Yeah, me too - maybe lower. But if we want to test differences up high, then files limited to 15KHz aren't the best. We can do 22KHz easily, so as not to affect what we are testing.
Whether you or I would hear the difference is another matter.
Thanks, guys for the suggestions and spectral plots. There were issues with sampling rate and the 20k peak is also erroneous in that the filter seems to alias downwards and increase lower frequencies as you get to the band edge.
I've redone the samples with a little more care this time. I've bumped to a +6 Q of 4 (previously Q of 2) at 10, 14, and 18KHz (which doesn't seem to alias). Saved them as 256k mp3 (still need Mp3 to keep the sample smaller) and confirmed that the mp3 is full bandwidth and gives the same spectrum as the orriginal. Also included a 14k dip for Simon.
Note that in Cool Edit you can grab a slider and move the filter in real time. It has a distinct pitch at any frequency, and for me I can clearly hear the pitch up to about 11kHz. A little above that the brightness remains but I don't sense the pitch of the peak. On sine wave my ears are good to at least 14kHz, perhaps 16k if I crank the level a bit.
I can't hear the 18k version, but I'm pretty sure I hear the 14k one. As before the "effect" is off for the first second, on for the second (1 to 2 sec), etc. Note that a 6dB peak at 14k reminds me of a lot of old Shure cartridges I used to own, at least curve wise.
Once again, these are mislabeled as .zip files. Save them to your desktop, change the .zip to .mp3 and play them back.
I don't want to hear any grief about mp3 files. ITS PINK NOISE!
Enjoy,
David
I've redone the samples with a little more care this time. I've bumped to a +6 Q of 4 (previously Q of 2) at 10, 14, and 18KHz (which doesn't seem to alias). Saved them as 256k mp3 (still need Mp3 to keep the sample smaller) and confirmed that the mp3 is full bandwidth and gives the same spectrum as the orriginal. Also included a 14k dip for Simon.
Note that in Cool Edit you can grab a slider and move the filter in real time. It has a distinct pitch at any frequency, and for me I can clearly hear the pitch up to about 11kHz. A little above that the brightness remains but I don't sense the pitch of the peak. On sine wave my ears are good to at least 14kHz, perhaps 16k if I crank the level a bit.
I can't hear the 18k version, but I'm pretty sure I hear the 14k one. As before the "effect" is off for the first second, on for the second (1 to 2 sec), etc. Note that a 6dB peak at 14k reminds me of a lot of old Shure cartridges I used to own, at least curve wise.
Once again, these are mislabeled as .zip files. Save them to your desktop, change the .zip to .mp3 and play them back.
I don't want to hear any grief about mp3 files. ITS PINK NOISE!
Enjoy,
David
The poster quoted 17,600 and 17,640.
All together now: "seventeen thousand six hundred and forty, take away seventeen thousand six...."
David
Sounds like the PEQ algorithm in the program could have been done a bit better... one way to work around that would probably be to do the initial generation of pink noise and the PEQ at a much higher sample rate like 96Khz, then resample it, but then you have to worry about whether the resampling algorithm is good or not...
Ah, thank you kind sirI've redone the samples with a little more care this time. I've bumped to a +6 Q of 4 (previously Q of 2) at 10, 14, and 18KHz (which doesn't seem to alias). Saved them as 256k mp3 (still need Mp3 to keep the sample smaller) and confirmed that the mp3 is full bandwidth and gives the same spectrum as the orriginal. Also included a 14k dip for Simon.
Well here are my results for the test, listening on my normal system which includes a ribbon tweeter, and with hearing that goes up to 17.5Khz and then drops like a stone. Scale of zero to ten, zero being no change, one being barely detectable with 10 being "are you kidding me, that's so obvious".
