Clipping internally

Forums Forums SQ Forums SQ general discussions Clipping internally

This topic contains 33 replies, has 11 voices, and was last updated by Profile photo of volounteer volounteer 1 month ago.

Viewing 15 posts - 16 through 30 (of 34 total)
  • Author
    Posts
  • #95725
    Profile photo of volounteer
    volounteer
    Participant

    @dave Meadowcroft

    Clipping is because of the digital age.
    If you have floating point arithmetic used, then it is only a theoretical possibility.
    When you have fixed length binary, then you need to keep the gain low and then fix it on the DAC out by turning the power amp knob to the right.

    I appreciate the 18dB headroom that AH provides.
    That was the minimum that I set when I do things on my own gear and DAW.
    Perhaps AH should have snuck in another 12 or more dB they don’t tell people about to avoid real clipping.

    #95809
    Profile photo of Dave Meadowcroft
    Dave Meadowcroft
    Participant

    I didn’t actually get any clipping – just a few occasional red flashes from the layer buttons and the chromatic meter for Main.
    Perhaps that was because I was a couple of dB below, or maybe it wouldn’t have clipped even if pushed more. Who knows!

    It’s not a problem as I stated. I’ll just have to tell any other engineers using the desk to pay attention to gain staging throughout the mix (not just at ADC and DAC as is the normal nowadays) and the 18 dB above ‘0’ on the meters is headroom available overall across gain, compressor make up gain effect, faders (for post fade), DCA etc…

    Thinking a bit more, having a DCA for all live channels and balancing the internal level by lowering that below unity and bringing up the Main Fader would fix the potential issue entirely as all post fader sends would be affected by the lowering of the DCA and the only concern would be Main/Pre Fader Aux final output levels not going above +18 (0 dbFS / +22 dBu) or the maximum level the crossover/amp/transmitter that follows can handle at its inputs which is entirely normal of course 🙂

    #95812
    Profile photo of volounteer
    volounteer
    Participant

    @dave Meadowcroft

    I have not noticed it either.
    It could be the red light comes on at 0dBFS and not when it is actually over yet.
    AH may have some secret extra headroom to protect people from themselves.
    Or it may be that the clipping is so minimal that our ears just dont recognize it.

    #95814
    Profile photo of MarkPAman
    MarkPAman
    Participant

    With a quick search, you’ll find…..

    From Qu Reference Guide:
    Pk – Lights red to warn that the signal is too hot and gain or trim should be reduced. It turns on 3dB before clipping to warn you before audible distortion. Pk senses the signal at several points within the channel.

    and from the SQ:
    Peak Indication -3dBFS (+19dBu at XLR out), multi-point sensing

    which I take to be the same thing.

    #95816
    Profile photo of volounteer
    volounteer
    Participant

    @markpaman

    Thanks for that info. I suspect that my red lights were not actual clipping yet. Possibly Dave’s too.

    I would still like to have that 4th orange/yellow light like our old analog AH had which was between the clip and the two for sensed and normal.

    #101222
    Profile photo of DocDocDocDoc
    DocDocDocDoc
    Participant

    I’d be interested in knowing more exactly where A&H consoles use which bit depth. The attached PDF I found in this thread makes sense to me. As far as I understand it, integer arithmetic guarantees fixed processing time, while floating point operations can vary depending on the data (=audio), thus needing more latency for safety. DAW users of open source plugins, some of which are programmed badly, know the issues of denormal numbers. In this special case, numbers very close to zero blow up the processing power requirement of a plugin and all of the sudden after the end of the song as the reverb fades away, CPU load pumps up as hell. It is a somewhat tricky issue that is for sure avoided with integer computations.

    What bothers me is precision for low volume stuff. Yes, 56 or 96bit accumulators. Right. But if you don’t feed these properly, how low will your bit depth drop?

