No, the digital noise floor is far far below analog noise floor. Analog noise floor is in any case limited by physics (thermal noise). Modern issue is gain mismatch, where you end up having back and forth attenuation / gain with additional analog stages, which causes increased distortion and dynamic range degradation. Cutting some of that out helps improving output fidelity.
Output levels between -10 dB and -20 dB usually give best performance figures from DACs. So if you can get gain matching around -20 dB that is a very good result.
Yes, it is good idea to have safety margin. In best case the maximum would be limited by actual DAC output plus amplifier gain. But there are other ways.
But preferably the āDAC volumeā is not digital volume inside the DACā¦
Good starting point figure is that it would need over 96 dB of attenuation, so the amount of dynamic range 16-bit has to begin with.
Your analog noise floor is limited by thermal noise anyway, so if you use analog volume control it doesnāt get any better. Analog volume control just adds itās own distortion and noise on top.
And the most stupid thing is that āpre-ampsā are āpre-attsā these days. So if you set preamp to -20 dB volume and typical power amp gain is 36 dB, then the net gain is 16 dB. But you just lost 20 dB worth of dynamic range by shuffling the gain back and forth. It would have been better to just have 16 dB of gain to begin with in first place. Then the power amp wouldnāt be also amplifying the noise by unnecessary extra 20 dB.
Low noise DAC can output correct level for the power amp, and the noise floor is well below that of the power amp.
That is still less than what most good DACs have. If you connect the two directly, the DACs noise floor will be anyway below that of the ampās. So your only limiting factor is the amplifier.
Anyway good test is to just listen if you can hear the noise. If I put my ear 1 cm from the tweeter, I cannot hear anything from the speakerā¦
To properly reconstruct 24-bit data, you need to have 144 dB attenuation before beginning of the first image. And without phase errors. Thatās the challenge. With analog filters, you begin to have phase issues way before frequency response issues. With hires data, you likely also want to have about 100 kHz pass-band.
Output needs to settle within Ā± half LSB within small fraction of sample period. This is step response of your conversion system, when it steps from one voltage to another in an instant. More bits you have, more accurately your D/A conversion output needs to settle. When you increase sampling rate, the sample period becomes shorter and shorter, so you have less and less time to settle to that half LSB value. So with PCM you cannot keep blindly increasing sampling rate, because while doing so you lose precision due to settling time. So your output value errors increase. With lower rates you have the analog reconstruction filter issue.
So for a long time PCM converters have been abandoned and instead we increase rate and use less and less bits. Ultimately reaching to DSD where you switch between two values where the allowable half LSB error is the biggest. At the same time we can improve dynamic range and get rid of zero-crossing and low-level signal distortions by using noise-shaped delta-sigma modulators.
However, it is still possible to get pretty high quality output from PCM ladders like Holo Audio by using linear region of the conversion stage combined with purpose designed noise-shaper to correct he D/A conversion linearity errors. You can see this as a sort of hybrid between traditional PCM and SDM (DSD) - use less bits to do more.
Thanks for the valuable insights! It is a shame settling time is not measured in most dac test suites, Iād be interested what kind of maximum PCM rates common dac settling rates would correspond to.
I assume the trick with the Holo dacs will keep working even further. Right now we can use 20 bits at 1.5 Mhz, which settling time wise should be the same as 19 bits at 3 Mhz. Of course if you keep doing that 18 more times you end up with DSD, but still it would be cool to see if thereās an optimum somewhere
It is not in the measurement suites, because it needs to be measured from the conversion ladder output, before the analog post-filter. This means that you need to open the box and hook to the pins on PCB. But usually DAC chip manufacturers specify it, although not in any easily comparable or consistent way. Last audio PCM ladder chip manufactured was TI/BB PCM1704 and it is specified on the datasheet:
IIRC, there was DAC chip from Analog Devices that was actually accurate to 24-bit, but due to the settling time constrains, the maximum sampling rate was around 0.1 Hz or similar. So essentially designed to provide variable accurate DC reference voltages. Not for audioā¦
Output impedance there specifies the theoretical maximum dynamic range (level of thermal noise).
It is just about balance of trade-offs. Which parameter you want to optimise. In order for the hybrid noise shaper trick to work, you need relatively high rate.
Thank you for the detailed response. Iām not sure if I understood the above correctly. Iām not sure how my Gustard X26 Pro does volume control but Iād like to think that it does it in the ESS9038Pro chips whose volume control is supposed to be good. I know that some DACs offer analogue volume control, but surely that would affect the dynamic range even more through analogue attenuation?
I know that Iām probably repeating myself, but @jussi_laako you deserve some form of an audio equivalent of the Nobel Prize for your Gauss filters, for me especially the -xla one. Iāve just yet again done some switching back and forth just to remind myself of why this one is so absolutely brilliant.
poly-sinc-ext3 - sounds āOKā, but quite ādryā and transients can be a little ādigitalā, i.e. harsh.
sinc-Mx - the opposite of ādryā, sounds rather bloated in a way which comes across as more physical but itās almost like the colour saturation setting on your TV turned up too muchā¦ and transients are still rather unnatural.
polys-sinc-gauss-xla - absolutely purrrrfect. Neither dry nor bloated. Transients are pleasant, I can turn up the volume as my brain is not forced to waste energy on blocking out unpleasant edges of sounds. The term āsoundstageā kind of ceases to exist as things simply āareā and music feels like a performance (can be a beautifully intimate performance with the -long filter), not a reproduction of individual sounds coming at you from seemingly unconnected points within the available space.
