If you had to explain that pre-amp processing headroom issue to a dumb noob (asking for a friend), and mostly what steps should be taken to alleviate it, what would that read like?
Yeah sorry, I came across as little arrogant there, I just really love this device and I'm a bit frustrated that I have been listening to it, posting about it and surprised now to find my EQ settings were out the whole time.
At the very start of this whole thread,
@wslee talked about the need to set the source volume to maximum for best sound results. I agree, but I assumed this meant this would be maintained throughout the chain before the final analog volume stage.
I didn't realise there was quite a lot of headroom on the ES100 side when applying EQ, I just did what I normally do, which is get as close as possible to 0 db without clipping.
My procedure for setting EQ just below 0dB is to set the different EQ bands the way I like, some higher than others, paying attention to the frequency with the highest boost value, taking that value and setting the pre-amp to subtract that much and a tiny bit more (0.1dB) away from the whole signal.
For example, if the EQ has a bass boost of 4.3 dB and this is the highest boost for all EQ frequencies, then I set the pre-amp for this whole EQ to be -4.4 dB
According to the math, this will ensure no clipping. I've done this many times before with wav files in the past and it's consistent - if I don't leave that tiny bit of processing headroom (0.1dB), then the math can round the wrong way and clip a few samples.
That should be the end of it, but Radsone are using either -6dB or -12dB for processing headroom as well, which is actually reducing the sound quality substantially when applying to a real analog amp stage at the end. The amp will respond best to a normalised signal.
This is why I proposed the Auto Gain Protector in the first place - I thought we were all on the same page about setting the volume to very close to 0dB all the way through the chain until applying analog volume at the end.
Sorry to be a nerd about it.
Edit: It depends on the Q value of the bands too, but it is a standard to do things this way with a default graphic EQ.
I'm pretty sure it's a digital volume control at the end, not an analog one, which means the sound quality change suggests the math is rounding badly throughout the whole process, not good.
On fw 1.32, I am using a source volume of 13 out of 15 and an EQ that doesn't go above +12dB when I sum the highest boosted band and the pre-amp and I'm getting no clipping on my loudest recordings. Use only sharp roll-off and higher than 1x oversampling and don't use the crossfeed at all.
