Ittaku Posted February 5, 2019 Author Share Posted February 5, 2019 And here is a link to a patch for sox to allow you to perform this operation. It's only a 4 line patch. http://ck.kolivas.org/apps/sox/megataps.patch Enjoy! 1 1 Link to comment Share on other sites More sharing options...
Ittaku Posted February 5, 2019 Author Share Posted February 5, 2019 You can't infer too much from these graphs but they're interesting to see how changing the filter changes an impulse nonetheless. Look for the shape of the sinc function, the duration of ringing, and the peak amplitude matching the impulse. The scale is the same on all of them. This is the standard upscaling by sox: This is with the very high quality selection: This is with the steep filter: This is with my suggested filter: And this is with the extreme 500m taps filter: Link to comment Share on other sites More sharing options...
Nada Posted February 5, 2019 Share Posted February 5, 2019 (edited) 4 hours ago, Ittaku said: And here is a link to a patch for sox to allow you to perform this operation. It's only a 4 line patch. http://ck.kolivas.org/apps/sox/megataps.patch Enjoy! Thanks for sharing your work so freely. Unfortunately for me a patch is something that goes on my bicycle tube after a puncture. I wonder if anyone has the skills to mod the Foobar dsp called Resampler-V which uses SOX , to add in a tap function, allow multithreading and widen the scale parameters? Resampler V : https://sourceforge.net/projects/resamplerv/ Here are two contrasting SOX plots from Resampler-V with very different settings. It would be brilliant to be able to have a slider for tap length too. Edited February 5, 2019 by Nada Link to comment Share on other sites More sharing options...
krebetman Posted February 8, 2019 Share Posted February 8, 2019 Thanks for the patch file, @Ittaku. Have been playing around but am a bit limited by the RAM on my machine (only 2GB). I have generally found I like minimum-phase filters and am currently enjoying some tracks I processed with: Quote sox --buffer 65536 -S --multi-threaded -V4 input.flac -b 24 output.flac vol -2dB rate -b 99.998 -M -d 24 -R 160 192000 1 Link to comment Share on other sites More sharing options...
eltech Posted February 8, 2019 Share Posted February 8, 2019 (edited) A story... Between 5 and 10 years ago I used to muck around with PC playback software and tried the upsampling etc. Of course, different upsampling sounded a bit different. To cut the story short, of course I used the one that sounded best to me with my DAC, and associated equipment. but these days I'm not so sure that it's possible to say that one is definitively better than another. Different yes, better; I'm not necessarily sure. I am convinced that anybody could probably pick a favourite in their own system, and many people may concur with those settings. But there's always going to be someone who disagrees in their own system and there's probably a fair reason for the disagreement. The impulses shown above don't actually exist in real music. But they're really good to show what the filter is doing. Not even a snare drum or a cymbal crash has that sort of impulse. Which leads to the obvious question why do different upsampling filters sound different? I don't have a definitive answer but I do think that electrical noise such as electromagnetic radiation and ground bounce in digital circuits plays a large part in the sonic signature of different oversampling or upsampling filters. so I think it has more to do with the speed at which you run the data through your digital chain and how your electronics deal with that, more so than the actual filter itself. When I think about it, my logic thinks that since we really can't hear at or above 20 kilohertz any effects of the oversampling filter even basic types are well above what we can hear. But we certainly can hear the effect of ground bounce and emi affecting frequencies within the range of our hearing. For example I note that in these sorts of discussions people often comment about sibilance in female vocals but this is around 7 khz. How possibly can different filters be affecting this frequency? I don't think it can. But electrical interference and noise and ground bounce can. So a lot of it I think it comes down to how the decoupling capacitors and ground plane in all parts of the digital chain manage to suppress the unwanted effects of digital switching noise. Long story short is to pick the one sounds best to you in your system, oh, and too much trying different things can make you go a bit mad and lose your reference point and start to become unhappy with how everything sounds because you can start to overanalyze everything rather than enjoy the music. I think the best thing is to pick a setting and stick to it Edited February 8, 2019 by eltech Link to comment Share on other sites More sharing options...
