Sorry - I'm going to have a little go at you 😉
Harbeth is just another brand that likes to "tell a story" - they don't "eat" anything. At best, they "charm" someone.
Normal drivers are what? Above 2kHz - why?
Phase details and "rapidity" ? What's next - silky highs and tight bottoms?
Which tweeters suffer - and why at 1kHz? Waveguides? SPL? Tweeter type? Filter slope? Lots of designs work great with around 1100Hz.
Drives do not "sing", they move, they potentially distort and break-up and then "ring" - resonate - all drivers do, at some point - physics.
Expensive drivers do not sound better - necessarily - not at all - in any way. It is the design of the entire speaker that makes the difference.
There are top measuring and absolutely loverly speakers out there, filled with "standard" line tweeter from the likes of SB and SEAS. You do not need exotic material or snake-oil talk to make a good speaker, you need an overall good, well-thought-out complete design.
My apologies again - but you tickled my BS meter a bit
Its not bs, it's a fact about what happens when. You xo a driver bellow 3k, it reduces dramatically its accuracy,
We talk about why single cheap drivers (and small) sound better with more ease than wifi setup , that is why
These full range also compromise the low end which can easilly creap into the midrange with the harmonics,
I was just saying some addressed that and thos is how they did it
If you had these loudspeakers at home you probably didn't placed them properly or had low noise amps
Yes seas is exotic materials, there aluminum magnesium are quite advanced and designed , they are highly suited for bi-ampimg, or electronic xos, i like a lot seas sound,
We talk about why single cheap drivers (and small) sound better with more ease than wifi setup , that is why
These full range also compromise the low end which can easilly creap into the midrange with the harmonics,
I was just saying some addressed that and thos is how they did it
If you had these loudspeakers at home you probably didn't placed them properly or had low noise amps
Yes seas is exotic materials, there aluminum magnesium are quite advanced and designed , they are highly suited for bi-ampimg, or electronic xos, i like a lot seas sound,
Not necessarily irreversible in the future:
https://hms.harvard.edu/news/scientists-regenerate-hair-cells-enable-hearing
If you're feeling actual pain I would suggest you may have damage to your ears. Do you enjoy loud music? Perhaps you have some hearing loss? Tinnitus? I mention this because I have all the symptoms I've mentioned. I love loud music and it's taken it's toll on my ears. Loud music, above about 95db at my listening position causes pain after a short period from mid range up so I'm very careful how much I turn it up now. I used to foolishly listen at concert hall levels. Took about 30 years to happen. My cut off is about 12khz. I have tinnitus that thankfully I'm able to ignore.
Thought I would give everyone an update.
I ended up purchasing a graphic equalizer. Actually, it was a second graphic equalizer, as the first one was lost in shipping!
I have used it to tune down by 1.5 to 3 db the frequencies between about one kilohertz to 8 khz.
This has had a favorable impact, but there is some degradation of the quality of the signal due to the additional equipment in the signal path. Bottom line here: a small Plus.
I did some additional searching, and found out that I had not taken into consideration room acoustics. I live in a very small apartment, and the listening area is small (12x17) and full of a lot of hard randomly placed surfaces. I seem to have had better luck in improving the sound by positioning my speakers in a near field orientation - about a 5-ft equilateral triangle. This appears to minimize the impact of the items in my listening area.
Thanks for all the ideas that were suggested.
I ended up purchasing a graphic equalizer. Actually, it was a second graphic equalizer, as the first one was lost in shipping!
I have used it to tune down by 1.5 to 3 db the frequencies between about one kilohertz to 8 khz.
This has had a favorable impact, but there is some degradation of the quality of the signal due to the additional equipment in the signal path. Bottom line here: a small Plus.
I did some additional searching, and found out that I had not taken into consideration room acoustics. I live in a very small apartment, and the listening area is small (12x17) and full of a lot of hard randomly placed surfaces. I seem to have had better luck in improving the sound by positioning my speakers in a near field orientation - about a 5-ft equilateral triangle. This appears to minimize the impact of the items in my listening area.
