True Peaks

[nobby]

I just saw an article about uploading files to Spotify and saw this phrase:

"files should peak no higher than -2 dBTP (True Peak)"

Which got me to thinking about a recent incident.

I was mixing a track in my DAW. Nothing at the master bus indicated any overs -- there's a "hold" function that shows the highest signal level, so if you turn you back and didn't see the highest level, you can see what it was.

I think it was -2.5 dB, but in any case, no higher than -2dB.

The output of my DAW goes to Lynx Aurora converters and then to a Tascam 112 that I use to control the line signal going to the monitor amp.

The peak lamps on the 112 were flashing. I'm pretty sure its (magnetic) VU meters weren't topping out at higher than -3 dB.

I traced the problem to a snare sample which was quite loud soloed but masked in the mix.

My point is, I hadn't heard it as a problem (not saying nobody would) and there was no visual indication in the DAW.

So... True Peak vs... Fake Peak?

[keks]

Look into "inter-sample peaks".
I guess the will be invisible for the DAW and VU will, well, VU...

[Tim Halligan]

Look into "inter-sample peaks".

Yep.

The ITU/EBU spec for TV programme delivery is -24LKFS, with -2dBTP.

Not complying is expensive.

Cheers,
Tim


PS. Nobby, your flashing peak LEDs happen because of level - duh - and rise time. I think that peak meters have a rise time of 300 microseconds, whereas VU is much slower.

[nobby]

Thanks, guys.

PS. Nobby, your flashing peak LEDs happen because of level - duh - and rise time. I think that peak meters have a rise time of 300 microseconds, whereas VU is much slower.

I know the electromagnetic VU meters in the Tascam are slow. I didn't know the meters in the DAW were, though. I didn't see why they should be.

I found an article (link at bottom)

The phenomenon of inter-sample peaks is one that seems to scare and confuse budding mastering engineers in equal measure.

I'm not any kind of mastering engineer, which is why I'm not scared, only confused

I'm not planning to handcuff the mastering engineer with high levels, but I want it fairly loud to listen to on consumer electronics.

I thought -2 or -2.5 0dBfs would be enough headroom for that, because it always is, with this one exception.
I was just surprised that nothing within the DAW showed these peaks but the analog peak lamps did.

And it only happened during snare hits. The obvious suspect, and guilty party (other than the "engineer" ) was the triggered snare sample, which was mostly masked by the snare in the overheads and the under snare mic, and which soloed proved to be loud as fuck. I turned that down and now everything is fine.

https://www.musictech.net/2012/09/10mm- ... ple-peaks/

[meloco_go]

There are plenty of plugins offering intersample peak control.
Personally, I have PSP Xenon on the master output. I don't drive into it, so I usually have no gain reduction happening, but it's there for dither and catching of occasional intersample peaks.

[Bob Olhsson]

The waves L1 offered intersample peak control only they called it "analog!"

D to A converters expand audio to peaks beyond "0" and lossy encoder filters are even worse.

[nobby]

There are plenty of plugins offering intersample peak control.

In this case, determining what was causing the peaks and addressing it at the source seemed to be the way to go.

[nobby]

The waves L1 offered intersample peak control only they called it "analog!"

D to A converters expand audio to peaks beyond "0" and lossy encoder filters are even worse.

In the article it said that was more of a problem with prosumer D/A -- I'm using Lynx Aurora.

In any case, It isn't a bad thing that my peak lamps are in the chain after the D/A conversion.

[meloco_go]

In this case, determining what was causing the peaks and addressing it at the source seemed to be the way to go.

Yes, but sometimes one wants big transient. It's a good thing to have control.

[upstairs]

The waves L1 offered intersample peak control only they called it "analog!"

D to A converters expand audio to peaks beyond "0" and lossy encoder filters are even worse.

It seems to me that a big problem with digital is just how messy it is. I've been trying to figure this stuff out for a while, but I really have no idea just what's going on. Even when I probably do, I feel like I really don't, and that messes with my mind. I wish things were more standardized to where I could just use it without worrying about this nonsense. Maybe I can already, but you know, I'm obsessive.

[Bob Olhsson]

It's simply math. When the numbers are filtered and converted to analog, there can be an increase in peak level because the final sample point is at a different place in the waveform.

[meloco_go]

It seems to me that a big problem with digital is just how messy it is. I've been trying to figure this stuff out for a while, but I really have no idea just what's going on. Even when I probably do, I feel like I really don't, and that messes with my mind. I wish things were more standardized to where I could just use it without worrying about this nonsense. Maybe I can already, but you know, I'm obsessive.

As Bob says, digital is plain math, so there is no inherent messiness to it. OTOH when crossing between domains there is potential for not so straightforward behavior.