Ittaku Posted February 8, 2019 Author Share Posted February 8, 2019 As I said earlier, different filters affect timing, phase, and amplitude of the audible signals. No need to go looking for other explanations. Link to comment Share on other sites More sharing options...
eltech Posted February 8, 2019 Share Posted February 8, 2019 (edited) @Ittaku How do filters affect timing or phase or amplitude within the audible frequencies? The filters only work above frequencies that we can hear. There are other explanations as I've put forth above. It's well understood on the DIY Audio forum. (And in the field of electronics design) Electrical noise is IMO the primary cause of audible differences. Phase is on the periphery of what we can perceive unless it's a gross change from zero degrees. Can you hear one degree of phase change? What degree can you percieve? Even if you can perceive a change, how much is important to the enjoyment of music? Edited February 8, 2019 by eltech Link to comment Share on other sites More sharing options...
legend Posted February 8, 2019 Share Posted February 8, 2019 (edited) I think timing and phase are important - and affect my enjoyment of music/audio reproduction. Timing is the underlying basis of Rob Watt's M-Scaler (leading to @Ittaku extreme filtering) - and of Bob Stuart's MQA. I hear significant practical improvements with all of them (better resolution of notes. more PRAT, better imaging etc). Phase/timing is also the underlying basis of DEQX (apart from its Xover work). One can correct for group delays/phase over the whole hearing frequency range of a passive speaker - from which I hear better focus of the image. PRAT etc and so more like being there at the musical performance/recording & its emotional impact, Edited February 8, 2019 by legend grammar Link to comment Share on other sites More sharing options...
Ittaku Posted February 8, 2019 Author Share Posted February 8, 2019 (edited) 9 hours ago, eltech said: @Ittaku How do filters affect timing or phase or amplitude within the audible frequencies? The filters only work above frequencies that we can hear. There are other explanations as I've put forth above. It's well understood on the DIY Audio forum. (And in the field of electronics design) Electrical noise is IMO the primary cause of audible differences. Phase is on the periphery of what we can perceive unless it's a gross change from zero degrees. Can you hear one degree of phase change? What degree can you percieve? Even if you can perceive a change, how much is important to the enjoyment of music? The amplitude rolls off at higher frequencies with a minimum phase filter; it does not result in a flat frequency response in the audible frequency range unless the original sampling rate is 88 or higher. The group delay and therefore phase is variable depending on the frequency therefore they do not occur at the same time, but it is minimised. In a linear filter the frequency response is flat to beyond the audible range, and the group delay is constant across the frequency response but large - if applied beforehand (such as software upscaling the file as I've done here) the latency is irrelevant at playback time. Electrical noise within the active upscaling filtering stage on a DAC during playback can surely affect sound but it can't have any effect on a software upscaled file. Edited February 8, 2019 by Ittaku Link to comment Share on other sites More sharing options...
Ittaku Posted February 8, 2019 Author Share Posted February 8, 2019 Here is a blind test comparing linear vs minimum phase filters, but NOTE these are with the default filters from sox, not the ridiculous ones I'm using here: http://archimago.blogspot.com/2015/04/internet-blind-test-linear-vs-minimum.html Link to comment Share on other sites More sharing options...
eltech Posted February 8, 2019 Share Posted February 8, 2019 (edited) @legend when i was referring to filters I wasn't specific, but was talking about the normal sort of linear phase filters found in normal cd players and DACs. Which shouldn't have any phase or frequency response issues. @Ittakuif there is a change in frequency response then there is always a chance someone can hear it. But I don't know why people would set out with that goal in mind except as an experimentation. I didn't suggest that noise in the electronics could effect the upscaled file, but there is much anecdotal evidence that changing settings (even OS settings) on a PC does effect sound. So the amount of processing being done could be having an effect (not necessarily always positive) There are no isolated systems in this world. Everything affects everything else. Some people think that if they only change one thing it will just affect the next thing in the chain that they intended, but there are also other effects. Think of the butterfly effect, and the double slit experiment Edited February 8, 2019 by eltech Link to comment Share on other sites More sharing options...
