Calibrating Input Level for Plugins

[James Freeman]

We should start asking for these values from manufacturers for their software plugins like Amplitube and NDSP.
The most accessible and convenient for the customer and manufacturer is to provide RMS values anyone can measure with a multimeter and dBFS values we can clearly see in the DAW or plugin.

The calibration value for HX/Helix/Native is (all the same):
1v Peak = -12dBFS
707mv RMS = -12dBFS
500mv RMS = -15dBFS

EDIT:
20*log(500/707) = -3dB.
In case you are wondering why they are all the same.

 

[MirrorProfiles]

Big up @James Freeman for not only this thread but also the old KVR one, and putting those clowns in their place who just weren’t understanding the importance of this.

I followed the guide last night for calibrating for Helix Native. My chain is a bit more complex as I’m not using a single interface, but rather a combination of separate DI, A/D, and D/A. Not only that, but the DI I’m using can directly feed the converters without the need for a preamp and the A/D has adjustable calibration for various dBu references.

My signal was coming in around 2dB lower than what would trigger the gate to open on/what I’d get using the HX Stomp, which seems right based on my experiences with using the HW. [EDIT: this is actually incorrect - I just tried calibrating again at 707mv RMS with the gate at -12.2 and my DI was calibrated absolutely dead on].

I then spoke to one of the devs who coded STL’s plugins (he’d previously offered to help me calibrate for using their software). I’m sure he won’t mind me sharing his procedure (which is very much what James suggested years back on KVR):

Sure, first thing to do is looping a sinewave in your DAW (better use a signal generator than a .wav file to avoid discontinuities in the output), something like 60Hz, with a maximum peak of 0dB, basically the clipping threshold.

Then set your multimeter in AC (1V range if possible) and connect a cable to the soundcard output (leave the other jack disconnect), then set your soundcard output level to the maximum.

Connect the multimeter scopes to the tip and sleeve of the cable jack and see what RMS value you get. Now tweak the DAW output level until you read exactly 0.707V on the multimeter, that's your settings for output unity gain.

Now connected the free jack of the output cable to the input of your soundcard in order to send the signal back into your DAW, just make sure the track receiving the input doesn't route the output to the master bus, or else they will sum and clip everything resulting in squeals.

Monitor the input and tweak the soundcard input level in order to see a peak of exactly 0dB: that's your input unity gain level. Note the soundcard input level settings somewhere (or take a picture) then decrease the level in order to read -6dB peak, note it somewhere then proceed to -12dB.

Now you know exactly the input settings for unity gain, -6dB and -12dB.

Connect your guitar straight to the soundcard input set to unity gain and play. Does it clip? No: ok, yes: set the input to -6dB and repeat. Does it clip? Set to -12dB and repeat...

Now you know what kind of attenuation the soundcard is applying to your guitar, so when you hit the first guitar plugin of your rig, make sure to boost the input to compensate the loss: if you set the input to -6dB, boost by 6dB, if -12dB, boost by 12dB.

Now the plugin sees exactly the same signal your amplifier sees when playing.

Keep in mind most amp sims have an internal boost to compensate for the fact that most users have their soundcard input level set low, for AmpHub this boost is +6dB, so if your soundcard is set to -6dB, set the AmpHub input level to 0, or to +6dB if your soundcard is set to -12dB. Disregard the Input Level Listener at this point, you have your levels set perfectly.

The final paragraph here is really the critical piece of information that we need to know from developers - there’s no standard amount of boost for any modeller or emulation. Emulations with less of an internal boost will favour a hotter signal, ones with a larger boost will want a lower DI signal.

For instance (if I’m understanding this right), STL Amphub wants a 6dB hotter signal than Helix Native.

It boggles the mind why this information isn’t freely available or even mentioned anywhere for the most part - do we as audio engineers calibrate tape machines and consoles by ear? Or are they regularly calibrated for any change in circumstances and components? Why should we treat any other gear any differently, particularly when there is so much amplification of a signal involved?

Developers understandably assume that it’ll go over the head of most users, but ultimately it just means we’ve set the bar too low. There’s already been far too much misinformation about modelling and blanket statements made without conditions being made fair.

