Calibrating Input Level for Plugins

[James Freeman]

seems like any time something passes through an input block, including a plugged-in guitar, it gets a +18dB boost

Yes, only digital inputs have +18dB internal boost, digital outputs are unaffected, so it seems.

 

[Whizzinby]

When I:

• Record a guitar DI track into my DAW (USB channel 5)
• Record the stereo output of the FM9 (USB channels 1/2) at the same time
• Run the DI track back out to the FM9 via USB to re-amp it…

it requires that built-in +18dB boost to sound the same as the recorded stereo output. Just a heads-up to people who are re-amping via USB and using the amps in the box.

It almost seems like any time something passes through an input block, including a plugged-in guitar, it gets a +18dB boost, but the DI output on USB5/6 (for the FM9) does not get that same boost.

I use this setup to record and ultimately reamp. Should I be altering any of the IO anywhere, because to this point I have been leaving everything stock.

 

[SwirlyMaple]

I use this setup to record and ultimately reamp. Should I be altering any of the IO anywhere, because to this point I have been leaving everything stock.

If you're re-amping Fractal amps in the box (not re-amping real amps) then nope, you don't need to change anything. Default values will give you correct, consistent results. This is because it applies that "hidden" +18dB boost to anything through an input block, so whether it's your guitar or your DI, they will come out sounding the same.

 

[SwirlyMaple]

Yes, only digital inputs have +18dB internal boost, digital outputs are unaffected, so it seems.

I wonder if what we're seeing with that +18dB passing through input blocks is the internal compensation for the input trim. Maybe that hidden +18dB value varies depending on the A/D sensitivity we set. Edit: hmm, nope. Tested it with the A/D sensitivity max and min and it doesn't seem to have any effect on the +18dB boost happening at the input.

 

[James Freeman]

I think only the instrunent input is auto compensated so the amp models always have correct gain.

 

[Wasbeer]

Color me an idiot. I can't find the spec I need for the Behringer 202hd.. Ugh, I are the stupid.

I finally found something that says that the Max Output is +3. That seems not right though.
I have been leaving it's preamp levels at 0 and adjusting the input gain on the plugs to where
I think they should be by ear. That isn't the optimum method though.

The Behringer UMC202 and 204 indeed have almost no headroom @gain 0 (max. instr. input level -3dBu). You have to use the PAD on the input to engage a ~20dB pad to actually have a usable input headroom of -23dBu and work from there.

 

[DLC86]

I don't get how you guys came to the conclusion that it's the inputs that add the 18dB boost and not (some) outputs.

- Recording the DI signal via usb inputs 5/6 doesn't have that boost, while doing it by directly connecting IN1 to OUT1 blocks in the grid and recording usb inputs 1/2 (or 3/4 if using OUT2) has the boost.
- Recording the DI signal via SPDIF by selecting "INPUT 1" as the SPDIF/AES out source doesn't have the boost, while selecting "OUTPUT 1" or "OUTPUT 2" has it
- Feeding back a signal from a PC to the grid via any digital input (USB input block or other input blocks set to "digital") doesn't add any boost
- OUT3 and OUT4 (the "unity gain" outputs) are exactly 18 dB lower in volume than OUT 1 and OUT2

All leads me to think the boost is only applied to outputs 1 and 2 to bring the signal to line level, boosting at the inputs and attenuating at some outputs would be a kinda cumbersome implementation... unless there's a good reason to do so.

I think only the instrunent input is auto compensated so the amp models always have correct gain.

Nope, all inputs are auto-compensated and have exactly the same gain (in my tests at least)

 

[James Freeman]

All leads me to think the boost is only applied to outputs 1 and 2

Nah, put a Gate/Expander (Modern Gate) block and check again.
The +18dB boost is definitely on the Input.

EDIT:
You can't output more than 0.0dBFS from your DAW so why the 18dB of headroom in the Fractal?
This totally screws with D.I tracks for reamping.

 

[DLC86]

Nah, put a Gate/Expander (Modern Gate) block and check again.

What should I see with a gate block?

