Audirvana Studio trial questions

Couple of questions about Studio.

I have 2 devices that I use frequently, one Headphone via Wasapi and 2 DACs via ASIO.

Is there any way to switch quickly between these devices only?
Because when I click on the speaker icon on the bottom right I can only pick another ASIO or Wasapi output.
I have to go to settings and change output type and pick the correct one.
Wonder if there’s a way to have a favorites list to do a quick switch instead.

About DSD output if I understand correctly there’s no way to do DSD upsampling or DSD resampling to the same rate. Is that correct?

DSD re-modulation (conversion) to a higher DSD sample-rate is done offline with a only a couple of standalone applications, because of the computational overhead needed to do this function relatively quickly.

Resampling DSD is irrelevant because this involves a D/A → A/D conversion to DSDxxx, which is counter intuitive, because you do not create more sonic information than was captured in the original DSD A/D encoding.
:notes: :eye: :headphones: :eye: :notes:

I don’t get it, why should it be done offline with Audirvana?
foobar2000, HQPlayer and I think also JRiver are doing it in real-time.
Sure there’s a computational overhead but that’s irrelevant, same for SoX or r8rain converting PCM to DSD.

There are reasons to prefer the non native DSD output, even if the DAC does not allow full bypass.

HQ Player and AuI ConverteR 48x44 do offline re-modulation of DSD.

I personally down-convert stereo DSDxxx to DXD so to apply HRTF DSP on these recordings…

When resampling, you are at the mercy of the DAC platform D/A, etc… there is nothing to gain in resampling DSD to a higher DSD sample-rate in this context… In the digital domain it is a different story, because you get the benefit of a bit-stream waveform that appears to the D/A output circuitry, more like an analog signal waveform.

:notes: :eye: :headphones: :eye: :notes:

HQPlayer does it real-time, there’s a small or big latency that depends on the settings.
Aul is offline.

Up-sampling and re-sampling on DSD is used to shift noise.
You are not adding anything, you are moving data using an SDM with a different order or filter type.

You should get a decent ADC and check yourself with MTA or RMAA.

This is a 1 kHz Sine analysis on my DAC with PCM:

This is converted to DSD256 via Audirvana SoX 8th order type B:

This is with native DSD256 output:

This is with native DSD256 but converted first with 5th order type D:

Even if I can’t bypass the DAC filters there’s already quite some difference and with a better DAC could be much more.

You are talking about “modulating” PCM to DSDxxx… I am speaking to your question regarding DSDxxx → DSDxxx conversion…

I down-convert DSDxxx to DXD and apply HRTF and then “modulate” (up-sample) this 24/352.8kHz PCM signal to DSD128 (5.6MHz) using r8Brain in Audirvana.

:notes: :eye: :headphones: :eye: :notes:

No I’m talking about DSD to DSD.
The first screenshot is just as reference for pure PCM output.
The Audirvana screenshot is about PCM to DSD cause I couldn’t get it to do re-sampling of DSD.
Which I’m able to with foobar2000 and HQPlayer.
Audirvana and HQPlayer have both a similar higher quality DSD output than native on my DAC.
But with Audirvana is limited to PCM content while on HQPlayer can be used also with all DSD content as long you disable native playback (DirectSDM).

Can you supply documentation from the HQ Player or foobar2000 user guides, describing real-time DSDxxx → DSDxxx conversion?

I don’t think there are user guides.

If you want to test with foobar2000, you need the ASIO+DSD output or the ASIO DSD Transcoder with the DSD Processor.
It can be added as a DSP module in the DSP manager with the latest interim version.

With HQPlayer you just need to set in Outputs the Default mode as SDM (DSD), set in SDM tab the Bit rate limit and be sure the DirectSDM in DSD Sources is not selected.
First time you play a DSD file a loading bar could come up for a few seconds, depends on your system performances.

Okay thanks…

Maybe you can give insight as to why one would re-modulate DSD64 (2.8MHz 1-bit DSD) to DSD128 or DSD256, etc.? The Nyquist frequency of DSD64 is 1.4MHz… even at a Nyquist of 705.6kHz the noise spectrum is well beyond human hearing… The 1-bit 2.8MHz DSD A/D encoding will not be changed by any re-modulation to 1-bit 5.6MHz, 11.2MHz or 22.4MHz DSD… no harmonic and dynamic-range information captured in the original 1-bit 2.8MHz A/D will be changed… Likewise, this is the same for DSD128 and DSD256 recordings… The only thing changing by re-modulating to a higher DSDxxx sample-rate is the resolution of the 1-bit signal and the potential dynamic-range of the file as a result of the noise-shaping algorithm being applied in the re-modulation function…

In the context of sample-rate conversion… Re-modulaton of the majority of DSD recordings to a higher DSDxxx sample-rate seems like “slicing hairs” and completely dependent on the resolution of the playback system being auditioned and the subjective listening skills and hearing acuity of the listener, even if this these things could be tangibly audible.