Thanks for all the ideas that were suggested.
Vinyl has high distortion > 10Khz, may be you can actually hear that? A digital source will not have that distortion.
Did you check your upper hearing limit?
Digital EQ is far better these days. There are some good apps that can do parametric EQ
However, digital, especially sigma-delta modulators, may have bizarre correlated noise issues that affect sound perception in some humans. Such effects don't normally exist in purely analog amplifiers, which were mostly what were in use back in the day when most research into distortion audibility was done.
Can you give an example and quantify at what level? I can't imagine that this could be louder that the noise and distortion generated by Vinyl..
At the most extreme it can be louder than the loudest pop due to a bad scratch on a vinyl record. Or can be at any volume level less than that. And what's more, you may not even see it an a typical audio FFT. What looks like low level white noise on an FT can be a Dirac impulse in the time domain. IIRC the math explaining that was worked out back in the 1930s, maybe 1938?
In a less extreme case, some people were affected by a correlated noise below a level where the noise couldn't be clearly discerned, yet high enough to still influence perception in some people. Here I am using the example of ESS Hyperstream II modulator Hump distortion. Its not really a distortion in the same sense as weakly-nonlinear analog amplifier distortion. Its probably better classified as correlated noise, since it is extremely nonlinear, non-time invariant, and insidious in a way a analog amplifier HD/IMD could never be. It changes drastically with audio signal's more or less instantaneous amplitude.
Even for other types of dacs, and all dacs really, there are two critical analog signals that can pollute the audio signal with noise. They are the clock, and Vref. They are multiplicative with the dac audio output, so any noise on either one (and there always is some), intemodulates noise with the audio signal. The noise times audio signal intermodulation products are also sometimes called modulation sidebands. They are always moving around in amplitude and frequency because they are noise or noise-like signals. Thus they do not show up very well in FFTs, since the FT is an integral transform. FFTs do not integrate noise or other non-PSS signals over all time, but they do integrate or average those effects over the time the FFT dataset is acquired by an ADC. Since the effects are averaged, an FFT cannot show the peak volume level, only some average. How often do high level signals occur? It depends on the physical mechanism that is the source of the noise.
Moving along then, there is in theory an infinite number of types of noise and or of correlated noise. If you want to know more about how that could be, there are various things that have to be understood. Some if it is math, some is how sigma-delta modulators work, some is how human perception especially brain processing can be quite different in different people. (The brain learns what is auditory signal and what is noise. People talking in a crowded restaurant is noise, unless maybe the signal of interest is one person out of the crowd, maybe your friend trying talk to you from across the table. The brain has to filter out the noise to hear the signal. There is much more which could be said on the brain processing factor, but its too much for one post.)
In a less extreme case, some people were affected by a correlated noise below a level where the noise couldn't be clearly discerned, yet high enough to still influence perception in some people. Here I am using the example of ESS Hyperstream II modulator Hump distortion. Its not really a distortion in the same sense as weakly-nonlinear analog amplifier distortion. Its probably better classified as correlated noise, since it is extremely nonlinear, non-time invariant, and insidious in a way a analog amplifier HD/IMD could never be. It changes drastically with audio signal's more or less instantaneous amplitude.
Even for other types of dacs, and all dacs really, there are two critical analog signals that can pollute the audio signal with noise. They are the clock, and Vref. They are multiplicative with the dac audio output, so any noise on either one (and there always is some), intemodulates noise with the audio signal. The noise times audio signal intermodulation products are also sometimes called modulation sidebands. They are always moving around in amplitude and frequency because they are noise or noise-like signals. Thus they do not show up very well in FFTs, since the FT is an integral transform. FFTs do not integrate noise or other non-PSS signals over all time, but they do integrate or average those effects over the time the FFT dataset is acquired by an ADC. Since the effects are averaged, an FFT cannot show the peak volume level, only some average. How often do high level signals occur? It depends on the physical mechanism that is the source of the noise.
