Just wondering, as I now have some PEQ going on and this has for the first time caused clipping on certain tracks, what’s the best way to prevent that… volume levelling or headroom management?
If both these features use the same code and often folk say they can hear a lowering of SQ using volume levelling, then surely if I apply -5db (seems to be the magic number with my EQ settings to prevent clipping) then isn’t that going to effect SQ on all tracks?
A lot of my stuff is 24bit and my DAC as far as I know accepts the 32bits that Roon seems to be sending it - so maybe this isn’t an issue?
I think you could also apply some across the board gain reduction in the EQ to mitigate any clipping risk. If your maximum peak gain is +5.0, for example, you could try to reduce everything by -3.0 (adjust this depending on your circumstances) and see if that helps with the clipping issue.
Volume levelling doesn’t always reduce level; what you should be looking at is (as I think you’ve already decided… ) is headroom management.
As 16 bits is (pretty much) enough to encode the entire audible dynamic range, a 5dB reduction in level calculated as 64bit float and output to a DAC at 32bit resolution really shouldn’t be audible! If you really want to prove it to yourself, make sure you turn the level up at the preamp so it’s level matched. A sound level meter on a smartphone and some pink noise will help. The actual calibrated level doesn’t matter, it just needs to be the same.
If you raised the level on one or more filters above 0dB, then it’s clear that material with levels approaching 0dB in that frequency region will clip the output.
One additional countermeasure is to just grab the PEQ curve in the editor and pull it down so it’ll stay below 0dB, hence no clipping…
Thanks everyone for the advice and info. So @Marin_Weigel if I’m understanding this correctly, if I drag that right hand slider down in PEQ so the 0dB line is around level with the highest peak of my boost adjustment (6dB seems to do it) I’m applying (with a useful visual aid of the whole graph moving down) the same effect as an overall 6dB headroom reduction? Which should avoid all clipping as there will be no level increase in my EQ?
I honestly haven’t bothered with EQ for room or system correction before, except some speaker placement and balance adjustments, but after an afternoon with REW it was great to see on the measurement graph exactly what I was hearing, and then set about correcting it! The brightness I had was clearly visible as a significant raised area way above the ideal curve and just letting REW carefully pull this down a little really improved things. But yes it did lift a couple of small areas as well, ignoring some giant bass suck outs which I thought was quite clever.
So gathering, is something like the MiniDSP hardware better at this than Roon, in that it won’t effect SQ? Or is the principle exactly the same regardless?
Did you read @Magnus excellent guide on doing measurements, generating convolution files and applying them in Roon?
You ought to try that as a next step, which will probably clear up your sound even better.
I have a friend with an older Mini DSP DDRC unit with Dirac, which does a really good job with his Magnepan/Tube amp system, you wouldn’t want to hear it without!
I guess, you always need to check specs against your goals and desire for future proofing your investment.
In my opinion, software solutions seem to fare better in the long run, compared to hardware with limited upgradability.
There’s several paid and potent software solutions, as well as paid filter generation services (after you did the measurements).
Or, you can use REW and then Rephase for even better corrections, with the penalty of a learning curve.
Sound quality wise, with today’s computation power and resolution, I honestly think, there’s no audible degradation with any of those solutions, as long as best practices are considered.
Any system will gain the most by optimising the loudspeakers and their interaction with room acoustics.
DACs, amps, upsampling, reclocking, cabling, mechanical and electrical isolation -in no specific order- are orders of magnitude less important.
Have fun with your tweaking and enjoy the enhanced quality of your music!
Hi - yes I started with the Magnus Opus! Read it months ago, every single post, started off thinking I’d grasped it then realised there were a hundred way to do everything he mentioned!
Someone posted a very simple video to that thread and I used that to get the gist (typical setting and so on), then grabbed ideas from those who had used RTA and went with the ‘wave the mic about’ technique! In the end I tried various methods and the measurement results were all very similar, typing EQ settings into Roon was favourite as I thought I might tweak afterwards and sure enough I did want to.
I’ve done myself a tutorial as well, step by step as though I was explaining to a complete idiot (it’s so easy to miss telling someone that when you say ‘turn the volume down’, you mean the physical hi-fi amplifier volume, not some volume that might or mightn’t be buried somewhere in the REW interface!) - because when I get to do this again I will have forgotten completely I’m sure…
just checked out your nice system with those full range speakers.
Just like with my DIY full range line source speakers, you’d have no need to use Rephase in addition to REW, since we’re blessed with a nice time-coherent step response anyway.
I think, you’re doing fine with your technique!