If we are talking about tangible qualitative differences in native DSD recordings of 2.8MHz vs 5.6MHz vs 11.2MHz, this is a different animal and you will get no argument from me about those tangible differences in contextual harmonic and dynamic elements that are captured by the different A/D conversions of analog source signals.

:notes: :eye: :headphones: :eye: :notes:

Because it’s not PCM, it’s 1-bit DSM.
The noise is not added above the Nyquist rate but overall the whole frequency range.
The noise shaping is moving it away from the audible range, above the 20-20000 Hz range.

This is DSD64 with HQPlayer:

This is DSD256 with HQPlayer:

Closer that noise is to the audible range and lower is going to be the efficiency and higher the distortion of the power amplifier and the speakers. They will have to reproduce it, even if you can’t hear it.
In my case with a Top-ART Yamaha Amp that goes up to 100 kHz and AMT tweeters that can go up to 24 kHz, it could be an issue.
Not really cause I just listen to very low volume but higher is the volume, higher is the impact of that noise.
This is also an ideal situation. Some IIR filters in DACs put that noise at 17k and above which is very well within the audible range for many with good ears and/or young.

Noise is also only partially shifted from the passband, some of it it’s still there.
This is causing a lot of other small audio distortion issues.
Using a higher order and different filter type can reduce these.
But then it’s increasing a lot the probability of overload of the DAC SDM.
I guess that’s why Audirvana and HQPlayer have a -3 dB attenuation.
If you set HQPlayer volume to 0 dB it will overload immediately.

There are very good reasons why DSD 1-bit is criticized and Sony/Philips blamed for not starting with 3/5-bits instead of creating Wide-DSD years later.

This is an interesting reading:

It’s in 3 parts, here’s an excerpt:

A third problem is that SDMs are found to have any number of unexpected error or fault loops in which they can find themselves trapped, which are not yet adequately explained or predicted by any theoretical treatment. These include phenomena known as “limit cycles ”, “birdies ”, “idle tones ” and others.

This is why it’s better to re-modulate and up-sample, especially DSD64.
The resolution of DSD is in the time domain and higher is it, better the filter can do its job, just like with PCM that needs up-sampling first.
About the qualitative differences with higher rate DSD, my best ones are at DSD64.
A higher native resolution is better but the audible range is always the same…
It’s more about how well the DAC can output a specific DSD rate.
In my case DSD256 is the sweet spot, the DAC is cheap not a hi-end one.
DSD512 is already too much and the output is measurably worse.

All these issues together with the wide-spread noise is funnily what makes the “analog feeling” so much enjoyable for many.

There are many reasons why DSD recordings are sometimes better than PCM.
In many cases the master is analog but more often is about the quality of the equipment and the skills of the people who did the mastering.
On a purely technical level a DxD master is just better; I personally don’t think the higher time resolution is doing any difference compared to 352,4/384 kHz and for sure it’s not compensating the drawbacks of the 1-bit modulation.
But even there you can find better DSD versions than the same in DxD.

Tacet’s SACD are exceptional, much better than the Redbook version, while just being DSD64 at 44.1 kHz fs. The recordings are made all analog but then mastered in PCM at 96 kHz.
The conversion to 44.1 kHz for the CD version suffered a lot while to DSD it was done properly, the very high rate definitely helped big time.

Better is the system and your ears and higher is the likelihood you can appreciate these small quality improvements with fine-tuning DSD reproduction.
But even there if you don’t have amazing ears or a very hi-end system, you can make some fine recordings more enjoyable than they are already if you know them very well.

I’m not sure what you are trying to tell me, because I totally understand the Direct Stream Digital, Pulse Density Modulation, Delta-sigma 1-bit encoding scheme and the realities of a 1-bit signal dynamic-range without noise-shaping…

The issue is with noise in the audible-band… and this is a matter of the Nyquist frequency vs band-pass filter design vs real-world human hearing…

An SACD is by nature 1-bit PDM (DSD) encoded at 2.8MHz… If it is derived from a modulated 16/44.14kHz PCM recording to PDM (DSD) 2.8MHz, the Nyquist filter cut-off of the 16/44.14kHz PCM file is the limiting factor in the contextual frequency response of the SACD… The modulation to 1-bit 2.8MHz and it’s inherent noise-shaping increases the perceivable dynamic-range that is masked by the 16/44.14kHz dynamic-range, affectively lowering the noise-floor of that encoding in the audible band. The 16/44.14kHz PCM file theoretically has a dynamic-range of 96dB and a functional frequency response of approximately 22kHz, where dynamic range of PCM is calculated as 6db/per bit (6dB x 16 = 96dB) and the Nyquist frequency is calculated at 1/2 the Fs (44.14kHz ÷ 2 = 22kHz).