I think it would be very handy if we can find the calibration levels from as many pieces of software and HW as possible and have it easily accessible. I can already anticipate some companies will be more willing to help than others, but the reality is everyone is likely to be using a combination of different software with a whole range of different HW.

I’m sure I read a post from IK‘s rep on KVR saying that they modelled their amp sims with different calibrations for each model….. if this is the case then (but also please tell us what each of these values are, there’s enough space in the manual for it).

 

[James Freeman]

Big thank you for sharing this @MirrorProfiles !
And big kudos to STL for giving solid information instead of PR mumbo-jumbo.

For instance (if I’m understanding this right), STL Amphub wants a 6dB hotter signal than Helix Native.

As I understand from what they say, yes, for STL AmpHub the correct calibration level would be 1v Peak (707mv RMS) = -6dBFS when leaving the plugin Input level at 0.
That's hotter by 6db than Native and leaves less headroom for hotter guitars unless you do what they suggest and lower it further by -6dB (to -12dB) on your audio interface and compensate with +6db in the plugin which should give you 12db of "hard strumming" headroom like Native does.

Talking about headroom;

Let's assume a 4.5v Peak output from standard guitar gear like Pedals or EMG pickups etc., that's half the voltage of a fresh ideal 9v battery and the absolute maximum standard 9v gear can produce (excluding voltage boosters and charge pumps like in the Klon Centaur).
In reality the peak voltage is less than 4.5v, nothing is ideal.

20*log(4.5/1) = +13dB.
If we want to leave digital headroom for standard 9v guitar gear we need 13db of headroom if we calibrate with 1v Peak as reference, assuming the audio interface allows that.
My Focusrite Saffire produces -8dBFS with gain at zero so it's already slightly too hot to allow 9v gear without clipping.

I think it would be very handy if we can find the calibration levels from as many pieces of software and HW as possible and have it easily accessible. I can already anticipate some companies will be more willing to help than others, but the reality is everyone is likely to be using a combination of different software with a whole range of different HW.

This is very important and we should nag them until they put some clearly understandable numbers in their manuals.
I would even go as far as standardizing a format like "1v Peak (707mv RMS) = -x dBFS" so everyone's on the same bandwagon.

I’m sure I read a post from IK‘s rep on KVR saying that they modelled their amp sims with different calibrations for each model….. if this is the case then

If that's true, then WTF.
What about some consistency and accuracy?
They have OFFICIAL models ffs.

 

[MirrorProfiles]

well unless their position has changed, I'd classify this in the "mumbo jumbo" category. Peter can be a hard nut to crack when it comes to trying to help IK make better products.

 

[TheTrueZoltan!]

I wonder what level Mr. DSP at Neural codes his plug-ins for. I tried Nolly with the Helix calibrated level and it sounds quite good to me. Perhaps Neural and Line 6 are modeling with similar levels?!

 

[MirrorProfiles]

As I understand from what they say, yes, for STL AmpHub the correct calibration level would be 1v Peak (707mv RMS) = -6dBFS when leaving the plugin Input level at 0.
That's hotter by 6db than Native and leaves less headroom for hotter guitars unless you do what they suggest and lower it further by -6dB (to -12dB) on your audio interface and compensate with +6db in the plugin which should give you 12db of "hard strumming" headroom like Native does.

I think I may have interpreted it differently - as I understand it here, if I'm calibrated for Helix at -12dBFS, then to calibrate for STL Amphub, I need to use their internal input level at +6dB. From memory, this is roughly what their input calibration does when I strum hard. But I might re-read a few times....

 

[James Freeman]

I think I may have interpreted it differently - as I understand it here, if I'm calibrated for Helix at -12dBFS, then to calibrate for STL Amphub, I need to use their internal input level at +6dB. From memory, this is roughly what their input calibration does when I strum hard. But I might re-read a few times....

STL's reference is "1v Peak = 0dBFS" from what he writes in the first few paragraphs.
He says that AmpHub has +6db internal boost and I quote: "if your soundcard is set to -6dB, set the AmpHub input level to 0," basically undoing the +6dB boost back to reference 0dBFS, and effectively giving you 6db of headroom before clipping.

