Hmm, as 27 model, first repeated may 4..8 of the samples and afterwords did the spline over...
Newer models had some more/different algo's. The latest models even wit SACD but cannot comment the change from 16bit source against 24bit source.
Also the 27 model with the outdated Motorola DSP, while today we have more horse power. For me the question rises why no chico's started with new gear even up to DXD or even 768kHz..
Hp
This i recall indeed...
Actually i would like to have the oportunity to listen to it again as it has been a while now...
In the mean time a lot has changed while developing my own (i feel for the better lol)...
Best wishes,
Lampie519 aka Frank
Last edited:
... I have a DIY AD1865 based DAC which I designed and built specifically for conducting various experiments in NOS. Well, I recently performed one experiment too many on the PCB of that DAC and trashed it.
Would an AD1862 based NOS DAC be a good "lens" for NOS vs OS listening?
Perhaps the more interesting question is not about NOS vs OS but NOS vs which OS.
Foobar has a very interesting DSP called resampler-V with sliders that display the LPF and Impulse as the filter changes while music is playing. The changes can be subtle but good PC OS can sound better then NOS.
I recall seeing two concurring reports, one from dddac and the other from Marcel, on my proposal to eliminate the 20KHz droop item from the suspects list. In addition, while I don't wish to speak for anyone else, some past posts of Abraxalito's leads me to believe that he would also concur. Which would make 4 of us. Should you have a dissenting report regarding this suggestion, please post it immediately. Otherwise, we will press forward.
As Marcel correctly pointed out, we ideally would perform an finely calibrated ABX listening test to verify this decision. However, lacking resources for a properly controlled ABX test, unless someone wishes to volunteer to conduct one, I feel that it is more safe than not, to assume that no merging of OS/NOS subjective character would reveal itself at a calibrated less than 0.1dB matching to flat at 20KHz, which does not also reveal itself at a uncalibrated less than 1dB matching to flat at 20KHz.
In addition, I find that the OS versus NOS subjective difference is well apparent even with lower register instruments and the human voice. Which, do not have overtones approaching anywhere near 20KHz.
As Marcel correctly pointed out, we ideally would perform an finely calibrated ABX listening test to verify this decision. However, lacking resources for a properly controlled ABX test, unless someone wishes to volunteer to conduct one, I feel that it is more safe than not, to assume that no merging of OS/NOS subjective character would reveal itself at a calibrated less than 0.1dB matching to flat at 20KHz, which does not also reveal itself at a uncalibrated less than 1dB matching to flat at 20KHz.
In addition, I find that the OS versus NOS subjective difference is well apparent even with lower register instruments and the human voice. Which, do not have overtones approaching anywhere near 20KHz.
Last edited:
I actually disagree. I think the droop is the most obvious difference between non-oversampling DACs with only a zero-order hold as reconstruction filter and any DAC with a steep filter, such as oversampling DACs. If you want to make it plausible that there are more subtle effects playing a role, you have to find a way to get rid of this difference first - for example by applying an equal roll-off to the digital signal going to the oversampling DAC.
There are computer programs (foobar with foo_abx) that you can use to do a simple ABX test between two audio files, so it would be great if the whole thing could somehow be reduced to comparing two computer files via the same equipment. Hence my earlier proposals.
There are computer programs (foobar with foo_abx) that you can use to do a simple ABX test between two audio files, so it would be great if the whole thing could somehow be reduced to comparing two computer files via the same equipment. Hence my earlier proposals.
Last edited:
Here i have to disagree (sorry Marcel , i hate to disagree with you ! lol)...
The droop is (in my experience) not the main difference (yes it is audible).
But unless you do not have a very very good FIR you will hear the difference in other aspects more.
Like dark black backgounds and pin point precision of instruments, etc. all you will also hear from other NOS evangelists.
The droop is (in my experience) not the main difference (yes it is audible).
But unless you do not have a very very good FIR you will hear the difference in other aspects more.
Like dark black backgounds and pin point precision of instruments, etc. all you will also hear from other NOS evangelists.
I don't recall any DAC chip datasheet then or now specifying flat response to 22.05KHz, only to 20KHz. The original specification requirements of CD undoubtedly included the assumption of 20KHz as the maximum flat signal frequency. Since Philips and Sony engineers well understood that the sampling theorem required band-limiting of the signal by the Nyquist frequency, and that there were no practical true brick-wall filters, Nyquist would need be set somewhere above 20KHz to allow for a practical brick-wall filter to have a transition band. Which, for CD, is barely enough at 2KHz. Many have rightly questioned whether CD's 2KHz transition band is wide enough.
The 44.1KHz sample rate was, supposedly, settled upon, for a combination of reasons. Providing a 20KHz signal band, plus a sinc filter transition band. Plus, providing a total data compactness which enabled some certain classical music composition (One of Beethoven's symphonies, perhaps) favored by the then president of Sony, to fit on a single 5 and 1/4-inch optical disc. In addition, I seem to recall something about the final sample rate having to be compatible with Sony U-matic studio videotape machines, which were to be adapted to record digital audio.
@marcel Mmm, what you are comparing are the reconstruction filters, where as most “NOS” DACs at not using a filter at all. I would say compare dacs with and without FIR - over sampling and without reconstruction filter
After that you can add several reconstruction filters and compare filters. That is a totally different cup of tee …
My 2 cents are that a NOS DAC with a theoretical steep reconstruction filter does not sound so natural and anlog as one without. So if you want to compare DACs with their reconstruction filters, I know who will “win” but still you will not know what makes NOS so special
After that you can add several reconstruction filters and compare filters. That is a totally different cup of tee …
My 2 cents are that a NOS DAC with a theoretical steep reconstruction filter does not sound so natural and anlog as one without. So if you want to compare DACs with their reconstruction filters, I know who will “win” but still you will not know what makes NOS so special
Last edited:
Marcel, I agree, that the un-equalized droop is the most obvious audible difference between OS and NOS. However, my suggestion is not regrading that. Instead, my suggestion is that once the NOS droop is EQ'd flat (in analog), the NOS sound character remains apparent. Such as, an immediate sense of ease, and long-term low listening-fatigue, more 3D depth effect.
