Thursday 16 July 2015

The Linear vs. Minimum Phase Upsampling Filters Test [Part III]: SUBJECTIVE IMPRESSIONS

This is part 3 of the results from the test of digital filters which began with Part I (RESULTS) and follows Part II (ANALYSIS).

As usual with these tests, I provided the respondents with an opportunity to describe what they heard when comparing the two samples (linear phase vs. minimum phase upsampling). In this portion I will collate the responses into what people described as the sonic difference and try to classify the responses to the type of upsampling setting for each of the samples... By doing this, I think one can get a sense of whether there is any subjective consistency and also an idea of the variability of subjective opinion. I have put my editorial comments in brackets and italicized.

I. "Mandolin" sample:

Linear phase impressions:
- I would guess its mostly placebo, but i think (this one had) a bit more/better presence. I would describe it a bit like having a bad quality mp3 compared to a better one.
- more diffuse and smeared, with less resolution
- high frequency transients were more noticeable
- The key difference I heard was in the attack / initial transients. It was very audible in the "pluck" of the mandolin and guitar, audible to a lesser extent in the piano and cymbals. For example, once excerpt sounded "snappy", with very precise / quick initial transients, while the other track sounded more "rounded". I assumed (perhaps incorrectly) that the more "snappy" sample was the minimum phase and the more "rounded" sample was linear. [This person preferred linear phase in 2/3, preferred minimum phase with "Give It Up".]
- sharper, clearer
- the (linear phase version) seemed to easier differentiate (instruments) from one another. (the minimum phase version) squashed them together and made their individual contributions to the ensembles performance harder to hear
- More realistic.
- Spatial separation of the music and more precise instrument placement.
- (Linear) sounded warmer with less 'zing' in the mid - upper frequencies.

Minimum phase impressions:
- a bit bright
- more natural sound
- more 'breath'
much more clearer and doesn't sounded "dirty". also the volume at the end of each note goes way more down and raised faster at the beginning of the next one.
- The mandolin (minimum phase) was more expansive in its sound and seemed more lifelike.
- Slightly more natural less processed sound
- just sounded better
- In Mandolin (minimum phase), the violins sound more natural. The reflections of the environment are more full. (Linear) is a shallower presentation. There's an edge to the violins in (linear) that is unpleasing. However, I somewhat dislike all of the reverb/reflections in (minimum phase) (even though they're more natural sounding).
- More dynamic.
- Less bright / harsh

II. "GrandPiano" sample:

Linear phase impressions:
- a bit bright
- sounded like it was recorded in a room with harder walls, but it did a better job of placing me amongst the instruments.
- deeper soundstage and better separation of the individual instruments. It sounded as if a veil was lifted off the music. . This was best audible with the GrandPiano piece.
- Being very sensitive to timing, I find in most cases linear phase sounds a little smeared and not sounding as alive, but linear phase does sound most natural regards frequency response and soundstage height. Minimum phase, to my ears, better with transients, timing and sound-stage depth. I'm quite confident the B group of all samples has Minimum phase filter applied. [This person despite believing he prefers minimum phase filtering, actually chose linear phase 2/3 times as the better sounding.]
- just sounds more like a real piano. (Minimum) has a flat presentation... It was easiest to tell which sounded most lifelike (having been in the presence of a real piano).

Minimum phase impressions:
- more natural sound to high notes - less tinny
- sounded bit more relaxed to me, also a bit slower and a bit not so tight in the bass.
- sounded "strange"
- I'm not sure at all, but thought there may be small differences in texture, dynamics and sound stage. I listened to each pair several times and in the end chose the one I preferred. [This person preferred 2/3 minimum phase.]
- more reverberant body to the piano and a more "liquid sound". I don't which might have sounded more real, i picked a flavor I liked.
- transient attack (better)

III. "Give It Up To Love" sample:

Linear phase impressions:
- seemed like it was missing something in the high freq spectrum (compared to minimum phase).
- sounded a bit more tight in the bass, but less sound stage and more glued to the speakers
- more correct in pitch
- More natural highs and more details in the singer's voice.
- the detail of the drum / percussion seemed a little crisper in (linear phase sample)

Minimum phase impressions:
- drum strikes less tinny
- For the third sample (Give it up to Love) I "thought" A [minimum] was more open than B [linear]. I failed the trial though as "guessing". [This person tried ABX and failed the trial for consistency - good try!]
- Sam sounded more like a real human being singing is the best way I can describe the difference I heard.
- Sound was very similar on all. Sam McClain was more distinctive, perhaps because I am familiar with that song. Images were more distinct and focused. Less of that extended "shimmering" sound in the high frequencies.
- More natural.
- Better transient attack and decay
- There's a harshness to the hi-hat in GIUTL (minimum) right from the beginning, and the vocals have a harsh edge. Cymbal strikes seem harsh and muddy.

