It's just your opinion, you should add "I think" at the beginning of your statement. Here you have two links, if you want to read them, you will see that you cannot summarize things in a totalitarian and fundamentalist way. There are greys, not everything is black and white.
http://www.soundstagenetwork.com/vinyl/vinyl041998.htm
https://www.vox.com/2014/4/19/5626058/vinyls-great-but-its-not-better-than-cds
I hope not. As sound reproduction fidelity is not a problem of the DAC. It lies in recordings and sound excitation. What its going on currently is just a compensation game... where the sentenced party is not guilty.Patience. It will happen. If not tomorrow, definitely by next millennium
//
Not only not guilty; utterly incapable of solving the problem. To talk about femto second level phase noise/coherency as a target, when you are standing in a room, surrounded by objects that reflect/absorb/delay sound at a level so far above that before we even talk about the speakers, is baffling.
What is better is hardly the point. More interesting and telling is our general miscomprehension of human perception. Despite a lot of handwaving, neither Hafler could make his amps emulate a decent amp, neither any of the tube emulators can make sand sound like a Jadis or a Kondo. Digital, despite endless dsp power also fails to emulate the lesser analogue.
In response
The dialog presented by Markw4 is a statement of opinion being "tested" by his personal experiences. In absence of further evidence, extraordinary or otherwise, this statement can be weighted of merit by any mechanism a reader chooses. There is no need of consensus. It seems you are suggesting that all readers ought to weight such statements as worthless in absence of further necessarily extraordinary evidence.
I made some measurements with both my AK4490 and ES9038Q2M dacs. I used Autoranger between DAC and ADC at -6dB setting to lower the DAC output to avoid clipping at ADC. I used Sox to play the files.
At 44k1Hz FS only 11025Hz was visible as predicted with both files. So I made more measurements at 176k4Hz fs.
Here is 11025Hz at 0dBFS.
So no clipping artefacts visible. The strange noisefloor is probably due to original file being resampled by Sox.
Next 11025Hz normalized at 0dBFS
As MarceldvG predicted heavy clipping is visible.
Then I lowered the DAC digital volume to -3dBFS.
Here is 11025 at -3dBFS
So more or less the same as at 0dBFS.
11025Hz normalized at -3dBFS
Clipping artefacts are gone which indicates that the digital volume control operates before filter with AK4490. Same applies to ES9038Q2M so no need to repeat those measurements.
So IMO this intersample overshoot is a real problem but can be quite easily avoided with lowering DAC digital volume to e.g. -3dBFS at least with AK4490 and ES9038Q2M (probably applies to all modern AKM and ESS dacs).
You know - not everybody prefers the analog sound - some of us actually think that that digital technology has progressed sound reproduction to be more lifelike and enjoyable. Maybe it is us who regularly visit non amplified live music events.... and know how reality actually sounds.... and appreciate that reality - which is not always beautiful
//
Indeed it is a real problem and one that can easily be defeated for ALL recordings, by reducing level by default and accepting 3dB 'lower' measured performance specs on your headline spec. I plan to do that with 9028. performance of ~120dB THD+N is still totally acceptable IMO ...
I don't fully understand you. Do you use a program called Sox to interpolate the signal to 176.4 kHz sample rate and then put it into your DAC or do you only run the ADC at 176.4 kHz sample rate to see the ultrasonic distortion products?
The step in the noise floor looks like the 16 bit dithered noise of the file, filtered by an interpolation or reconstruction filter. Nothing alarming anyway.
Try the intermodulation test of post #127, https://www.diyaudio.com/community/...quality-between-some-dacs.386815/post-7040483
I thought I was running both the DAC and ADC at 176.4k and Sox resampling the file to 176.4k but actually I did not check so I'm not sure if it was only ADC running at 176k4. I could repeat the test to make sure that DAC is running at 44k1 and ADC at 176k4 but would it really change the outcome?
All the above is confusing to me - sorry. Emulate what with how powerful? Analogue experience with 5 cm wire of what?I was not discussing personal preference but our general inabilty to emulate one type of tech with another, even if the latter is much more powerful. Forget the full analogue experience, we don't even know how to emulate the sound of a 5cm silver wire
Complete jibberish... 🙃
//
BTW, the subject of intersample overs came up years ago in one of the Blowtorch threads. It wasn't long after the Benchmark article appeared. IIRC it was me that brought it up for discussion.