Moving along then, there is in theory an infinite number of types of noise and or of correlated noise. If you want to know more about how that could be, there are various things that have to be understood. Some if it is math, some is how sigma-delta modulators work, some is how human perception especially brain processing can be quite different in different people. (The brain learns what is auditory signal and what is noise. People talking in a crowded restaurant is noise, unless maybe the signal of interest is one person out of the crowd, maybe your friend trying talk to you from across the table. The brain has to filter out the noise to hear the signal. There is much more which could be said on the brain processing factor, but its too much for one post.)
Last edited:
Ok, if the time-domain difference is that big (like a full scale Dirac pulse) it should be easy to measure and seen in a difference between the two recordings isn't it?
Actually, the problems are more seen in dacs than in ADCs. That's likely for various reasons, including the requirements for a dac internal to an ADC as opposed to dac for final reconstruction. And the noises are not usually so extreme as a Dirac impulse (which would be the very worst case).
Mostly what can be measured in an FFT can only be seen in a very high res FFT. The artifacts are referred to as "noise skirts" which are seen at the base of a spectral line.
Another interesting example of the effects of noise that hardly anyone realizes is there is in standard dacs such as SMSL or Topping. Its simply that people have never heard a recording played back without that particular correlated noise. But now that's changing. There is a new dac from Millennia that is impressing some serious professionals. Around 160dB dynamic range, and that's for real. People are now hearing what low level signals are actually encoded on a good CD recording for the first time. https://www.diyaudio.com/community/threads/27bit-dac-162-db-dynamics.406498/post-7998112
Mostly what can be measured in an FFT can only be seen in a very high res FFT. The artifacts are referred to as "noise skirts" which are seen at the base of a spectral line.
Another interesting example of the effects of noise that hardly anyone realizes is there is in standard dacs such as SMSL or Topping. Its simply that people have never heard a recording played back without that particular correlated noise. But now that's changing. There is a new dac from Millennia that is impressing some serious professionals. Around 160dB dynamic range, and that's for real. People are now hearing what low level signals are actually encoded on a good CD recording for the first time. https://www.diyaudio.com/community/threads/27bit-dac-162-db-dynamics.406498/post-7998112
Last edited:
Perfectly fine. Many other people will feel similarly. At least you are willing to admit you have a strong expectation bias.
Depends on the particular proposition. For flat Earth theories, I am highly skeptical. For better dacs existing than most people have ever heard, I think I would be skeptical of a claim that it couldn't be possible. My reasoning is that we don't know how to measure the localization cues for soundstage, so it doesn't get measured. OTOH, what does get measured tends to get fixed over time. Even at 16-bits it could be possible that localization cues are not be reproduced as well as they could be. That combined with the fact that ITD localization in humans involves inter-channel phase coherence down to within a few microseconds, there is no guarantee right now that level of performance is being obtained in "standard" dacs at any number of bits. IOW, it reasonably remains within the realm of possibility that it could be done even at 16-bits. It is not ruled out by sampling theory.
Last edited:
According to multiple studies ITD sensitivity is in the order of 10us. With 96kHz sampling rate this would mean about 1 sample timing difference between channels. Can you name any DACs that have such timing differences between channels? Not to mention that such difference would be easily detected in measurements.
Its not simply a matter of theoretically perfect sample timing. It includes AN and PN combined. A sample point is defined in a Cartesian space in terms of amplitude and time. An error in either one or both makes the sample point in error. An FFT can measure averaged periodic errors of that type on the assumption that the error is PSS. It can't measure instantaneous non-PSS errors other than as shown in averaged noise skirts. There is no meaningful way to disentangle noise skirts back into inter-channel phase coherence at a dac's analog output.
Last edited:
