Threshold detection on stereo compressors is very broken

dLive dLive General Discussions
1

I’m posting this in DLive General because it seems like a problematic engineering decision rather than a troubleshooting/bug issue.

TLDR; On stereo compressors or limiters (eg bus limiters), the DLive takes a copy of the L&R input signals, subtracts 6db from each then sums them to mono and uses that mono signal as the detection circuit. This means the signals that are common to both L&R may trigger the compressor as expected, but signals that mostly/only exist in one of the sides can easily make it through the compressor undetected, peaking well above the threshold, which means true limiting is impossible etc.

I noticed that it was impossible to do true brackwall limiting with any of the built-in dynamics processors on a DLive. Some peaks always leaked through by unpredictable amounts, and I was very confused as to why. This makes limiting for digital broadcast (where peaks should normally be just under 0dbfs) impossible on the DLive, and created other issues for any stereo compression or limiting not behaving as expected.

I noticed the input meter on the compressor only shows a single mono bar. I sent signal generator tone to the bus on only the Left side, and noted that it was sending out at a level that appeared approximately 6db higher than the input meter on the compressor showed coming in. However, when sending the signal generator to both sides, the input level on the compressor matched the output level of the generator.

What does this mean?

The vast majority of stereo compressors and limiters, especially those intended to be bus compressors, will detect left and right inputs separately, and if either side exceeds the threshold on its own, this will trigger the compressor for both sides.

On the DLive, a signal on just one side must go 6db over the threshold to trigger the compressor, while a mono (or correlated stereo) input will trigger it at 0db. This means highly-correlated signals will trigger too much compression, or uncorrelated signals will trigger too little compression. It makes any stereo compression behave in a very unusual and not-industry-standard way. It also makes brick wall limiting on stereo sources impossible unless you are willing to sacrifice 6db of headroom, and overcompress correlated signals.

A workaround for getting a true brick wall limiter is to split the stereo bus/matrix into two separate mono busses/matrices, and run a separate mono limiter on each one. However, of course, this has the huge downside of severely distorting the stereo image whenever one limiter kicks in from an uncorrelated signal, and the other does not.

I am hoping A&H may read this and release a fix in the form of true stereo detection algorithms in their dynamics processors.

1 Like
2

If you think thats bad, just wait until you find out that ganged stereo channels are dual mono…

3

This is really helpful (and disappointing) information. Thank you.

4

Like you have 2 stereo channels that are ganged and in the end you got 4 mono channels?

5

What I meant was that if you use ganging to create stereo channels like the manual suggests, then all processing is dual mono instead of stereo. This can result in having gain reduction on one side of the stereo signal, but not the other causing the stereo image to get warped.

6

But ganging doesn’t make channels stereo, this is just a way to control multiple parameters on several channels at once. If you want a stereo channel, you have to do that in the mixer configuration.

7

If you read the dLive reference guide, it is suggested that you use stereo channels for fixed stereo inputs such as an iPhone input or an fx return, and use ganged channels for stereo inputs ‘on the fly’. And yes, waiting for the console to reconfigure is not always something you want to do.strong text

8

Does it even make a difference though? If true stereo compression doesn’t exist on stereo busses, why would it exist on stereo input channels? It seems like their compressor threshold detection algorithms were just designed to be summed to mono, and there was a huge oversight in the consequences of doing this.
I suspect the same detection circuit problem exists on stereo input channels. I’ll test this later to find out.
If so, ganged dual-mono channels will at least actually catch independent peaks properly, even if it distorts the stereo image… while their so-called “stereo” compressors on a stereo channel will only detect correlated signals properly - true stereo (uncorrelated) signals will blow past the threshold.