Ken,
I would like to repeat that settling time can be caused by an LPF, which is a linear process without adding distortion to the signal, while slew rate is a non linear process potentially causing distortion when not fast enough.
Whith the 1.5 usec settling in all occasions for the TDA1541 would be caused by slew rate, and let’s say the I/V converter goes from +4Volt to -4Volt, that would mean ca 5V/usec which as we have seen is exactly on the edge when using a 2nF cap and would asks for a passive I/V converter.
Sofar everything under control.
It’s obvious that with the much faster 1794 you don’t even have to worry at all and an active I/V could be used, but as I’ve reported in my case, when output impedance is not low enough, you better use a passive I/V.
So to conclude, I’m now all the sudden in favour for the passive V/I solution, no issues with slew rate, GBW or output impedance, without any need for adding a S&H.
Hans
I would like to repeat that settling time can be caused by an LPF, which is a linear process without adding distortion to the signal, while slew rate is a non linear process potentially causing distortion when not fast enough.
Whith the 1.5 usec settling in all occasions for the TDA1541 would be caused by slew rate, and let’s say the I/V converter goes from +4Volt to -4Volt, that would mean ca 5V/usec which as we have seen is exactly on the edge when using a 2nF cap and would asks for a passive I/V converter.
Sofar everything under control.
It’s obvious that with the much faster 1794 you don’t even have to worry at all and an active I/V could be used, but as I’ve reported in my case, when output impedance is not low enough, you better use a passive I/V.
So to conclude, I’m now all the sudden in favour for the passive V/I solution, no issues with slew rate, GBW or output impedance, without any need for adding a S&H.
Hans
Settling-time is not only comprised of slew-rate error. It's complex, and also includes glitch energy and may even reflect a briefly incorrect code value being output from the quantizer. Because of those additional factors, the energy/area of each sample is always in some amount of non-linear error. Whether, or not that error is subjectively significant, I do not know. Your point is that they are not significant, I believe. To which, I can only paraphrase something which Frans asserts in one of his S/H papers. That, it's undesirable to simply low-pass filter the settling error, as it's not simply a linear error. Once integrated in to the signal via filtering, it permanent alters the sample in a way which can not be removed later. A properly implemented S/H circuit completely gates the error, blocking it's passage to later analog stages.
Regarding the PCM1794, or any other modern OS audio DAC, don't you think it arbitrary to declare their settling-time to be an irrelevant subjective factor just yet?
I should make clear that I'm not promoting the adoption of Fran's S/H ideas. I do, however, find the technical issues he highlights intriguing. So, I'm planning to experiment with a S/H circuit in my next DAC. I wouldn't be surprised to find that I didn't hear any subjective benefit of it's inclusion, but that's what diy audio experiments are about.
Last edited:
Ken,
This is all about DIY, nothing is more rewarding then to experiment with alternatives, so let nobody stop you to try the S&H.
And as you mentioned yourself, the ulimate test is by listening to your favourite music.
I made a small detour from your thread because I thought it could be a vital factor in the difference between NOS and OS, where many NOS systems seem to be using passive I/V converters.
Now I will sit back and follow your nice thread with great interest.
Hans
This is all about DIY, nothing is more rewarding then to experiment with alternatives, so let nobody stop you to try the S&H.
And as you mentioned yourself, the ulimate test is by listening to your favourite music.
I made a small detour from your thread because I thought it could be a vital factor in the difference between NOS and OS, where many NOS systems seem to be using passive I/V converters.
Now I will sit back and follow your nice thread with great interest.
Hans
I have some NatSemi LF398H S/H circuits, could they be used as de-glitcher after a DAC? The output transient does not look very nice:
http://mdc.custhelp.com/euf/assets/images/KB_856_NatSemi_SampleHold_circuit.pdf
Or maybe even a 74HC4066 analog switch could do (as Frans Sessing recommended)?
http://mdc.custhelp.com/euf/assets/images/KB_856_NatSemi_SampleHold_circuit.pdf
Or maybe even a 74HC4066 analog switch could do (as Frans Sessing recommended)?
Thanks, Hans.
I much appreciate having your inputs here in the investigation. In fact, I'm especially encouraged to see you showing attention toward suspect category, D) Sample-Period Related Quantization Errors. While I've made no secret of my suspicion that equiripple filters are causing, at least, a siginificant portion of the disappointing 'digital sound' which we're now investigating. That suspicion may, however, easily prove completely wrong.
Should it actually be wrong, then the quantization-error category would seem most likely to hold the actual cause, among the remaining suspects on our list. A list which, undoubtedly, isn't comprehensive anyhow. So, it very much helps us that you are already applying some good thinking there.
Also, have a look at the interesting approach to a post audio DAC sample-and-hold circuit that is shown in the attached Burr-Brown PCM56 data sheet photo.
Attachments
Coming fresh from perusing the DS of the LT1468 (thanks Hans, looking into the Bel Canto DAC was instructive) I just wanted to mention that its not a linear relationship that 8X OS means 8X higher percentage of the time taken to settle. That's because OS reduces the step size between samples and LT helpfully shows a couple of plots that show that settling time increases for larger steps. Unfortunately the plots don't show below 2V steps which would be the area of interest so we'd just have to guess that the trend of lowering settling time for smaller steps doesn't hit a brick wall at 2V.
Having given this some thought in times gone by I also reckon the shape of the settling transient is probably important. I've never seen a pure exponential settling waveform from an opamp but that is what would be needed for distortion-free performance. I think we can safely infer that post-opamp the distortion is going to be higher with faster sample rates - not necessarily harmonic distortion though.
I doubt your probe / probing technique was correct enough to make your results valid. Could easily be radiated pickup.
Yes. I am not trying to pick on Mark here. I just see this sort of thing all the time, especially up past tens of MHz.