You can't output more than 0.0dBFS from your DAW so why the 18dB of headroom in the Fractal?
This totally screws with D.I tracks for reamping.

By recording the DI track via usb inputs 5/6 or SPDIF (with source set to "INPUT 1") you get the exact signal level that's needed to send it back to IN1/USB IN. I really don't get what kind of issues you're having with reamping levels, care to explain a little more?

 

[SwirlyMaple]

I don't get how you guys came to the conclusion that it's the inputs that add the 18dB boost and not (some) outputs.

- Recording the DI signal via usb inputs 5/6 doesn't have that boost, while doing it by directly connecting IN1 to OUT1 blocks in the grid and recording usb inputs 1/2 (or 3/4 if using OUT2) has the boost.

Yep, that's right. The boost is happening after a signal passes through an input block on the grid.

Recording the DI via inputs 5/6 does not require any blocks on the grid. It's a true DI and thus lacks the boost.

- Recording the DI signal via SPDIF by selecting "INPUT 1" as the SPDIF/AES out source doesn't have the boost, while selecting "OUTPUT 1" or "OUTPUT 2" has it

Same reason as above. Output 1 and Output 2 necessarily require an input block on the grid to connect them.

- Feeding back a signal from a PC to the grid via any digital input (USB input block or other input blocks set to "digital") doesn't add any boost

It does if it passes through an input block on the grid.

- OUT3 and OUT4 (the "unity gain" outputs) are exactly 18 dB lower in volume than OUT 1 and OUT2

On my FM9, you can't route OUT3 in via USB at all, so I can't test this one.

What should I see with a gate block?

You can use this to see what the internal signal level is on the grid, between the input and output blocks. If I send a -22dBFS signal from my DAW to any of the inputs, and then put a gate after an input block on the grid, the gate pops open at -4dBFS. That proves the +18dB boost is happening at the input block and not at the output, because the gate is prior to the output block.

 

[SwirlyMaple]

You can't output more than 0.0dBFS from your DAW so why the 18dB of headroom in the Fractal?
This totally screws with D.I tracks for reamping.

It adds that boost once the signal passes through an input block on the grid. So if you're playing straight in with your guitar, it gets that +18 before any other blocks.

When you record a DI via its USB5/6 (dedicated DI) sends, it doesn't have the boost. But when you send that DI back to the unit for re-amping, you have to send it through an input block on the grid, where it gets the +18 boost, the same as playing a guitar straight-in.

So you still get consistent results -- you don't have to boost the track from your DAW. But it is still weird that it works this way. It's like all of the models have been calibrated to an input signal boosted by +18dB from what the input A/D produces. I'm sure there must be a good reason for it -- maybe it reduces input noise or something.

 

[DLC86]

Yep, that's right. The boost is happening after a signal passes through an input block on the grid.

Not if you use IN1 connected to OUT3, so it's either that or OUT3/4 having a -18dB attenuation

Recording the DI via inputs 5/6 does not require any blocks on the grid. It's a true DI and thus lacks the boost.

Correct

Same reason as above. Output 1 and Output 2 necessarily require an input block on the grid to connect them.

Correct as well

It does if it passes through an input block on the grid.

At this point we can't still be sure it's due to the input block or the output block

On my FM9, you can't route OUT3 in via USB at all, so I can't test this one.

I currently have an FM9 too, you can test this by looking at the output clipping, boosting OUT3/4 level by 18dB gives you exactly the same clipping point as that of OUT1/2.
So, again, the boost could be applied at the input and compensated at OUT3/4, or just applied at OUT1/2

You can use this to see what the internal signal level is on the grid, between the input and output blocks. If I send a -22dBFS signal from my DAW to any of the inputs, and then put a gate after an input block on the grid, the gate pops open at -4dBFS. That proves the +18dB boost is happening at the input block and not at the output, because the gate is prior to the output block.

I tried to test this but I don't get consistent results. Anyway the threshold value you set in the gate block is expressed in dB and not dBFS, it might be just an arbitrary internal dB scale (just like the VU meters), so that's more a clue than a proof to me.

