That definitely happens.
Like this example of a NOS DAC, the well regarded Holo May measured by JA (bless him), with IMD from high frequencies mirrored around 22.05kHz and folding back into the midrange:
HoloAudio May, NOS mode, wideband spectrum of white noise at –4dBFS (left channel red, right magenta) and 19.1kHz tone at 0dBFS (left blue, right cyan) into 100k ohms with data sampled at 44.1kHz (20dB/vertical div.).
https://www.stereophile.com/content/holoaudio-may-level-3-da-processor-measurements
This is an extreme torture test with a full dB 19.1kHz tone. The 44.1kHz sample rate creates a ~6kHz IMD clearly visible above. But its >100dB down. I doubt anyone could hear that in normal playback.
To make things even less of a concern for NOS lovers, most files I have are -50dB down over 15kHz. I would speculate that even if a DIY NOS DAC measures much worse then a Holo, with say 30dB worse IMD so what? With most tracks the IMD's from aliasing IMD products will still be -120dB and lost in the noise floor.
I cant see a happy NOS user being worried by IMD folding down into the audible band when its inaudible.
I still think OS with a great filter has the potential to markedly improve listening enjoyment. Anyone?
What if you use a serie notch filter in your tweeter filter at 21 khz ? All the break-ups harmonics and the ringing will not go back in the ears frequency range. There are acoustical notch for tweeter more complex than the usual lense for metal tweeters but they are difficult to acheive (Boston Acoustic Lynnfield serie)
it also affects the performance of amplifiers and preamplifiers, mostly negative
jfet op.amps are better with HF noise then bipolar in I/V stage ,that helps too especially when it's NOS dac
You may be missing my main point, which is that the majority of image-suppression for 'filter-less' DACs comes from band-limiting by the human ear. There is additional acoustic suppression provided by the a loudspeaker's tweeter, but it's not as much as provided by the ear. Even together, they don't anywhere near approach the -100dB@22kHz (or better) image attenuation target for the CD format, and afforded by brick-wall digital filters. That's all I'm saying.
Last edited:
Okay, I believe that I understand now.I want to say that ultrasonic content and noise affecting whole audio spectrum.
I tried (among other) filter solutions, Your idea with C-(double L)-C for the example and differences was significant...
It was in some other topic but i couldnt find the link.
Would you mind describing the difference in sound character you are hearing with the ultrasonic filter in place?
Nice post. JA's chart captures the effect which I believe Zoran is alluding to. My take, on what we see is that the 6kHz IMD spur isn't fundamentally due to the sampling process or to the DAC image frequencies being present. It's an artifact of the IMD generated within the DAC's own analog stages. That chart showing a 19.1kHz test tone inter-modulating with it's 25kHz image tone within the DAC's analog stages.
(44.1kHz - 19.1kHz) = 25kHz image
(25kHz - 19.1kHz) = 5.9kHz IMD spur
If analog stages were so linear as to not exhibit IMD, there would be no 5.9kHz spur, despite the 25kHz image frequency being present. At least, that's what I read Atkinson's chart as appearing to show.
Last edited:
I will try with emphesize the mid and even deeper to the "core" of tone if i may said.
(must say that never disturbing the phase in end of top BW, confirmed with measurements and sims...)
- Even if the measurements are flat somehow specific feel os that the "line" is going slight to top.
That is corresponding to visual domain when You look at the FFT "density", from LF to HF, without the signal the peaks are in the same range BUT somehow we recept it like it is wider in HF region
- finishing of musical events are better and more clear.
- reverberations are more natural and connected with the beguining of event LF spectrum.
- more space between the events.
- somehow more fundament in all tones and global outcome.
- slight less of hissing constant "preasuure", which is already low
I will post Sin(x)/x function. Because more people hear and described what i spot. Probably is something about the imaging.
Next I will try multi RLC notch filter @nfs points up to say n=1..5.
Something as Zanden style passive I have calculated and preapred pasive elements from local store FB like inductances (not recc. because of the Hi Q...) but just to "feel" the diff. and measure.
That will not change the phase more than few degs in 20KHz region, but will optionaly att. to -24db nFs point frequencies.
I will report
Next I will try multi RLC notch filter @nfs points up to say n=1..5.
Something as Zanden style passive I have calculated and preapred pasive elements from local store FB like inductances (not recc. because of the Hi Q...) but just to "feel" the diff. and measure.
That will not change the phase more than few degs in 20KHz region, but will optionaly att. to -24db nFs point frequencies.
I will report
Here is a great paper on NOS DACs, that explains why they measure badly.
It seems our own hearing system acts as a filter. So what is measured at the DACs output, is not what we perceive once we hear the music through our ears.
"Because our hearing naturally functions as a strong filter, our brains tend to interpret the signal from the NOSDAC as if it has passed through a FIR-filter. This is due to the limited bandwidth of our hearing."
Anyway... This is pure psychoacoustics... mumbo jumbo... Voodoo science. So the measurements and all the writings on the subject in favor of why (or against) this DACs sound so good is pointless.
It seems our own hearing system acts as a filter. So what is measured at the DACs output, is not what we perceive once we hear the music through our ears.
"Because our hearing naturally functions as a strong filter, our brains tend to interpret the signal from the NOSDAC as if it has passed through a FIR-filter. This is due to the limited bandwidth of our hearing."
Anyway... This is pure psychoacoustics... mumbo jumbo... Voodoo science. So the measurements and all the writings on the subject in favor of why (or against) this DACs sound so good is pointless.
Last edited:
This actually speaks to a wider point in home music reproduction. Which is that, IMHO, all that matters is the psychoacoustic result produced by a given component when it’s comes to music reproduction. Be that a DAC, or an amplifier, or what have you. While music reproduction isn’t Voodoo, neither is it (as yet) a completely objectively defined human experience. I’m hesitant to reopen this can of worms, but I think it fair to say that many of us have observed that excellent specifications often don’t lead to commensurately excellent subjective sound. In other words, the specifications sheet isn’t the reliable proxy for sound quality that we all wish that it was.…Anyway... This is pure psychoacoustics... mumbo jumbo... Voodoo science. So the measurements and all the writings on the subject in favor of why (or against) this DACs sound so good is pointless.
Last edited:
Digital filter is math. It is calculating missing points (like the average from existing above/below values). The calculation works well for a clean sine wave but it may not hit the value in a complex music pattern (in music is maybe better if something is missing and faded out than artificially incorrectly added )
Passive mechanical filters may work better/more natural (ear, speaker) + passive filters from capacitor, coil, resistor
)Passive mechanical filters may work better/more natural (ear, speaker) + passive filters from capacitor, coil, resistor
I find, that the trick to gaining an intuitive understanding of what a digital interpolation, oversampling filter does to create new samples is to look at how it effects the frequency-domain (like spectrum analyzer view), rather than the time-domain (like an oscilloscope view). When viewed in the frequency-domain, it’s much easier to recognize that a digital filter doesn’t guess, or estimate the new, in-between sample values needed for oversampling. It merely filters away the signal’s image-bands. That’s it. What image-band filtering looks like in the time-domain only falsely appears as though it must be estimating the new, in-between sample values. In reality, the new, in-between samples have EXACTLY (within the digital filter’s computational limits) the values necessary to make the images go away (be suppressed) by the process of low-pass filtering.Digital filter is math. It is calculating missing points (like the average from existing above/below values). The calculation works well for a clean sine wave but it may not hit the value in a complex music pattern (in music is maybe better if something is missing and faded out than artificially incorrectly added
Passive mechanical filters may work better/more natural (ear, speaker) + passive filters from capacitor, coil, resistor
Last edited:
Knowledge represents the synthesis of reasons (concepts) and sensory experience. Not just concepts, not just perception, but their synthesis.
Scientific knowledge has its limits, and those limits are determined by the forms of human sensibility and reason.
Therefore, there is no compelling reason to claim that the limits of our scientific or theoretical knowledge are identical to the limits of reality.
When someone reads these words on a post, for example, photons from the letterforms on the monitor travel and hit the retina, and their energy activates an electrical signal in the light-detecting cells in the eye.
This electrical signal spreads as a wave along the long strands of the so-called axons, which are part of the connections between neurons. When a single formed electrical signal reaches the end of an axon, it triggers the release of chemical neurotransmitters at the synapse, the chemical junction between the axon tip and the target neuron.
The target neuron then responds with its own electrical signal, which, in turn, spreads to other neurons. Within a few hundred milliseconds, the signal has spread to billions of neurons in dozens of interconnected areas of the brain, which receive these words in it.