    I do a lot of live streaming these days and this means I go digital from the ADs right through to the receiver’s device on the other side of the Internet (except for monitoring). I use a digital broadcast limiter that boosts the signal by 15dB internally, then limiting safely to exactly-1dB (so basically it’s a 15dB headroom thing). It turns out that some mixing engineers who grew up with analog hesitate to drive their digital consoles “hot enough” in order to a) seize the ADs not only in like 12 bits of range and b) provide enough output for the broadcast.

    So for those shy guys it would be nice to know how the A&H consoles fare when not actually using their 56 or 96 bits, but way below. Floating point would just make the issue go away, shifting the precision behind the comma, so to say.

    Concerning me & myself, no trouble getting hot on the inputs. 😀

    #101223
    Profile photo of volounteer
    volounteer
    Participant

    @doc

    not sure I understand the problem. Our MD was an analog guy and tends to way overdrive the digital.
    maybe he thinks he has to do that to keep the SNR up. dont know cant say.

    anyway as I think you asked the question there is absolutely NO problem unless you turn the signal so low that you cannot even see it on the lights or hear it at all.

    And if you look at the AH answer earlier, they change the set point for their fixed point computations differently in different parts of the mixer. In short you cannot get too close to zero unless you are trying to totally sabotage things. And then you may still fail as I read the AH document.

    As to streaming, we adjust the signal in the PC that the mixer sent to them. They have problems because they adjust by ear and wont use a visual LUFs meter. When at times the sound is low at home then it is because the DR was too big and nobody compressed it properly.

    #101347
    Profile photo of Konrad
    Konrad
    Participant

    I recently got an SQ 6 to check out and my tests seemed to indicate that it does not have any headroom and clips internally at 0dBFS. I did two tests, detailed below. Source was Pro Tools at 96 kHz via Dante. Result were confirmed by recording the L/R output fo the SQ back into Pro Tools.

    1. 20 – 40k sine tone sweep at -10dBFS. Input to channel 1, channel fader at unity. Master fader at -20. Did a 12 dB boost at 230 Hz, Q of 0.5. The sweep clipped, the red clip light on the EQ lit, and I could audibly hear clipping on the output even though the master fader was at -20.

    2. 400 Hz sine tone at 0 dBFS, multed to channels 1 – 8 on the console. All 8 faders at unity. Master fader at -20. With only one channel unmuted the tone sounds clean, adding in a second channel, the tone is pushed into clipping and audibly clips at the output even though the master fader was at -20.

    This does not seem right, at least not for a modern digital console. But I am new to the SQ and perhaps something was set strangely on the demo unit? What am I doing wrong?

    #101352
    Profile photo of SteffenR
    SteffenR
    Participant

    How does you recorded signal look like?
    Screen shots?

    Where did you set the source for the recording? Pre or post fader?

    #101353
    Profile photo of KeithJ A&H
    KeithJ A&H
    Moderator

    @Konrad –

    0dBFS is by definition the maximum value that can be represented in the digital system.
    So on point 1 – you have sent in a -10dBFS signal and then boosted part of it by 12dB = +2dBFS (not possible).
    On point 2 – you are combining much (much!) hotter signals than you should be.

    Please check out the following article about levels, metering and gain staging – http://support.allen-heath.com/Knowledgebase/Article/View/levels-and-metering-in-qu-and-sq

    0dB on the SQ meters = -18dBFS.
    So there is 18dB of headroom, and for testing you should be starting with a -18dBFS signal.

    Hope this helps!
    Keith.

    #101356
    Profile photo of volounteer
    volounteer
    Participant

    @Konrad
    you are new to the digital world which is far different from analog, not just new to the digital console.

    EVERYTHING clips at 0dBFS. You do not ever want to get even close to that.
    You do not need to do it as digital SNR and DR is SO much greater than analog ever was.

    Digital requires an 18dB headroom, if you want to be sure not to clip or have other problems. AH provided that.
    That is 0 on the AH readouts. You do not want to go over that level ever,
    but if you do go over a dB or 2 then you will not clip but you gain nothing by pushing it hotter like you did with analog.

