Recommended Song Snippets

[Amphiprion]

Hi,

I will be posting some work on my website which will be of great interest to speaker DIY'ers. I need recommendations for song snippets (10-15 seconds) of high quality reference recordings. I would like keep it to one song snippet each per genre. I'll obtain the CDs if I don't already have them.

So what is your favorite 10-15 seconds of your favorite song? Well-known audiophile favorites are preferred (Krall's "Peel me a Grape" or Miles Davis "Kind of Blue" for example).

Thanks!

[cjd]

Audiophile "favorites" don't tend to be particularly difficult, but instead are already over-processed to be warm and fuzzy. At least, that's my experience.

I'll try to remember this and toss together a collection I find particularly revealing.

C

[Amphiprion]

Just in case it helps, I'm doing phase shift audibility tests. I'm creating several all-pass filters in Matlab. Each all-pass filter will have a different phase response (2nd order L-R, 4th order L-R, 8th order L-R, 16th order L-R, etc). Frequency response, phase response, impulse response, and pole/zero plots will be posted for each filter, but the songs will NOT be identified as to which is which. I'd like 10-15 second snippets so people can listen to them back-to-back quickly, as well as to keep my website bandwidth within reason since I will be leaving these in .wav format for purity.

My Matlab code is appended in case anyone wants to check it for errors. I had to rename the .m file to a .txt so the board would accept the upload.

[cjd]

I can tell you one album i personally use a lot: Hillary Hahn playing Bach unaccompanied partitas and sonatas. But I'm a violin player. Single instrument though, and some passages run from ~200Hz up to ~3kHz. And many of the things i listen for are instrument character, not something most folks have an ear for.

Mahler's 8th symphony has some absolutely brutal passages with so many layers of instruments, finding a system that can keep them distinct is surprisingly more difficult than I would have thought. That should take a little more than 10-15 seconds though.

C

[don9146]

What are you trying to show? There are some passages from "The Complex" by The Blue Man Group that will show weaknesses in the bottom end of any speaker and/or sub (notably track #3 "Sing Along"). I like using this track since it shows off a more "musical" bottom end instead of using a big movie explosion. Many of the other tracks on the disc show off excellent stereo imaging and subtle details that crappy speakers can't reveal. Its about my favorite recording from recent years.

Later,
Don

[Amphiprion]

What are you trying to show?

Whether or not crossover-induced phase distortion is audible. I am making digital filters in Matlab that have flat frequency response, but the phase response of a typical crossover filter (such as a 4th order Linkwitz-Riley at 2kHz simulating a typical two-way speaker crossover). By filtering songs through these all-pass filters and listening to them with headphones, one can determine for one's self whether or not crossover-induced phase distortion is audible, and if so at what point it sets in.

[minkuni]

I read a similar report in one of this years AES journals. The experiment was performed at a university in Belgrade if I am not mistaken. The conclusion from the experiment was that the phase distortion was not audible, however I do not think they tried to see how much phase distortion was necessary for it to be audible. The subjects were university students and staff, with no specification of "audiophileness". It has been a while since I read the report and unfortunately I do not have the journal here.

Interesting read on phase distortion at Audioholics

[cjd]

The Bach I mentioned may actually be a good option, if you run it mono. It's actually recorded in stereo with some interesting shifting in a few spots. I first noted this on headphones, and only on my better speakers was it audible at home also. So that could be yet a different interesting test. I can find the specific spot in the track if that's of interest.

[tubbsbright]

How about some Miley Cyrus?...if your speakers can make that trash sound good I think you're off to a great start.

[Hdale85]

There is a drum solo in this Charlie Byrd self titled cd on track #2 that's just awesome If your looking for a drum solo. Great quality recording too.

[Gerald Jansen]

off-axis response matters

Whether or not crossover-induced phase distortion is audible. I am making digital filters in Matlab that have flat frequency response, but the phase response of a typical crossover filter (such as a 4th order Linkwitz-Riley at 2kHz simulating a typical two-way speaker crossover). By filtering songs through these all-pass filters and listening to them with headphones, one can determine for one's self whether or not crossover-induced phase distortion is audible, and if so at what point it sets in.

Dear Mark,

I encourage your effort to show the (in)audibility of crossover-induced phase distortion. Do i understand it correctly, that you're primarily focusing on the direct sound (wave)? How about the power response into a real-live acoustic enclosure (if it matters - I think so)? Wouldn't that eventually take a Dolby Headphone or other HRTF processing to hear differences?

Kind regards,
Gerald

[Amphiprion]

Hi Gerald,

Thanks for the encouragement. Yes, I am only focusing on the direct sound wave - I'm intending for these song snippets to be listened to on headphones. Some of the existing work done already on phase audibility found that it was easier to detect phase distortion with headphones than with loudspeakers, plus doing anything beyond that (like the more involved techniques you mention) is a bit beyond what I intended for this small experiment.

I'm not intending to address the issue entirely - if I were I would save that for an AES journal publication This is just meant for those inquisitive souls as something to add to the general "internet knowledge" and give folks an accessible first-hand experience at listening to the possible differences in phase distortions.

Right now I'm having trouble getting Matlab to give acceptable results beyond 10th order filters using the invfreqz() command. Up through 10th order my all-pass creations yield less than +/-0.01dB variation, but when I try 12th or higher I get as much as +5/-10dB variation.

The method I am using to obtain the all-pass filters is a bit roundabout, but it lets me quickly change filter characteristics with the built-in functions and use invfreqz() to create my filter. It's just failing when it gets to 12th order or more. Probably getting poles to close to the unit circle, but still - this is 32 bit floating point math. I would think it should work. Up through 10th order the pole/zero plots look like nice perfect reciprocals of the pole/zero pair radii. When I hit 12th order, it doesn't look anything like it should.

I've also decided to just pick songs that I like. You guys will be getting some Miles Davis, Natalie Merchant, Pink Floyd, and (acoustic) Alice in Chains.

[---k---]

I like Jack Johnson's In Between Dreams CD. Good mix. Well recorded.