Is there any SQ loss from using both volume levelling and headroom management? I’d been running VL before getting into room correction. After I started using a room correction filter (generated with REW), I did start to get clipping occasionally (with VL still on). So now I’ve been running VL, HM and a RC filter all together (in that order, as shown under signal path).
Since all these manipulations are computed with 64 bit floating point precision, there should be no possibility for sound quality degradation, even when using 24 bit source material.
Anyway, what matters in the end, is your effectively enhanced musical enjoyment.
I’ll add my thoughts as a MiniDSP DDRC-22D owner. I played with REW to good effect but a desire to correct other sources like the TV tempted me. The Dirac software offers a bit less control than REW but it’s easier to use and uses FIR filters as well as IIR filters for phase correction. MiniDSP support IIR only boxes and provide a good, balanced comparison of Dirac vs REW and FIR Filters vs. IIR. For me the biggest audible Dirac advantage is imaging accuracy and a marginally wider soundstage, but I’d not swear to the latter .
I probably could be happier with the results but there’s not much wiggle room. I use DSP for headphone listening also via Roon and while there’s still differences the overall character of the sound across devices is pretty consistent. Bass controls the big one for me and the first thing I listen for when testing. Once that’s in place everything sounds better. There’s no point listening for splashy cymbal noises if you can’t hear them properly. Martin’s advice seems sound also BTW, you have better speakers than my AE1 actives.
Yes it’s a shame you can’t reroute other stuff through Roon for DSP. I’ve just popped an optical toslink cable into the DAC from the telly for listening to the BBC Proms concerts this year (catchup on iPlayer) and straight away I realised my Roon balance and speaker distance DSP settings of course couldn’t be implemented - so that’s where the MiniDSP hardware certainly comes in.
Here’s a pic of my Mark Audio drivers, hugely revealing but I have a sneaky feeling that’s where the brightness originates - still EQ has certainly dealt with that, pulling down that excessive treble response REW showed from around 500 to 2Khz. I’m still not too sure about the bass, mines all over the place and REW pretty much left it alone - just pulled down a couple of peaks and left the deep troughs alone (which I presume are down to room problems anyway). It sounds ok though, goes very low and no obvious issues when sat in my listening position. It’s amazing how changing the interconnects between the DAC and amp dropped the bass right down and opened up the top end no end! It was really useful seeing that displayed in the REW measurement graph and I’m pretty sure I’ve cured the brightness without losing any of the detail - might work on the bass a bit more one day though!
PS.The drivers have been upgraded to the Alpair 11MS (more copper coloured than the Pluvia’s silver finish) this speaker has no spider so travels without any resistance and it sports a very flat profile and a novel slight shoulder around the cone perimeter - all designed to better reproduce the human voice so I’m told. As an engineer Mark Fenlon is always innovating, I was just lucky these latest one’s dropped into my Pluvia Pensil 11 box I already had without any adjustment - I might try building the Frugal Horns next, supposed to be very good!
I am in the same situation
I need to use PEQ for my room and it clips
So I need to use HM as well
But when I occasionally listen to playlists, turning on VL is nice but then I have VL, HM and PEQ on all together
The sound is good but I don’t know if theoretically this isn’t the thing to do
Who am I to doubt the CTO’s assessment…but he seems to say ‘theoretical’…
If you are using boosts in your PEQ then of course clipping can occur.
Instead of adding HM you can also pull down the slider to the right of the PEQ graph to keep the curve below 0dB, but the net result will still be the same reduction in overall level.
An adjustment of 6dB is worth a bit of resolution, but calculation is performed using 64 bit, so ‘theoretically’ you shouldn’t be worried about your 24bit file.
I understood headroom management is quite the same than moving the whole level down by some dBs in the PEQ. The fact that these processes are not in same order of the signal path (ie for example HM comes before sample rate converter and PEQ comes after) does have any theoretical impact on precision and in general in SQ?
Your questions can properly be answered only by someone who has in-depth knowledge of the mathematics involved in the underlying code and/or who has the means to scientifically test the various possibilities.
Lacking both, my unscientific dabbling and listening shows no discernible differences.
I also prefer discovering new music instead of worrying about the n-th degree of refinement, repeatedly listening to the same ‘reference’ tracks over and over while tweaking things.
Regular exposure to live acoustic music, impossible during the pandemic, helps to judge the fidelity of my system. I’m in the lucky position that there are many recordings of Munich’s Symphonieorchester des Bayerischen Rundfunks in the Herkulessaal venue, which I know quite well.
.