This is why you'll often see such wild variations in peoples' opinions about gear. We can't listen to anything without it being just one part of a pretty complex system. Some gear is less sensitive to interface effects and that makes life a bit easier but most statements about one piece being "better" than another are actually pretty meaningless unless something is very flawed.

I use modeled plug-ins a fair amount simply because I know from experience what to expect and some limit my choices which forces me to be both more decisive and more creative.

It is up to you to decide this - of course. I know what I think as a designer in both fraternities. Either it is an accurate emulation - or it is not an emulation. The fact that this is audio does not exempt it from this logic?

If there is no reference - and a reference can't be obtained because it's variable - and you still consider that there's some compelling truth in the argument that it is still an 'emulation' of something old and revered - and this still has some intrinsic value to you, then yes for you it is worth it, if you're personally are satisfied with it.

But it is not an emulation by definition

But by the same token, you must understand the frustration of people seeing their designs apparently re-sold as digital emulations, knowing that these must be missing the point in most important senses - and then having the very same people coming back and asking you publically why their ITB mixes 'sound different' - LOL - and quoting pseudo-science relating to digital failings as the fundamental reasons, when in fact the logical reasons are quite obvious and simple :-)

But if the DAC is basically "analoging-out," how can it produce overs in the first place?

It produces "over the limit of maximum voltage" overs. Like any "over" it manifests as distortion.

When you run your car-engine at its speed limit and pump even more gasoline into it then it will *not* drive faster (=more voltage), but choke (distort) or break (burn out). Usually the circuitry/build wont allow the latter to happen though, instead it will limit the maximum gasoline you can pump into your engine.

So it doesn't literally PRODUCE the intersample peaks, it just kind of flubs them out as distorted noise?

I'd assume this, yes.

That is, unless the converters and connected analog circuitry leave some headroom over what is digitally 0 dBFS. But that would either mean that they offer a bigger dynamic range (to add something on "top" of the 24 bit 144 dB range) or offer less dynamic range up to 0 dBFS (lose the bottom to have some spare for "over the top", aka don't output maximum voltage at 0 dBFS).

This is where my detailed knowledge of converter lore ends. As far as I know even high priced premium converters don't offer full 24 bit/144 dB range yet. If this is due to them leaving some for headroom (the latter of the above cases) or due to physical restrictions (which would not allow the first of the above cases) I don't know.

But even if we have/had that headroom. You still need analog circuitry after the converters that can deal with the higher voltage. It's not like your speakers are connected to the converters directly. There at least one gain-stage inside the audio interface to bring it to nominal levels (-10 dBv, +4 dBU or higher) and you can only push its input so far.

Another chain of gain-stages may come with your mixer/level knob and amp. If you push +6 dBU into a mixer/amp that is set to receive +4 dBU signals you may get additional distortion. And even if the mixer/amp offers enough headroom, what about your speakers (unless you turn the output of the mixer/amp down)?

And now consider that your rendered output is not played via professional gear, but over home stereos and mp3 players (after lossy encoding and distortion inducing decoding). Distortion can easily get ugly when you push these cheap systems towards their limits. On the other hand I heard that todays youth is used to exactly this kind of artifacts.

My train of thought was that the DAC basically mirrors the digital domain in every aspect, and shared the whole characteristic of "perfect, perfect, perfect, perfect, CLIPPING".

So are converters using technology that is just too damn expensive to perfect, kind of like steep analog anti-aliasing filters?

"Perfect" clipping is a straight line or better a plateau, but analog waves are based on sinusoids. There is no such thing as a "perfect square wave" in analog, only a more or less close approximation (which in case of clipping sounds "bad" enough).

And even worse, it's analog, so there is no such thing as "perfect" anyway. Slight variations in PSU voltages and chip and conductor temperature leads to varying results.

This is why a good circuitry design and power-supply are mandatory complements of good converters and likely more often than not contribute more to how good an audio interface is than the choice of converters.

And here we also get back to the point of analog emulation, because the same is true the other way around. Digital is too "perfect" (aka controlled) to emulate all variances and imperfections of analog circuitry unless you create a mountain of rules and calculations that get away from a simple mathematical algorithm.

