Headroom in the digital domain

[Jonas_h]

I could use some input. I have the JBL SDP-55 which has a bug that the LFE SPL is not increasing correctly through Dante when you raise the volume above 62. (And you are not using Dirac and bass management).
I can mitigate this by doing some hacks in my signal chain but I am unsure if this has a negative consequence. Especially about headroom in the digital domain. I was hoping that guys like @amirm has input here.

Basically what I want to do:
1. Reduce sub-gain in the JBL SDP-55 with 16db
2. Send the signal digitally to my DSP through Dante (BSS Soundweb London)
3. Add 16db gain to the subwoofer channel in the DSP
4. Do all the processing/mixing/bass management/EQ in the BSS as I would normally do, with the assumption that all channels have the gain as intended
5. Right before the D/A conversion, reduce the gain to ensure that no clipping is happening on the output (I can adjust amp sensitivity as well to ensure everything is ok SPL-wise in the end)

My question: There is a lot of talk about 0dbFS in the digital domain and that anything above that will be hard clipped. On the other hand, I also know that DSPs (both floating point and fixed point) have a lot of digital headroom internally and as long as I ensure that the signal is not clipping when reaching the D/A, then everything is good.

Do anyone with knowledge about signal processing and the digital domain see issues by doing this? If no, then it will actually allow me to use Dante which would be great!

For visual people, here is a representation of my fix. I am doing other things in the DSP but the illustration keeps it simple.

 

[AnalogSteph]

Let me clarify:

I have the JBL SDP-55 which has a bug that the LFE SPL is not increasing correctly through Dante when you raise the volume above 62. (And you are not using Dirac and bass management).

That is the main volume you are referring to, right?

OK, this might get stupid question of the day then: Why don't you avoid the problematic volume range by giving the DSP some extra gain on all inputs and proceeding as usual? Dante supports at least 24/48 audio so you shouldn't exactly be short of dynamic range on the bottom end.

In other words, I'm suggesting the "then don't move your arm like that" fix.

I am not sure how you came up with your idea for a fix to begin with. Why 16 dB? BSS seems to be rather tight-lipped about the processing of their products, though it wouldn't be outlandish to assume 32-bit float in this day and age... if in doubt, you may want to confirm this with their support, or test it yourself with some absurd kind of DSP chain (say input gain +96 dB, output gain -96 dB - if that sounds fine you have nothing to worry about).

 

[DVDdoug]

My question: There is a lot of talk about 0dbFS in the digital domain and that anything above that will be hard clipped. On the other hand, I also know that DSPs (both floating point and fixed point) have a lot of digital headroom internally and as long as I ensure that the signal is not clipping when reaching the D/A, then everything is good.

Correct EXCEPT fixed-point DSP is limited to 0dBFS. With integer (fixed-point) 0dBFS is defined as the maximum you can "count to" with a given number of bits and you simply can't go any higher.

I ASSUME the data over the Dante bus is integer.

The numbers are bigger with 24-bit than with 16-bit but everything is automatically scaled to match the bit-depth of the DAC so the 24-bit file isn't louder.

With floating-point 0dBFS has a value of 1.0 and for all practical-audio purposes there is no upper or lower limit.

 

[Jonas_h]

@AnalogSteph, yes the main volume. And not a stupid question, but giving all channels gain will result in me asking the same question: Do I have headroom in the digital domain to give this much gain on all channels. The 16db is because this is enough sub-out gain reduction on the SDP-55 to get around the bug. (Might be 12 or 14.. But same same in regards to my question). In short, if you lower the sub-out gain you increase the number on the main volume where the LFE-bug is introduced. I tested this.

@DVDdoug, Dante is 16-bit on the SDP due to DHCP-restrictions. I cannot find a lot of information on the BSS and have only gotten a very short and non-usable answer :"Soundweb London use floating point and fixed point processing depending on the processing object used".

I guess I will have to just test it.

 

[AnalogSteph]

@DVDdoug, Dante is 16-bit on the SDP due to DHCP-restrictions.

(censored string of expletives) That would mean that my proposed approach would be pretty much out. My idea would have been using like 30 dB of attenuation... too much for 16 bit, generally.