1 Topic by anttipi (edited by anttipi 2021-12-14 22:00:55)

  • Registered: 2010-02-18
  • Posts: 51

Topic: Does the Post-FX meter measure "true peak"?

I've got certain classical music tracks that eat into the ADI-2's extra headroom after EQ'ing, peaking at over +2. There's no clipping at the source, Foobar2000 and ReplayGain take care of that.

I'm assuming the Pre-FX meter probably isn't detecting intersample peaks (because the reading is, well, "pre-anything"), but how about the Post-FX meter, after the signal's been EQ'ed?

The reason why I'm asking is that EQ'ing might very well introduce intersample peaks that ReplayGain naturally can't have prevented upstream. So... If I use EQ should I make sure the Post-FX reading never exceeds 0 dB, to accommodate intersample overs? Or are the intersample overs calculated by the RME and shown on the meter - in which case I can be relatively sure a +2.4 dB isn't actually going to clip?

2 Reply by MC

  • MC
  • MC
  • Administrator
  • Offline
  • From: Germany/Thailand
  • Registered: 2007-04-06
  • Posts: 24,877

Re: Does the Post-FX meter measure "true peak"?

Are we talking about the analog outputs? They have enough headroom for you not to clip. But if you want to to have more peace of mind simply set the volume to -2 dB instead of 0 dB.

The level meters do no calculate TruePeak.

Regards
Matthias Carstens
RME

MC's Website

3 Reply by anttipi (edited by anttipi 2021-12-15 11:05:04)

  • Registered: 2010-02-18
  • Posts: 51

Re: Does the Post-FX meter measure "true peak"?

MC wrote:

Are we talking about the analog outputs? They have enough headroom for you not to clip. But if you want to to have more peace of mind simply set the volume to -2 dB instead of 0 dB.

Yeah, I'm trying to optimize my RCA output to amp.

On the rare occasion with classical music, peak transients can go into the red on the Post-FX meter (because of the EQ I've applied). That said, I've adjusted Foobar2000's ReplayGain preamp so that the Post-FX meter never shows OVR though. So, the RME doesn't clip. But with a +7 dBu ref level and 0 dB fixed volume, my amp is overdriven if the output exceeds ~+1 dB. Still I'd rather not lower the fixed volume, as all rock/tracks remain way below max on the Post-FX meter because of ReplayGain (averaging closer to -10 dB).

In conclusion, as the > +1 dB peaks are quite rare, momentary and happen in the analog domain, I guess I'll just leave things as they are. If I ever hear distortion because of overvoltage to the amp, I'll just roll the fixed volume down a notch.

The level meters do no calculate TruePeak.

Understandable; I'm sure the computational resources are better spent elsewhere.

4 Reply by KaiS (edited by KaiS 2021-12-15 20:50:13)

  • KaiS
  • DAW Mastermind
  • Offline
  • From: Germany
  • Registered: 2020-02-05
  • Posts: 3,150

Re: Does the Post-FX meter measure "true peak"?

anttipi wrote:

I've got certain classical music tracks that eat into the ADI-2's extra headroom after EQ'ing, peaking at over +2. There's no clipping at the source, Foobar2000 and ReplayGain take care of that.

I'm assuming the Pre-FX meter probably isn't detecting intersample peaks (because the reading is, well, "pre-anything"), but how about the Post-FX meter, after the signal's been EQ'ed?

The reason why I'm asking is that EQ'ing might very well introduce intersample peaks that ReplayGain naturally can't have prevented upstream. So... If I use EQ should I make sure the Post-FX reading never exceeds 0 dB, to accommodate intersample overs? Or are the intersample overs calculated by the RME and shown on the meter - in which case I can be relatively sure a +2.4 dB isn't actually going to clip?

You have a misconception about InterSamplePeaks.

In the digital domain they do not exist.
Only during the DA-conversion ISPs are developed.

The current lines of DA-converter chips (and ADI-2) provide enough internal and analog headroom to handle all possible ISPs, as ISPs are absolutely common with the still raging loudness war.


Peaks generated by EQing are a different story.

EQ level boosts and phase shifts require some extra headroom.
ADI-2‘s PostFX Level Meter faithfully shows these effects.
ISPs, which are not shown at the meter by design, are +3 dB on top of the value shown, worst case, with or without EQ.

ADI-2 DAC has 1 dB (headphones out) or +2.5 dB (line outs) headroom above 0 dBFS (meter full scale), see manual page 30.
Headphones out doesn‘t usually need headroom, as it already has so much power that nobody with half-healthy ears can stand the full level and dials to full volume.

KaiS's Website

  • Registered: 2010-02-18
  • Posts: 51

Re: Does the Post-FX meter measure "true peak"?

There is no misconception, but apparently poor wording on my part. Of course intersample peaks can't be generated before the D/A conversion. But ReplayGain scanner does calculate true peak (if configured to do so) and that is reflected in the ReplayGain value. The highest overs in my collection are in the +6 dB range (one specific album).