Ittaku Posted February 8, 2019 Author Share Posted February 8, 2019 Whilst the PC is rendering data sure, but not on an offline file. There is no mechanism for any PC noise to get into a data file. As I said it's only of relevance in active live filtering. Link to comment Share on other sites More sharing options...
rand129678 Posted February 9, 2019 Share Posted February 9, 2019 3 hours ago, legend said: Timing is the underlying basis of Rob Watt's M-Scaler (leading to @Ittaku extreme filtering) - and of Bob Stuart's MQA. I hear significant practical improvements with all of them (better resolution of notes. more PRAT, better imaging etc). Interesting quote from HQPlayer's developer, Jussi Laako about this: "looked from two extremes, both Chord talking about transient accuracy with extremely long filters and MQA talking about transient accuracy with extremely short filters are both right in a way, but only looking at things from one point of view while ignoring others. As usual in life, truth is somewhere between the extremes..." https://audiophilestyle.com/forums/topic/49609-john-atkinson-yes-mqa-is-elegant/?do=findComment&comment=866897 Link to comment Share on other sites More sharing options...
Ittaku Posted February 9, 2019 Author Share Posted February 9, 2019 3 minutes ago, Music2496 said: Interesting quote from HQPlayer's developer, Jussi Laako about this: "looked from two extremes, both Chord talking about transient accuracy with extremely long filters and MQA talking about transient accuracy with extremely short filters are both right in a way, but only looking at things from one point of view while ignoring others. As usual in life, truth is somewhere between the extremes..." https://audiophilestyle.com/forums/topic/49609-john-atkinson-yes-mqa-is-elegant/?do=findComment&comment=866897 Yes I recall seeing this quote a lot, but then that's just their interpretation that the somewhere between the extremes is correct. 1 Link to comment Share on other sites More sharing options...
rand129678 Posted February 9, 2019 Share Posted February 9, 2019 (edited) 6 minutes ago, Ittaku said: Yes I recall seeing this quote a lot, but then that's just their interpretation that the somewhere between the extremes is correct. You are looking at things only from a digital filtering side... HQPlayer's developer looks at the digital filters side, plus analogue recordings/measurements perspective... https://audiophilestyle.com/forums/topic/49609-john-atkinson-yes-mqa-is-elegant/?do=findComment&comment=874422 Edited February 9, 2019 by Music2496 Link to comment Share on other sites More sharing options...
Ittaku Posted February 9, 2019 Author Share Posted February 9, 2019 1 minute ago, Music2496 said: You are looking at things only from a digital filtering side... HQPlayer's developer looks at the digital filters side, plus analogue recordings/measurements perspective... https://audiophilestyle.com/forums/topic/49609-john-atkinson-yes-mqa-is-elegant/?do=findComment&comment=874422 I understand, but for the purposes of this discussion and experimentation, I only have control over the digital filtering during upscaling and am confined only to that digital filters side. 1 Link to comment Share on other sites More sharing options...
rand129678 Posted February 9, 2019 Share Posted February 9, 2019 (edited) 2 minutes ago, Ittaku said: I understand, but for the purposes of this discussion and experimentation, I only have control over the digital filtering during upscaling and am confined only to that digital filters side. Noted. I'm just giving some background to where his comments "somewhere in between the extremes" may come from... he considers what you consider, plus does measurements with real instruments and music... (per the link above) Edited February 9, 2019 by Music2496 1 Link to comment Share on other sites More sharing options...
eltech Posted February 9, 2019 Share Posted February 9, 2019 1 hour ago, Ittaku said: There is no mechanism for any PC noise to get into a data file. I'm going to try to find you an article that actually says it can. I'll post it later. I and another technician from Sydney have anecdotally verified that the OS settings, and PSU power quality of the CD Drive do actually affect the sound quality of ripped CDs even though the checksum of the files are the same, the percievable sound quality is different. The same tests have been done with ripped DVDs (video) If you don't want to believe it, that's ok. I have no intention of getting into a debate. And I have no way to prove it except to suggest trying different things for yourself. But if I find the article I'll post a link. As for affecting an existing audio file on a hard drive, I wasn't implying anything like that at all. 15 minutes ago, Ittaku said: I understand, but for the purposes of this discussion and experimentation, I only have control over the digital filtering during upscaling and am confined only to that digital filters side. Understood. Link to comment Share on other sites More sharing options...