The advantage of a PDM signal is that simple high-frequency filtering is applied at output, as the signal more closely resembles an analog waveform when presented to the output circuit… The higher the PDM sample-rate ( number of samples) the more refined the signal waveform looks to the D/A output circuit.

In the case of a DSD recording, it is very simple… The higher the sample-rate, the more contextual harmonic and contextual dynamics are captured and encoded in the 1-bit PDM signal…

Your graphs are describing predictable yet spurrious harmonic artifacts at very, very low levels in the audible band and at ultra-high frequencies of 20kHz and up to 50kHz… Nobody can discern spurious musical elements at -130dB or lower… or even at -100dB as in your spike at 2K.

The real-world is where we make the assessments… Modulating natively recorded DSD64 to DSD128 or DSD256 or DSD512 is superfluous… Recording analog source as 1-bit PDM (DSD128, DSD256) will always capture more contextual harmonic and contextual dynamic-range than a DSD64 recording, depending on the nature of the analog source signal, whether it is a transfer from an analog recording media or of a live capture of a musical performance in the analog domain with microphones, etc,.
:notes: :eye: :headphones: :eye: :notes:

I’m not sure what you are asking me :slight_smile:

Yes, mostly. But also the noise in the non audible band can be a problem; you power amp section and speakers will get it and try to reproduce even if you can’t hear it.

This is true, if you are converting from a Redbook recording.
Not relevant for native 1-bit recordings.
Today most recordings are made with the Sonoma DAW at 1-bit and mastered with HQPlayer Pro.
There’s no modulation or conversion to PCM as far as I can tell.

This is true but it has to go through the SDM first.
And the SDM is a cascade loop of low pass filters, 5 to 10 depending on the order, each one with a signal and noise transfer function.
To me doesn’t look like anything simple as often suggested…
It does have some inherent advantages compared to PCM due to the high sampling rate but as being more resembling the original analog waveform I’m a bit skeptical.
For sure has some advantages about phase and linearity but it comes with the burden of the very high noise.
Overall most of the data in a 1-bit DSD recording is just noise and all the processing effort is to remove it.
Which is a good thing cause your really don’t want to hear all that noise but this means that you never really get the original analog waveform; same as PCM it’s a good approximation.
If it’s better or not than PCM I wouldn’t know.
Seems many are arguing that the increased time resolution is not that big, PCM needs one sample to change amplitude while DSD format is using bit patterns so it needs at least n samples.
Others are arguing the higher time resolution is useless cause we with PCM at DXD rate it’s already below the human hearing resolution.
As said, I wouldn’t know, they are both more than enough for me when done properly :smiley:

That’s a 1k sine test signal, the effect of a better output with real music is something else.
Think about what that means when many different instruments overlapping each other are being reproduced, the sum of their harmonics are not being so subtle.

I will quote NativeDSD description of their DSD re-modulated content, which costs some money because done by professionals with HQPlayer Pro:

These higher bit rate DSD 128 and DSD 256 releases are all pure DSD created. They are not up samplings, for there are no PCM or DXD conversions involved in their production. They are re-modulations of the original DSD 64 encoding modulation that produced the DSD 64 releases. The sonic advantage to these new Stereo and Multichannel DSD 128 and DSD 256 releases, as with all higher DSD bit rate releases, is the wider frequency passband prior to the onset of modulation noise. This results in the listener’s DAC using gentler and more phase linear filters for playback of the music

Don’t forget the Merging Technologies Pyramix systems, which are by far, the most common DSD recording/production platform.

The NativeDSD statement can also be construed to just acknowledge the reality of the modulation to a higher sample-rate, and this does not necessarily translate into tangibly audible elements greater than that, which was encoded in the 2.8MHz bit-stream… All DSD filters are gentle sloping and the Nyquist is far beyond human audibility… If there is problems with ringing in the filter or some other anomaly that does not reduce the ultra-sonics from reaching sensitive components, this is bad design… most DAC designers that support DSD understand these problems and design accordingly to avoid passing damaging artifacts along the signal-path.

The idea of the harmonics of overlapping instruments having the exact same timing and phase is unrealistic, in the perceived musical performance.

This is not true…

The process of decoding a PDM signal into an analog one is simple: one only has to pass the PDM signal through a low-pass filter. This works because the function of a low-pass filter is essentially to average the signal. The average amplitude of pulses is measured by the density of those pulses over time, thus a low-pass filter is the only step required in the decoding process.
Pulse-density modulation - Wikipedia

Sorry I thought I specified it before; this is about the ESS DACs like in my case and all others that doesn’t have a direct path for DSD.
Or they have but it’s not implemented, like some Topping or SMSL with AKM.

1 Like

The element that you should be most concerned with in any DAC platform, is the slew-rate of the output circuit components. :wink: In the case of DSD this becomes the most salient influence on perceived sound quality, beyond FIR filter design.

This topic was automatically closed 90 days after the last reply. New replies are no longer allowed.