Soft clipping or compression in reverbs

[j wazza]

DAC/ADC converters don't usually suffer from "saturation" (in the musical sense; they certainly have limited dynamic range, but you'd expect hard clipping and just about any defect makes them sound less analog and more bad; the biggest issue might often be noise though) and there's nothing really analog about them either.

All analog signals can saturate, especially at high levels, analog clipping sounds different from digital clipping, but even subtly when not intentionally saturating, analog systems aren't perfectly linear. Everything with analog outputs affects the sound a bit, and I'm sure lots of hardware emulation reverbs do use some saturation to emulate the analog character of hardware, even if it's not a noticeable effect. Hardware reverbs sometimes have analog components like filters, gain stages, mixers etc too. There are lots of reverbs with plugin versions and hardware versions with a subtle but still noticeable difference in sound

[j wazza]

The democratization of audio brings some people with no audio knowledge but with money to spend to the field

Notice that if you aren't placing the fx inside the reverb loop then that IMO doesn't count as part or the reverb.

PS: And now the democratization of building plugins is coming too...

That happened ages ago with SynthMaker and similar. I wouldn't worry about it too much.

tv-simpsons.png

"Back in my day, music was a dictatorship"
"Series processing doesn't count as part of the plugin"
"People coding their own reverb have no audio knowledge and must be rich"

There's always some guy like this on kvr

[mystran]

DAC/ADC converters don't usually suffer from "saturation" (in the musical sense; they certainly have limited dynamic range, but you'd expect hard clipping and just about any defect makes them sound less analog and more bad; the biggest issue might often be noise though) and there's nothing really analog about them either.

All analog signals can saturate, especially at high levels, analog clipping sounds different from digital clipping, but even subtly when not intentionally saturating, analog systems aren't perfectly linear. Everything with analog outputs affects the sound a bit, and I'm sure lots of hardware emulation reverbs do use some saturation to emulate the analog character of hardware, even if it's not a noticeable effect.

There is no such thing as "analog character" unless you count the lack of digital artifacts. Different analog circuits produce entirely different results, although we could even make the case that DA/AD converters aren't really truly "analog" in the first place (they operate in discrete time).

ps. For anyone truly interested in understanding analog, I highly recommend buying a soldering iron and ideally a scope and then trying to design and build some actual circuits. You'll quickly discover that things are much more complicated than they might seem and "saturation" is not a particularly good way to describe most of it except in some limited cases such as filters specifically designed to saturate in a nice way (or magnetic tape, but that's it's own thing then).

[j wazza]

There is no such thing as "analog character" unless you count the lack of digital artifacts. Different analog circuits produce entirely different results, although we could even make the case that DA/AD converters aren't really truly "analog" in the first place (they operate in discrete time).

ps. For anyone truly interested in understanding analog, I highly recommend buying a soldering iron and ideally a scope and then trying to design and build some actual circuits. You'll quickly discover that things are much more complicated than they might seem and "saturation" is not a particularly good way to describe most of it except in some limited cases such as filters specifically designed to saturate in a nice way (or magnetic tape, but that's it's own thing then).

There are definitely lots of different colours and sounds that analog can make, and the same with digital, and with acoustic. But I'll have to agree to disagree that there is no such thing as analog character, and likewise I think there is digital character and acoustic character even though acoustic instruments can be very different from one another. Sounds travelling through physical materials sound different from sounds generated or processed in a circuit, which sounds different from sounds chopped into 1s and 0s. A good digital emulation of an analog synth might sound closer to that analog synth than a completely different analog synth, but definitely for me there is such a thing as analog textures and digital ones, i can hear it although it can be subtle sometimes. Yes the lack of digital artifacts is part of it, id definitely count that, as character can be very subtle but still there, but analog has its own artifacts too.

I've used analog gear and done circuit modelling in code but never built my own audio circuits, I'd love to at some point. I know saturation is only part of the story, and most circuits cant be modelled with static waveshapers. Circuits have complex frequency dependent and dynamic effects. Noticeable harmonics are usually only heard with high gain or in highly nonlinear circuits, but a lot of the character im talking about is often very subtle, but as these subtle effects add up they can make the difference. The output of a da converter is still analog, even just running a sound out of an interface and back in changes the character slightly.

