How much attenuation...

[Bob Katz]

How much digital attenuation should I expect is going to occur when I insert Audiolense in a full 5.1 setup with two subwoofers? I figure 8 DAC channels are needed if we want to have the least digital attenuation by summing the LFE channel in the analog domain. My current system is level-calibrated for production purposes with a 1 dB/step analog-domain volume control (Cranesong Avocet). I calibrate the output SPL by using a multiturn passive attenuator in front of each power amp, but it has a certain maximuim gain and I want to make sure I'm prepared for Audiolense. 

I've read some threads here where you suggest you can make up for the loss in Audiolense by using gain in JRiver, but I can't believe that, since you can't beat the laws of physics. Digital is digital, so I can't see how you can make up any digital loss anywhere in the digital chain than would cause you to exceed full scale digital on any output channel. So, it's logical you should apply the attenuation that's needed in Audiolense, and keep JRiver at unity gain. So the gain for audio level calibration has to be performed in the analog domain between the output of the DAC and the input to the power amplifier. 

As far as I can see (tell me if I've got this right): If I use 8 DAC channels and sum the LFE to the subs in the analog domain. First of all, I only need to add about 4 dB analog gain in the two subs to get the 10 dB of additional LFE level. And because there's a DAC channel dedicated to the LFE, there's no loss of digital SNR or additional attenuation needed due to the LFE channel. The only digital attenuation that's needed in the other channels is to accommodate 3 channels worth of summing into each sub. And to accommodate any possible EQ boosts.

My off-the-cuff calculation is about 6 dB attenuation will do it, eh?

BK

[Brad]

I think this is what you are saying and what works for me using three subs and numbers adjusted accordingly.

For your two sub case using separate corrected channels for each sub:

1. Set the acoustic gain to each sub at +4db relative to your other speakers.

2. Turn off all bass management outside of Audiolense.

3. In Audiolense route the LFE channel to each sub with a multiplier of .5 (voltage multiplier)

4. Check the LFE 10 dB boost box.

5. Audiolense will make each sub +4 db relative to the other speakers but since the analog acoustic gain is already +4dB the convolution filters should end up peaking at unity gain.

6. Audiolense will automatically re-route bass to subs from other channels with proper gain adjustment. You can specify which speakers off-load bass to which subs. I've been treating bass as mono (reroute to all subs) but you have indicated you like some stereo content in the rerouted bass. Maybe you can say more about that as some mix engineers seem to advocate mono bass. My subs are symmetric left-center-right so I could try your approach.

7. I am paranoid, so I review the filter config text file for properly specified filter paths.

8. Instructions are here: convolver config file

Brad

[Bernt Ronningsbakk]

6-10dB is the typical attenuation figure, but there is huge variation here, depending on the frequency response of the speakers before  correction, the target response and how much correction boost you allow for. I see figures of 20dB or more regularly.

 

With sub woofers and 10dB LFE amplification enabled, everything will be leveled so the subs – after correction – with a full input – will be 10dB louder than the other speakers. If the subs don’t have +10dB sensitivity compared to the main speakers, this will lead to a higher attenuation in the main speakers.

 

So normally you will need 10-15dB of gain headroom in the analog domain to account for the sound correction.

 

While we’re at it, a good workaround that may actually be the best solution overall …. I think I’ll start a separate topic on that one.

 

 

Kind regards,

 

Bernt

--
Audiolense User Forum.
http://groups.google.com/group/audiolense?hl=en?hl=en
To post to this group, send email to audio...@googlegroups.com
To unsubscribe, send email to audiolense+...@googlegroups.com

[Bob Katz]

On Wednesday, January 9, 2013 3:40:49 AM UTC-5, BerntR wrote:

6-10dB is the typical attenuation figure, but there is huge variation here, depending on the frequency response of the speakers before  correction, the target response and how much correction boost you allow for. I see figures of 20dB or more regularly.

 

With sub woofers and 10dB LFE amplification enabled, everything will be leveled so the subs – after correction – with a full input – will be 10dB louder than the other speakers. If the subs don’t have +10dB sensitivity compared to the main speakers, this will lead to a higher attenuation in the main speakers.