I still tend to think it's at the output cuz on the axe fx II and ax8 it used to be applied at the amp block instead, so for this new generation they decided to move it to another place (for some good reasons I can pretty easily guess), and it surely makes more sense if that other place is the output block(s).

Anyway... I don't see how all this really matters in the end, since for all practical purposes it makes no difference if the boost is at the input or output

 

[SwirlyMaple]

Not if you use IN1 connected to OUT3, so it's either that or OUT3/4 having a -18dB attenuation

I currently have an FM9 too, you can test this by looking at the output clipping, boosting OUT3/4 level by 18dB gives you exactly the same clipping point as that of OUT1/2.
So, again, the boost could be applied at the input and compensated at OUT3/4, or just applied at OUT1/2

Yep, thanks. I see on the output level meters that Output 3 is 18dB 20dB less than Output 1 & 2 (so maybe it is using a different scale). But...

I tried to test this but I don't get consistent results. Anyway the threshold value you set in the gate block is expressed in dB and not dBFS, it might be just an arbitrary internal dB scale (just like the VU meters), so that's more a clue than a proof to me.

The attack, hold and release times need to be set to reasonable values for it to work consistently, and it also might depend on what kind of input signal you're feeding it. (Also, there is a hysteresis with the open/close of the gate. For an accurate result, start from a high enough threshold that the gate is fully closed, and then increment it down until it pops open. The value where it opens is equal to the signal level that is running into it.)

If you put gates after Output 1 and Output 3, they both still pop open at the same dB threshold value as a gate upstream of those output blocks. This is despite the Output 3 block level meter showing a -20dB lower level than Output 1's block. Maybe the signal passing through the output blocks is the same level, and it is in fact the Output 3 block that removes the extra input gain.

I agree on your point about the scales, but it'd be a real coincidence that I'm measuring a difference of exactly 18dB from the signal I'm feeding it from my DAW using the threshold of the Gate block. But the Gate block's threshold doesn't agree with the output meters, so maybe the output blocks are using a different scale (VU?) for those output meters vs. the scale referenced by the gate block. Honestly not sure; I'm scratching my head here.

Output block 1's meter reads 0dB for a signal that pops the gate open at -12.0dB. Output block 3's meter reads -20dB for that same input signal. And from my DAW, I'm feeding it a sine wave that is -30dBFS RMS (-11.2 peak).

 

[riffy]

The Behringer UMC202 and 204 indeed have almost no headroom @gain 0 (max. instr. input level -3dBu). You have to use the PAD on the input to engage a ~20dB pad to actually have a usable input headroom of -23dBu and work from there.

So pardon me if I am being obtuse. I promise it isn't intentional. At -23 with my pad engaged and I am using NDSP plugins, how should I set the input gain on the plugin itself. I am sorry for being ignorant, but I am trying to learn.

 

[SwirlyMaple]

My last reply got confusing, so to clarify, here is what I'm observing with my FM9:

• Signal from DAW is a -30dBFS RMS (-11.2 peak) sine wave.
• Signal goes to Input 1 on the grid.
• A gate placed after this pops open at an indicated value of -12.0dB.
• This signal reads 0.0dB on Output 1's meter and -20.0dB on Output 3's meter.
• Gates placed after Outputs 1 and 3 still pop open at an indicated -12.0dB.

That's all kinda weird, because the FM9 output meters are showing a difference of 20dB between outputs 1 and 3. But the value indicated by the gate shows a difference of 18dB from my input signal.

And, if I send a signal from my DAW into input 1 and take it right back out via Output 1 or 2, it comes back into the DAW +18dB higher than I sent it. This agrees with the offset we see between the DAW send level and the threshold that the gates pop open, but the scales on the output block meters don't seem to agree with either the scale nor the value from the Gate block.

 

[SwirlyMaple]

So pardon me if I am being obtuse. I promise it isn't intentional. At -23 with my pad engaged and I am using NDSP plugins, how should I set the input gain on the plugin itself. I am sorry for being ignorant, but I am trying to learn.

