I’m using Topping DX3 Pro + as my PC DAC. I also use a Fosi DS1 when listening on my iPhone, which has the same DAC chip as the Topping, and either use Qobuz or FLAC Box on the phone for playing FLAC files that I ripped from CD.
I agree making presets per album is OTT - I can be somewhat OCD - but thus far the EQ I made for the headphones sounds good.
I haven’t tried to replicate it on FLAC Box as its EQ isn’t so configurable. But FLAC Box seems to warm the sound on its own, and I only use the phone for casual listening in any case.
But my main focus is on Audirvāna on the PC, when I’m immersing myself in the music.
That I’m enjoying it and not seeking other apps/hardware tells me, subjectively speaking, I’m on the right path with it.
I’ll fiddle with up sampling soon. On the PC I’m struggling to enter specific values for that - the sliders work but it won’t accept keyboard input. I’ve raised a case with AV about it as it feels to me like a bug. Double clicking the fields, right clicking, etc etc select the value but don’t enable keyboard input…
For reference… I don’t experience this behavior on macOS…
Note:
The DX3 Pro+ uses an ESS chipset… ESS chipsets convert all input signals including DSD to PCM for output… By up-sampling all PCM to high sample-rate DXD 352.8kHz (multiples of 44.1kHz) or 384kHz (multiples of 48kHz) or higher (705.6 or 768kHz), you will remove processing load from the DAC platform where it is best done in Audirvāna via r8Brain or SoX so to reduce noise in the processing topology of the DAC… (I suggest r8Brain)… For DSD playback, setup Audirvāna to ‘convert DSD to PCM’. … The higher sample-rate in both cases presents a more refined signal to the D/A converter output circuitry.
This will most likely change your EQ strategy…
Regarding ESS DAC chips, here’s a post on another forum from Jussi Laako, developer of HQPlayer. He’s been working with digital filters, modulators, and DSP in general for a couple of decades or more, and worked for Intel before leaving to develop HQPlayer full time.
When you send DSD to ESS, it will still go through the ESS’ DSP. But it will bypass bunch of processing compared to PCM, since there are no rate conversions being applied to the DSD data (your ES9038PRO and earlier models). It also “fixes” some of the ESS related issues such as IMD hump.
The overall performance varies depending on implementation of the DAC’s analog stages. So not all ESS based DACs perform the same.
I’d recommend to initially try upsampling to DSD256 with ASDM5EC-light modulator. And possibly with ASDM7EC-light too. And compare how it works vs 705.6/768k. Using same filters etc. You can then switch output rate to DSD512 to check if it improves things.
Note! On DACs based on very latest ES9039 chip, the DAC should be run always at DSD256! Since they’ve changed how DSD is handled and now feeding the DAC with DSD512 seems to result conversion down to x256 rate in the DAC chip!
Edit: The 5th order modulator he recommends to try initially in HQPlayer would correspond to the B5 or A5 modulator in Audirvana.
Why can’t we just keep this in the context of the modulators supported by Audirvāna (r8Brain and SoX) ? Or at least refocus the concept of up-sampling/modulation as it applies to Audirvāna users and those settings? I know you are enamored with those proposed modulator settings… however, you have no idea how these settings will sound on a given ESS DAC platform output circuit architecture/topology… It’s one option… You know I agree with modulating all PCM to DSD… And it is most certain the ESS chipsets decimate DSD512 and all 1-bit signals to high sample-rate PCM before passing through the Hyperstream processing to a multi-bit D/A circuit topology.
I appreciate all the tips given and admire the passion that accompanies the advice.
Currently, I use AV’s EQ for correcting my headphones, BlueCat Liny EQ for adding some bass (I found it better, or easier, to get the sound I liked using it in addition to AV’s EQ than adjusting AV’s EQ.)
I use BlueCat’s free plugin to compensate for the fact that my left ear is more sensitive than the right, by about 1db.
I am now comparing BlueCat’s Re-Head and 112DB’s Redline Monitor.
Re-Head’s EQ correction for the HD560s sounds identical to AV’s EQ - as both use AutoEq’s settings I guess that’s not surprising.
But I’m not sure about the headphone adjustments between Re-Head and Redline Monitor. Re-Head seems more flexible, but I’m not sure if I’d use that flexibility.