Sorry, I'm still an Audiolense newbie! In your response above I assume you are assuming that all the bass is summed in the digital domain. Do you calculate the attenuation on an individual channel basis (or in pairs)? Or do you base the attenuation on the channel that needs the maximum attenuation? In other words, if the LFE is on a separate DAC channel, do you still "penalize" all the other channels by 10 dB? (which would be a bad practice)

I'm still getting a handle on this "automatic attenuation" business. I have to try out Audiolense in more detail but my authorization (license) file is not working. Basically I hope that ergonomically I can see that a given channel is too high and attenuate it first in the analog domain rather than in the digital domain. This would be a step-by-step process I assume, measure, adjust analog, remeasure, etc. 

BK

[Bernt Ronningsbakk]

Yes, everything else is attenuated by 10dB if there is one or more subs involved and you’ve chosen the 10dB option.

 

I agree that it could have been solved better.  I will look into this later. In the meantime there are usually workarounds – ways to amplify the signal before the convolution process to make up for the lost sensitivity.

 

Kind regards,

 

Bernt

--

[Bob Katz]

Yes, this needs to be handled better. I was actually counting on there NOT being any extra attenuation in any channels. I do not have much analog gain in my Lipinski power amps and it would be a drag to have to make up gain in the analog domain.  

Well, until you solve this, since it's floating point, couldn't we change the multiplier coefficients in the matrix page? This would not hurt any resolution, and actually I could use the multipliers to get a custom SPL trim for my purposes, etc.  Says Bob, pulling out his trusty old Hewlett Packard with RPN notation!

OR, what if I don't choose the 10 dB option? If I have a separate DAC channel for the LFE, and I manually raise the gain of the LFE in the analog domain by 10 dB, what happens when Audiolense takes its next measurement? Will it say the LFE is 10 dB too loud and turn it down again?  Probably so. So the only way out of this is to use the 10 dB option and change the multipliers on the matrix page where you choose how much of each input channel gets sent to which output channel. Instead of a multiplier of 1, change it to 3.16 for 10 dB boost. That will make up for the 10 dB loss in the filter designer. How's that, Bernt?

BK

[mojave]

I thought if I checked "10 dB of amplification for LFE," the attenuation of other channels took place only for the measurement purposes, but was reversed for the filters so I could use the analog gain on my subwoofer. If not, that is how I would like it to work. Also, the rerouted bass from other channels needs to be attenuated properly and added to the LFE so that energy is preserved. Matt at JRiver implemented this very well in JRiver's Room Correction DSP for bass management. Perhaps he could share the math.

 

[Bob Katz]

Mojave, I know what you mean but it's important to distinguish the channel term LFE from the loudspeaker term "subwoofer". A lot of recording Engineers I know are still mistakenly using the term "the subwoofer channel" which really is confusing as soon as you put bass management into the picture. So I've taken the liberty of editing your paragraph with (hopefully ) more exact language.

Revised quote: " I thought if I checked "10 dB of amplification for LFE," the attenuation of other channels took place only for the measurement purposes, but was reversed for the filters so I could use the analog gain on my subwoofer. If not, that is how I would like it to work. "

This is the way I thought it would work, too, but it bears repeating that is only practical when there is a separate DAC output to feed the LFE. Otherwise if the LFE is mixed digitally with the low pass filtered bass managed material from the main channels, that material must be attenuated 10 dB in the digital domain. However, technically, the high pass XO to the main speakers does not have to be attenuated as that can be handled in the analog domain.

Revised quote:
"Also, the rerouted bass from other channels needs to be attenuated properly and summed with the LFE channel into the subwoofer so that energy is preserved. Matt at JRiver implemented this very well in JRiver's Room Correction DSP for bass management. Perhaps he could share the math."

I'm sure Bernt has the math :-). It's just a complicated routine to get right yet simplified in an automatic/ergonomic way as Bernt has endeavored to do. Sounds like something for 4.7. And I'm hoping my proposed trick with the matrix page will take care of it all as a workaround for 4.6. In other words you'll see a trace in the correction graph at some level but the actual digital out can be 10 dB (or any amount) higher if you wish.

