TOPIC:

The level meters that appear in the in Routing and Outputs sections of the plugins are obviously displaying samples that are refreshed every so many milliseconds, rather than producing a continuous display. Which leads me to the question - is the value displayed:

a) the instantaneous value of the level, measured at the time of the sample,

or is it:
b) the maximum value that the level has attained since the time of the last displayed sample?

If the answer is a), these meters cannot be relied upon to give the user a visual indication of when clipping occurs, as there could be a higher level between two samples that never gets displayed. If the answer is b), then the user can watch the meters (particularly at the output) to see if the signal exceeds 0dB, and thereby detect if clipping occurs. This relates to the thread on clipping here:

www.minidsp.com/forum/dirac-series-suppo...l-eq-target-clipping

as b) may give us a useful clue as to where the master volume needs to be set in order to avoid clipping in a Dirac configuration that includes any filters that boost signal levels. It isn't the whole story, of course, because a given piece of program material doesn't necessarily reach 0dB, but at least it would be a help in diagnosis. The real solution, as I have indicated in various threads, would be for either the Dirac software or the plugin to report the maximum possible level that could be achieved if a 0dB signal is passed through the current set of filters.

@devteam - any comment on this? Can you shed light on exactly what it is that the level meters are showing us?

@Tony_J

We now have about 10 Dirac Live platforms @ miniDSP... Clarifying which one exactly you're refering would help us with a bit more context I think.. Thanks.

The generic answer is RMS meters so none of the above options here.. It's an average of the signal. Not peak.

No matter if it's a miniDSP, a 20k USD box, you "always" want to keep headroom inside your DSP chain. With miniDSP, we're giving you a LOT of flexibility in an effort to help. Maybe we can improve to remove that flexibility to give less options I guess... (e.g. similar AVR giving less room for tweaking). We can certainly look into this in the future...

in the mean time, without killing flexibility, there are some very simple rules that can help:
a) ALWAYS leave some headroom at your input
b) ALWAYS try to keep your target curve "below" the natural response = You will be applying downEQ rather than boost.

These 2 rules typically work in 90% of cases. We're however more than happy to help if we see.
a) A Dirac live project along the loaded DSP file loaded inside (e.g. DDRC24 XML).

DevTeam

DevTeam

@DevTeam

Thanks for the reply - apologies, I should have said I am talking about the meters on the DDRC-24 plugin. I know that these devices have considerable flexibility and I do appreciate that, so I wouldn't suggest that you constrain that flexibility; what I am looking for is additional tools that will help me understand what is going on within the Dirac filters, which to a large extent have to be treated as a "black box" by the user - the Dirac algorithm runs, it shows you what can only be described as an artist's impression of the idealized result, and you have to figure the rest out by yourself. At least in the 2X4 HD world, because you are setting up all of the filters yourself, you (should) have a pretty good idea how much headroom those filters are using, and you can act accordingly; in the DDRC-24 world, you are pretty much at the mercy of how the Dirac "black box" chooses to interpret the measured input and the target curves, and you have very little real clue as to what it has actually delivered to you in the way of filters, and how much headroom they may need. Hence my interest in a peak meter. But as with all things electronic, the main thing is to have a clear understanding of what the information you are being given actually means, hence the question.

Your points a) and b) above are well taken. I presume point a) really only applies to the analog inputs, as a digital input cannot exceed 0dB...?

After writing the original post it occurred to me that your reply might be "none of the above" - I was right! Unfortunately, for the purposes that I have in mind, a meter that indicates an average (presumably an average over the time interval since the previous sample) is probably the least useful option that I could think of. A peak meter would be much more useful, or even just a "clipping" indication - a signal "LED" that indicates that a peak value has occurred that exceeds 0dB. Is this a possible enhancement to the plugin(s)?

Hi Tony

Thanks for the feedback as it's always great to hear how we can improve.
Headroom is a tricky one as you can imagine..

We can't really say how much "headroom you should leave" within Dirac as we just don't know how much signal you're going to feed into the box. It's not a Dirac issue actually, you can get in trouble with DSP with simple PEQ. Let me rephrase with an example.
- Input signal = -40dBFS on RMS meter => 30dB of boost inside a PEQ => You're "less" than 0dBFS (-40 + 30 = -10dBFS)
- Input signal = -20dBFS on RMS meter => 30dB of boost inside a PEQ = > You're "above" 0dBFS (-20 + 30 = +10dBFS), you're in trouble.