Lots of high end analog mastering gear is designed to minimise artifacts and be as subtle and precise as possible in analog, but if they had no character at all people wouldn't spend 10s of thousands on it, they would just use plugins which are more precise and have no noise floor etc, but those unavoidable, subtle effects are what make them sound good, analog can sound close to acoustic because of those imperfections

[j wazza]

This thread wasn't intended to become another debate about analog vs digital, its about saturation in reverbs

I couldn't find any more open source reverbs with soft clipping etc
but I found that waves H-Reverb "applies analog modeling on the reverb output, modeling the A-to-D components of a hardware unit."
I think waves also have a reverb with a transient designer inside

And I found this about Relab lx480:
"We've given you the option to add the same, much sought-after saturation sound from Random Hall and added it to both the Hall and Plate/Room algorithms. Emulations of the different analog and digital input/output stages. Everything from internal clipping to quirks in the feedback loops. Gain settings above 160 introduce saturation with various intensities in both the diffusion and the reverberator sections to create a more upfront reverb sound."

[rafa1981]

If I'm correct an lx480 doesn't saturate in the reverb loop. It will have cummulative truncation/quantization effects and hard clipping inherent to fixed-point DIGITAL arithmetic. This reads like a marketing blob to me.

An ADC doesn't saturate in the analog sense. If a DAC gets a digital value at the input it outputs its totally-in-spec voltage for that digital value. As boring as that.

It might use a companding curve at the output and the opposite at the ADC side to distribute more of the digital range near zero to a bit the hide digital effects, be noisy, have a lowpass effect etc. This is an effect in series that can be added to any reverb with an external plugin.

Usually when someone emulates something digital the hardware owners will claim that whatever they own is better because it had an ADC. The reasons they do who knows, resale value depreciation This is pretty obvious with the dsp56300 project. Then of course some plugin devs might follow.

Compact disc players of the mid 90's also had an ADC and no one is trying to emulate them. They were mostly transparent or at least subtle enough already (while high end converters were better). This is not the case for the 80s though.

But the TL; DR, the reverb block is mostly LTI excluding the fixed-point arithmetic. Non linear processes might happen in series.

Now I'm getting the itch to experiment with companding and using a smaller fixed-point acumulator at TurboPaco (and probably doing the compander stuff too).

[j wazza]

If I'm correct an lx480 doesn't saturate in the reverb loop. It will have cummulative truncation/quantization effects and hard clipping inherent to fixed-point DIGITAL arithmetic. This reads like a marketing blob to me.

An ADC doesn't saturate in the analog sense. If a DAC gets a digital value at the input it outputs its totally-in-spec voltage for that digital value. As boring as that.

Usually when someone emulates something digital the hardware owners will claim that whatever they own is better because it had an ADC.

I think you mean the 480L, which the relab lx480 plugin is based on. Relab says the lx480 does saturate in the feedback loop so I'd believe them, and it also saturates outside the loop. The saturation in the loop could be digital clipping and quantization if they went for accuracy but they dont always have to. I think they also modelled analog clipping of the converters, but even without clipping, converters aren't perfect and affect the sound.

Relabs plugins sound good and I appreciate when companies talk about what's inside their plugins, I don't think its just marketing, it's informative and what they said is probably true about whats in the lx480. I don't think hardware is necessarily better but a plugin will never sound exactly the same, even if its the same software in hardware with analog ins and outs. That doesn't always mean it's worth the extra price for the hardware though. Running a reverb send out of an interface could achieve the same effect if the only analog part is the converters. But that's also less convenient and tactile.

[rafa1981]

The lx480 saturates inside the reverb loop in the way that saturating integer arithmetic can saturate: if the input is too hot at some internal point doing an arithmetic operation the accumulator (wider variable) can clip the value at the maximum representable value in the integer range. Nothing analog, but undesirable digital hard clipping of the same sort as if your DAW hard clipped tracks at 0dbFS with aliasing. If they went for something else then it's not an emulation.

Outside of the reverb loop:

The output DAC can't clip as it gets digital in-range values.

The input ADC can hard clip the signal it's trying to sample if it's too hot, like any soundcard or digital recorder does, but again, this is an effect in series as simple as "clamp(x, min, max)".

Now I don't know if the goalposts are being moved, as this thread's title is about soft-clipping, like e.g. a tube.

[j wazza]

The lx480 saturates inside the reverb loop in the way that saturating integer arithmetic can saturate: Nothing analog, but undesirable digital hard clipping of the same sort as if your DAW hard clipped tracks at 0dbFS with aliasing. If they went for something else then it's not an emulation.