BK (still a newbie but trying hard)

[Bernt Ronningsbakk]

Unfortunately, there’s a limit of +/-1 in the io matrix in Audolense. But the solution with adjusting the gain in the DACs feeding the subs will only work if you don’t use the subs to offload the main speakers.

 

If you use JRiver I think you can use one of their embedded DSP devices to raise the LFE by 10dB.

 

Or… as I said in a long post a few days ago … how about simply adding the subs to the front left and front right speakers, run Audiolense in a 5.0 configuration, and let the player remix the signal to 5.0? I have a good feeling about this alternative.

[Bernt Ronningsbakk]

Unless you’re running several subs and the subs are a bottleneck for getting enough gain you have nothing to worry about.

 

Kind regards,

 

Bernt

 

From: audio...@googlegroups.com [mailto:audio...@googlegroups.com] On Behalf Of mojave
Sent: Monday, January 14, 2013 10:38 PM
To: audio...@googlegroups.com
Subject: Re: [audiolense] How much attenuation...

 

I thought if I checked "10 dB of amplification for LFE," the attenuation of other channels took place only for the measurement purposes, but was reversed for the filters so I could use the analog gain on my subwoofer. If not, that is how I would like it to work. Also, the rerouted bass from other channels needs to be attenuated properly and added to the LFE so that energy is preserved. Matt at JRiver implemented this very well in JRiver's Room Correction DSP for bass management. Perhaps he could share the math.

 

 

 

--

[Bob Katz]

Thanks for your help, Bernt, my replies:

On Monday, January 14, 2013 7:08:13 PM UTC-5, BerntR wrote:

Unfortunately, there’s a limit of +/-1 in the io matrix in Audolense. But the solution with adjusting the gain in the DACs feeding the subs will only work if you don’t use the subs to offload the main speakers.

Help! I don't understand. I think you're trying to explain it below and also in another thread, but I'm probably too newbie to grasp the new concept you describe. 

 

 

If you use JRiver I think you can use one of their embedded DSP devices to raise the LFE by 10dB.

So you are saying not to set the +10 dB LFE gain in Audiolense 4.6, and that this would work because Audiolense does not make its measurements through JRiver, so it will never know it was cheated up. However, how do you keep the LFE from clipping in this case? With the SMPTE RP200 SPL calibration and widescreen movies and with many other dynamic sources, I run my volume controls fairly close to the 0 dB mark. So I don't understand what you are advocating, Bernt. 

 

 

Or… as I said in a long post a few days ago … how about simply adding the subs to the front left and front right speakers, run Audiolense in a 5.0 configuration, and let the player remix the signal to 5.0? I have a good feeling about this alternative.

I remember that post, but #1, I can't find it right now, and #2 I didn't understand it when I read it (as a newbie). What do you mean by "adding the subs to the left and front right speakers"??? Can you please show an image of the setup screens with this option so I (we?) can understand. Regardless, I do not trust any player (especially standalone DVD and Bluray players) to do any matrixing or recalculating. Who knows what their internal resolution is?  And besides, having three different 5.1 sources, including live mixes that I create in the mastering studio coming directly from a DAW, I cannot do any matrixing in the player---I have THREE players, and maintaining three custom setups is a pain. Also, as a mixing engineer from tracks, I have to see the levels and evaluate the level of the LFE on its own meter. 

I hope this makes it clear why we eventually have to have a solution in Audiolense. Pity the matrix is limited in its gain. Until, let's say, Audiolense 4.7, since the coefficients are floating point, I'm willing to go with separate DACs for the LFE channel, let Audiolense calculate for +10 dB LFE, and then make up the gain in some section of JRiver. Philosophically I would prefer to do all this inside Audiolense, but technically there is nothing wrong with distributing the gain structure between two different floating point entities in a signal chain, the filter and the convolver. 