I had other interesting comments like someone who did well with the 24-bit "heard nothing" in comparison here. Here's another detailed response:
The Sam McClain song was easier to identify IMHO. But, the piano sounded absolutely divine-the best piece- thanks to JR. I used two metrics to make my selections. One, I identified which sample caused 'irritation' to my ears. This is a tickling/tingling like feeling in the auricle that I am not able to identify clearly but is what I get when something is amiss in the high frequencies, for eg, a tweeter off axis or sibilance. Second, the samples I chose were those where I had a tough time focusing on the details and becoming more absorbed into the music. Having said that, both criteria are related because if I don't feel any ear irritation, it is easy for me to get absorbed into the music. Thank you for this. I had a great time and eagerly await the results.
This person preferred the linear phase filter in all 3 samples!

Here's another one from someone listening with their partner:
I don't have a strong feeling for naturalness in the recordings. that is, what I am liking hearing could be an artifact! I was surprised to always prefer b and went back through each pair for a relisten. I always felt b had a wider soundstage, better low end and especially in the third piece, better - more consistent - placement of sound, especially the vocals. on the second pass, my partner listened with me and also always preferred b, also felt the sound was fuller and spatially deeper. I've never seen her use a hand gesture indicating a spatial feature after listening until tonight - I thank you for conveying the idea of a soundstage with these lovely recordings! honest, I wasn't trying to coach her, I think this is what she actually heard
Remember, the linear filter was A/B/B so this tester and his partner thought the linear filter had a "wider soundstage" 2/3 times.

A number of listeners admitted there really wasn't much difference as expressed by these testers:
- I found the quality of the samples / music very high. If I heard any difference, which my imagination made me believe I very well may have, I was unable to define that difference.
- I doubt I'd be bothered about any of the differences such as they were, in fact, I'm pretty sure I wouldn't.



As you can see, from a subjective perspective, we have lots of opinions, and people do indeed hear and describe things differently... There are enough positions and impressions in these comments to selectively promote any position we wish. If I wanted to promote my new DAC with minimum phase filters, I could easily (and honestly) say with some confidence that "many listeners find our minimum phase upsampling algorithm more natural, improves the 'air' around instruments, improved texture, dynamics, soundstage, and better reproduces the room acoustics" just based on the comments in favour of the minimum phase setting for the "Mandolin" sample above. As we know last week, statistically, indeed there was a preference for minimum phase filtering, but not strongly significant based on statistical data.

I do believe it is worth remembering this when we read subjective reviews and comments in forum posts for example. Certainly, as it pertains to subjective preference for digital filters (even with as much ringing as the test here), it does seem that differences are likely subtle at best for most audio lovers even with what I believe is significantly above average audio equipment used in this cohort of 45 responses.


As usual, have a great week ahead everyone! Hope you're enjoying the music... Whether through linear or minimum phase filters :-).

BTW: Keep an eye out for the results of the Ars Technica / Randi Foundation "Million Dollar Challenge" this weekend (July 19) using AudioQuest Vodka (~$350?) vs. Amazon Basics Cat5 ethernet cables (~$5?)! Of course, readers of this blog know where I stand: MEASUREMENTS: Ethernet Cables and Audio...

The video they promise to post should be interesting in any case! I really do hope they do it right with good equipment, a nice treated sound room, low ambient noise, good speaker positioning, and an adequate sweetspot for the listener. Otherwise, it could get nasty with the spiritualistic audiophiles pointing fingers at "flawed tests", "poor statistics" and all manner of crying foul (objectivists could point fingers at poor design as well). Of course, at the end of the day, no matter how rigorous or properly set-up a listening test may be, I doubt some could ever accept even the possibility that there just isn't a difference. Beliefs immune to evidence are like that.

No comments:

Post a Comment