Fet probe with short ground spring; noise pickup is very low. The noise was switching noise generated by the modulator and shifted up out of the audio frequency band. It has to be there or the dac will not work.
Frank, there is something which I find very interesting about popular music today - which is dominated by rap and hip-hop. I'm not refering to their supposed musicality, or lack thereof, but that popular music has been stagnant since their rise to the top. When I think back, popular music has experienced generational change about every decade, since at least, the 1930s or 40s. This seems likely related to the coming of age for each generation of we humans. From puberty to adulthood spans roughly a decade. One specific genre, or two has dominated each decade ever since then. Roughly approximating:
1930s > Dixie land style Jazz
1940s > Big-band Jazz
1950s > 'Modern' Jazz > Doo Wop
1960s > Rock & Roll
1970s > Disco
1980s > Punk
1990s > Rap & Hip Hop
2000s > Rap & Hip Hop
2010s > Rap & Hip Hop
2020s > So far, Rap & Hip Hop
Each decade has always also held other popular genres, but it seems that one always was dominant. I'm uncertain of why this stagnation has now occurred for the past 30 years, but I do primarily suspect the following two reasons:
1) The big corporatisation of the music industry. Meaning, where once creative executives made the key talent and production decisions, they have gradually been replaced by lawyers, MBAs, and lawyers with MBAs.
2) Perhaps, popular music by it's very nature of being popular (Meaning, the most easily accessible and appreciated of music entertainment forms) will always seek and appeal to the greatest number of ears. Whereas popular music once required, at least, some instrumental or some vocal talent to create or mimic, even that is no longer necessary. Just about any average person on the street can passably join in the singing of a rap song. I don't want this to seem like a criticism of rap as an musical art form. Many rap songs speak about deep societal problems experienced by their creators, and are sometimes much more than just musical entertainment. I'm just speaking of the seeming creative stagnancy of the popular music industry.
Where does popular music go from here? Perhaps, it goes no where, doomed to repeat the past three decades. Which wouldn't be that unprecedented because classical music, it could be argued, has remained stagnant for hundreds of years now. Or, perhaps, popular music comes full-circle and essentially repeats past popular genres? I have no idea.
Last edited:
FYI - For the benefit of those who may be interested. Attached, are two short papers regarding how to, and whether or not to, simply RC filter a DAC's output glitching energy. Glitching, is one element among several comprising the settling-time error for each sample update. The answer is, that it comes down to the application. The second page of the Intersil paper (the bottom attachment) gets quickly to the point.
Last call for alcohol - er, Analog Filter Assessments
While I've solicited the opinions of those who have DACs featuring strictly analog (no digital OS) reconstruction filters, we have not received much response as yet, with the notable exception of Abraxalito. I've attached another screen shot showing some additional feedback from him. I was, however, hoping to have developed a much more clear picture of the subjective character which purely analog image-band suppression utilizing steep analog filters produces, by having multiple reports posted by now. With almost no reports to consider, the risk of drawing incorrect conclusions is very high.
Abraxalito's positive subjective experience with passive analog filtering does, however, support the suspicion that some aspect of digital OS interpolation-filters performance produces listening artifacts. If subjective reports were, instead, that steep analog reconstruction sounded, essentially, the same as digital OS reconstruction, then the general topic of signal reconstruction along with it's methods would remain equally suspect between analog and digital. Which is why I view it as important that we obtain even just a few more confirming listening reports.
As I had mentioned earlier, the purely analog reconstruction filter approach was utilized by Sony in their first CD-player, and may prove to be the best sounding practical DIY solution. Our understanding of the negative sound implications of active analog filters, undoubtedly utilizing relatively slow op-amp ICs back then, and the processing of high bandwidth DAC signals certainly has come a long way since 1983. Just as it has for other DAC sub-system elements and physical components. So, who knows, we may end up coming full-circle in that regard.
Therefore, I once again ask for anyone having experience listening to a DAC featuring steep, purely analog reconstruction filters to please share your subjective assessment with us. Your subjective assessments are an important part of the puzzle, and valued.
While I've solicited the opinions of those who have DACs featuring strictly analog (no digital OS) reconstruction filters, we have not received much response as yet, with the notable exception of Abraxalito. I've attached another screen shot showing some additional feedback from him. I was, however, hoping to have developed a much more clear picture of the subjective character which purely analog image-band suppression utilizing steep analog filters produces, by having multiple reports posted by now. With almost no reports to consider, the risk of drawing incorrect conclusions is very high.
Abraxalito's positive subjective experience with passive analog filtering does, however, support the suspicion that some aspect of digital OS interpolation-filters performance produces listening artifacts. If subjective reports were, instead, that steep analog reconstruction sounded, essentially, the same as digital OS reconstruction, then the general topic of signal reconstruction along with it's methods would remain equally suspect between analog and digital. Which is why I view it as important that we obtain even just a few more confirming listening reports.
As I had mentioned earlier, the purely analog reconstruction filter approach was utilized by Sony in their first CD-player, and may prove to be the best sounding practical DIY solution. Our understanding of the negative sound implications of active analog filters, undoubtedly utilizing relatively slow op-amp ICs back then, and the processing of high bandwidth DAC signals certainly has come a long way since 1983. Just as it has for other DAC sub-system elements and physical components. So, who knows, we may end up coming full-circle in that regard.
Therefore, I once again ask for anyone having experience listening to a DAC featuring steep, purely analog reconstruction filters to please share your subjective assessment with us. Your subjective assessments are an important part of the puzzle, and valued.
Attachments
I can not say that my PCM 601 ESD is my best DAC ....
I still use it for recording though.
Sony PCM-601ESD
I still use it for recording though.
Sony PCM-601ESD