Both can get close to each other, especially with the listener's ear reaching its own limit at some point. But archiving this is exactly what makes the design a quality design.

Which is a good hook to finally give the mic back to Paul. Sorry for taking so much space (I'm very bad at keeping it short)!

PS: Maybe some mod can split this thread, so that we only keep Pauls answers and have the discussion/explanations of users like myself in another thread?

I didn't mean to imply that I thought DACs and other analog circuitry actually reflected the literal perfection of the digital realm, but rather I didn't know that converter technology was put under such strain.

In that case the answer to your last question is: Yes!

It's the very same problem as with most things: The last 5% cost a lot more and are a lot harder to get towards perfection than the first 95%. There are physical and monetary restrictions.

Thanks for the clear and simple explanation Paul. It raises an interesting question -- will digital emulations ever be able to behave in the same seemingly infinitely complex way that analogue systems can? The infinitely complex aspect of analogue is probably part of what we all love about it. Personally, I find it fascinating that something like emulated distortion still "sounds digital" (though, admittedly, it seems to be getting harder and harder to tell the difference these days). It's amazing that we can hear the differences in such a complex signal. It's like our ears can distinguish between real, infinitely complex noise vs. mathematically-derived noise. Ears are cool.

I imagine everyone in this thread discussing this subject in lab coats and spectacles.

holding clipboards.



hope that's not creepy

It's more the brain than the ears. It likely has to do with how we are still influenced from the early ages of mankind where eyes/ears/brain have to quickly distinguish *change* from "background noise" in order to quickly realize potential danger and benefits.

Anything algorithmic usually is repetitive, which in turn quickly becomes "boring" or better to say blended out into "background noise" by the brain. *Change* on the other hand is "interesting" because it marks something new and potentially "dangerous" or "benefitial" and thus keeps your brain's attention.

UggUggUffUff

Myth.

{FACT}
As long as all the harmonics we can hear within the distortion profile of a distorting analogue circuit are generated in the correct proportions, phase and dynamics, it can sound the same. There is nothing stopping digital systems doing this faithfully, if all variables are satisfied and enough effort is made.

What is not reasonably possible for a digital plug-in (as I said in the last post) is to emulate the bi-directional dependancies in analogue, caused by wide variety of external gear driving and interacting with each other. Some of the character of older devices is affected by these factors..

So it's not that the distortion itself under specific conditions is beyond digital processing - it's more a question of the environment that surrounded analogue gear that's missing from the equation.

{OPINION}
A far more sensible approach to distortion sound art (IMVHO) is to forget the notion of blindly trying to emulate past analogue systems - and instead make digital systems that are capable of new subtle and artistic distortion effects - which are completely under the user's control and are reliable and repeatable in the digital environment.

From the point of view of someone who has made and extensively used analogue systems since the 1950s, the concept of 'emulation' seems to have been taken out of context and may be serving more to satisfy a need for justification and validation than it provides a rigorous solution

I gotta answer this - yes you are so very right. I've experienced lots of this stuff and used it to good effect. What our ear/brain combination is capable of detecting is astonishing, especially given the crude nature of the ear itself.

Much of what we have discovered simply by listening and comparison is so unexpected and even illogical that it has to indicate the amazing power of the brain to process information from the ears. It's clear that hearing must have formed a large part of our survival in the distant past.

I don't think the formal scientific community has got close to explaining very much of what can be demonstrated and I have had some pretty epic diagreements about it over the years..

Awesome! So if the "bi-directional dependancies in analogue" end up being an important missing element, then the new "digital systems that are capable of new subtle and artistic distortion effects" might need to be set up differently all the way back to the AD converter stage?

When you say that digital systems can produce the same distortion as an analogue system, does that mean that a distortion plug-in can be as infinitely complex and non-periodic as my Rat pedal? For some reason I just can't believe that. How can 1's and 0's reproduce the behaviours of electrons tumbling through wires and capacitors? Of course I have no expertise in this whatsoever, but this is so interesting that I can't help but jump in...

SNR is still an issue on today's digital systems, because you'll always convert back to 16 bits at the end. That's why you should keep the master bus as near as possible to 0db.

(That was actually said by a mix professor at my university...)

Truth or Myth?