Does this make sense?

With regards to the peak, it's something we can always consider in the future but will need to contact engineer to confirm. Some blocks have a "peak hold" (i.e. showing the highest peak during a certain period)... Might be useful..

DevTeam

Hi @DevTeam

Thanks for the response - headroom is indeed a tricky area.

I'm not suggesting that you get into second-guessing what is happening upstream (at the inputs); that is, of course impossible, because you have no clue as to what the user is doing. Different users do different things; some control the overall volume upstream of the DSP, some use the DSP Master Volume, some control the volume downstream of the DSP. Also, the programme material itself has differing levels, depending on the type of music and the choice of the guy that did the mastering. So in reality, a digital signal entering the DSP could be at 0dBFS (loud material, volume on the source cranked right up) or lower. That isn't the point.
What I am after is the answer to the question "If the input to the DSP is at 0dBFS, is it possible for any of the filters in the DSP to boost that signal above 0dBFS, and if so, by how much?" If I am using a 2X4HD (for example), that is a question that I can answer myself, because I know what filters I am defining for it, and whether they apply a boost or a cut, and how much boost/cut is applied. So, for example, if I define one filter that boosts by 3dB, centered on 1000Hz, and I leave all other settings alone, I know that a 0dBFS 1000Hz signal at the input will get clipped unless I apply a corresponding 3dB cut before the signal reaches the DAC. That is all well and good, and is all controllable by the user, knowing what they are feeding in and what filters have been defined. So if I am controlling the overall volume via the Master Volume or by controlling the volume of the source, clipping would only be an issue if I crank everything up to absolute max volume - unlikely, as it would make my ears bleed. However, if I am using a downstream volume control (after the DSP's DAC), then it becomes an issue, because I had better not ever set the Master Volume above -3dB because it can clip loud signals around 1000Hz.
However, if I am using the DDRC-24, I have no accurate information about what is going on internally, because Dirac and/or yourselves don't give visibility of the filters that the Dirac algorithms generate, so even if I am able to completely characterize the behaviour of the system before the inputs and after the outputs of the DSP, I have no real clue as to what the Dirac filters are actually doing to the signal. My only option in that case is "trial and error" - play programme material at 0dBFS through the DSP and see whether it clips, adjusting the Master Volume down accordingly until it goes away. That is far from ideal, given that it would be perfectly possible for Dirac and/or yourselves to analyze the filters it generates and report the maximum boost that the filter set can generate. Failing that, but really a poor second best, would be meters that report peak levels and/or a clipping indicator.

The other thing that would greatly help here is for there to be a discussion of these issues in the product documentation, in the same way that you discuss "gain structure". The issue is basically the same - you want to be able to utilize the full dynamic range of the equipment, but without introducing any danger of overload/clipping at any point in the chain.

aDevTeam - would be nice to get a response to my previous post

Tony

"Above 0dBFs" doesn't exist. Quite simple rule to follow..

We're unfortunately a bit unsure where is the confusion after our above explanation and how we can explain in more details..
Try to keep it simple if all the above is confusing you... So just look at your RMS meters. If you see you're feeding -10, you can boost 10dB.
What does 10dB boost mean? Look at the "before measurement" if you see a big hole/dip and your target curve is 10dB above that curve, well there isn't 1000 ways to boost that dip. It can be done by time or gain. But overall, if there is a large dip, there will be some level of boost.

Sometime, trying to make things simpler (i.e. not trying to overthink) may actually work fine.

DevTeam

@DevTeam - OK - maybe I phrased the question badly. Let me try one more time.

We've already established that the RMS meters are showing an average level, not a max peak level. So looking at the RMS meters and then acting accordingly (RMS meter says -10 so you can boost 10dB) isn't actually true, because anything in the input signal that is above -10dB (and there will be some, unless you are listening to a single tone at a constant level) is in danger of clipping to the degree with which it exceeds -10 at the input. So actually, the meters don't give you the answer that you want. The safe option: don't boost, only cut. Anything else is just trial and error.

