Add pre-amp for Direct Decoding mode
Status: Inactive
Brought to you by:
manisiutkin
Menu
▾
▴
#13 Add pre-amp for Direct Decoding mode
Some tests have revealed that the Direct Mode produces slightly lower volume levels (between 1.2 and 2db on the tracks tested). It also produces a slightly worse dynamic range (0.5 - 0.7db). I've only tested it with the parameters given in the attached file, though. Unfortunaltey, this would mean all values which I've submitted with the metabase files are relative and not absolute.
For now I recommend those that want to use my replaygain values to reproduce my decoder settings:
Asio Driver Mode: PCM
Volume: +0db
PCM Samplerate: 352800
DSD2PCMmode: Multistage (Floating-Point)
Discussion
Anonymous
Seems that tests on 1000 Hz sine wave needed to see about output level
At high samplerates (176400, 352800) there will be significant difference in levels due to high frequency noise if multistage converter is used. Better to do level estimations with direct converter because it removes everything above 30kHz. I tested converter on 1 kHz sine wave. Levels match except 352800 mulistage (when it's -1dB), but that was made on purpose and will be removed in 0.4.1
Is this an issue of 0.4.0 or are all my values calculated with 352800 on previous versions wrong?
Unfortunately, it's the second. I think, it's better to calculate at the low samplrates (44100 or 88200). Or calculate with the direct converter. With the multistage you'll have difference due to high pitch noise in output.
Dang... what's the reason I'm getting better dynamic rating values with Multistage/352800 only? All other settings with either ms or direct yield slightly lower values. Noise couldn't cause that, could it?
Hard to say... Noise seems to be the only difference. Does direct decoder give different values when samplerate is variating?
Nope, only at MS/352800 is the DR higher.
Noise then. Actually, "multistage" at 352800 consists of the only one stage:) And it's DSD->PCM decimation (factor 8) with the very short (order 79) FIR filler. So, noise has the very high level. Why don't use 44100 or 88200 samplerates where noise is filtered out?
Well, using 44kHz kind of defeats the purpose of using a HiRes format, and anyway, there wasn't any documentation that would have helped me avoid these pitfalls before :)
I've meant doing replaygain at low samplerates. Who new, you were the first who did:)
I've been using the multistage (fixed-point) and a PCM samplerate 44.1khz. Now there is a new method available (Direct). Is this the prefered method now? What is the difference? Also, if I use samplerate 44.1 will that not be a effective filter at 22 khz (nyquist)? Please advise!
"Direct" method is actually 30 kHz lowpass filter, so it's output doesn't contain high frequency DSD inherent noise. But "Multistage" doesn't filter out this noise. DSD noise spectrum lies above 30kHz and if you do some measurement (like replaygain estimation) this noise should be suppressed. Although there's no preferred method I can suggest the simple rules of using decoder setup:
1. Use floating-point because it uses SSE and saves performance.
2. For replaygain (and other estimations) use 44100 or 88200 with multistage converter and any samplerate with direct converter.
3. At lower samplerates direct decoder should have better performance.
About listening to the SACDs you should to decide by yourself whish one is better for you.
So in point (3) lower samplerates would be 44.1khz? What I mean is that by choosing this samplerate you already filter out everything above 22.05 khz (even when choosing multistage). Do you agree?
And finally another question! How do you think the sacddecoder stacks upp against the much more complicated process that I have seen:
ISO->Scarletbook-> DSDIFF->Weiss Saracon->PCM
Is there any difference in quality nowadays (assuming you don't want to do on-the-fly conversion, but rather convert and archive as e.g. FLAC files)?
Yes, sure. Everything above 22kHz will be filtered out (and even direct converter uses SRC extra step to get 44100 from 88200). But this filtering can only slightly lower the average volume.
I didn't compare conversion quality. Some people think that Korg AudioGate is better. Someone is yet to figure out which one is the best for DSD->PCM.