1 (edited by MstrC-117 2022-12-11 19:33:18)

Topic: Internal processing & output-level scaling of ADI-2 -series

Here and there pop ups constanly allegation, that user can push ADI-2 Pro/DAC to clip its outputs by screwing too much EQ. Great uncertainty prevails, how much this actually is.

I have conceived, that ADI´s smart DSP scales levels down in background same pace as user screws EQ up, until its 24 dB internal processing limit is reached.
Also I have come to understanding, that normally it can´t clip its outputs below 0.0 dB(r) volume level (that can only happen in Phones outputs, on <32 ohm load).
Hence ADI could be normally pushed to "digital suicide" only by exceeding 0.0 dBr volume level, and/or exceeding in combined FX-boosts, its internal 24 dB total limit.


Could the MC himself enlighten, how ADI actually performs is such a usage situation? Can user push it to "digital suicide" i.e. to actually clip its outputs if those limits (24 dB total FX-boost, 0.0 dBr -volume setting, no <32 ohm  -limiter active in PH outs) are not exceeded? If yes, what are the limits then?

2

Re: Internal processing & output-level scaling of ADI-2 -series

MstrC-117 wrote:

Here and there pop ups constanly allegation, that user can push ADI-2 Pro/DAC to clip its outputs by screwing too much EQ. Great uncertainty prevails, how much this actually is.

I have conceived, that ADI´s smart DSP scales levels down in background same pace as user screws EQ up, until its 24 dB internal processing limit is reached.
Also I have come to understanding, that normally it can´t clip its outputs below 0.0 dB(r) volume level (that can only happen in Phones outputs, on <32 ohm load).
Hence ADI could be normally pushed to "digital suicide" only by exceeding 0.0 dBr volume level, and/or exceeding in combined FX-boosts, its internal 24 dB total limit.

Could the MC himself enlighten, how ADI actually performs is such a usage situation? Can user push it to "digital suicide" i.e. to actually clip its outputs if those limits (24 dB total FX-boost, 0.0 dBr -volume setting, no <32 ohm  -limiter active in PH outs) are not exceeded? If yes, what are the limits then?

With the current firmware and current units a combined digital/analog headroom exists which allows one to dial in +2.5 dB volume without clipping (+1 dB for phones, noted and explained in the manuals). The Post OVR indicator in the level meters has been adjusted to follow this.

With EQ amplification and full scale signal it will be necessary to reduce volume so that the level meters no longer show max levels or overs. The digital headroom of 24 dB ensures that this scheme always works, and no digital clipping appears prior to the level reduction by volume.

In AutoRef level mode the Ref Levels get shifted according to the EQ amplification, so would switch to a higher one much earlier, to make sure that clipping the analog output stage will not happen.

It is possible to dial in EQs that raise the level unexpectedly (lots of Google hits on how attenuation in EQ can do that). This could then exceed the offset on the AutoRef level. But that issue would be shown on the Post level meter, asking the user to reduce volume by one or two dB, so again there is no issue.

One of the things that might help - and we did this with the ADI-2/4 already - is to set the level meter into Dual mode as default. This is not the case with DAC and Pro as the dual meter was a later addition, and we are afraid to overwhelm the new user with too many features. But it seems to make sense to change that attitude, at least for the level meter:

SETUP - Display - Hor. Meter - Dual

Regards
Matthias Carstens
RME

3 (edited by KaiS 2022-12-12 10:42:45)

Re: Internal processing & output-level scaling of ADI-2 -series

MC wrote:

In AutoRef level mode the Ref Levels get shifted according to the EQ amplification, so would switch to a higher one much earlier, to make sure that clipping the analog output stage will not happen.

If this is the intention, it does not fully work now in conjunction with the EQ.

I analyzed the way Auto Ref Level takes into account EQ boosts:

Only the highest gain in any single band is used to determine the Auto Ref Level’s switch threshold.
The effect of overlapping bands is not at all considered.
Same applies to the effect of the Q-Factor on the gain, it’s not considered at all.


MC wrote:

It is possible to dial in EQs that raise the level unexpectedly ... This could then exceed the offset on the AutoRef level.

Aside from phase shift related raise in peak level:

This is not only possible, but currently very likely, as overlapping EQ band are quite normal.
Fortunately, with music, the result on the peak level mostly is lesser than the full EQ gain.
This, and the DAC’s 2.5 dB headroom, is the reason why the problematic practically is not quite obvious.

MC wrote:

But that issue would be shown on the Post level meter, asking the user to reduce volume by one or two dB, so again there is no issue.

With Auto Ref Level active, for a certain listening level, the user has no way to manually switch to a higher Ref Level and dial down the volume, for the same loudness result.

Switching off Auto Ref Level is the only, not fully satisfying option in this case.


I think Auto Ref Level should use the result from the Frequency Response Graph to determine it‘s switching threshold.



BTW:
The manual uses the term “Bode Plot” for the frequency response graph.

The special thing about a Bode Plot:
It displays both Magnitude AND Phase in a double-graph.

ADI-2 just shows a Frequency-Response-(Curve).
https://en.wikipedia.org/wiki/Bode_plot
https://de.wikipedia.org/wiki/Bode-Diagramm

One could argue ADI-2 shows a “Bode magnitude plot”, so the manual isn’t completely off.

4 (edited by MstrC-117 2022-12-12 19:10:01)

Re: Internal processing & output-level scaling of ADI-2 -series

MC wrote:

With the current firmware and current units a combined digital/analog headroom exists which allows one to dial in +2.5 dB volume without clipping (+1 dB for phones, noted and explained in the manuals). The Post OVR indicator in the level meters has been adjusted to follow this.


Is that +1 dB reserve of Headphone outputs available on Hi-Power mode too, when output is about to exceed it specced maximum of +22dBu, i.e. its secret, real maximum output level is +23 dBu?
The ADI-2 Pro manual won´t mention this any way in part of Headphone outputs. It tells unambiguously that Headphone outputs max level is +22 dBu. For line outputs it does mention it (that their real maximums are +21.5 and +26.5 dBu), but not for HP outs.

These +1 / +2.5 dB must come from analog domain, little analog extra gains applied to outputs, right?


What about OVR-indication, does it light up when

A) only reaching the end of these reserve limits i.e. ADI is about to really clip next

B) or already when level is yet only 0.0 dBFS, and these +1 / +2.5 dB -reserves are still available?


MC wrote:

With EQ amplification and full scale signal it will be necessary to reduce volume so that the level meters no longer show max levels or overs. The digital headroom of 24 dB ensures that this scheme always works, and no digital clipping appears prior to the level reduction by volume.

With EQ on, the level meters no longer show max levels or overs?? Please elaborate...


MC wrote:

It is possible to dial in EQs that raise the level unexpectedly (lots of Google hits on how attenuation in EQ can do that). This could then exceed the offset on the AutoRef level. But that issue would be shown on the Post level meter, asking the user to reduce volume by one or two dB, so again there is no issue.


Ok, so complex filter combinations, causing complex waveforms, can trick ADI´s AutoRef -offset calculation in some cases, even when only attenuating. Must it, though, be very extreme level attenuation in question? Tickling that 24 dB limitation, but from negative side? Or can it happen on any (milder) filter values too?



MC wrote:

One of the things that might help - and we did this with the ADI-2/4 already - is to set the level meter into Dual mode as default. This is not the case with DAC and Pro as the dual meter was a later addition, and we are afraid to overwhelm the new user with too many features. But it seems to make sense to change that attitude, at least for the level meter:

SETUP - Display - Hor. Meter - Dual

What related confusion usage of Dual -level meters should solve?

5 (edited by KaiS 2022-12-13 01:01:17)

Re: Internal processing & output-level scaling of ADI-2 -series

Pt. 1:
The headroom of 2.5 dB is built into the DAC, it‘s meant as a reserve to handle Inter Sample Peaks:
https://musictech.com/wp-content/uploads/2013/01/Inter-sample-diagram-copy.jpg

The 1 dB headroom in the headphones amp is using the same mechanism, but less of it‘s range (different gain staging).


The meter’s OVR indication reliably shows the clipping limit in the analog path.
No OVR, no clipping.
It‘s even automatically adjusted for low impedance headphones, where the max. output dBu / power is limited.

Max. output for XLR Balanced Line Out is indeed 26.5 dBu.
Headphone‘s max. out is 21.5 dBu with high impedance loads.



Pt. 2:
EQ doesn‘t change the Meters or OVR indication.
The meters show the level in the digital domain.
If EQ boosts or cuts, this is reflected.



Pt. 3
All filters, cut or boost, usually shift the phase relationship of different frequencies (*1).
This lifts the peak level, as higher frequencies now “ride“ on top of the lower frequencies’ waveforms.
The steeper the flank of an EQ, the more the phase is shifted.
The theoretical max. peak level change is ca. 6 dB, practically up to 3 dB.

This might be unexpected for a cut filter, as you take something away.

The higher level is easily handled by ADI-2’s internal 24 dB headroom.

I have elaborated where’s the problem with Auto Ref Level and EQ two postings above.
It does not calculate based on combined gains of 2 or more EQ bands.



REMARK:
(*1)
: There are so called “Linear Phase” filters, a special design that preserves the phase relationship of different frequencies.
This is on cost of impulse response, they have a significant, unnatural pre-ringing.
This type of “filtering” does not exist in natural sounds.
In the bass range the pre-ringing can become very obvious, sounding like a fade-in for percussion instruments.