KVR Audio

plexuss · Post by **plexuss** » Tue Dec 05, 2017 6:30 pm

Here are my tests. Perhaps there is a methodology error. However, I don't think you can realiable measure ISPs unless you have something that can acually measure ISPs. ProL (not 2) was notorious for letting ISPs through - the manual even warns against this. So be careful of your own methodology. AppleRoundTrip is designed to measure ISPs specifically.

ProL2 with 0dB ceiling showing ISPs

ISP ProL2 0.jpg

ProL2 with -0.5dB ceiling showing no ISPs. The fact that it needs a ceilings means its not a true peak limiter.

ISP ProL2 -0.5.jpg

An example of a true peak limiter Nugen ISL2 with 0dB ceiling and no ISPs

ISP ISL2 0.png

audiosabre · Post by **audiosabre** » Tue Dec 05, 2017 6:34 pm

https://www.fabfilter.com/help/pro-l/us ... aklimiting

(look under 'Preventing true peak overshoot'). Same as before really: use oversampling & lookahead.

audiosabre · Post by **audiosabre** » Tue Dec 05, 2017 6:43 pm

Possibly a more scientific way to test:
https://techblog.izotope.com/2015/08/24 ... detection/

HSum · Post by **HSum** » Tue Dec 05, 2017 6:48 pm

audiosabre wrote:Possibly a more scientific way to test:
https://techblog.izotope.com/2015/08/24 ... detection/

Seems to be one of the methods FabFilter is already using and why they're recommending 4x oversampling & lookahead when disabling isp detection.

audiosabre · Post by **audiosabre** » Tue Dec 05, 2017 8:24 pm

I'm not a dev or a super technical wizard. My concern with the above claim is that there's a certain amount of blind faith put into any ISP meter. I personally can't determine which true peak meter gives a more accurate reading, without some kind of scientific methodology. There is no evidence to suggest anyone here knows the actual true peak value of the signal(s) they used to test for ISPs, and are simply relying on what software/meter Y tells them. Additionally, how can anyone know the inner workings of Pro-L 2's algorithms enough to state accurately what true peak level should come out of Pro-L 2 at x amount of gain, lookahead & oversampling?

Thus, unless someone can do better than screen-capping a meter that claims an over, I can only take the claim that Pro-L 2 isn't a true peak limiter with a massive pinch of salt. The claim that Pro-L (v1) is notorious for letting ISPs through doesn't gel with my own experience. I'd have thought interpolation was understood enough for fabfilter to calculate ISPs fairly accurately, with a satisfactory margin of error.

Lesha · Post by **Lesha** » Tue Dec 05, 2017 9:13 pm

I usually add -0.3 dB to be sure no peaks go through after conversion to mp3 for example.

Neon Breath · Post by **Neon Breath** » Tue Dec 05, 2017 9:18 pm

The quest to the loudest possible mix seems to cause lot of headaches and migraines to some…

Magnut · Post by **Magnut** » Tue Dec 05, 2017 9:23 pm

There was a nice explanation from a dev why it's hard to say anything about true peaks and their values. He showed an example where samples alternated between 0.5 and -0.5. The more you've got in line of these the greater was the isp at the end. He also showed depending on how big your oversampling factor is the true peak value got greater. Idk if there's a specification what exactly can be defined as true peak and how to measure its value, but it doesn't seem to be trivial.

I got quite "reliable" results with FabFilters limiter (reliable in the sense of I tweaked its settings and different other meters show me these values too). Additionally I generally use an output gain of -0.8 to -1. That doesn't hurt anyone and it's save for encodings as well as playback.

btw: I often see master engineers complaining about audio software on mac because for example the "core audio sdk" and itunes seem to change samplerates during playback to something fixed (I thought 44.1) no matter what the source is. There were more reasons I think. So I wouldn't trust mac only meters or something that is dedicated to itunes.

plexuss · Post by **plexuss** » Tue Dec 05, 2017 9:33 pm

Magnut wrote: btw: I often see master engineers complaining about audio software on mac because for example the "core audio sdk" and itunes seem to change samplerates during playback to something fixed (I thought 44.1) no matter what the source is. There were more reasons I think. So I wouldn't trust mac only meters or something that is dedicated to itunes.

I think you are talking about the iTunes app itself. There is no reason to suspect Apple's ISP meter is not registering ISPs correctly. If someone has something better than anecdotal evidence or simple dismissal, please share it. Otherwise...

Soundplex · Post by **Soundplex** » Tue Dec 05, 2017 9:43 pm

Checked with the hofa meter. No ISPs after limiting with ProL2 (true peak enabled).

camsr · Post by **camsr** » Tue Dec 05, 2017 9:50 pm

plexuss wrote:
Magnut wrote: btw: I often see master engineers complaining about audio software on mac because for example the "core audio sdk" and itunes seem to change samplerates during playback to something fixed (I thought 44.1) no matter what the source is. There were more reasons I think. So I wouldn't trust mac only meters or something that is dedicated to itunes.
I think you are talking about the iTunes app itself. There is no reason to suspect Apple's ISP meter is not registering ISPs correctly. If someone has something better than anecdotal evidence or simple dismissal, please share it. Otherwise...

Well, it did come from the same company that codes things like root access without a password

plexuss · Post by **plexuss** » Tue Dec 05, 2017 10:02 pm

Soundplex wrote:Checked with the hofa meter. No ISPs after limiting with ProL2 (true peak enabled).

Yes there are discrepancies. I wonder if there are different methods of attempting to meter ISPs and this determines what a given meter will report. The jury is still out. Not that it matters - use whatever limiter you like. It would be nice to have a reliable method for testing for ISPs however. I do know that RoundTripAU is accurate at least in that when I create a file I know that will not generate ISPs, it does not report them. Perhaps the Apple approach is using a more sensitive method than HOFA or other meters? Just something to consider... the occossional ISP is not going to be audible probably, but an ongoing stream of them will be. So far worst case with Pro-L2 it only throws them occassionally as reported with RoundTripAU.

HOFA and DPMeterII both throw ISPs with Pro-L2 in my testing. Not a continuous stream of them, but L2 is throwing them with 4x oversampling. *shrug*

ISP hofa dp.jpg

Endor-8o8 · Post by **Endor-8o8** » Tue Dec 05, 2017 10:17 pm

I had an overshoot during a mix too, monitoring true peaks with Nugen VISLM.

plexuss · Post by **plexuss** » Tue Dec 05, 2017 10:40 pm

Endor-8o8 wrote:I had an overshoot during a mix too, monitoring true peaks with Nugen VISLM.

Well, lets create a new class of limiter: the "mostly true peak" limiter.

Well... I wouldnt use ProL2 as a mastering limiter. But as a track/buss limiter it would be good. I may upgrade to it even though for me it wouldn't be used for mastering. The character modes are pretty good. Some seem to add saturation. Regardless they sound nice. So with some headroom, ProL2 works great. Maybe even as a first stage mastering character limiter and then a real true peak limiter on the output.

DJErmac · Post by **DJErmac** » Tue Dec 05, 2017 11:22 pm

I'm not in love
With this plugin
Even if modern algo sounds really good
But the thing is
I still prefer
Limitless and, Eventide Elevate
I'm not in love
I'm not in love

Pro-L 2 by FabFilter