As users, we have full control over the crossover/PEQ etc. settings for the output channels, so we know precisely how much boost or cut we are applying. However, that isn't true of the filters that are generated by the Dirac calibration software; all we have to work on is the before/after curves, from which with a bit of imagination the user may be able to get a vague idea of what Dirac is actually putting in its filters, but the graphs are almost certainly "averaged"/smoothed, so again, the user has no accurate information to say what the maximum boost can be. Again, safe option: keep all of the target curve above all of the measured curve, and hope that the smoothing on the measured curve hasn't been too aggressive. Anything else is trial and error.

What I am suggesting is that you are in a position to provide some better help to the user here. You could do one or more of the following:

- Change the way the plugin's meters work so instead of showing an average value, they show the maximum peak value over the last measurement interval. This would show whether the signal at the output is really in danger of clipping. You could even, processing power permitting, offer two alternative views (RMS vs Max/Peak) and allow the user to switch between them.
- Add a real-time "clipping" error signal, that shows whether clipping is occurring.
- Extract from the Dirac filters exactly how much boost they are capable of generating, and make that visible to the user.

Of course, I would ideally like all three...

I apologize for bringing back an old thread, but I'm facing the same exact question as the original post. Specifically I'd like to understand in the signal/gain chain, how much total gain Dirac may add. From reading all of the forum posts, it appears that this isn't reported out by Dirac, via the Mini DSP UI, and it is unclear if Dirac even provides that information from their API. Reading other posts people say you should leave 6db of headroom for Dirac, others say 10db to be safe. I can interpret that to mean Dirac will increase gain up to 10db over unity. Is that a correct interpretation?

To provide more context, my goal is to setup my system for maximum output without clipping on any material at any stage of the system. prior to Dirac I'd perform the following steps to achieve this. With my playing test tones at 0db I would pass signal through my c-dsp 8x12 at 0db. I would then setup the gain on my amplifiers to be just shy of clipping when handling that signal. I would then tune using REW only applying cuts and boosts that sum to <= 0db, to get a response curve that I like. With this process no output from my headunit (since it can't exceed 0db) would turn into a signal that would result in clipping in my amps (ignoring the possibility of the reactive load of a speaker at resonance frequencies or other acts of god or nature beyond the amps stepping-in and causing clipping, right?)

Now that I'm using Dirac, from what I'm reading online, I now have an element in my signal chain that can generate a boost over 0db. This can now result in an input signal of 0db into the DSP being boosted to something that will drive my amps to clip. I want to avoid this. If I knew the maximum boost that Dirac could generate, my plan would be to apply the same steps above, but add that amount of gain temporarily in my c-dsp outputs while I setup my amp gains, then remove it prior to running Dirac. That feels safe, but probably too conservative (for example, if I'm told max boost is 10db, and I use that as my temp value, but Dirac after analysis only generates 3db of maximum boost, I'm leaving 7db of gain unused. Ideally I'd like to either set the cap for Dirac to boost to a value and know it's hitting exactly that value, or if that's not possible, I'd like to know the maximum value of boost Dirac generates for my environment so I can adjust accordingly (that sounds as I'm writing this like an ugly two pass process)).

If I'm on the wrong path in any of my thinking above, I'd sincerely appreciate any feedback or corrections anyone has to offer.

If the answer is in the manuals, or another post, I'd also appreciate if someone could be specific in how I'm missing what is being stated elsewhere. I've read the manual multiple times, as well as the forums, and I'm still not clear.

Aside from my lack of clarity on this, I'm really pleased with the product and having a great time.

@bmeek AFAIK you are pretty much on the money with your description of the problem and your approach to fixing it - one minor point is that if you apply filters within the miniDSP, either via DIRAC or via the per-channel filters (or both) that sum to >0dB, then it will be the DACs at the miniDSP outputs that do the clipping, not the subsequent amplification stages, although the practical result will be the same.

I have taken the view that in a practical setup, you are unlikely to actually fall foul of this problem, because if your amplification is sufficient, your ears will start bleeding before you drive anything into clipping, so it may not be a problem in reality. However, I completely agree that it would be useful to know what the "gain structure" within the miniDSP unit is, whether there is any headroom internally, what order the different gain controls/filters are applied, ... etc., so that the user actually has a chance of understanding what is possible and what is not. I read somewhere in this forum that internally, the processing is done in floating point, and that as a result, it is possible for the internal "signal" to be boosted above 0dB as long as a corresponding cut is applied before the DAC, but without clarification from @DevTeam, it is hard to tell whether that is the case or not.