All digital EQ's can be easily made to sound alke, making many emulation eq's bogus.

Truth or Myth?

Myth.

If you think about it, the fact that you can record the sound of it digitally and play it back again shows that there's nothing about the sound that somehow cannot exist in the digital domain.

By the same token there's nothing about this sound that cannot be generated in the digital domain either :-)

There's no such thing as 'infintely complex distortion' - all sounds and waveforms are a mixture of sinewave harmonics added together in the appropriate proportions and phase - as long as the signal is generated accurately it has to sound the same, there is no other possibility :-)

However, yet again there's an important issue with how the box is actually used: Any musician would realise that the performance consists of the musician playing through, with, and into his gear.

This means that every performance is a complex interplay between the artist and the gear he is using that form part of the actual sonic instument he or she is using, intuitively and proactively. In fact it is a crucial part of the performance itself.

So for instance, even if you had Jimi Hendricks himself alive and well in your studio right now, it would be impossible to recreate his performance by taking away his amps and pedals, DIing his guitar straight into you DAW and sticking on his gear in the mix after the event - even if it was his very own stack and all his pedals set exactly as he would have played it..

And so the big missing factor for any mix plug-in, however great it may be, is that it does not form part of the performance itself.. Therefore the performance and our impression of it will be entirely different without the artist 'playing' the distortion subtley to get the right emotion and expression from it all.

This has nothing to do with whether an effect is analogue or digital - it is a physical artistic work flow issue - and it tells us that no DAW plug-in will ever completely replace the function of your pedals - anymore that it would ever replace your instrument...

Myth - practically speaking.

A properly dithered 16 bit signal produces around 93dB of SNR. I.e. it only adds noise to the signal at -93dB down (forget the misused term 'resolution').

Whereas in a fully treated sound proofed control room with good monitors turned up loud, we know we can hear this noise in the absence of anything actually playing, if we stick our ears right up the speakers - in practice under normal listening conditions playing real music even at loud levels it's very very difficult to get the ambient acoustic noise in the listening space down to 93dB below the music level - let alone actually hear the noise while the music is playing.

If you add to that the fact that even the most dynamic of music struggles to cover a 60 - 80dB actual range (even entire orchestras), we can appreciate that 16 bits is more than enough for almost all situations of normal playback.

As for the religion of thrashing full level in permanance to overcome the 'limitations' of 16bit audio? This is one of the biggest causes of bad sounding music in the current era.

Ignore your 'professor' and do what sounds right :-)

This is a very important point that has implications beyond emulating a Rat pedal. For example Vinyl records versus CDs.

More on the topic of how closely a digital algorithm can emulate an analog circuit:

Even one analog circuit cannot emulate another to the level of precision some here are mentioning. But it's not necessary to duplicate every infinitesimal nuance to achieve a successful emulation. All that matters is for the emulation to be audibly close enough. We can measure differences 40+ dB lower than what we can hear, and differences below the threshold of audibility, based on masking and Fletcher-Munson, are irrelevant. Sort of like the difference between 0.001 percent THD and 0.002 percent. Yes, those are measurably different, but nobody can hear distortion that low so they sound identical.

--Ethan

Myth - but we are courting controversy here, because it's the 'easily' part that's crucial.

If 2 EQ's produce exactly the same freq transfer characteristic, noise, distortion and phase response at all levels they will sound the same.

So - if 2 EQs are using the same internal algorithms and precision, if despite their controls you can arrive at exactly the same responses they are going to sound alike.

If they are using different internal processing (such as in the phase domain), or there are other additional factors, they will not sound the same (obviously).

However the biggest massive oversight that this question makes is to assume that the character of EQs is somehow defined only by their static responses. This is completely missing the point of what a great EQ actually is :-(

In fact the perception and feedback of an EQ we actually take from it in use is as much about the way it is presented, the laws, ranges and interactions of the controls etc.. It is manipulating these factors that allow us to usher the user into the intended character we are trying to encourage so that the user is more easily able to get the effects he most wants and get the jobs he most needs done quickly - and avoid things that will immediately blind and put him off.