In those days Scott Wurcer was still very active, and Billshurv was one of the regulars there. IIRC the reaction I got from those guys was not insignificantly to the effect that it was mostly a Benchmark marketing gimmick. The reasoning was because natural music doesn't have harmonics at 0dBFS @ 11.25kHz. Natural music has diminishing harmonic signal levels as frequency increases.
Also IIRC at some point maybe in another thread, @1audio reported that some dac chips were more linear if run at peak levels a few dB below 0dBFS, in some cases maybe as low as -6dBFS peak.
Given the foregoing it was my practice for quite some time after all that to recommend to people that they set their digital volume controls in their playback apps to somewhere around -3dBFS to -6dBFS, whatever they thought best for their dac. Then do any additional needed volume adjustment in or after the dac chip.
While all the above was going on and although it was generally known to mastering engineers by then that peak levels should stay at or below -3dBFS, there was also a continuation of the digital volume wars with people trying to figure out how high peak levels could actually be before it mattered to consumers. Slowly over time clients and mastering engineers crept up peak levels closer to 0dBFS waiting for the market to finally complain, but it didn't. Instead a musical trend emerged that used clipping distortion as a cue to the expression of strong emotion (Sound Garden, Green Day, Nirvana, etc.). Another musical trend that emerged from public acceptance of high levels of clipping distortion and yet still more demand for louder releases, music was mixed and mastered to eliminate the natural HF roll of instrument harmonic frequencies. In other words, they couldn't just set low and midrange frequencies to 0dBFS, all frequencies had to be at that level. Enough consumers accepted the change, once again.
While all the above was going on, I was experimenting with how to best reproduce the illusion of space and preserve reverb decays (best as I could figure out, anyway). Turns out DSD worked out better for me than PCM. Also turns out the DSD modulator sound changed some depending on the PCM input drive level...
I'll leave off with the historical reminiscing here.
Bottom line for me: My recommendation to Marcel about the relative importance of fixing intersample overs remains my opinion for now.
In those days Scott Wurcer was still very active, and Billshurv was one of the regulars there. IIRC the reaction I got from those guys was not insignificantly to the effect that it was mostly a Benchmark marketing gimmick. The reasoning was because natural music doesn't have harmonics at 0dBFS @ 11.25kHz. Natural music has diminishing harmonic signal levels as frequency increases.
Also IIRC at some point maybe in another thread, @1audio reported that some dac chips were more linear if run at peak levels a few dB below 0dBFS, in some cases maybe as low as -6dBFS peak.
Given the foregoing it was my practice for quite some time after all that to recommend to people that they set their digital volume controls in their playback apps to somewhere around -3dBFS to -6dBFS, whatever they thought best for their dac. Then do any additional needed volume adjustment in or after the dac chip.
While all the above was going on and although it was generally known to mastering engineers by then that peak levels should stay at or below -3dBFS, there was also a continuation of the digital volume wars with people trying to figure out how high peak levels could actually be before it mattered to consumers. Slowly over time clients and mastering engineers crept up peak levels closer to 0dBFS waiting for the market to finally complain, but it didn't. Instead a musical trend emerged that used clipping distortion as a cue to the expression of strong emotion (Sound Garden, Green Day, Nirvana, etc.). Another musical trend that emerged from public acceptance of high levels of clipping distortion and yet still more demand for louder releases, music was mixed and mastered to eliminate the natural HF roll of instrument harmonic frequencies. In other words, they couldn't just set low and midrange frequencies to 0dBFS, all frequencies had to be at that level. Enough consumers accepted the change, once again.
While all the above was going on, I was experimenting with how to best reproduce the illusion of space and preserve reverb decays (best as I could figure out, anyway). Turns out DSD worked out better for me than PCM. Also turns out the DSD modulator sound changed some depending on the PCM input drive level...
I'll leave off with the historical reminiscing here.
Bottom line for me: My recommendation to Marcel about the relative importance of fixing intersample overs remains my opinion for now.
Last edited:
Mark. I wasnt involved in any of those discussions and have been setting my levels at about -3db on the dac for years; so i'm afraid i'm not going to credit you with that decision to make it -3db by default with 9028 unless overridden. I did know of Benchmark doing it, but they also do other things like running everything aync, which I will not be doing.
- Status
- Not open for further replies.