SOTA in digital room correction and device correction is convolution… The better convolution engines are about as accurate as you can get… with FIR filters its like having (32,768) a graphic slider at every 0.7Hz from 20Hz to 22kHz, so more than 100 sliders below 100Hz where it really counts…
Following this digital room or device correction it is certainly fine to add to your small signal pipeline any number of additional surgical or colorful mastering quality equalizers to suit your taste…
Heck I even run it thru my hardware harmonic generator AKA a tube preamp…
The accuracy of the theoretical math is one thing… the ‘real-time’ computational resources available in the computer platform like accumulator/register capacity is another thing in regard to maintaining accuracy… not all DSP implementation is created equally… fewer interpolations is better.
Not theory… My theses is that my old macbookpro from early 2015 with an i5 processor and 8 GB ram runs the HangLoose Convolver with 2 FIR filters, L&R, along with one of several stereo mastering EQs, stream from qobuz and surf the web without any glitches… And my fiancé often surfs the web or plays an online interactive game and still no glitches…
I am sure most around here have faster computer kits than this… Stop trying to scare everyone from trying things @goldenear… And on most vector and/or parallel processor systems convolution is not and issue… Even this old i5 is a parallel vector processor…
It’s not about the processor… it’s about the calculation(s) accumulator/register allocations … doesn’t matter how fast the processor is or how many processors are be applied… Maybe inconsequential., maybe not…
The only thing scary is the recommendation(s) to pile a bunch of non-vertically integrated DSP on any given file playback in Audirvāna… It is one thing to utilize a bunch of processing on tracks or stereo sub-mixes to achieve a sound, but it is another thing to influence the playback of a pre-recorded digital-audio file… I don’t argue with DSP based playback acoustical space management and I certainly don’t argue with individual prerogative in regard to one’s rationale to color the playback signal to suit a personal sonic aesthetic… So take it easy baby…
I guess you missed the concept of vector math and vector processing and what that has to do with convolution maths… Time for you to go back to computer science skool… The other area of concern is feeds and speeds… This is a pretty lazy 2.7 GHz Dual-Core Intel processor… Not the fastest processor on the block… Again my theses is most, if not all, modern processors should be able to handle convolution non-vertical or not…
Time to retrain your AI and see if you can’t get it to stop hallucinating…
I don’t doubt that they can do the processing… How accurate the results are is not just a result of the calculation(s)… Clock interrupts from error-correction request wait-times, and buffer (accumulator/register) under-runs or over-flow and wait-times across the memory bus, signal flow gremlins in association, like BLER… all have influence on the result(s)… So, this is why I say fewer interpolations is better.
“Accuracy” is a relative term… it’s a spectrum of interpretation… sort of like the measuring Quantum Entanglement… From my perspective, nothing is perfect… . what you believe, is what you believe… I’ll take fewer interpolations in my playback of digital-audio files…
Is your AI suffering from computer paranoia???
Many years ago one might worry about these issues with interrupts, ECC, buffer under-runs/over-flows, wait times across the various I/O busses and signal flow issues… It is only as fast as its slowest component… Todays consumer, purpose built, personal computers not so much… Maybe if you were building your own computer from the component level, you would at least need to understand the various components and ensure they all play together via compatible specs… So, this is why I say that things like convolution, when done correctly, is pretty much a no brainer on todays consumer computer platforms…
That pro EQ that can do anything you want, like for me just adding a little low end bass for my liking, compare to a new EQ into a software that is just beginning to be implemented…
I work as a C++ programmer in the semiconductor industry, writing control systems that monitor and control machines and automation robots, and so performance is rather critical. Thankfully, as is mentioned by @Ddude003, modern CPUs have the ability to perform multiple operations over data simultaneously. Of course, the program has to be written so as to utilise this, but modern C++ tools generate very good instructions.
That said, I don’t know what Audirvana is written in, but given that it doesn’t install .NET (on Windows, at least) and has many C/C++ DLLs in its program folder, my money would be on it being a C++ program.
In any case, I did an experiment using the two setups I mentioned above: one using AV’s new EQ + 112DB’s Redline monitor + BlueCat Gain + BlueCat Liny EQ. The other replaced AV’s EQ and Redline Monitor with BlueCat Re-Head.
Both used the same CPU resources - 5% when playing music but 30% when doing upsampling as well.
I wasn’t surprised by the results - upsampling is a complex algorithm, much like enlarging a photograph. It’s here significant processing occurs.
EQ adjustments in comparison are relatively simple, and the CPU usage bears that out. Turning off all the Processing and just playing the music ‘as is’ used about 4% CPU time. Therefore, the EQ (etc) processing has a tiny impact on the CPU, on my Ryzen 5 PC at least.
To be honest, both setups sound great, with slight differences between Re-head and Redline Monitor. I’ll spend more time listening to favourite pieces with these setups to see which I prefer…