By the same logic Native has +12dB of internal boost.

if I'm calibrated for Helix at -12dBFS, then to calibrate for STL Amphub, I need to use their internal input level at +6dB

Yes, you get exactly the same result.
After calibrating for Native you can boost by +6dB with an EQ plugin before AmpHub, as long as you feed AmpHub with 1v Peak = -6dBFS you'll get accurate results.

 

[MirrorProfiles]

After calibrating for Native you can boost by +6dB with an EQ plugin before AmpHub, as long as you feed AmpHub with 1v Peak = -6dBFS you'll get accurate results.

Excellent. I just tried their auto calibration feature set to hot pickups, and it basically stayed at 0 (in an ideal world that would adjust to +6dB based on my set up and how it’s calibrated).

Going to make a spreadsheet with these written down (and also note down my DI and interface settings).

 

[James Freeman]

I tried Nolly with the Helix calibrated level and it sounds quite good to me. Perhaps Neural and Line 6 are modeling with similar levels?!

I'd like to know too, can you ask them?
I wouldn't be surprised if it's -12dB or even -18dB, they need that headroom for the type of modern guitars their customers are typically using.

 

[MirrorProfiles]

FWIW I've messaged Softube and Mercuriall. Some amount of prodding is going to be useful, but if the same companies keep getting the same questions they may go a bit mad.

I wonder if its possible to use noise gate's to get an idea of what calibration they might be using. Its not as straightforward as assuming all gates are equal because there can be things like the knee and hysteresis to consider. But some do offer values with dB readings, so presumably getting those to correlate to the same input value could help in determining it?

Likewise, its not uncommon for (non guitar) plugins to have various calibration settings within the plugin. This just seems like the ideal solution for guitar based plugins, because then you'd be able to set them all to the same (correct) value for your particular rig, and you can use the input controls purely for creative uses (or to preserve the calibration if other things are in the chain). UAD tend to have a Headroom screw and I believe their calibration values are written online somewhere (or perhaps in their manual too).

 

[MirrorProfiles]

I wonder if its possible to use noise gate's to get an idea of what calibration they might be using. Its not as straightforward as assuming all gates are equal because there can be things like the knee and hysteresis to consider. But some do offer values with dB readings, so presumably getting those to correlate to the same input value could help in determining it?

If I do the gate trick with Amplitube, then it seems the calibration is the same as Helix Native (both gate's kick in at exactly -12.2dB on my system). This assumes that the gate behaves as simply as it seems to, and that IK's calibration is consistent across the board though.

 

[TheTrueZoltan!]

Another method could be to use a Tubescreamer pedal to optimize the input level. Pretty much every amp sim has a TS model, so if a sine wave is sent through the TS pedal and simultaneously through the amp sim's TS model, the input level could be adjusted until the the level of the sine wave's frequency as well as the created harmonics overlap (= gain matched). That should at least get you in the ballpark, if the TS model of the amp sim is any good...

Unfortunetaly I don't have a TS pedal :-(

(Edit: Tried to do it with my Boss SD-1 and Helix Native's Stupor OD. Found out that the Helix' model seems to be a little bit flawed, it has approx. 3,5dB less output than my pedal. But after putting a +3,5dB gain block behind the Stupor OD, I could match them quite nicely.)

 

[MirrorProfiles]

Trying with Nembrini's Cali Dual now. Their gate also has a dB scale, if I feed it at -12dB sinewave, the gate shuts at -16.6dB (compared to -12dB on Helix and Amplitube). I tried at different input levels, and its consistently 4.6dB lower.

This means that Nembrini needs 4.6dB more level than Helix to have them calibrated equally. Or (correct me if I'm wrong) 1VAC=-7.4dBFS

EDIT: I'm on a roll here. Neural DSP (Fortin Cali) needs 2.3dB more level than Helix to have them calibrated equally. Or (correct me if I'm wrong) 1VAC=-9.7dBFS. Same seems to be true for Fortin Nameless. The gate's arent as straightforward though, seems to be hysteresis and some low level stuff passing through.

 

[James Freeman]

If I do the gate trick with Amplitube, then it seems the calibration is the same as Helix Native (both gate's kick in at exactly -12.2dB on my system).

Wait, that doesn't mean they have the same calibration value.
It means -12db is -12db and their db meters are working correctly, nothing more.

This means that Nembrini needs 4.6dB more level than Helix to have them calibrated equally. Or (correct me if I'm wrong) 1VAC=-7.4dBFS

No, we can't assume that.

