PCM->DSD128 conversion in A+ = amazing SQ improvement

I have been experimenting with Audirvana PCM->DSD128 conversion for my SGCD over the weekend and I’d like to report my findings. Notice I was not told to do this and just tried it on a whim. I have tried PCM upconversion in Audirvana in the past and was underwhelmed so I wasn’t expecting much. Well I am shocked at the SQ improvement here.

Audirvana uses Izotope 64bit SRC to do upsampling/conversion to the DSD sample rate and then uses a modulator to go 1-bit DSD.

Here are my Izotope settings :

  • Slope 10db
  • Max filter length 1,250,000
  • xNyquist 1.00
  • Anti-aliasing 200
  • Pre-ring 1.0 (this results in a pure linear phase filter)
  • DSD filter is set at C(7th order)
On my system there is no question there is a BIG SQ improvement with the DSD conversion. The sound is smooth, realistic, improved vocals, vastly improved depth, transparency, note decay, ambiance, you name it . Sounds like a different system running the DSD conversion. The sound is much more REAL!

I have experimented with the Pre-ring which lets you change the balance of minimum vs. linear phase filtering. I have messed with this a lot and I find anything other than pure linear to have too many negative trade-offs. For example, sibilance goes up significantly and reduces the lifelike quality, depth decreases, and there is too much of an edgy quality.

Signal path is Audirvana to Ultrarendu (powered by Sbooster Linear Power supply) via DLNA (router and FMC’s/powered by linear power supply) to SGCD via WW silver starlight USB. For those unaware, the Ultrarendu uses similar USB hub circuitry as the new Uptone Iso Regen which produces excellent signal integrity/very clean signal.

Just updated above settings… moved cutoff back to 1.00x Nyquist. This is to avoid aliasing leaking through. To me this sounds more natural, raising the cutoff at all increases sibilance and distorts cymbal transients noticeably.

Updated again to 10db slope… this increased musicality hugely.

Fun reports. Your experience supports the decisions made in the design of the DirectStream.

I use iZotope’s SRC, bit-depth conversion/dither, and their mastering suite in my own recordings. Wonderful software.

More experimentation tonight… cause why not.

Goal : determine if performance gains can come from moving S300 to the wall and move the SGCD to the High-Current section in parallel with MC.5 magnet on the Dectet.

I have two dedicated 20A circutis with powerport outlets.


(1) AC-5 to Dectet/MC.5 on unused outlet on wall. S300 to HC section via AC-5. SGCD in its own isozone next door with AC-5.

(2) S300 direct to wall via AC-5. AC-5 to Dectet. SGCD to HC via AC-5 / parallel MC.5 magnet in spare HC outlet.

Setup 2 above sounds much more live and realistic to me… more body to drum kits, better holographic imaging. Cymbals have more air, vocals are a little more dimensional and realistic. Music is a bit more relaxed in presentation.

Went back and tested PCM vs. DSD conversions on SGCD. The power cable switching makes a big difference for both but DSD converted PCM is still amazingly live sounding compared to straight PCM.

1 Like

Rapid fire update…

I decided to go more bold and move the SGCD to the wall and bam… I’m in heaven :slight_smile:

BTW this magnet is a no go on the dectet… sounded cool for 20 min but it produced a fatiguing sound forcing highs a little too forward and exaggerated.

I really like where we are now!

New settings for A+ DSD up-conversion. So it took some digging since I don’t have full Izotope software but using lower than ~31 for filter steepness introduces aliasing. Therefor to prevent this, here are my settings.

Slope : 31
Filter length : 500,000
Cutoff : 0.94 (this is critical and will give you 31db attenuation At the Nyquist) if you use 1.0 you will get aliasing unless you go steeper.

Anti-aliasing : Use 150 to 200.

Pre-ring : 1
DSD filter : C 7 th order

Recently I changed speakers and just sat around with the same DSD thing convinced it was the way to go.

Well I jumped up from $600 polk monitor towers up to the GET TRITON 3+.

Considering the SGCD/S300 has a full warm bodied sound, I think it was just to much for the Polks. The DSD conversion gave me much better instrument separation and was clearly an improvement. Fast forward to the Tritons and I kept tweaking and thinking they sounded a little suppressed and a little thin. Recently I tried PCM again with A+ and bam - the dynamics and musicality in spades.

Now I’m back to DSD. Why?

I finally got my asymmetric room layout to sound better via asymmetric toe in of speakers (extreme toe-in of boundary wall, and toe-in to my ear on the open side speaker). I’ve posted on a separate method regarding this.

Anyway I tweaked The speaker angles till I got it nearly optimal via PCM and checked DSD again and now I’m back on the DSD bandwagon :slight_smile: