Forgot about those. Just finished my 130dB gain instrumentation preamp. The 1nV noise almost rails my A/D and no spurs above the noise over 700Hz on heavily averaged 65K FFT's. Haven't tried batteries yet but the line noise residue is pretty low on a cheap supply (5nV spur at 180Hz) I'm ready to begin my search.
I bet you can't wait.I used the THAT 1512 based pre-amp from one of our sponsors. At $79 complete I couldn't see spinning a board and sourcing all the parts. Very few mods were needed (It had a 20MHz whistle on the LM317 when not in their manifold due to probably some external bypassing or not using their external PS.)
Last edited:
Forgot about those. Just finished my 130dB gain instrumentation preamp. The 1nV noise almost rails my A/D and no spurs above the noise over 700Hz on heavily averaged 65K FFT's. Haven't tried batteries yet but the line noise residue is pretty low on a cheap supply (5nV spur at 180Hz) I'm ready to begin my search.
Scott,
Just since you don't remember I have done the microdiode copper wire experiment. I measured the distortion in a copper strip then heated it until lots of oxides formed, let it cool and the distortion stayed the same.
In mechanical contacts it turns out the issue is not gaps but sulfur contamination. The silver or copper sulfide causes the molecules to grow a bit and hold off the mechanical contact. Application of voltage if the growth is not too thick causes ion migration and restores the contact. This shows up as no conduction until enough current and voltage to restore the contact and then if left unused the sulfur ions grow back.
ES
Yes, I will repeat some of these experiments. The above explanation does not work for the sine wave plots with a dead zone at each crossing, and you should at least have a disclaimer on the speculative nature of this.
I would oxidize two bare wires and twist them together to see any effect and I STILL don't picture what you mean by the "distortion in a copper strip". Where is signal forced and where is voltage measured. I find it hard to believe a four wire measurement on a bar of copper shows any distortion.
Last edited:
Yes. Remember, if you pick 123 you've eliminated 456. As long as you don't care which group is which, those choices are degenerate.
Let me know and I will send you the actual part I used in that test. (Assuming I can still find it!)
I also have samples of the old 1N4004 diodes that made so much switching noise that people started to notice it!
I would like to reiterate what I know, and did not know about monaural phase over the decades.
Before I was born and then into the 1960's, most people did NOT believe that the human ear was monaural phase sensitive. This was called: 'Ohms Law of Acoustics' up till then, and was held in veneration by people like Paul Klipsch and many others who made multiple loudspeaker systems with large time differences between drivers.
You see: Scientists had a model of the human ear as some sort of resonance detecting device, and like equilizers based on the same principle, they do not give proper phase information. For over 100 years, people believed this.
Back in the late 1960's, Richard Heyser, then owner of a K-horn, (just like me) for ultimate reproduction, found it problematic, and he gave a paper on path length differences between loudspeaker drivers at the AES, and got an active protest, on the spot, by Paul Klipsch, who was in the audience, just like I was.
Then, a few years later, 'TIME ALIGN' became a common term. This meant trying to get the speaker drivers to line up acoustically, along with the corresponding crossover in order to get the arrival time closer to ideal. This is when the SLOPED CABINETS became popular, especially on 2 way speakers, attempting to align the drivers.
Many of us, including John Meyer, started making 'time or phase aligned' speaker systems, and much research was done in Japan on this as well. All that I can say is that once you live with a 'time aligned' speaker for a period of time, you may never want to go back to K-horns for most material.
About 1975, Manfred Schroeder, then a researcher at Bell Labs and elsewhere, wrote a paper in the IEEE 'Proceedings' that gave more recent research by Bell lab's, etc. and how 'Ohm's Law of Acoustics' was now superceeded by new results, BUT these results had relatively little ABSOLUTE predicability as to when we might hear a specific kind of phase shift and when we might not. This leads to confusion, if someone attempts to either PROVE that a particular phase shift is audible, or that a similar phase shift is inaudible.
Much of the reasoning for hearing monaural phase is based on IM byproducts made in the middle ear. However, TIME ARRIVAL of certain portions of the audio signal might be just as much or more important. What we were discussing, recently, was TIME ARRIVAL, not phase differences, and I still think that that is worth serious consideration still. (enough for now)
Before I was born and then into the 1960's, most people did NOT believe that the human ear was monaural phase sensitive. This was called: 'Ohms Law of Acoustics' up till then, and was held in veneration by people like Paul Klipsch and many others who made multiple loudspeaker systems with large time differences between drivers.
You see: Scientists had a model of the human ear as some sort of resonance detecting device, and like equilizers based on the same principle, they do not give proper phase information. For over 100 years, people believed this.
Back in the late 1960's, Richard Heyser, then owner of a K-horn, (just like me) for ultimate reproduction, found it problematic, and he gave a paper on path length differences between loudspeaker drivers at the AES, and got an active protest, on the spot, by Paul Klipsch, who was in the audience, just like I was.
Then, a few years later, 'TIME ALIGN' became a common term. This meant trying to get the speaker drivers to line up acoustically, along with the corresponding crossover in order to get the arrival time closer to ideal. This is when the SLOPED CABINETS became popular, especially on 2 way speakers, attempting to align the drivers.
Many of us, including John Meyer, started making 'time or phase aligned' speaker systems, and much research was done in Japan on this as well. All that I can say is that once you live with a 'time aligned' speaker for a period of time, you may never want to go back to K-horns for most material.
About 1975, Manfred Schroeder, then a researcher at Bell Labs and elsewhere, wrote a paper in the IEEE 'Proceedings' that gave more recent research by Bell lab's, etc. and how 'Ohm's Law of Acoustics' was now superceeded by new results, BUT these results had relatively little ABSOLUTE predicability as to when we might hear a specific kind of phase shift and when we might not. This leads to confusion, if someone attempts to either PROVE that a particular phase shift is audible, or that a similar phase shift is inaudible.
Much of the reasoning for hearing monaural phase is based on IM byproducts made in the middle ear. However, TIME ARRIVAL of certain portions of the audio signal might be just as much or more important. What we were discussing, recently, was TIME ARRIVAL, not phase differences, and I still think that that is worth serious consideration still. (enough for now)
John,
Time Align was as far as I remember a phrase coined by Urei, for a concentric speaker based on Altec Lansing stuff (the small horn through the middle of the bass driver). Urei's trick as far as I remember was to have crossovers with an all pass network to maintain absolute phase (what you call monaural phase). It could even reasonbly well reproduce square waves. I don't know if there is any passively filtered loudspeaker in production today that is truly time aligned in this way.
Later, the concept of 'time align' was hijacked to mean something else: to compensate for the different acoustic centres of the different drivers, by placing all acoustic centres on the same vertical axis. Subsequently, if the Xover maintains the same phase between the two drivers in the cross over region, this will place the dominant lobe into the direction of the listener.
However, this does not mean that absolute (or monaural) phase is preserved, 90 degrees of phase shift is added for each filter order. In a 4th order LR, for example, both drivers experience a 360 degree phase shift, but an equal one at all relevant frequencies.
Janneman's test is whether absolute phase can be detected by ear, an issue that is really undetermined as far as I am concerned. I tend to favour the theory that, within bounds, absolute phase cannot be perceived.
Therefore, I hope that the best ears on this site will be doing this listening test. It is no trick, and a null result is no shame. For me, it is about possibly falsifying a theory I adhere to, based on my own listening experiments from long ago.
In order to avoid misunderstanding: the ear is extremely sensitive to binaural phase differences. So, the two stereo channels should maintain phase identity to the highest possible degree. Otherwise, stereo image will suffer.
vac
Time Align was as far as I remember a phrase coined by Urei, for a concentric speaker based on Altec Lansing stuff (the small horn through the middle of the bass driver). Urei's trick as far as I remember was to have crossovers with an all pass network to maintain absolute phase (what you call monaural phase). It could even reasonbly well reproduce square waves. I don't know if there is any passively filtered loudspeaker in production today that is truly time aligned in this way.
Later, the concept of 'time align' was hijacked to mean something else: to compensate for the different acoustic centres of the different drivers, by placing all acoustic centres on the same vertical axis. Subsequently, if the Xover maintains the same phase between the two drivers in the cross over region, this will place the dominant lobe into the direction of the listener.
However, this does not mean that absolute (or monaural) phase is preserved, 90 degrees of phase shift is added for each filter order. In a 4th order LR, for example, both drivers experience a 360 degree phase shift, but an equal one at all relevant frequencies.
Janneman's test is whether absolute phase can be detected by ear, an issue that is really undetermined as far as I am concerned. I tend to favour the theory that, within bounds, absolute phase cannot be perceived.
Therefore, I hope that the best ears on this site will be doing this listening test. It is no trick, and a null result is no shame. For me, it is about possibly falsifying a theory I adhere to, based on my own listening experiments from long ago.
In order to avoid misunderstanding: the ear is extremely sensitive to binaural phase differences. So, the two stereo channels should maintain phase identity to the highest possible degree. Otherwise, stereo image will suffer.
vac
Last edited:
I don't quite understand how the question of human hearing and monaural phase relates to multi-driver speakers. Once you divide a mono signal with a crossover, it's not technically monaural any more. You then have two different signals being sent to two different drivers, and the phase relationships most certainly determine the tone of the signal heard by one (or two) ears.
While it might be interesting to investigate and confirm or disprove whether the human hearing system can detect phase in a monaural sound source, I don't think this has any relevance for speakers which employ a crossover. The original source signal might be monaural, but the output of the speaker is really multiple signals. Just because the desire is to reproduce a single channel does not mean you can ignore the fact that there are two (or more) completely separate transducers involved in moving the air. Phase most certainly plays a part when mixing multiple signals (bands) in the air.
While it might be interesting to investigate and confirm or disprove whether the human hearing system can detect phase in a monaural sound source, I don't think this has any relevance for speakers which employ a crossover. The original source signal might be monaural, but the output of the speaker is really multiple signals. Just because the desire is to reproduce a single channel does not mean you can ignore the fact that there are two (or more) completely separate transducers involved in moving the air. Phase most certainly plays a part when mixing multiple signals (bands) in the air.
Thorsten,
This is not a meaningless test, it actually only gets better. At stake is the question if the human ear is able to reliably pick up absolute phase. It might well be beyond human capacity, which I suspect.
So, on this forum, we have all these highly trained ears, like yours, in one nice coral. You can play these wav. files Janneman posted any way you like. If the best ears cannot pick up a difference, that would settle an important issue in loudspeaker design. But if they can reliably, that would point (at least me) into a new direction.
The way to do it with 6 files to be divided in 2 groups is just a way of getting statistically relevant results.
Vac
Well, I'm pretty sure I heard the difference in those tracks despite not being a golden ear and using a 5 way speaker system. When Prof. Hawksford provides the key, I'll at least know if it's something I need to worry about. I hope that he's amused by some of the weird paranoia that has been exhibited about a simple listening test...
Phase etc.
Forgive me if I am mis-understanding the discussion, but I think these two issues are somewhat different in operating mechanism.
The tests I have done and seen regarding multiple driver alignments has to do with the resulting merging of wavefronts and resulting comb filtering in the crossover regions due to the phase/time difference between the drivers. Within the Haas limit the ear perceives the algebraic sum of the two driver's wavefronts with corresponding frequency domain irregularities.
In my experience monaural phase issues are usually due to all-pass filtering or other frequency dependent delays and cause a different effect which is more like pulse smearing, where the fundamentals cease to correctly align with harmonics, and transient pulses are time stretched. This effect may or may not be accompanied by frequency domain irregularities.
Both are of understandably undesireable.
Just my $0.02 worth (or less)
Howie
Howard Hoyt
CE - WXYC-FM 89.3
UNC Chapel Hill, NC
www.wxyc.org
1st on the internet
Forgive me if I am mis-understanding the discussion, but I think these two issues are somewhat different in operating mechanism.
The tests I have done and seen regarding multiple driver alignments has to do with the resulting merging of wavefronts and resulting comb filtering in the crossover regions due to the phase/time difference between the drivers. Within the Haas limit the ear perceives the algebraic sum of the two driver's wavefronts with corresponding frequency domain irregularities.
In my experience monaural phase issues are usually due to all-pass filtering or other frequency dependent delays and cause a different effect which is more like pulse smearing, where the fundamentals cease to correctly align with harmonics, and transient pulses are time stretched. This effect may or may not be accompanied by frequency domain irregularities.
Both are of understandably undesireable.
Just my $0.02 worth (or less)
Howie
Howard Hoyt
CE - WXYC-FM 89.3
UNC Chapel Hill, NC
www.wxyc.org
1st on the internet
This is certainly a valid difficulty in making this test with multi-driver loudspeakers. All-pass networks are not "constant voltage" (the term used for constant group delay for summed response "on axis") except the special case of single pole filters. Any networks that aren't all pass have reverberant room total magnitude responses (sometimes called "power response") that aren't flat with frequency. And any networks that aren't constant voltage have varying group delay with frequency "on axis".
Baekgaard has shown a way, using an extra driver, to make a two pole constant voltage crossover with all pass characteristics, and G. R. Koonce has shown how to implement the same thing with state-variable filters without the extra driver. But I've never heard of the latter being done commercially.
Thanks,
Chris
Phase distortion
One of the main "disasters" for audio engineers, according to my view, is the Fourier Theorem; it's a quite complex argument, so it has attracted the interest of many. FFT tools and software are easily available today, so everybody use them and most of the discussion are based on their outputs.
But I don't believe they are really useful to understand the listening difference in the audio reproduction products. I stopped to use FFT as the main design tool, for my own diy realizations, from long time: it has been replaced by a good analog storage scope.
I want to highlight an interesting article come to my attention only today; inside it some discussions related to phase audibility are worthy of interest. It appeared on Audio magazine November 1961, nearly fifty year ago, but they are still very actual and neglected today!!
I completely agree with the theory of John Campbell (see my past posts): here the link Sineward distortion in High-Fidelity amplifiers.
One of the main "disasters" for audio engineers, according to my view, is the Fourier Theorem; it's a quite complex argument, so it has attracted the interest of many. FFT tools and software are easily available today, so everybody use them and most of the discussion are based on their outputs.
But I don't believe they are really useful to understand the listening difference in the audio reproduction products. I stopped to use FFT as the main design tool, for my own diy realizations, from long time: it has been replaced by a good analog storage scope.
I want to highlight an interesting article come to my attention only today; inside it some discussions related to phase audibility are worthy of interest. It appeared on Audio magazine November 1961, nearly fifty year ago, but they are still very actual and neglected today!!
I completely agree with the theory of John Campbell (see my past posts): here the link Sineward distortion in High-Fidelity amplifiers.
Is this the reason why Quad ESLs (with their multilple allpass delay line) sound the way they do?
Hi,
FFT has it's uses, but just like many other tests there is no inherent value in an FFT test result. It requires interpretation. Interpretation requires the ability to correlate the FFT output with listening experience, which one can only if one has listened to many types of systems AND looked at the FFT results.
I derive much more useful information from a set of FFT's made at low level, at operating level and at maximum level than I would from just having the THD at each point.
The real problem is that engineers and customers alike are attracted by single number figures of Merit (be it THD in amplifiers, horsepower or maximum speed in cars, accuracy of 100 Years in Watches), even though it should be clear to anyone with more than three braincells that the world does not work this.
But the engineer will be happy if I tell him "Make me an amplifier with 0.02% THD at full power and 20KHz (or generalised - make this needle of this instrument point at this number), yet if I ask him "make me an amplifier that sounds good" he will be asking me for a "good sound meter" and for the number I would like to see on it.
Equally, the customer may between two cars of the same price be tempted to buy the one with more power and speed (or more miles per gallon etc.), but on test drive may find the car with the worse numbers is the better car to drive...
So complex multidimensional analysis is unavoidable, but try to get a bunch of people to understand such or to agree on a "single number figure of merit" that is derived from such analysis.
Generally I find people prefer false certainties (the amplifier that has 200W at 0.01% THD is better than the Amplifier that 180W at 0.05% THD) to real uncertainties (without doing a serious listening comparison we cannot conclude if one amplifier is better for listening to music than the other or not).
Ciao T
FFT has it's uses, but just like many other tests there is no inherent value in an FFT test result. It requires interpretation. Interpretation requires the ability to correlate the FFT output with listening experience, which one can only if one has listened to many types of systems AND looked at the FFT results.
I derive much more useful information from a set of FFT's made at low level, at operating level and at maximum level than I would from just having the THD at each point.
The real problem is that engineers and customers alike are attracted by single number figures of Merit (be it THD in amplifiers, horsepower or maximum speed in cars, accuracy of 100 Years in Watches), even though it should be clear to anyone with more than three braincells that the world does not work this.
But the engineer will be happy if I tell him "Make me an amplifier with 0.02% THD at full power and 20KHz (or generalised - make this needle of this instrument point at this number), yet if I ask him "make me an amplifier that sounds good" he will be asking me for a "good sound meter" and for the number I would like to see on it.
Equally, the customer may between two cars of the same price be tempted to buy the one with more power and speed (or more miles per gallon etc.), but on test drive may find the car with the worse numbers is the better car to drive...
So complex multidimensional analysis is unavoidable, but try to get a bunch of people to understand such or to agree on a "single number figure of merit" that is derived from such analysis.
Generally I find people prefer false certainties (the amplifier that has 200W at 0.01% THD is better than the Amplifier that 180W at 0.05% THD) to real uncertainties (without doing a serious listening comparison we cannot conclude if one amplifier is better for listening to music than the other or not).
Ciao T
- Status
- Not open for further replies.