Discovered that some devices may not do a good job with the 32-bit depth DAC Input in Kernel mode. I am using a SMSL PO100 as a digital to digital converter and noticed that notes sounded plucked with accentuated sibilance in treble like T and S sounds. Switched to 24 bit depth and this cleared up. May just be the PO100 but maybe worth experimenting to see if it helps you as well. Anyone else experience this?
Are you sure it isn’t your DAC being overdriven by the dynamic-range of a 32 bit signal? Managing your gain structure is important …
Theoretical maximum of LPCM encoding:
6 (dB) x (number of bits) + 1.75dBv = Dynamic-Range of LPCM…
6 x 24 + 1.75 = 145.75 dBv
6 x 32 + 1.75 = 193.75 dbv
Not realistic for even the best DACs… So managing gain structure is critical in eliminating digital overload clipping of any given DAC architecture.
Also if your DAC is not capable of 32 bit reproduction, it may be truncating the 32 bit signal to 24 bit and introducing some distortion due to round-off error…
1 Like
-
The PO100 uses a 32-bit chip, so it’s not failing to handle a 32-bit signal.
-
Since recordings aren’t made at 32 bit resolution, you’re talking about Audirvāna taking a 16 or 24 bit signal and sending 32 bits. It does this by “zero padding,” that is, it simply adds zeros to the lowest bit registers. This does absolutely nothing to change the signal, any more than writing 1 as 1.00000000 changes the quantity.
-
The thermal noise of the electronics kicks in at around 20-22 bits (-120 to -132dB) in really good electronics, so even if there were signal components that far down (which there aren’t, see previous point about zero padding), you wouldn’t hear them.
So either there’s no actual difference or something else is going on besides Audirvāna’s zero padding.
Yes… however, it is passing the 32 bit output signal along to the DAC, where the distortion is occurring related to signal clipping… We don’t listen to the output of the DDC.
Just an FYI… What you are speaking-to in the your statement below, is the real-world slew-rate dynamic capabilities of the physical components like operational amplifiers (op-amps), capacitors, resistors, etc, utilized in the output circuitry and amalgamated platform topologies…
Noise Power of -174 dBm/Hz is the reference for any noise power calculation when designing RF systems working at room temperature.
I think you are probably correct and the truncation was causing the problem. I use hang loose convolver and dial the gain down so it doesn’t trip the clipping indicators. If you have any tips on filter construction to avoid having to bring down the gain - even with no boost on eq I still have to drop db.
The PO100 is passing the signal on to a benchmark DAC3-l which is rated at 24/192 max input. Maybe this is where the issue was occuring. I am going to try going direct as well - kept the PO100 in to help with Jitter but Benchmark has its own method as well. PO100 was reviewed as very quite so I didn’t think it would impact the chain.
The PO100 is just passing on the signal it receives quite adroitly … however you must be very careful with your power/ground/earthing scheme so not to introduce noise into the components and transmission signal… These should be sharing a common grounding point on a common power-circuit (Best practice is to have all your components on a common power/ground/earth circuit, with no other electro-mechanical devices on the circuit.)
You can lower the gain before upsampling in the ‘Upsampling’ module settings… This works well with your plug-ins so to get just the right amount of gain for your particular DAC… Also depending on which SDM filter you are using, these filter settings can have a dramatic impact on the output sound-quality.
Adding any plug-in alone, increases the gain by approximately 1.7dBv before any DSP is performed…
1 Like
That is some sound advice. I use emotive CMX2s to help with line noise. System is dead quite volume up 3/4 no music playing - no speaker hiss whatsoever. Good to know about the 1.7db jump. I upscale but I found that the benchmark has no trouble with taking the native resolution and it felt like a slight veil when upscaling, no proof of this just what I think I hear. Definitely an improvement in transparency with Kernel mode, this and using the convolver with REW and rephase has taken my system next level - I have never heard a stereo system with this much detail and imaging accuracy. It is awesome and well surpasses anything I was able to achieve with roon.
1 Like
Just remember this… If you upsample a 16/44.14kHz PCM file to 24bit or 32bit dynamic range, you can dump some of this gain (1-bit = 6dB)… a 16/44.14kHz LPCM encoding has a theoretical 96dB of dynamic-range… What you gain by upsampling to a higher sample-rate, is a higher Nyquist frequency and a more refined signal waveform being presented to the output circuitry of the DAC… This can reveal audible non-linearities in the playback system configuration. and room response characteristics
2 Likes
Thank you. What are your thoughts on settings in SOX or R8rain? Bandwidth, stopband, phase. I want as close to source as possible - nothing added or lost - not interested in coloring the sound.
If I have 24/192 native music which matches the max input for my dac does upscaling matter? Filters still applied or would it make sense to downscale to apply the filters?
Also, do you know if vst3 plugins operate before or after upsampling? In the past I believe control panel showed the upscaled bit rate under that step but not sure that is actually what is happening but maybe I am mistaken (not at my setup to confirm).
Thank you!
