Skip to main content
Topic: INVITATION: Linear Phase vs. Minimum Phase Listening Test (Read 5038 times) previous topic - next topic
0 Members and 1 Guest are viewing this topic.

INVITATION: Linear Phase vs. Minimum Phase Listening Test

Hey guys. Just want to see if I can get some data on whether folks are able to detect linear phase vs. minimum phase upsampling digital filters using default SoX settings.

Potentially of value when DAC manufacturers discuss and show images of the "importance" of the lack of pre-ringing using impulse response images.

As usual, data collected anonymously. Naturalistic survey of what "audiophiles" hear.

44kHz sample converted to 176kHz so need high-resolution DAC to play back the samples, of course.

Instructions and test samples here:
http://archimago.blogspot.com/2015/04/inte...vs-minimum.html

Thanks in advance to the participants.

INVITATION: Linear Phase vs. Minimum Phase Listening Test

Reply #1
Why upsampling?  Isn't this the less interesting (and much easier) case vs. downsampling?

INVITATION: Linear Phase vs. Minimum Phase Listening Test

Reply #2
This way you can rule out the filter in the DAC at 20-22kHz and have the same amount of ringing at these frequencies for all testers. Since there is no music content above 22kHz you can also rule out IM of some tweeter plays a role. I wonder if this is the much better attempt to test the audibility of a filter as some famous study tried.
Is troll-adiposity coming from feederism?
With 24bit music you can listen to silence much louder!

INVITATION: Linear Phase vs. Minimum Phase Listening Test

Reply #3
This way you can rule out the filter in the DAC at 20-22kHz and have the same amount of ringing at these frequencies for all testers. Since there is no music content above 22kHz you can also rule out IM of some tweeter plays a role. I wonder if this is the much better attempt to test the audibility of a filter as some famous study tried.


But an upsampling filter only has to reject imaging, whereas a downsampling filter has to reject aliasing.  Because imaging is much harder to hear, the design of the two filters can be quite different, and ringing is much less likely to be at audible frequencies.    What do the transition bands on these two options actually look like?

INVITATION: Linear Phase vs. Minimum Phase Listening Test

Reply #4
To my understanding it uses a typical SoX settings people would use for upsampling like at playback in foobar. After the test has finished i guess we get the details. These samples have hefty pre-ringing against the minimum phase magic.
Most likely the ringing distribution looks like that.
picture at Slimdevices forum
I guess the SoX downsampling ringing would be different but at least alike in shape.
Is troll-adiposity coming from feederism?
With 24bit music you can listen to silence much louder!

INVITATION: Linear Phase vs. Minimum Phase Listening Test

Reply #5
To my understanding it uses a typical SoX settings people would use for upsampling like at playback in foobar. After the test has finished i guess we get the details.


Ok so the VHQ settings are a very steep filter at 22kHz with very nearly no imaging allowed.  That makes a little more sense in terms of the test (although I would not personally use a filter like this since it would have way more ringing then would make sense). 

Still, this basically reduces to a test to see if very long minimum vs. linear phase ringing at ~22kHz is audible.  The resampling step seems uncessary, the whole test could have been done at 96kHz, or maybe even 48k with constant rate.

INVITATION: Linear Phase vs. Minimum Phase Listening Test

Reply #6
Ok so the VHQ settings are a very steep filter at 22kHz with very nearly no imaging allowed.  That makes a little more sense in terms of the test (although I would not personally use a filter like this since it would have way more ringing then would make sense). 

Still, this basically reduces to a test to see if very long minimum vs. linear phase ringing at ~22kHz is audible.  The resampling step seems uncessary, the whole test could have been done at 96kHz, or maybe even 48k with constant rate.

I wouldn't use such a steep filter either but until lately we even had such recommendations here at HA. It is good as having very strong ringing to test here. If this gives already no clear result what does?
88.2kHz would be enough but many people claiming the superiority of minimum phase are likely to only accept 192kHz already and hopefully accept these 176kHz as enough so they can't blame a to low samplerate.
Is troll-adiposity coming from feederism?
With 24bit music you can listen to silence much louder!

INVITATION: Linear Phase vs. Minimum Phase Listening Test

Reply #7
If this gives already no clear result what does?


The downside I see is that the ringing will be somewhat higher frequency then a lot of filters, so there is less likelihood of audibility.  I am quite certain that I cannot hear infinitely long ringing at 22 kHz because I've tried test tones that high and they cut out around 18 kHz for me

INVITATION: Linear Phase vs. Minimum Phase Listening Test

Reply #8
Hey guys. Just want to see if I can get some data on whether folks are able to detect linear phase vs. minimum phase upsampling digital filters using default SoX settings.

Potentially of value when DAC manufacturers discuss and show images of the "importance" of the lack of pre-ringing using impulse response images.

As usual, data collected anonymously. Naturalistic survey of what "audiophiles" hear.

44kHz sample converted to 176kHz so need high-resolution DAC to play back the samples, of course.

Instructions and test samples here:
http://archimago.blogspot.com/2015/04/inte...vs-minimum.html

Thanks in advance to the participants.


It appears to be a preference test without a qualifying same/diffence test.


 

INVITATION: Linear Phase vs. Minimum Phase Listening Test

Reply #10
http://www.hydrogenaud.io/forums/index.php...24&hl=phase

I only saw 2 positive responses to my maximum phase filter at 20kHz test above.

Still this test may have these positives due to the hardware changes things in the audible range because of the content above the filter.
One more advantage in Archimagos test is people can't cheat easily with some spectral analyser because the samples look very similar.
Is troll-adiposity coming from feederism?
With 24bit music you can listen to silence much louder!

 
SimplePortal 1.0.0 RC1 © 2008-2019