BK

[Bernt Ronningsbakk]

So you are saying not to set the +10 dB LFE gain in Audiolense 4.6, and that this would work because Audiolense does not make its measurements through JRiver, so it will never know it was cheated up. However, how do you keep the LFE from clipping in this case? With the SMPTE RP200 SPL calibration and widescreen movies and with many other dynamic sources, I run my volume controls fairly close to the 0 dB mark. So I don't understand what you are advocating, Bernt. 

[BR: ] JRiver has a separate “Room Correction” tab in the DSP studio. You can place this before the Convolver. Then you can use it to attenuate all input channels except the LFE by 4dB. In the Audiolense IO matrix you specify that each subwoofer should be fed with LFE without attenuation, in other words use the value 1.0 instead of 0.5.

 

If you have some headroom up to 0dB in the correction filter of the woofer, you can amplify the LFE channel accordingly in the Room Correction tab. You need a total difference of 4dB here, so whatever is left after you’ve amplified the LFE you have to make by attenuating the other channels. If your subs don’t have higher sensitivity than the main speakers, you need to attenuate all channels but the LFE with 4dB.

 

 

I remember that post, but #1, I can't find it right now, and #2 I didn't understand it when I read it (as a newbie). What do you mean by "adding the subs to the left and front right speakers"???

[BR: ] 2^nd  solution: With your 5.2 system you start by specifying a 5.0 system. Then you enter the crossover tab and add a driver to front left and front right speaker. See attached image. Note that this is an alternative solution to the above. If you chose this solution you also need to configure JRiver to feed a 5.0 configuration and prepare Audiolense to receive one. The Audiolense part in the second attachment.

 

Can you please show an image of the setup screens with this option so I (we?) can understand. Regardless, I do not trust any player (especially standalone DVD and Bluray players) to do any matrixing or recalculating. Who knows what their internal resolution is?  And besides, having three different 5.1 sources, including live mixes that I create in the mastering studio coming directly from a DAW, I cannot do any matrixing in the player---I have THREE players, and maintaining three custom setups is a pain. Also, as a mixing engineer from tracks, I have to see the levels and evaluate the level of the LFE on its own meter. 

[BR: ] A 3^rd alternative: Keep the 5.1 format in the player, and the 5.2 format in Audiolense. Attenuate all input channels except the LFE in the format channel matrix. See 3^rd attachement.

 

I hope this makes it clear why we eventually have to have a solution in Audiolense.

[BR: ] I think I’ve got the message J

Pity the matrix is limited in its gain.

[BR: ] That would be easy to change, but I’m a bit concerned that it may promote accidential digital clipping.

Until, let's say, Audiolense 4.7, since the coefficients are floating point, I'm willing to go with separate DACs for the LFE channel, let Audiolense calculate for +10 dB LFE, and then make up the gain in some section of JRiver. Philosophically I would prefer to do all this inside Audiolense, but technically there is nothing wrong with distributing the gain structure between two different floating point entities in a signal chain, the filter and the convolver. 

[BR: ] I agree about the philosophy part. However I am not sure what is the best GUI solution in Audiolense for this. I am open to suggestions.

[Bob Katz]

For those who are following along, I probably need only an extra 4 dB to get the 10 dB boost in the LFE because I'm using two subs. They are stereo subs, and so the LFE will hopefully double (6 dB). This is of course dependent on room acoustics, time delay between the subs, etc. But the effect is close to 6 dB in my room at low frequencies. 

That's why he attenuates the main channels by 4 dB (a factor of 0.63 in the I/O matrix). 

If I do not have a separate DAC channel for the LFE, then I have to use this option (Bernt's third image), called "format channel routing 5.1.bmp". 

However, my sub amps have plenty of analog gain, but my main amps do not. So if I do not get enough SPL for my SPL calibration out of the main speakers, I have to have a separate DAC channel for the LFE (and sum LFE and low pass bass in the analog domain inside the subwoofer) in order to run the I/O matrix at 1.0 for all channels. 

I am beginning to understand the language that Audiolense uses and my learning curve has been cut down with Bernt's help in this thread. Many thanks,

Bob