In many ways this mirrors the issue raised in the post about the distortion box - using an EQ in a mixing session is actually an artistic process, involving a human operating a high level of discernment and creative brain function.. It is very easily disturbed - one can very easily end up in that dreaded situation all engineers fear where 'it's just not happening' - and never quite knowing why.

People who claim they have found something important by painstakingly spending ages finding a null between 2 EQs are completely missing the point :-(

So an EQ is not only about what's under the hood technically (whether analogue or digital) - a large amount of art and experience is involved in the design too. :-)

If this were not the case a simple GUI with only one darned great drag graph and nothing else would be the only universally accepted model for an EQ - and no one would ever want to use anything else!..

Therefore emulations of EQs do not have to be bogus in principle - even though the reasons claimed for their 'historical credentials' might be completely misleading for marketing reasons.

HI Ethan.

I understand what you are saying here :-)

But from experience I would strongly contest the above - but everyone knows that's my opinion so there's no point in arguing it yet again here.

One could argue though that there are a bunch of non-linearities and random elements in analog circuitry and the interaction with it which are at least not easy to implement in algorithmic code. The designer may need to do some "manual" work that go beyond the general algorithms and maybe even add one or the other randomizer.

What's wrong with playing through a (n emulated) plug-in on a low latency setup? As long as I can work with the effect while I perform I don't care if it comes out of a stomp box or anywhere else.

This may well be the most important factor for "good" emulations vs. "bad" ones. If the emulation satisfies the users expectations it's likely good (enough). And that expectation is more than just sound (especially with the lack of a unique reference), it's also character of usage and behavior. In that a good emulation is a good copy of the original design that comes as close as possible to the original experience (but obviously somewhat hampered by the use of a mouse or generic Midi controllers).

Example: When you go to a Pultec EQ emulation you want it to give you the feeling of using a Pultec EQ. And even if you never used the analog Pultec yourself it's still a good thing to get access to the "Pultec experience" when it's called for. And with "experience" I mean the whole package: sound, behavior and usage.

I knew it! Thank you

If two analog audio signals are measured exactly the same on today's most precise audio test and measuring equipment, they will sound exactly the same to the human ears. I mean, the impulse, frequency, dynamic (non linear) response, distortion figure, phase / group delay are measured exactly the same. We do not miss any of the parameter of the audio signal that we have never even tried to measure in human history.

Truth or Myth?

Thanks Paul. You are a great asset to the proaudio
community and a great contributor to gearslutz!

I'm not arguing, just trying to put this stuff into perspective because it's so important. I'm surprised you disagree since masking and Fletcher-Munson are so well understood and accepted! So let's try this:

My AES Audio Myths video plays a DAW project containing both a loud pop tune and gentle classical music, with a nasty buzzing sound mixed in at various levels below the music. That section starts 32:00 into the video. Once the nasty artifact is 60 to 80 dB below the music I can't hear it at any playback volume. For reference, -60 dB is the same as 0.1 percent, and -80 dB equals 0.01 percent. Maybe others can hear artifacts that soft, though I doubt it.

So for anyone (not just Paul) who believes stuff 60 to 80 dB below the music matters, I'd love to see an example. Post a short clip of any music you'd like, and also post a Wave file of whatever artifact you think best makes the point. Then people here can mix the two together in their own DAW, and report at what level the artifact is no longer audible.

A set of files like this would be a good addition to the demos in my video, and will only increase our knowledge. I really hope someone will see the value and post some files.

--Ethan

It's one thing mixing some kind of uncorrelated noise in with another audio stream, it's another thing mixing in some amount of "noise" that is correlated to the other audio stream, such as distortion.

I would imagine that the threshold of perception for the "noise component" in each case is quite different, hence dither.

So I guess I'm hypothesising that perception of something like THD (which would be an actual distortion of a signal), and the summed result of adding noise to another signal, are not the same; and perhaps that section of your AES presentation isn't relevant when talking about correlated distortions of a given signal.

Perhaps. That's why I said people should use any combination of music and artifacts they feel best reveals the artifacts. Again, I really hope one or more people will create some files and post them here. Even if people do something less formal and don't post, it's a real eye-opener to discover what you can hear at various levels. At least it sure was for me. But it takes only a few moments to create a set of test files, and then we'll all benefit.

--Ethan