Don’t know if I’m simply missing something, or being a complete amateur, but setting the output celling of all channels at -2 I’m still getting overs of -1.8 and -1.9… is there somthing I need to enable to have “True” true peak limiting?
Should be impossible. I’d like to see your settings and even session with audio files if that’s possible.
Are you sure you dont have some extra processing happening after Stemcell?
This first instance is the same across all our main stems, then the second image,
is what we have on our 51 master, with no other gain changing plugins anywhere else and Youlean at the end of the chain reading overs of 1.9 on our center channel. I’m sadly not able to send any other sort of information, but the sound in question is a sword being drawn from its sheath. If there’s any other tests you need me to run let me know 
First of all - make sure your 2.0 is actually 2.0 by changing it, then manually typing in 2.0
Next, it sounds like you’re using some other 'true peak" metering tool, rather than the actual sample peaks.
The truth is that there’s no such thing as “true peak”. You’re trying to limit or measure a peak which doesn’t actually exist, but MIGHT appear in some other downstream processor like a radio transmitter or MP3 encoder, when processing is applied to the audio.
A “true peak” tool tries to guess where these imaginary peaks may appear by oversampling the signal a number of times, then applying a low pass filter which “reconstructs” inter-sample peaks. The problem is that different amounts of oversampling will give different results, and oversampling gets more and more expensive as you do more stages. The filters used for up and downsampling also make a difference, and these filters necessarily affect the limited signal, so that when you get back to 48k, you now have a few new “overs”.
So true-peak limiting is like chasing a ghost.
It’s a ridiculous standard which should never have become our responsibility as the content creators. IMHO, if a radio tower or an MP3 encoder generates “overs”, is THEIR responsibility to limit them. It’s absurd for the sound mixer to pre-empt and predict any signal overs which might occur anywhere in some unknown process in the future. We’re fixing the “problem” in the wrong place.
For the record, Stemcell uses the standard true peak approach defined by EBU or SMPTE or whoever it was. I’m gonna guess YouLean uses much more intensive oversampling and filtering to do a “better” job, so it sees some imaginary peaks which Stemcell does not. Or it’s summing channels in some way…
Totally agree! Our main concern is that setting the limiters to 2.0 following your procedure as mentioned still creates overs that both Youlean, wlm and Nugen LM correct all pickup on, thus we no longer comply with A85 Specs. Is there some form of test you’d recommend us do, or any settings on our end so we can verify Stemcell is functioning correctly? We’ve even gone and set the limit to -10 and get overs of 9.7 which sort of defeats the purpose of Stemcell for us.
Here’s the dilemma…
I can switch to using more oversampling and most expensive filter types and it will probably match those metering tools. But then it would be a very CPU intensive plugin which would not let you use lots of them in a session. It would also have a really long delay compensation value which makes realtime mixing horrible due to the lag in the faders. We want this tool to be viable for mixing.
I could give you a switch to turn on an “over eager true peak” mode - but that would have to be automatable and would change the delay comp value at runtime - which is a recipe for nightmarishly hard to locate snaps and pops (looking at you, Fabfilter)
I can take another look at it but we went thru all this for v1 and followed the spec. When I get a second I’ll see what the meter plugins are doing and maybe there’s an easy “fix”.
For now, if you’re regularly seeing a 0.2 - 0.3 dB difference, you could just set your threshold that much lower.
Good point, I was just more curious to see if this was some regularly reported behaviour for stemcell. I was looking all over the internet and hadn’t seen any users mentioning this issue, so I was interested to see if you’ve noticed ever getting overs in sessions. It seems that theres a weird combo of frequencies that lead to just a bit of audio getting let through the limiter stages.
doesn’t surprise me TBH. I’ve been mulling over ideas for solving this without adding massively to the CPU/delay of the plugin. I agree it would be good to sort it out.