Ittaku Posted February 9, 2019 Author Share Posted February 9, 2019 1 minute ago, eltech said: I and another technician from Sydney have anecdotally verified that the OS settings, and PSU power quality of the CD Drive do actually affect the sound quality of ripped CDs even though the checksum of the files are the same, the percievable sound quality is different. The same tests have been done with ripped DVDs (video) If you don't want to believe it, that's ok. I have no intention of getting into a debate. And I have no way to prove it except to suggest trying different things for yourself. But if I find the article I'll post a link. Yeah if two data files are identical, then I can't move into the faith/belief stage that they sound different. Best we stop this discussion here and stick to filtering. As for audible quality differences, I've already pointed out that linear vs minimal phase have differences in the audible frequency spectrum so again I'm not even sure why you want to blame something else for them sounding different. Link to comment Share on other sites More sharing options...
zenelectro Posted February 9, 2019 Share Posted February 9, 2019 1 hour ago, eltech said: I'm going to try to find you an article that actually says it can. I'll post it later. I and another technician from Sydney have anecdotally verified that the OS settings, and PSU power quality of the CD Drive do actually affect the sound quality of ripped CDs even though the checksum of the files are the same, the percievable sound quality is different. The same tests have been done with ripped DVDs (video) If you don't want to believe it, that's ok. I have no intention of getting into a debate. And I have no way to prove it except to suggest trying different things for yourself. But if I find the article I'll post a link. As for affecting an existing audio file on a hard drive, I wasn't implying anything like that at all. Understood. This has been observed by certain mastering engineers years ago when CD's were still the main medium. There were various scientific explanations to do with increased servo activity of replay drive due to inferior burn quality. You probably should re do this test, copying files from said burned CD's to HDD and playback with an isolated, re clocked USB interface. The differences will then most likely disappear. T 2 Link to comment Share on other sites More sharing options...
Ittaku Posted February 9, 2019 Author Share Posted February 9, 2019 44 minutes ago, zenelectro said: This has been observed by certain mastering engineers years ago when CD's were still the main medium. There were various scientific explanations to do with increased servo activity of replay drive due to inferior burn quality. You probably should re do this test, copying files from said burned CD's to HDD and playback with an isolated, re clocked USB interface. The differences will then most likely disappear. Ah CD transports in a realtime relatively synchronous environment playing the same data is totally different. I can understand how they'd affect the sound quality. Different CD transports ripping the same data which is then played asynchronously at a later time though cannot be any different. 1 Link to comment Share on other sites More sharing options...
eltech Posted February 9, 2019 Share Posted February 9, 2019 (edited) 1 hour ago, zenelectro said: You probably should re do this test, copying files from said burned CD's to HDD and playback with an isolated, re clocked USB interface. Actually, this test has been done and the improved sound quality persists. This test has been verified by enthusiasts around the world, but so far none of us have figured out why. We just accept it. If I find the article I'll post the link. But I won't discuss it further in this thread because this thread isn't about this topic. It was a passing comment. Anyone can PM me for more details, but I've told most of it already. The reason I mentioned it was in reply to something someone said, and is only there as an explanation for my thoughts on things. I'm not passionate to prove it, or defend what we've observed. If you don't agree that's cool. I won't hold it against anyone, or be surprised if 99.9% don't believe. That's the logical thing to do. But if you did hear it, like us, then you'd have to believe, even though it can't be explained with normal digital theory or measurements. But it can be perceived. Please, back to the topic of upsampling. Edited February 9, 2019 by eltech Link to comment Share on other sites More sharing options...
eltech Posted February 9, 2019 Share Posted February 9, 2019 The article I mentioned (for completeness) http://www.enjoythemusic.com/hificritic/vol6_no1/audio_networking.htm I will let you read for yourselves. Please see the section rewards the bottom entitled "New Zealand LOG Rips " The person mentioned - Kethal in the article is my penpal technician friend from Sydney. Please though, for respect to the OP, if you want to discuss send me a PM. I didn't want to distract from this thread. Link to comment Share on other sites More sharing options...
Ittaku Posted February 9, 2019 Author Share Posted February 9, 2019 Unscientific unblinded experiments need not apply. 1 Link to comment Share on other sites More sharing options...
eltech Posted February 9, 2019 Share Posted February 9, 2019 Just now, Ittaku said: Unscientific unblinded experiments need not apply. Just one question. Do you listen to your filters or just measure them? Link to comment Share on other sites More sharing options...
Recommended Posts