The input ADC can hard clip the signal it's trying to sample if it's too hot, like any soundcard or digital recorder does, but again, this is an effect in series as simple as "clamp(x, min, max)".

Now I don't know if the goalposts are being moved, as this thread's title is about soft-clipping, like e.g. a tube.

Again, you're confusing the relab lx480 with the lexicon 480l that the relab emulates. Relab discuss the saturation in their reverb which I quoted. Emulations can emulate the sound and don't always need to emulate everything scientifically. Maybe as hard clipping is undesirable like you said, they decided to saturate it instead

Analog clipping in an ADC is not the same as a digital hard clamp, that's the most basic model of it

I've said in my posts the thread is generally about these related topics: saturation, clipping, compressing, generic dynamics effects used in reverbs and that hasnt changed. Talking about all forms of clipping in a thread titled soft clipping in reverbs is not exactly moving the goalposts very much. That's just looking for things to complain about.

[j wazza]

It just occurred to me that the airwindows saturating the input with sin before the reverb, then using the inverse arcsin on the output to undo it, could be a way to increase modal density using harmonics, while still keeping it relatively clean

[JustinJ]

It just occurred to me that the airwindows saturating the input with sin before the reverb, then using the inverse arcsin on the output to undo it, could be a way to increase modal density using harmonics, while still keeping it relatively clean

It wouldn't necessarily be that clean. For decent results it's likely you'd need oversampling.

[j wazza]

It just occurred to me that the airwindows saturating the input with sin before the reverb, then using the inverse arcsin on the output to undo it, could be a way to increase modal density using harmonics, while still keeping it relatively clean

It wouldn't necessarily be that clean. For decent results it's likely you'd need oversampling.

Yep oversampling and antialiasing are usually needed for clipping. I didnt check if the airwindow reverb used oversampling or not

The sin and arcsin in his reverb also wouldn't be completely clean because the reverb is in between so it won't perfectly undo the sin clipping. I think when I tried sin clipping it produced fm-like or wavefolder-like waveforms.

As well as oversampling, other antialiasing techniques can be used, these are the ones I know of:
Lowpass filtering the output
Lowpass filtering the waveshaper shape itself
Polyblep, usually for sound generators
Interpolation, spline fitting or smoothing of a waveshaper function itself
ADAA

I'm not sure they'd all apply to a sin clipper. But basically the smoother the cllipping function is, the less aliasing there will be. As sin is already smooth, it might not make sense to smooth it further
Maybe some people more experienced than me can shed some light on these or other antialiasing techniques

[mystran]

Compact disc players of the mid 90's also had an ADC and no one is trying to emulate them. They were mostly transparent or at least subtle enough already (while high end converters were better). This is not the case for the 80s though.

Many of these DACs (at least cheaper ones) were 1bit delta-sigma and those are not really transparent, because you need a multi-bit converter to avoid numerical saturation... but this has nothing to do with "analog saturation" as such. Modern oversampling converters would typically use a few bits (eg. 4?) to avoid the control loop saturation and those are basically as good as their analog-side filters.

R2R and equivalent converters can suffer from uneven quantization steps due to tolerances, but again.. this won't really produce musically useful distortion.

[rafa1981]

but again.. this won't really produce musically useful distortion.

Yes, the rabbit hole goes deep indeed

[rafa1981]

Again, you're confusing the relab lx480 with the lexicon 480l that the relab emulates. Relab discuss the saturation in their ...

It's more possible that you are confusing saturating integer arithmetic (as many fixed point DSP have, and I assume these units too) with (soft?) saturation. This saturation is a side effect of doing a fixed point implementation, as letting the integers wrap around results in garbage. You have to have it. It's not optional, specially with low resolution accumulators (20 bit?).

I have looked at the product. My guess is that they aren't adding the arithmetic saturation, but letting the algorithm run with higher headroom accumulators (e.g. 32 bit) to remove it, as it might be undesirable most of the time.

The clamp model gets you very far. One can put a lowpass at nyquist before and some noise. The point being: this is not even a bandlimited hardclip, but a digital one caused by the ADC outputting consecutive streams of the min/max value: aliasing. There isn't a lot of stuff worth emulating there.

But you seem very convinced that saturarions inside reverbs (decorrelators) are good. Let me reframe this: digital reverb is a very mature field, why would putting waveshapers inside reverbs be a novel idea? Everyone here has tried that already and if you don't belive what I (or we) are saying you can just try it yourself.