Attenuating The Dirac Target Filter?

[wanta911]

I have been messing around with Dirac in my setup using one of the MiniDSP units (DDRC-22D).

 

I was curious about whether or not attenuating the target curve would theoretically be of any benefit? For example once you get the Dirac target curve or tailor it to your preference, if you drop all the target points 3dB and then optimise, you will get a higher % of cutting versus boosting of frequencies (at least that is the case in my particular scenario). In effect the resultant curve or filter is exactly the same, albeit 3dB lower overall.

 

Everything I read on EQ adjustment suggests that cutting is better than boosting, I am not sure exactly as to why, other than than the obvious potential problem of higher levels producing clipping?

 

Anyway, I would be interested in any thoughts on this. Maybe Flavio can provide a perspective?

[wgscott]

Does this differ from adjusting the "DSP Gain"?

 

[wanta911]

Does this differ from adjusting the "DSP Gain"?

 

 

Thanks - I don't know as the MiniDSP/Dirac unit doesn't have "DSP Gain" as such, I also don't have the clip indicator (which would be handy). I have an output volume slider in the filter export screen but I am not sure it is the same? I assume your screenshot is from the computer/Mac software? I haven't used it so I am unaware of how it works. Here is mine showing the slots for 4 saved filters:

 

I would also assume "output volume" is exactly that, as it is a step after the creation of the filter and is only used during and after exporting. Everything I read suggests that output volume should be 0dB when exporting created filters?

 

So I guess my question on manipulation of levels before the creation of the filter still stands and whether or not cutting is better than boosting?

[wgscott]

I don't know the answer, but my guess is "Output Volume" is the same thing. The screen shot I think is windows, but the Mac version is fairly similar. Dirac for the computer used to set it to -8 dB by default, so it might be different, or normalized differently.

[ecwl]

It would be nice to have someone from Dirac/miniDSP to comment. I have only played around with the DDRC-22A. I don't think "Output Volume" is the same as DSP Gain. I agree that "Output Volume" in miniDSP is likely a post-Dirac digital volume attenuation. I strongly suspect DDRC-22D may not have a true DSP Gain setting. My guess is that it just defaults to the PC/Mac equivalent of -8dB for Output Volume. So the digital signal that gets fed is automatically reduced by -8dB digitally and then goes through Dirac to be digitally corrected and sent to the output. That way, at the minimum, you have 8dB headroom to boost the trough frequencies. And if you really, really need that 10-20dB frequency boost, it'll only clip by 2-12dB if the music amplitude is maxed out in those frequencies. If I were correct, and if your Dirac target curve does not contain any signal boost beyond 8dB, you will likely gain nothing from lowering your target curve except by reducing the digital volume, you're actually reducing your SNR that's coming out of your DAC, because you won't be using the full 24-bits but might end up using 21- to 23-bits of dynamic range. And even if the Dirac Live curve makes a small correction to a trough frequency of 12dB, my suspicion is that most of your music would not have maximal amplitude at that frequency. So the advantage of lowering your target curve by 4dB to avoid digital clipping in that particular frequency will likely not outweigh the disadvantage of losing almost 1-bit in SNR. But all this is completely speculative.

[wanta911]

It would be nice to have someone from Dirac/miniDSP to comment. I have only played around with the DDRC-22A. I don't think "Output Volume" is the same as DSP Gain. I agree that "Output Volume" in miniDSP is likely a post-Dirac digital volume attenuation. I strongly suspect DDRC-22D may not have a true DSP Gain setting. My guess is that it just defaults to the PC/Mac equivalent of -8dB for Output Volume. So the digital signal that gets fed is automatically reduced by -8dB digitally and then goes through Dirac to be digitally corrected and sent to the output. That way, at the minimum, you have 8dB headroom to boost the trough frequencies. And if you really, really need that 10-20dB frequency boost, it'll only clip by 2-12dB if the music amplitude is maxed out in those frequencies. If I were correct, and if your Dirac target curve does not contain any signal boost beyond 8dB, you will likely gain nothing from lowering your target curve except by reducing the digital volume, you're actually reducing your SNR that's coming out of your DAC, because you won't be using the full 24-bits but might end up using 21- to 23-bits of dynamic range. And even if the Dirac Live curve makes a small correction to a trough frequency of 12dB, my suspicion is that most of your music would not have maximal amplitude at that frequency. So the advantage of lowering your target curve by 4dB to avoid digital clipping in that particular frequency will likely not outweigh the disadvantage of losing almost 1-bit in SNR. But all this is completely speculative.

 

Your thoughts make sense but I wonder about the 8dB headroom because I did a software update (2.16 to 2.17) and it was definitely louder.

 

I am a little paranoid about all this after using a particular DAC in my chain some time back (before the Dirac) that had a light to indicate clipping and I was somewhat shocked by the amount of times it came on. It wasn't a lot in the grand scheme of things but it did highlight something that I had never considered before.

[ecwl]

Sure. But the clipping you saw (without Dirac) is from the dynamic compression and mastering process of the music itself. You can even argue it is intentional, possibly approved by the artist. So if you're not hearing any worsening in sound quality with the Dirac, I probably wouldn't worry about it. But if you still want to know, it may be worthwhile asking miniDSP/Dirac directly.

[LBob]

I don't know the answer, but my guess is "Output Volume" is the same thing. The screen shot I think is windows, but the Mac version is fairly similar. Dirac for the computer used to set it to -8 dB by default, so it might be different, or normalized differently.

 

The Mac version doesn't have the clipping indicator (boo!). You have to fiddle with it to get the right level.

[flak]

Hi wanta911,

 

yours is a good question... the maximum filter gain is appx. 10dB so some dips in the frequency response will not be fully corrected.

This is intentional because sharp peaks in the filter can sometimes be heard as resonances, especially outside the sweet spot.

If you lower the target curve then more dips will potentially be corrected.

In a few words lowering the target curve is not generally recommended but advanced users can experiment with it.

 

Flavio

 

 

I have been messing around with Dirac in my setup using one of the MiniDSP units (DDRC-22D).

 

I was curious about whether or not attenuating the target curve would theoretically be of any benefit? For example once you get the Dirac target curve or tailor it to your preference, if you drop all the target points 3dB and then optimise, you will get a higher % of cutting versus boosting of frequencies (at least that is the case in my particular scenario). In effect the resultant curve or filter is exactly the same, albeit 3dB lower overall.

 

Everything I read on EQ adjustment suggests that cutting is better than boosting, I am not sure exactly as to why, other than than the obvious potential problem of higher levels producing clipping?

 

Anyway, I would be interested in any thoughts on this. Maybe Flavio can provide a perspective?

[LocutusEstBorg]

Sometimes the default target curve clips so you have to lower it. The -8 dB gain is insufficient in such cases.