The Gauss filters are the only ones that sound like a cohesive whole, not a sum of individual parts. Like cereal with milk (or your choice of available substitutes), not just dry cereal in a bowl And cereal without milk is just not much fun plus the experience is likely to end much sooner - even if it tastes good, your mouth will get dry quickly and the enjoyment will be short-lived (my attempt at a transient analogy - sharpness and incisiveness can catch my attention but it hardly ever allows me to actually enjoy my listening sessions).
My silly metaphor aside - your creations have surely made a huge contribution to the world of digital audio, they have definitely transformed my experience.
If the attenuation is not big, that is one way. ESS chip DSP has pretty limited resolution, but smaller attenuations donāt yet generate much degradation.
Optimally there would be some analog gain pad, such as adjusting gain of analog stages or similar. If DAC offers switched resistor ladder attenuator, then that may be optimal way. But these always depend on the case. There are so many ways to implement such.
ADI-2 Pro has analog reference level setting. On T+A HA 200 and Ferrum Oor I have set analog volume control to maximum possible listening level and then turn down from there using HQPlayer.
Another cool thing about using HQP for volume control is utilising the Fletcher Munson loudness thingy.
But I wish there was a way to setup a Microsoft Surface dial to control HPlayer volume so that I have a physical volume knob and donāt need iPad open
Like rooextend does with Roon
This would really change the HQP listening experience with more people using it for volume control - especially for non Roon users
HQPlayer got a review from I must say one of the best headphone-related audio reviewers on YouTube. I donāt think the review did the software justice, if you scroll down youāll see a comment I posted under the video to express my disappointment (I have usually really enjoyed the content). I wish that the author had read this thread before posting the review.
Anyway, if according to him a $500 DAC fed closed-form filter in PCM by HQplayer already matches what the Chord TT2 + MScaler offer - gee, thank goodness I didnāt fall for the Chord stuff before I discovered the HQPlayer! That would mean that my set-up must be light years ahead
I have the moon430 had, it has a M-eVOL2 volume control. Wich is explained like this:
A pair of R2R DACs are used in each audio channel which separately alter the incoming audio signalās amplitude via a series of resistors. Despite the inclusion of these āDACsā to control the volume, the audio signal remains in the analogue domain at all times and thereby doesnāt degrade.
Should i set it to max and then adjust volume in hq player?
That is typical switched resistor ladder volume control. For safety reasons, you could tune the combination. First turn it down and leave HQPlayer volume at -3 dBFS. Then turn the analog volume control up as much as you would ever want to listen. And then dial down from that using HQPlayer volume to comfortable listening volume. If you ever find lacking volume while reaching -3 dBFS on HQPlayer, you can easily turn up the analog volume and then leave it there.
That way you are safe from volume accidents on computers while retaining best possible SNR.
I agree, he really should have spent some time on this thread. Would have helped his review significantly. Lots of little errors sprinkled throughout.
But I do think this review is good in one significant way: itās the reaction of a committed, experienced audiophile to trying HQPlayer. I think this review does a good job of showing the friction of using this great tool for newcomers. Heās basically plugging it in and going for it. Gets some benefits, but far from optimal. And clearly doesnāt understand everything heās doing.
At any rate, nice to see a mainstream HiFi reviewer cover this great piece of software. Hopefully this leads to more people trying it and getting the benefits.
Many times it would help if people would at least look at the bundled manualā¦
For comparison, for someone coming from Windows or Linux, starting to use macOS is really confusing. And Apple provides practically zero documentation for their things. For example people coming from other OS would think that clicking the red X button in the corner of main window would close the application. Which is of course not the case on macOS.
Why would closing a window necessarily close the app? Thinking that it would is what is confusing. Now, with an app with a single window that is always open that makes sense.
Remember, macOS is document-based while Windows is app-based. So the paradigm of the windowing system is different. This the different behavior.
Yeah, I have no issue with this, but I struggled with this already in the early 90ās when I was new to Macā¦ Now writing this from my MacBook Pro.
Point was that at least Iām trying to provide a manual. While Microsoft, Apple or Adobe donāt even bother. I think last OS I had that had proper manual was IBMās OS/2.
Iāve had my challenges when I try to offer Macs to people who have never used a Mac before and are used to only Windowsā¦
For me, Mac feels more home, since anyway first thing I usually do is to open terminal to access command prompt. And Mac is just a regular Unix-style system like Linux. While Windows is thing of itās own.
(I donāt use any IDEās for software development)
@jussi_laako i do Get some quick popping/tapping sounds (only in the right side of the headphones, happens no matter what headphone I use) right before song starts. Happens when switching to different sample rates and also when changing to the same sample rate. Sometimes it doesnāt happen. What could I do? This is my settings
I stopped watching his videos after he tried to compare an expensive usb cable vs a standard one and basically confirmed his confirmation bias while being unable to tell which cable was which and which one sounded better.
There are some errors, like of course HQPlayer works with any source, although not the best option if you want to use it for video though due to FIFO buffer delay. And you know if you should be using apodizing filter and not for example halfband from the Apod counter figure.
The interesting part are the listening impressions, no surprises there. Filter preferences are personal, and using Chord stuff as reference to compare against (if one is looking for closest sound) it is not surprising what what he ends up with. Although I personally think sinc-Ll is closest to M-scaler from objective perspective. But he didnāt go much off. Technically best? No. But probably sonically close to what he compares against.
And not surprising about source material dependency either. Different genres have different requirement emphasis.
And cereal without milk is just not much fun plus the experience is likely to end much sooner - even if it tastes good, your mouth will get dry quickly and the enjoyment will be short-lived (my attempt at a transient analogy - sharpness and incisiveness can catch my attention but it hardly ever allows me to actually enjoy my listening sessions).
- your creations have surely made a huge contribution to the world of digital audio, they have definitely transformed my experience.