    You do not get louder by raising the dBFS, you get louder by turning the knob on the audio power amplifier to the right after you convert the digital to analog.

    #101358
    Profile photo of RS
    RS
    Participant

    Digital requires an 18dB headroom, if you want to be sure not to clip or have other problems. AH provided that.
    That is 0 on the AH readouts. You do not want to go over that level ever,
    but if you do go over a dB or 2 then you will not clip but you gain nothing by pushing it hotter like you did with analog.

    from my experience I would like to add to this quote: it depends
    If you are talking about inputs, I am totally with you. If we are talking about outputs, it depends on the summing of the console you look at. I don’t know how SQ handles this. But my dLive, for instance, has excellent summing that includes compression/harmonics effects I remember from way back in my analog days. Will say, if I push the console hard I get more pleasant sound.
    Would love to here more technical insight on that from A&H experts.

    #101361
    Profile photo of volounteer
    volounteer
    Participant

    @RS

    digital is digital, and different from analog. many people do not fully understand it.
    even one big textbook by a famous prof had (minor) errors. the internet is full of errors about it (not all minor).

    If you are adding f/x , or are clipping with summing, to add new frequencies,
    then you have distortion or possibly other issues.
    some people like distortion. some do not.

    Now in theory you should be able to push the console right up to 0dBFS,
    and as long as you never touch 0dBFS on any channel or sum of channels,
    you would not notice any change in the sound except the digital level.
    but you gain nothing and risk distortion if you try.

    In practice, some converters do have (very small) issues when you push them near the maximum,
    and those sound more accurate when you give them some headroom too. again practice not theory there.

    Whether pushing AH converters near the max is audible or not I do not know.
    But it should still be far less than what you got from analog when you pushed it to the max.

    I have to suspect that anything you think you hear is apophenia or golden ears syndrome,
    and that in a double blind ABCX test the statistics would say the results were random and not actual.
    AES did tests like that some decades back to test things like magic cables,
    and found what people claim to hear is more delusion than real.

    #101362
    Profile photo of Konrad
    Konrad
    Participant

    Thanks for the quick reply and the link to the gain staging document.

    @volounteer, I am not new to digital. I’ve been working on digital consoles for more than 25 years and almost exclusively in the digital domain since 1994. I understand that 0dBFS is an absolute during A/D & D/A conversion and when outputting a digital signal via AES or Dante, But most digital consoles allow for signal greater than this by using either floating point processing or fixed point with an accumulator. That way internal clipping is avoided. This is easily tested as I described in my original post. On a Behringer X32 which uses 40 point float, You can combine something like 32 channels of 0dBFS tone without clipping the internal mix bus. On a Yamaha DM2000 which uses 32 bit fixed with a 56 bit accumulator, you can combine 8 channels of 0dBFS tone before the internal mix bus clips. (of course in both cases the master fader has to be pulled down to -20 or greater to prevent clipping the 24-bit output).

    Do I understand correctly that the SQ does not have this additional headroom?

    #101363
    Profile photo of volounteer
    volounteer
    Participant

    @Konrad

    I understand that all the AH mixers have still have 18dB headroom at their 0 LED/meter.
    And have 3dB more after the red peak light on the channel.

    Some digital gear uses FLP for (near) ‘infinite’ headroom.
    I understand AH is all binary but adjusts the range to ensure enough bit depth that it is using at various levels in various places in the mixer to be roughly equivalent to wider bit depth. I believe it is quite possible to make the mixer clip internally if you have too much gain at various locations depending on the signal level you got to start with. Obviously at the final DAC you must be under 0dBFS at that stage of the processing whether you used FLP or binary. So with really bad internal settings you could well clip on the output too.

Viewing 15 posts - 16 through 30 (of 34 total)

You must be logged in to reply to this topic.