Optimal input level for highest accuracy when using NDSP plugins?

I’ve been posting online a bit recently about this subject, and it would be amazing to have some concrete numbers on what internal reference level the Neural DSP plugins are designed to work at.

I believe that the Quad Cortex DI input is 9.5dBu - if someone was to use this as a DI input for the Neural plugins, and with the plugin input at 0, is this the most 1:1 level for using them?

I’m not looking for responses like “adjust until it sounds good” or guessing - there will be a fixed value used to model/capture the amplifiers that gives the most accurate response (as in the same as plugging directly into the amp).

I believe it to be somewhere in the 9.5dBu-12dBu range but anything more concrete would be greatly appreciated!

1 Like

Just thought I’d update this.

The very helpful and friendly support guys gave a perfect answer that makes sense for me to share (in case anyone else wants to find this information).

TLDR is, if you use a UAD or (most) Focusrite interfaces, set your preamp gain to 0, and plugin input level at unity/0 and you have the perfect gain response (the same as plugging your guitar straight into the amp). If you use something else, check your specs sheet (it’ll give your value with gain at 0). If it’s a number bigger than 12.2, you’ll need to boost by the difference, if the number is smaller then you’ll need to subtract by the difference.

Our plugins are made with the following audio interface gain:

  • Analog domain sine waveform 1 Vp = 0.707 VRMS = -0.79 dBu equals -13 dBFS in the digital domain.
  • In other words, when feeding the audio interface input a sine 1 Vp from a signal generator, it results in -13 dBFS in the digital domain.

If you want to calibrate your interface in order to mimic the input gain our engineers use when creating and testing the plugins, I would advise you to feed a sine waveform 1 Vp = 0.707 VRMS = -0.79 dBu to the interface and set the interface gain to such level that the DAW peak meter shows -13 dBFS. Feeding a sine waveform on different interfaces will result in different values (again, this is the reason why we cannot provide a concrete value). Check these examples of feeding a sine waveform 1 Vp:

  • UAD Apollo x6: -13 dBFS
  • UAD Apollo Twin: -12.9 dBFS
  • Quad Cortex: -15.1 dBFS (1M impedance) Input level at 0 on QC.
  • Focusrite Scarlett: -13.1 dBFS
  • Focusrite Clarett: -14.6 dBFS
  • Apogee Duet: -14.5 dBFS

However, I have to tell you that can be achieved by connecting your guitar to the Hi-Z input of a UAD interface with the gain at minimum (to ease the pain of doing that with all your interfaces and electric guitar combinations). If your interface features a Hi-Z input, leaving the gain input by default (minimum) is more than enough. Add input gain if one of your guitars lacks output level (as our support team suggested, increase it as much as you can without clipping).

What if you use Quad Cortex as an audio interface?

If you use Quad Cortex as an audio interface and you want to get its input close to a UAD interface (-13dBFS), just connect your instrument with the QC’s instrument input at 0.0 dB. If you wanna match them exactly, you have to boost Quad Cortex USB output by approx +2.3 dB before reaching the plugin’s input.



I read your answer, your considerations and I also saw your video on Youtube where you explain the problem.

I find the topic very interesting, so since my sound card is a motu m2 I went to check my max Input level:
Maximum Input Level Mic Inputs: +10 dBu (Min Gain)
Line/Hi-Z Inputs: +16 dBu (Min Gain)

So being the level you advise me to set the gain knob of my sound card to zero and set to +3.8

Now my current setup for using the quad cortex is as follows:

The configuration I am using allows me to enter the quad cortex with the instrument from input 1. To be able to insert a series of effects exit with the USB 3 of the quad cortex and enter the input of a plugin from the plugin I exit and enter the cortex again with the USB 5/6 at this point I exit the Cortex again and enter a Motu M2 sound card

The strange thing is that I also asked support directly and the answer they gave me was completely different from the one you report and which I paste for clarity:

First email:

There’s no specific target for this parameter, but a good practice is to play as hard as possible and increase the input gain until the signal clips. Once it’s clipping, dial it back a bit so it no longer clips. This will give you the best signal-to-noise ratio.
Alternatively, you can just leave it at 0.0dB, which should be good for almost everyone unless your DI signal is too loud.

Second Mail

Indeed, the gain level of both inputs will mainly depend on the signal level, going out of the QC to your interface’s input.
In both cases, the same rule applies: to get the best signal/noise ratio, always make sure to keep the signal in the green range, so before any clipping starts to happen; there’s no other predefined rule.

What do you think? can you give me your point of view?

The answer they gave you there is essentially what customer support is trained to say, which is unfortunately quite vague advice (but potentially leads to less confusion at the cost of less accuracy). It took me a lot of work to get a direct answer from the developers with more specific information. Their advice of strumming hard and avoiding clipping would lead to single coil guitars being far too overgained, and you’d lose the nuance of different pickups between different guitars.

If you use your QC as a USB interface for NDSP plugins, the most accurate input level in the plugins will be +2.8dB.

If you use your MOTU M2 as an interface (with gain at 0), your value will be 3.8.

If you are doing more complex routing combining both bits of gear, I would recommend sending a sine wave through each chain and comparing the levels you get. Use M2+3.8dB or QC+2.8dB as your target.

Hi, thank you again for your reply. What you say makes sense, but I would like to be able to resolve my configuration with your help if you are available.

So I’ll try to be more clear.

First step quad cortex input instrument: here all input values ​​must be left at zero.

Second step from the quad cortex to the input of the neural dsp plugin here the output values ​​of the cortex USB must be left with the factory settings while the input on the plugin must be adjusted to 2.8.

Third step from the plugin to the quad cortex, in this case the plugin output is different for each scene so I imagine it shouldn’t be touched, while what values ​​should the inputs on USB 5/6 of the quad cortex have?

Fourth step, from the quad cortex to the motu m2 card, also in this case the gain of the card must be at 0 but the output of the quad cortex?

I hope I have been clearer and have your support in this regard

I don’t own a Quad Cortex so I can’t verify levels. You’d probably want to make sure that things are level when going in and out of the quad cortex. But also, what exactly are you trying to achieve?

If you want an accurate gain response, you might just find it easier using just plugins or just the quad cortex, or at most one round of conversion. Latency will probably be more annoying to have to deal with, and combining interfaces generally isn’t ideal for stability.

It’s certainly possible to do what you’re suggesting and have it perfectly calibrated level wise, but it seems like more hassle than it’s worth (unless i’m missing something?).

Maybe this video can you to understand better:

I follow the Tom suggestion, but i add 1 step before for insert the effect like pitch shifter or compressor before the plugin.