By the ">1 dB peaks" I was referring to the Post-FX meter reading. I do know the DAC has a 2.5 dB headroom.

6 Reply by KaiS (edited by KaiS 2021-12-15 22:56:18)

  • KaiS
  • DAW Mastermind
  • Offline
  • From: Germany
  • Registered: 2020-02-05
  • Posts: 3,150

Re: Does the Post-FX meter measure "true peak"?

anttipi wrote:

There is no misconception, but apparently poor wording on my part. Of course intersample peaks can't be generated before the D/A conversion. But ReplayGain scanner does calculate true peak (if configured to do so) and that is reflected in the ReplayGain value. The highest overs in my collection are in the +6 dB range (one specific album).

By the ">1 dB peaks" I was referring to the Post-FX meter reading. I do know the DAC has a 2.5 dB headroom.

ISPs should be physically limited to about square root of 2 = 1.41 times = 3 dB if the reconstruction filter doesn’t boost it up by phase shift.
So 6 dB sounds like a miscalculation to me.

Which album / track is it?
I’d like to measure it’s real world peak level in the analog domain.


Besides: a signal capable of peaking beyond +3 dBFS must be very strange sounding anyway, does it matter if it’s clipping too?

KaiS's Website

7 Reply by anttipi (edited by anttipi 2021-12-16 06:45:18)

  • Registered: 2010-02-18
  • Posts: 51

Re: Does the Post-FX meter measure "true peak"?

It's been a while but I remember deliberately playing with passband, phase and oversampling in the ReplayGain scanner to be on the safe said, a.k.a. to avoid possible true peak UNDERestimations. Because it's just an estimation anyway and can vary across different implementations. I'll need to check what the problematic album(s) was(/were).

  • Registered: 2010-02-18
  • Posts: 51

Re: Does the Post-FX meter measure "true peak"?

Rush: Stick It Out (from album Counterparts, AMCY-608, Universal, Japan, 1993):

Foobar2000 ReplayGain scanner (SoX 4x oversampling): +5.28 dBTP
Izotope Ozone 8: +5.3 dBTP
Cockos Reaper (using the Loudness extension): +5.1 dBTP

This was the worst calculated ISP in my collection (FLAC).

  • KaiS
  • DAW Mastermind
  • Offline
  • From: Germany
  • Registered: 2020-02-05
  • Posts: 3,150

Re: Does the Post-FX meter measure "true peak"?

anttipi wrote:

Rush: Stick It Out (from album Counterparts, AMCY-608, Universal, Japan, 1993):

Foobar2000 ReplayGain scanner (SoX 4x oversampling): +5.28 dBTP
Izotope Ozone 8: +5.3 dBTP
Cockos Reaper (using the Loudness extension): +5.1 dBTP

This was the worst calculated ISP in my collection (FLAC).

The original CD format (44.1 kHz 16 bit) isn’t available on Tidal.

Tidal offers 2 versions, a 1993 dated one in 96 kHz, and a “2004 Remaster” in 88.2 kHz, both “Tidal Master”.

So unfortunately no way for me to check the original file.

I can imagine:
• the 1993 version is already unusual hot in the 20 kHz band,
• the 2004 version has even more, extreme strong treble up to the top band.
This is what unsally causes high ISPs.

KaiS's Website

  • Registered: 2010-02-18
  • Posts: 51

Re: Does the Post-FX meter measure "true peak"?

KaiS wrote:
anttipi wrote:

Rush: Stick It Out (from album Counterparts, AMCY-608, Universal, Japan, 1993):

Foobar2000 ReplayGain scanner (SoX 4x oversampling): +5.28 dBTP
Izotope Ozone 8: +5.3 dBTP
Cockos Reaper (using the Loudness extension): +5.1 dBTP

This was the worst calculated ISP in my collection (FLAC).

The original CD format (44.1 kHz 16 bit) isn’t available on Tidal.

Tidal offers 2 versions, a 1993 dated one in 96 kHz, and a “2004 Remaster” in 88.2 kHz, both “Tidal Master”.

So unfortunately no way for me to check the original file.

I can imagine:
• the 1993 version is already unusual hot in the 20 kHz band,
• the 2004 version has even more, extreme strong treble up to the top band.
This is what unsally causes high ISPs.

I could check it with a spectrogram. The waveform is severely clipped, as can be expected.

11 Reply by KaiS (edited by KaiS 2021-12-17 23:09:39)

  • KaiS
  • DAW Mastermind
  • Offline
  • From: Germany
  • Registered: 2020-02-05
  • Posts: 3,150

Re: Does the Post-FX meter measure "true peak"?

To me the whole ISP thing is overrated.

Strong ISPs happen with overdriven, clipping, “broken“ signals only.
They can, but mustn’t result in some more, analog domain, clipping, which can’t audibly worsen anything.


A lot of DAC chips and DACs like the ADI-2 are well prepared to handle a significant amount of ISP level.

So why worry?!


The mentioned Rush recording, BTW, plays fine here.

KaiS's Website