Little experiment with L6 Helix re; quiet playing & bit depth of input signal

Cirrus

I mentioned in the "has modelling taken over" thread that if I found time I'd hook up my Line 6 Helix rack to my laptop and actually measure what sort of signal level the A/D input converter's dealing with, just out of curiosity. The question that prompted this in the other thread was "could the problem with modelling be that when playing quietly there's not enough bit depth to accurately reproduce the quiet parts, causing bit-crushing style artifacts when that low-volume input signal gets pushed up in level by the modelled amp compression?"

Note that I don't really agree with the premise, but it still sparked my curiosity. So I found time by staying up until 1:30am.

Plugging in my Strat with fairly moderate output single coils, loading an empty patch and using the Helix as the interface into my DAW, this is what I see (all levels are dBFS (decibels full scale, where 0dB is the maximum level that can be recorded);

The noise floor is at about -105dB with the guitar plugged in + volume rolled down. This is pretty much standard for an A/D converter. Note that i didn't actually test it without anything plugged in, it might be a few dB lower if I did that but what's the point in knowing how quiet it is if there's nothing plugged into it? 

When I roll the volume up but don't play anything, I see about -85dB. I'm close to my laptop, so there's EMI causing noise in the pickups.

When I did some really quiet finger-picking, I got a signal where the attack of each note rests somewhere between -40 & -50dB.

When I strum open chords hard enough that I might break a string, I see -13dB.

Interestingly, I can get louder if I do funky triads up the neck - max signal I could get was -11dB.


Next up, my Explorer with Oil City Firebird style humbucker pickups.


Noise floor the same, obviously, -105dB, -95dB with the volume rolled up so much less EMI in these pickups (which makes sense). Gentle fingerpicking peaking around -40dB, when I played as hard as I could it got up to -4dB. So I can see that if you had loud, active pickups, you'd probably want to use the -10dB input pad to avoid clipping.



So, let's see what the volume of a sustained, quietly picked note is on the Strat. That's probably the worst-case scenario.

Turns out, if I play a single note quietly and get a peak of about -50dB, the sustain evens out around -70dB. Obviously eventually it's going to fade down to the noise floor, of course, but the sustain portion of the signal hovered around -70 for a few seconds.

A -70dB signal's using somewhere between 12 & 13 bits. But it's worth noting that this sustain was only 15dB above the analogue noise floor of the guitar's pickups anyway. So I think, while a 12-bit signal will have a digital noise floor 72dB below the level of the signal which might be audible in ideal conditions, that digital aliasing noise will be at least 55dB below the level of hiss/ hum/ interference coming from the pickups, and that'd probably do a pretty good job of totally masking it.

In conclusion, I just don't think low level signals not using all of the available bits is going to impact the resulting sound audibly in any practical scenario. Also, this really is a worst case scenario. I can't think of any time in an actual performance on stage I'd play as softly as I was.

So, there we go. Curiosity sated, and I'm happy that it's just not an issue with 24 bit converters.
 

Cirrus

@p90fool & @ICBM , I know this post is pretty boring but it was based on what we were talking about last week in the other thread.

p90fool

Not boring at all, and that was a pretty noble sacrifice, staying up to do that! 

I realised after I'd posed the question that I'm using a digital 24bit wireless system night after night with none of the same issues I get with modellers, so I'm back to my EQ snapshot theory, that straying too far from the tone and amplitude of the original algorithm just becomes too artificial sounding. 

It's obviously not about having super sensitive ears or high standards (I have neither), but maybe my technique of running the guitar volume low in conjunction with treble pass caps asks too much of the software.