Agreed. anyone attempting to identify the reason for the subjective in character between OS and NOS via listening comparison, would need to EQ the frequency response difference to flat, so as to remove that as a variable. The preferable way to do that would be to EQ the NOS DAC to be flat (quite simple to do, either passively or actively in analog), rather than to roll-off the OS DAC to be non-flat.
"most “NOS” DACs at not using a filter at all"
As well in my case...
"I would say compare dacs with and without FIR - over sampling and without reconstruction filter"
Yes there is a difference just in my case the FIR (the one i am now using, unfortunately non DIY) will increase the audio quality... ( i know that is not what you would like to hear but it is "true" , because Lampie519 says so LOL )
You are (both) invited to hear for yourself !
Best wishes, Frank
As well in my case...
"I would say compare dacs with and without FIR - over sampling and without reconstruction filter"
Yes there is a difference just in my case the FIR (the one i am now using, unfortunately non DIY) will increase the audio quality... ( i know that is not what you would like to hear but it is "true" , because Lampie519 says so LOL )
You are (both) invited to hear for yourself !
Best wishes, Frank
Last edited:
dddac,
Yes, this is one of the more intriguing questions which I'm hoping that we can answer along the way. Your observation on the sound of steep-slope analog reconstruction filters is in alignment with Abraxalito's, I believe. He felt that steep analog reconstruction filters sounded superior to digital reconstruction filters. Until he began experimenting with the SoX resampling utility within Foobar. He now reports that the sound of the two filters is much closer, with him actually preferring, if I correctly recall, the sound via the Foobar digital upsampling engine on some music tracks.
For me, such observations as yours is leading towards a long suspicion that certain aspects of typical on-chip digital oversampling reconstruction filter implementations are, and always have been, at fault. I'm attempting, however, to arrive there methodically, and not rush to prejudge the conclusions.
Hi, kazap,
The AD1862 has an excellent reputation. I believe that it is the best non-SDM audio DAC ever produced by Analog Devices. It was also the favorite DAC chip of 'Jock Homo' (R.I.P.), who many of you will remember for his grounded-base, discrete I/V circuit. So, if you have a few, you should probably hold on to them.
Yes, that, is also an interesting question. I've expressed throughout this thread my expectation that it's specific implementation faults (for reason of cost) of typical on-chip oversampling digital filters which is the primary cause of the subjective character difference between OS and NOS which we are investigating.
Don't know how useful this might be? One possible avenue of investigating FIR Vs IIR interpolation filters is to use a DAC that can be easily switched between these two filters.
The PCM5102A DAC has both a normal FIR filter & low latency filter which is IIR - don't know what other DACs may have this function also?
A pin set to low or high switches between the two filters so can be changed within seconds.
From the datasheet, the FIR filter has passband ripple of +/- 0.02dB (pre echo & group delay of 22ts, if this is relevant). The IIR filter has passband ripple of +/- 0.0001dB (no pre-echo on graphs & group delay of 3.5ts)
Although the 5102 is SDM OS, if given a 384KHz input sample rate it bypasses the filters.
Page 17 & Pg 21 of the datasheet show the graphs & details
https://www.ti.com/lit/ds/symlink/p...https%3A%2F%2Fwww.ti.com%2Fproduct%2FPCM5102A
The PCM5102A DAC has both a normal FIR filter & low latency filter which is IIR - don't know what other DACs may have this function also?
A pin set to low or high switches between the two filters so can be changed within seconds.
From the datasheet, the FIR filter has passband ripple of +/- 0.02dB (pre echo & group delay of 22ts, if this is relevant). The IIR filter has passband ripple of +/- 0.0001dB (no pre-echo on graphs & group delay of 3.5ts)
Although the 5102 is SDM OS, if given a 384KHz input sample rate it bypasses the filters.
Page 17 & Pg 21 of the datasheet show the graphs & details
https://www.ti.com/lit/ds/symlink/p...https%3A%2F%2Fwww.ti.com%2Fproduct%2FPCM5102A
In fact, the redbook contains next to nothing wrt audio quality requirements. It only specified the digital format and all the details about coding, error correction, and mechanical restrictions.
Everything else was left to the manufacturers; Philips thought that 50 dB attenuation at nyquist would be sufficient in their oversampling filter and the antialias filters used for recording - and reproduction gear as well - were usually not offering even this number. In other threads, I've posted some of the datasheets from typical examples of these analog filters.
Everything else was left to the manufacturers; Philips thought that 50 dB attenuation at nyquist would be sufficient in their oversampling filter and the antialias filters used for recording - and reproduction gear as well - were usually not offering even this number. In other threads, I've posted some of the datasheets from typical examples of these analog filters.
This is a semantic issue, really. For reasons of mathematical convenience, discrete-time signals are usually regarded as series of infinitely narrow pulses. Anything that results in the DAC pulses having a non-zero width is then regarded as a kind of filter.
MarvelvdG is absolutely correct; the mentioned droop already results from the filtering action introduced by any ZOH at the DAC output.
Could be an interesting experiment to change any ZOH to RTZ by shortening the impulse length without changing the clock rate.
Of course, careful implementation would be required to avoid secondary effects due to glitches and ringing.
Could be an interesting experiment to change any ZOH to RTZ by shortening the impulse length without changing the clock rate.
Of course, careful implementation would be required to avoid secondary effects due to glitches and ringing.