10Khz +6dB: Blatantly obvious (10)
14Khz +6dB: Still quite obvious (8)
14Khz -6dB: A little bit less noticeable than +6 but still easy to discern (6)
18Khz +6dB: Subtle but not difficult to hear (3)
I'm surprised I can still hear the change at 18Khz subtly but distinctly when my hearing doesn't go that high, obviously the peaking filter has width that extends down into my hearing range and that is enough for me to notice it right at the extremity.
Of all of those the only one I found subtle was 18Khz, the others were pretty obvious right away. Bearing in mind of course that its a fairly large 6dB change we're testing, it might be a lot harder to notice the higher ones with 3dB or less.
Last edited:
I'd rank them in the same order you posted them. I found all to be less objectionable than the earlier 20k version. I found the last three to be easier than before to not focus on.
I'm not sure that this kind of distortion wouldn't colour the sound though.
I should also probably forget about 20k in general. Although I proved to myself yesterday that I can still hear it, the cat hasn't been quite himself since
I'm not sure that this kind of distortion wouldn't colour the sound though.
I should also probably forget about 20k in general. Although I proved to myself yesterday that I can still hear it, the cat hasn't been quite himself since
Well, leaving aside the question of whether phase shift on its own is even audible in the first place, (eg does an all pass response sound the same as the original provided that the phase shift is smooth and gradual so as not to cause a huge spike in the group delay) a true pink noise source has phase relationships between the different constituent frequency components which are both random, and randomly changing in time.
So whether there was phase shift introduced or not you are going from one completely random set of phase relationships to another set of completely random and indistinguishable phase relationships which are also changing in time as the signal plays.
So I would posit that its purely the amplitude response change that we are perceiving in this test, and that pink noise is an ideal test signal to single out the effects of just the amplitude response by itself...
Last edited:
I think that's kind of a separate, second question. The first question is can we actually hear any change at all under test signal conditions like pink noise with various response changes added. If we can't then we can probably conclude that it would also be inaudible on music, at least for that particular listener with their hearing ability and the speakers they tested on.
This helps give us a feel for what things might influence the perceived sound.
However if we can hear a difference the next logical question is does it matter in music and what would it do to the sound of the music. That is a much harder question to answer and a much more subjective one which there would probably never be agreement to, even amongst those who can hear a difference.
My own impression is that small response errors in the 10-20Khz range (considerably smaller than those in this test) don't necessarily sound like a tonal imbalance, but have noticeable effects on the imaging. In fact I would extend that remark to lower frequencies as well - small frequency response errors that are small enough to not sound like a tonal imbalance can often still affect the imaging significantly, and I suspect that these small lumps and bumps in the frequency response that we take for granted in a non-dsp design account for a large part of the difference in perceived imaging between one speaker and another. (The other major factors probably being directivity and diffraction)
Last edited:
Can you hear a distinct change in the sound of the noise which lasts for the whole one second interval, or are you perhaps just hearing a small click at the transition due to a switching transient ?
I'm assuming the test signal has an abrupt switch between the two states rather than being ramped more slowly, which would generate some frequencies within the lower audible range during the transition.
According to research done by Heerens & de Ru (search here or google) the physical part (well before any neuronal action) of our hearing is INSANELY non-linear, the SPL is differentiated and then... get that, squared(!!). Their approach is a real eye-opener (well, more like an ear-opener) to quite a lot of phenomena. Any speaker or audio designer should read it -- for example it backs up nicely the empirical findings of Earl on transducer and amplifier distortion audibilty.
Is this the paper you're thinking of ?
http://igitur-archive.library.uu.nl/med/2011-0204-200555/Book Heerens de Ru EN.pdf
http://igitur-archive.library.uu.nl/med/2011-0204-200555/Book Heerens de Ru EN.pdf
AArggh..
My HF driver evaluation write up is now going to take a few more days.
I had finished the testing I planned to do, including hundreds of magnitude, RTA and recordings of dual tones and music at various levels, took the drivers off the horns, and put away the test set up that had occupied most of my time for the last couple weeks.