No need for apologies! I can try to help. Here's the spec for your interface:

Under "max input level," it says Instrument: -3 dBu. (That is kinda weirdly low for an instrument input, but...) What this means is that, with that input's gain knob set to min and the selector for instrument engaged, it will convert a -3 dBu analog signal to a 0 dBFS digital signal.

For Neural's plugins, we have this conversion info:

I have official confirmation from Neural DSP that their plugins are made with this reference:

-0.79 dBu equals -13 dBFS in the digital domain

Since -0.79 dBu = -13 dBFS for Neural's models, then 12.2 dBu = 0 dBFS (add 13 to each side). This is the reference level they used for making the models.

Going back to your interface, it converts -3 dBu to 0 dBFS, so we have to compensate: set the input to the NDSP plugin to a value of - (12.2+3) = -15.2 dB. The reason we have to drop the input level is because your interface is converting a smaller analog signal to a larger digital one than the reference they used, so we lower the input in the plugin to correct this offset.

TL;DR #1: With your interface set to instrument level, the gain knob at minimum, and the pad disengaged, set your Neural plugin's input trim to -15.2dB.

Now, about the pad: I don't see a value for it anywhere in the spec sheet, so we're guessing here. Let's say it really is a 20dB pad. With the pad engaged, now your interface converts 17 dBu to 0 dBFS. (-3+20 from the pad = 17). So to get it to the reference level used by Neural, you need to add 4.8 dB (17-12.2) of input trim in the plugin.

TL;DR #2: IF the pad on your interface really is 20dB (I can't confirm that!), with your interface set to instrument level, the gain knob at minimum, and the pad engaged, set your Neural plugin's input trim to +4.8dB.

 

[James Freeman]

I think the dB scales/meters of the Output blocks have +12dB of headroom to clipping, it's an arbitrary scale where 0 dBFS is actually -12 dBFS.

 

[riffy]

No need for apologies! I can try to help. Here's the spec for your interface:

View attachment 14655


Under "max input level," it says Instrument: -3 dBu. (That is kinda weirdly low for an instrument input, but...) What this means is that, with that input's gain knob set to min and the selector for instrument engaged, it will convert a -3 dBu analog signal to a 0 dBFS digital signal.

For Neural's plugins, we have this conversion info:


Since -0.79 dBu = -13 dBFS for Neural's models, then 12.2 dBu = 0 dBFS (add 13 to each side). This is the reference level they used for making the models.

Going back to your interface, it converts -3 dBu to 0 dBFS, so we have to compensate: set the input to the NDSP plugin to a value of - (12.2+3) = -15.2 dB. The reason we have to drop the input level is because your interface is converting a smaller analog signal to a larger digital one than the reference they used, so we lower the input in the plugin to correct this offset.

TL;DR #1: With your interface set to instrument level, the gain knob at minimum, and the pad disengaged, set your Neural plugin's input trim to -15.2dB.

Now, about the pad: I don't see a value for it anywhere in the spec sheet, so we're guessing here. Let's say it really is a 20dB pad. With the pad engaged, now your interface converts 17 dBu to 0 dBFS. (-3+20 from the pad = 17). So to get it to the reference level used by Neural, you need to add 4.8 dB (17-12.2) of input trim in the plugin.

TL;DR #2: IF the pad on your interface really is 20dB (I can't confirm that!), with your interface set to instrument level, the gain knob at minimum, and the pad engaged, set your Neural plugin's input trim to +4.8dB.

I read it all!!! Thank you! You make a great instructor!

 

[Wasbeer]

I'd really use the pad anyways and compensate NDSP with +4.8dB otherwise you're still stuck with a -3dBu input which clips crazy fast. Back when I had the UMC204HD myself I did several tests with the help of MirrorProfiles and we concluded the pad is ~20dB on the Behringer

 

[Wasbeer]

So pardon me if I am being obtuse. I promise it isn't intentional. At -23 with my pad engaged and I am using NDSP plugins, how should I set the input gain on the plugin itself. I am sorry for being ignorant, but I am trying to learn.

Sorry, I was wrong on this one. Like Swirly said, it's not -23 with the pad engaged, but +17