Inside the neurons in the brain are also small cylindrical structures called microtubules. Gravity creates curvature of space-time. Quantum oscillations spread along microtubules in all parts of the brain. Every 25 milliseconds there is a reduction by which the brain receives information encoded in the geometry of space-time. Gravitational quantum collapses create immeasurable quantum computations that process information through massively parallel computations in the brain. The product of this is consciousness.
This is called Orchestrated objective reduction (Orch-OR)
https://en.wikipedia.org/wiki/Orchestrated_objective_reduction
@miro1360
Finally got time enough to complete AD1862 DAC, chips courtesy @Paddy Garcia (who is a genuinely honest and patient guy, informed me that the chips he dispatched to my address were delivered back to him. He posted them again and thankfully received them)
PCBs were made from JLCPCB with a purple solder mask. For PSU I used the 7805/7905/7812/7912 version, I have had enough with LT1963/3015 version.
the IV stage is standard OPA627/AD797, the final version will use OPA627AM or BM version (found the nasty buggers rotting in a small box in my lab).
USB interface was generic XMOS (chinese source) and the input was from RPi 4B (4GB version) having operating system of LibreElec, and the audio files were both FLAC and MP3s.
Amplifier's were Yamaha A1000 and Lo.D HMA-6590 (both have been completely recapped by me in the past)
First impression of the AD1862 sound was crisp, sharp and pleasant to my ears. Although I have enough test equipment to measure noise characters but I don't indulge with the esoterics.
After listening for a few hours I changed the DAC to AD1865N with the same USB/Input/AMP setup, what really surprised me was that even AD1865 performed much better than the previous experience. So I went on to test the RPi based input with other DACs like AKM4490, ES9038Q2M and ES9038Pro, TDA1541 and PCM63. Every DAC sounded better than before. I will share the picture tomorrow (hopefully).
I Understand @Vunce uses RPi as input source for his DAC. Next stage I tried to use the DAC directly interfaced with RPi over the I2S (required some ssh editing to include RPi DAC in the config file as PCM5102) (Data at pin 40 of RPi, LRCK at Pin 35 of RPI and BCK at Pin of RPi), although there was some sound but it was low and there was static and noise as well, which led me to conclude that I must do some serious reading to understand if there is a bitrate mismatch.
Anyways I have used PC, laptop, IPad, Samsung S22+ as input sources but trust me nothing did sound as crisp as the RPi setup, and merits further exploration.
On another note AD1862 DAC definitely is a notch above AD1865, AKM4490,ES9038Q2M, ES9038Pro and PCM63.
Finally got time enough to complete AD1862 DAC, chips courtesy @Paddy Garcia (who is a genuinely honest and patient guy, informed me that the chips he dispatched to my address were delivered back to him. He posted them again and thankfully received them)
PCBs were made from JLCPCB with a purple solder mask. For PSU I used the 7805/7905/7812/7912 version, I have had enough with LT1963/3015 version.
the IV stage is standard OPA627/AD797, the final version will use OPA627AM or BM version (found the nasty buggers rotting in a small box in my lab).
USB interface was generic XMOS (chinese source) and the input was from RPi 4B (4GB version) having operating system of LibreElec, and the audio files were both FLAC and MP3s.
Amplifier's were Yamaha A1000 and Lo.D HMA-6590 (both have been completely recapped by me in the past)
First impression of the AD1862 sound was crisp, sharp and pleasant to my ears. Although I have enough test equipment to measure noise characters but I don't indulge with the esoterics.
After listening for a few hours I changed the DAC to AD1865N with the same USB/Input/AMP setup, what really surprised me was that even AD1865 performed much better than the previous experience. So I went on to test the RPi based input with other DACs like AKM4490, ES9038Q2M and ES9038Pro, TDA1541 and PCM63. Every DAC sounded better than before. I will share the picture tomorrow (hopefully).
I Understand @Vunce uses RPi as input source for his DAC. Next stage I tried to use the DAC directly interfaced with RPi over the I2S (required some ssh editing to include RPi DAC in the config file as PCM5102) (Data at pin 40 of RPi, LRCK at Pin 35 of RPI and BCK at Pin of RPi), although there was some sound but it was low and there was static and noise as well, which led me to conclude that I must do some serious reading to understand if there is a bitrate mismatch.
Anyways I have used PC, laptop, IPad, Samsung S22+ as input sources but trust me nothing did sound as crisp as the RPi setup, and merits further exploration.
On another note AD1862 DAC definitely is a notch above AD1865, AKM4490,ES9038Q2M, ES9038Pro and PCM63.
Last edited:
