64 bits?-- ie to leave some extra bits available for the 10dB of digital headroom? Is that how they avoid digital clipping when summing the bass from several channels into one sub channel or to the L&R "Large" speakers?
Most grateful for your advice!
Rgds, Martin.

 

[Kvalsvoll]

Hello,
I have a Marantz pre-pro (AV7705) and sometimes I dis-engage my large sub (in the 5.1 speaker config settings) and set the front L/Rs to "Large" and all the others to "Small".
This way the bass-management routes the bass below 80Hz to the mains -- ie on top of the bass that they would already normally handle.
Doing so I can definitely hear more bass coming out of the front L/R speakers. And they seem to be fine (no distortion).
Would you happen to know whether in its digital DSP Marantz has extra bits reserved specifically as headroom for bass-management operations?
For example, say a Flac with a 0dB peak in all channels simultaneously, pre-DSP, would be a 24bit PCM word (for each channel) composed entirely of ones, ie -- 111111111111111111111111
Use-case / Question: Would the Marantz's Bass Management DSP scale this 24bit word (the source) into a 32bit (or 64bit) word, and do so in floating-point math, and without using-up all of the 32 or 64 bits?-- ie to leave some extra bits available for the 10dB of digital headroom? Is that how they avoid digital clipping when summing the bass from several channels into one sub channel or to the L&R "Large" speakers?
Most grateful for your advice!
Rgds, Martin.

How the processing is done inside Marantz I do not know, you have to ask Marantz about that. But I have measured Marantz and Denon processor/avr's, and they can handle a full 0dB signal on all 7.1 channels with no clipping if trim level on sub out is set to -12dB.

The 0dB on all channels is a very rare situation, I can not recall any movies with such a signal present. In my systems, I am confident all practical situations are handled with a trim level of say -8dB on sub out.

 

[Frontino]

LFE channels in some soundtracks already come with saturated headroom and peak limited just below full scale digital, so I seriously recommend to not use Bass Management ever, if you care about your subwoofers.
I know that setting all mains to Large might compromise clean output from small speakers, but so does downmixing.
Besides, you already know the saying "pick your poison".

 

[JStewart]

Its even better the processor manufacturers will allow the LFE channel to be boosted via the trims.

I am fortunate to have Monolith HTP-1 which shows digital signal level on all channels in real time and a user accessible setting for reserved digital headroom. 12dB is default.

I infer from this the capabilities are baked into the dsp chips and manufacturers are aware. I also expect they look for the best compromise in default settings and further assume they don’t provide clipping indicators because they don’t want to have to explain what we’re discussing to the average owner and or explain the loss of SPL the owner might experience to fix a clipping problem.

 

[Frontino]

Don't forget that the coherent summing of 2 channels in amplitude is +6 dB, not 3.
And just because you can, does not guarantee it won't sound overly bloated.

 

[sarumbear]

The problem is when bass from other channels is routed to the sub channel, and added together. In total, a full worst case signal ends up at 125dB for 7.1.

May I ask how you reached to that 125dB level?

 

[Kvalsvoll]

May I ask how you reached to that 125dB level?

115dB lfe + 7x 105db (lcr + 4x sr)

 

[sarumbear]

115dB lfe + 7x 105db (lcr + 4x sr)

How do you add seven different signals together. They are not phase coherent sine waves?

Besides, no soundtrack ever has 0dBFS bass signal on all channels. It’s absurd to think that.

 

[Kvalsvoll]

How do you add seven different signals together. They are not phase coherent sine waves?

Besides, no soundtrack ever has 0dBFS bass signal on all channels. It’s absurd to think that.

They are in-phase if meant to produce sound, since wavelengths are so long they would cancel out if not. If you mean how the calculation is odne, that is a simple math excersise summing 7 log values.

Some movies have full level on several channels simultaneously, though all 7 at once is extremely rare, if ever found. LFE + LCR can be found.

 

[Andysu]

well i have the equipment to analyse align the full range channels sub bass extension and LFE.1 ,
late 1960's hi-fi sub bass , rare few far between i guess and then found its way into cinemas , universal pictures , sensurround , earthquake that did 70mm six track in uk , then Dolby Stereo came about and star wars close encounters of third kind being first two , newest where sub is recorded rather than using control tones generated for sensurround
use few tools for alignment
cinema booth monitor
voltmeter
oscilloscope
spl dB meter
and our ears and body senses for memory recall which i'm good on

^ so if i wanted levels spot on , or if i simply didn't care for using any of the above , its not that hard , yes it is that hard to do , unless you have the test gear to make it less harder then enjoy that mono sub bass , often i just listen to the sub , why not to see if or what has been overlooked ?

 

[sarumbear]

They are in-phase if meant to produce sound, since wavelengths are so long they would cancel out if not.

Summation of, say a -6dBFS signal at 25Hz and another one at the same level at 30Hz does not generate 0dBFS. It generates a composite signal with a spectrum range of 25-30Hz range at -6dBFS. Not to mention the fact that your premise is clipping of the electrical signal chain where wavelength does not enter the calculation.