EDIT: I'm on a roll here. Neural DSP (Fortin Cali) needs 2.3dB more level than Helix to have them calibrated equally. Or (correct me if I'm wrong) 1VAC=-9.7dBFS. Same seems to be true for Fortin Nameless. The gate's arent as straightforward though, seems to be hysteresis and some low level stuff passing through.

Before you get too much ahead with this, I'll have to stop you.
These values tell us nothing about how the real world 1v Peak translates to virtual 1v Peak in the plugin.

 

[MirrorProfiles]

Wait, that doesn't mean they have the same calibration value.
It means -12db is -12db and their db meters are working correctly, nothing more.


No, we can't assume that.


Before you get to much ahead with this, I'll have to stop you.
These values tell us nothing about how the real world 1v Peak translates to virtual 1v Peak in the plugin.

I don't doubt you, but can you explain why that is?

Or, put another way:

If I run a -12dB peak sine wave into various plugins, why is it that some of their gates work at exactly -12dB, while others need to be offset slightly by some amount? Assuming that they're standard gates and not spectral or complex in some way. Doing the same thing, I had to adjust the STL Amphub gate by exactly 6dB (like I'd expect) for the gate to behave the same as Amplitube or Helix.

 

[JT]

FWIW I've messaged Softube and Mercuriall. Some amount of prodding is going to be useful, but if the same companies keep getting the same questions they may go a bit mad.

I wonder if its possible to use noise gate's to get an idea of what calibration they might be using. Its not as straightforward as assuming all gates are equal because there can be things like the knee and hysteresis to consider. But some do offer values with dB readings, so presumably getting those to correlate to the same input value could help in determining it?

View attachment 2516


Likewise, its not uncommon for (non guitar) plugins to have various calibration settings within the plugin. This just seems like the ideal solution for guitar based plugins, because then you'd be able to set them all to the same (correct) value for your particular rig, and you can use the input controls purely for creative uses (or to preserve the calibration if other things are in the chain). UAD tend to have a Headroom screw and I believe their calibration values are written online somewhere (or perhaps in their manual too).

View attachment 2517

Ehhh difference with Mercuriall is you plug your guitar in, fire up Spark and have the Fists of Godzilla in about 3 seconds. No shenanigans.

 

[James Freeman]

If I run a -12dB peak sine wave into various plugins, why is it that some of their gates work at exactly -12dB, while others need to be offset slightly by some amount? Assuming that they're standard gates and not spectral or complex in some way.

Their meters and math sucks, I don't know.
Amplitube can have a totally different calibration value than Helix Native AND have accurate db meters, one has no correlation with the other.

You assumed that accurate meters and same noise gate db levels = same calibration value?
Why?

 

[James Freeman]

I don't doubt you, but can you explain why that is?

I'll try.

Plugin A: Real 1v Peak = -12dBFS = Virtual 1v Peak.
Plugin B: Real 1v Peak = -12dBFS = Virtual 2v Peak.

Both of these plugins can have very accurate db meters and noise gates.

Now this is truly professional.
It's the same as the Input level we have but they had the wit to put actual meaningful value with it.

 

[MirrorProfiles]

Their meters and math sucks, I don't know.

Is it not more likely that they’re just basing the gate values off the “boosted/virtual” signal level, and any discrepancies are caused by how they’re calibrated? I feel like it’s more likely they landed on the offset values for some calibration related reason rather than just winging it.

My brain’s struggling to keep up, but besides an input level, output level and the internal levels the plugin is operating at, what else could the values relate to?

Are you saying it’s possible that the values could relate to the original (“pre boosted”) signal OR the internal one, and we can’t be sure?

In the cases of STL or Nembrini or Neural where the values are offset by some amount, would that not suggest that they’re scaled off in the internal levels?

This is hard to put into words ha.

 

[James Freeman]

Is it not more likely that they’re just basing the gate values off the “boosted/virtual” signal level, and any discrepancies are caused by how they’re calibrated?

Nope.

Are you saying it’s possible that the values could relate to the original (“pre boosted”) signal OR the internal one, and we can’t be sure?

Exactly.
Don't assume anything, ESPECIALLY when it is convenient.