Now this discussion made me think of another test that may explain some of the differences that are clearly audible in the high SPL music test recordings.
When two tones are presented to a HF compression driver, a subharmonic is generated at a frequency that is the difference between the two tones, for instance 2093 & 3729 Hz tones produce a distortion component at 1636 Hz as I noticed looking at the RTA screens, exactly as you and Earl had mentioned.
The 1636 Hz distortion tone is far below the other tones in level, and does not seem to present an audible problem, the upper odd order harmonics are far more objectionable.
However, a subharmonic distortion component of 16kHz and 20 kHz (both frequencies which are inaudible to me and others of a mature age :^) would be at 4K, where ears (without noise induced hearing loss) are most sensitive.
So I now must drag out the test gear again and listen and hear what the various drivers do in this regard before sending the evaluation units back.
Since some of the drivers are real "bat killers" with decent response to around 24 kHz, I will also test if beat frequencies from two ultrasonic tones reproduced by two discreet sources create an audible beat frequency.
I'll post that result in this thread proving whether it is a "canard" or not
.The rest of the compression driver evaluation will be posted in a new thread, hopefully no later than mid week.
Art Welter
I hear a difference in the noise for the whole second. Didn't pay attention to the transition, but it's probably audible too.
I'm just surprised because I can't hear pure tones that high, and I can't hear noise that is very steeply high passed above about 15K. So I don't know why I can hear this. Does having the low end content help me hear the high end? Or is it something else?
I've done tone test (for other reasons) with noise behind it about 30dB down. It helps me identify the tone better than just a tone alone.
The LF sound effects that accompany video seem to deliberately use "beat" to enhance the apparent bass heard.
The German aircraft that overflew Britain during WW2 had deliberately miss-set engine rpm that created a "beat" that made the bombers more audible.
I suspect this "beat" thing is not just an LF effect that our ears detect. I would expect our ears to detect "beat" at most other frequency ranges.
It's back to that comparison thing I mentioned earlier. We are good at comparing, not so good at absolute.
The German aircraft that overflew Britain during WW2 had deliberately miss-set engine rpm that created a "beat" that made the bombers more audible.
I suspect this "beat" thing is not just an LF effect that our ears detect. I would expect our ears to detect "beat" at most other frequency ranges.
It's back to that comparison thing I mentioned earlier. We are good at comparing, not so good at absolute.
I just checked out a few blind(!) pink noise tests these minutes (with Audition/CoolEdit), speakers with AMT tweeters that run without issues to beyond 30kHz.
1dB, Q=4, 16kHz : clearly audible in the difference with gapless A/B-switching, it doesn't matter for the relative change whether the EQ is +1dB or -1dB. It takes about 0.5 seconds for the change to notice but after 3 seconds or so the perceived spectrum returns to "flat" again. No chance to detect the change when there is a gap of some seconds, in this quick test.
6dB, Q=6, 19kHz : clearly audible as well.
My hearing (at 46yrs) starts to fall off at 16kHz..17kHz with sines for normal listening levels (80dBSPL avg), above that it is a question of level up to ~21k
My conclusion is that I hear dominantly the lower skirt of the bell curve for the 19k EQ, as 1dB difference @ 16k was detectable.
1dB, Q=4, 16kHz : clearly audible in the difference with gapless A/B-switching, it doesn't matter for the relative change whether the EQ is +1dB or -1dB. It takes about 0.5 seconds for the change to notice but after 3 seconds or so the perceived spectrum returns to "flat" again. No chance to detect the change when there is a gap of some seconds, in this quick test.
6dB, Q=6, 19kHz : clearly audible as well.
My hearing (at 46yrs) starts to fall off at 16kHz..17kHz with sines for normal listening levels (80dBSPL avg), above that it is a question of level up to ~21k
My conclusion is that I hear dominantly the lower skirt of the bell curve for the 19k EQ, as 1dB difference @ 16k was detectable.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.