1 Like
The HLC uses R8Brain for resampling… A picture may be worth a thousand words, so see this chart of resampler settings at Hang Loose Convolver (HLC) - Accurate Sound may be of help…
Plugins come before SOX or R8Brain…
BTW… I use the HLC plugin for room correction, another EQ plugin to add some warmth and sparkle,… And do not upsample… YMMV…
2 Likes
If your source file is 24/192kHz… r8Brain does nothing to the file if you have set the maximum sample-rate to 192kHz or using ‘Power of Two’ and the ‘Bridge’ max sample-rate limit’ to 192kHz…
Just remember the Nyquist frequency (Fs) of the original encoding will always be the upper frequency limiter and upsampling to a higher sample-rate will not change the bandwidth of the file…
However, in the case of ‘Power of Two’ upsampling of 16 bit/44.1kHz files, you will realize more dynamic-range going to 24 bit, and a more refined digital-audio waveform with a higher sample-rate of 176.4kHz [176.4kHz being the proper ‘Power-of-Two’ upsampling target/result, when the DAC is limited to 192kHz.] the Nyquist Fs of the resulting file gets pushed out to 88.2kHz (The Nyquist Fs of 176.4kHz).
The Audirvana team has provided good insights into the behavior of the filter settings of r8Brain and SoX, in the online ‘User Guide’ found at the top right corner of the Upsampling settings pop-up window… I find the Stop-band setting of r8Brain to be extremely useful in fine-tuning the output for my DAC… just make very small adjustments and don’t go higher than the default setting. r8Brain does a great job and is simple to tweak. I prefer Linear Phase and modulate all PCM files to DSD128.
1 Like
If I use HLC to upscale and load only a 24/192 filter to a slot and play a 44.1 (or any other sampling under 192) Audirvana shows the original sample rate to my dac and I confirmed the same on the DAC. Something isnt working right or I am doing something wrong?
I don’t know how HLC handles upsampling in the signal flow through the plug-in… I personally would just use the Audirvana r8Brain SDM to do the upsampling after the HLC DSP… @Ddude003 may have more insight on this…
Post a screen-shot of your main Settings window showing the signal path through the ‘Buffered’ modules to the output to your DDC or DAC during playback. Show us your DAC input settings.
1 Like
I personally have zero experience using HLC…
The only thing that I can see in these screenshots is that you have two filters inserted and may have the upsampling function of HLC bypassed and that you are upsampling to 32bit… Also, you don’t have the ‘Maximum sample rate’ for the PO100 DDC ‘Bridge’ set to 192kHz.
I suggest contacting Mitch at Accurate Sound regarding this behavior (he is on this forum) and wait for @Ddude003 to provide some insights… From my perspective it makes most sense to use r8Brain in Audirvana to upsample the results of your HLC DSP.
2 Likes
Very sorry, I could have been more clear with my comments above…
The HLC automatically up or down scales the FIR filters to match the input signal using R8brain… HLC does not up or down scale the input signal, although that might be an interesting feature request…
The reason I referenced the HLC page was for you to see the chart that shows the range of the R8brain resampler and the information about linear or minimum phase settings. Notice that the orange trace is what I would call incisive while the others are more warmish… This works the same as R8brain imbedded in Audirvana… Which @ko99021 was asking about…
Of the two filter banks only filter bank 4 is selected… The reason for the multiple filter banks is to be able to switch between the various filters, in more or less real time, so you can listen with little to no switching time, so your echoic memory remains intact as you decide which filter you prefer… Or keep multiple filters handy for your various speakers and or rooms…
Also, there is nothing stopping one from using the HLC plugin for Room Correction along with other plugins… You would most likely use HLC in AU slot 1 and others following… And then using Audirvana’s upscalers, SOX or R8brain, following in the signal chain… I don’t use Audirvana’s upscalers because I prefer the sound of my Chord DAC doing that upscaling internally…
2 Likes
Thanks for the great feedback and input, really appreciate it.
I emailed Mitch to get further clarity.
I think I misunderstood regarding HLC and upscaling. Sounds like the filter is scaled to the input rather than the other way around and the %, db, phase is related to the filter scaling (assumption) or in the case of a zip file applied in a slot it selects the matching filter to the input. I verified the filter is being applied when there is a mismatch between the input and filter by using one of the filters with high frequency roll off that Mitch provided to audibly confirm.
Ultimately means using Audirvana for upscaling duty. I will revisit this but in the past I didn’t hear perceptible improvement and leaned towards degradation vs letting the benchmark apply its own internal upscale and noise reduction. However worth another try.
Thanks again.
1 Like
The ESS DAC chipset employs internal SDM processing on all PCM signals and it will be a subjective assessment when juxtaposed to the Audirvana/r8Brain output, in regard to appreciable, contextual sound-quality elements, when upsampling 16/44.1kHz files to 24/176.4kHz… The primary issue being the Audirvana/r8Brain output is subsequently passed through the ESS DAC chipset SDM processor before the multi-bit D/A conversion.
1 Like
This topic was automatically closed 90 days after the last reply. New replies are no longer allowed.