RichP714 3,151 Posted November 16, 2011 Share Posted November 16, 2011 Some ideas on why upsampling sounds 'better' http://www.mlssa.com/pdf/Upsampling-theory-rev-2.pdf Link to comment Share on other sites More sharing options...
oldtexasdog 2,362 Posted November 16, 2011 Share Posted November 16, 2011 So a cheap upsampler is better? I'm confused now in what to look for Link to comment Share on other sites More sharing options...
RichP714 3,151 Posted November 16, 2011 Author Share Posted November 16, 2011 So a cheap upsampler is better? I'm confused now in what to look for I don't think the article means cheap when they say 'poor' Rather, my understanding is that a 'rich' (well-behaved) filter should eliminate all aliasing artifacts, which means a steep slope filter, which means time smear. The author posits that a 'poor' filter (shallow slope, less time smear, but allows more aliasing artifacts to pass through) might actually sound better, because the aliased signal is so un-correlated that it might as well be random, and can act like high frequency dither that's signal level dependent. Link to comment Share on other sites More sharing options...
oldtexasdog 2,362 Posted November 17, 2011 Share Posted November 17, 2011 Thanks Rich-Shoot I got all excited that Poor meant Cheap no getting around it I guess-you get what you pay for. Link to comment Share on other sites More sharing options...
schultdw 393 Posted November 17, 2011 Share Posted November 17, 2011 Rather, my understanding is that a 'rich' (well-behaved) filter should eliminate all aliasing artifacts, which means a steep slope filter, which means time smear. One of the neat things about digital filters is that you can design one with linear phase. All frequencies are delayed equally so there is no "time smear". Although the article never defined that term so I can only assume what they meant. Link to comment Share on other sites More sharing options...
RichP714 3,151 Posted November 17, 2011 Author Share Posted November 17, 2011 Rather, my understanding is that a 'rich' (well-behaved) filter should eliminate all aliasing artifacts, which means a steep slope filter, which means time smear. One of the neat things about digital filters is that you can design one with linear phase. All frequencies are delayed equally so there is no "time smear". Although the article never defined that term so I can only assume what they meant. That seemed odd to me too; I can see a steep slope analog filter causing phase artifacts, but thought that digital filters weren't susceptible Link to comment Share on other sites More sharing options...
RichP714 3,151 Posted November 17, 2011 Author Share Posted November 17, 2011 Rather, my understanding is that a 'rich' (well-behaved) filter should eliminate all aliasing artifacts, which means a steep slope filter, which means time smear. One of the neat things about digital filters is that you can design one with linear phase. All frequencies are delayed equally so there is no "time smear". Although the article never defined that term so I can only assume what they meant. That seemed odd to me too; I can see a steep slope analog filter causing phase artifacts, but thought that digital filters weren't susceptible After reading the article again. I'm getting two vibes 1. They believe that the 'time smearing' aspect of upsampling (the theory that the gentle slope filter has less phase artifacts as a reason for improved fidelity) is a red herring, and NOT responsible for the alleged increase in fidelity 2. I agree, they should have included a passage indicating that digital filters weren't susceptible to phase anomalies (as are anlog filters) regardless of slope. Perhaps they omitted this detail in the belief that it would obfuscate the issue (e.g. they had already addressed their belief that any possible reduction in 'time smear' wasn't the issue) The gist of the article (IMO) is the belief that an increase in fidelity is felt not through a reduction in phase anomalies, but in a purposeful permission of high frequency content past the anti-aliasing filters (through use of a shallow slope). They believe that this high frequency content is un-correlated to the audio signal in so much as it can be considered random, and behaves as a sort of dither. Finally, they appear to believe that, because of this, the entire upsampling DAC can behave as several lower bit DACs in parallel, with their outputs integrated, to reduce differential non-linearity? Link to comment Share on other sites More sharing options...
schultdw 393 Posted November 17, 2011 Share Posted November 17, 2011 That seemed odd to me too; I can see a steep slope analog filter causing phase artifacts, but thought that digital filters weren't susceptible Not all digital filters are linear phase. Analog filters are never linear phase. 1 Link to comment Share on other sites More sharing options...
RichP714 3,151 Posted November 17, 2011 Author Share Posted November 17, 2011 That seemed odd to me too; I can see a steep slope analog filter causing phase artifacts, but thought that digital filters weren't susceptible Not all digital filters are linear phase. Analog filters are never linear phase. What is your opinion of the article's premise? Link to comment Share on other sites More sharing options...
schultdw 393 Posted November 17, 2011 Share Posted November 17, 2011 >What is your opinion of the article's premise? The more I think about it the more I think it is slightly better than a fetid dingo's kidney. :-) Link to comment Share on other sites More sharing options...
RichP714 3,151 Posted November 17, 2011 Author Share Posted November 17, 2011 >What is your opinion of the article's premise? The more I think about it the more I think it is slightly better than a fetid dingo's kidney. :-) I'm not familiar at all with up-sampling; I used to pay attention to these types of things (back in the day, before the anti-aliasing filters became digital(oversampling, then single bit, etc.)) I have no experience with up-sampled playback, so I can't say whether it sounds better or not from any personal experience. I've heard many theories that the noise floor is thrown lower because all the 'unused' bits are in the dirt that the filter can be more shallow and this article, which implies that the presence of ultra-sonic content is a good thing I've read that DSD (SACD native encoding) has ultra-sonic content also, and some prefer it to DVD-A, even though DVD-A has greater bandwidth/sampling rates, and is usually more quiet than DSD Perhaps there's a connection..... Link to comment Share on other sites More sharing options...
Recommended Posts
Create an account or sign in to comment
You need to be a member in order to leave a comment
Create an account
Sign up for a new account in our community. It's easy!
Register a new accountSign in
Already have an account? Sign in here.
Sign In Now