"...don't see how you can dismiss it's importance."
I'm not dismissing it exactly. My view is that its a factor, but not necessarily as critical a factor as some may suppose. Why? From experience: I have seen what some people can do with audio design despite not having teenage-girl SPL threshold FR. Apparently, the brain learns to compensate in some way. Give it some time and the science will keep advancing though.
I'm not dismissing it exactly. My view is that its a factor, but not necessarily as critical a factor as some may suppose. Why? From experience: I have seen what some people can do with audio design despite not having teenage-girl SPL threshold FR. Apparently, the brain learns to compensate in some way. Give it some time and the science will keep advancing though.
Last edited:
I can't believe that in all my years on this forum, I haven't made it over to this thread yet. Well here I am. Prepare to be schooled. Just kidding.
However, I do have an observation with regard to measurements that I would like to bring up.
Over the years, I have had a subscription to Stereophile, and have read each issue pretty much cover to cover, as is my way. John Atkinson does measurements of most components under review, and these measurements are included as part of the reviews. A little while back, I had a look at their online review database for amplifiers, which allows for quick comparisons of the measurements between many different amplifiers.
Atkinson measures THD at output wattage from .1 watts to beyond the clipping point for each amplifier. I looked at these measurements and noticed that certain topologies tend to have distortion that rises as the output wattage drops towards zero, and other topologies have distortion that continues to drop as output wattage drops.
Atkinson measures down to .1 watts. In my setup, using La Scala speakers close to the wall, and at 8 feet away from the speakers, I get around 92 DB at the listening position with .1 watts of input. At .001 watts, I get around 72 DB at the listening position. 70 DB would be equivalent to a"loud radio" or "city street" or "vacuum cleaner at 1M" according to a variety of SPL charts online. Distortion at this SPL is obviously relevant, but not measured.
No one seems to measure distortion at low wattage. This is no doubt due to the high power amplifiers and low sensitivity speakers that were typical for the last part of the 20th century. Many ask what the output at clipping is, few if any ask what the low output linearity is like.
In Atkinson's measurements, you can look at the trend measured near his cutoff of .1 watts, and extend the chart to see what could be happening below his cutoff. I looked at a few dozen of his measurements, and certain trends became apparent. One is that push pull topologies tend to have rising distortion as the output drops. Another is that single ended topologies tend to have distortion that continues to drop as output drops. There are exceptions of course, but that is the fairly obvious trend shown in the dozens of measurements of his that I looked at. Please go over to Stereophile.com and have a look for yourselves.
Is this where the SQ VS Measurements division will be made whole?
However, I do have an observation with regard to measurements that I would like to bring up.
Over the years, I have had a subscription to Stereophile, and have read each issue pretty much cover to cover, as is my way. John Atkinson does measurements of most components under review, and these measurements are included as part of the reviews. A little while back, I had a look at their online review database for amplifiers, which allows for quick comparisons of the measurements between many different amplifiers.
Atkinson measures THD at output wattage from .1 watts to beyond the clipping point for each amplifier. I looked at these measurements and noticed that certain topologies tend to have distortion that rises as the output wattage drops towards zero, and other topologies have distortion that continues to drop as output wattage drops.
Atkinson measures down to .1 watts. In my setup, using La Scala speakers close to the wall, and at 8 feet away from the speakers, I get around 92 DB at the listening position with .1 watts of input. At .001 watts, I get around 72 DB at the listening position. 70 DB would be equivalent to a"loud radio" or "city street" or "vacuum cleaner at 1M" according to a variety of SPL charts online. Distortion at this SPL is obviously relevant, but not measured.
No one seems to measure distortion at low wattage. This is no doubt due to the high power amplifiers and low sensitivity speakers that were typical for the last part of the 20th century. Many ask what the output at clipping is, few if any ask what the low output linearity is like.
In Atkinson's measurements, you can look at the trend measured near his cutoff of .1 watts, and extend the chart to see what could be happening below his cutoff. I looked at a few dozen of his measurements, and certain trends became apparent. One is that push pull topologies tend to have rising distortion as the output drops. Another is that single ended topologies tend to have distortion that continues to drop as output drops. There are exceptions of course, but that is the fairly obvious trend shown in the dozens of measurements of his that I looked at. Please go over to Stereophile.com and have a look for yourselves.
Is this where the SQ VS Measurements division will be made whole?
At 0,1W for 70 dB at listening position and from a "high" distorsion amp, say -80 dB down..., ehh that's 10 dB under "silence".... what's the problem...?
"A sound at 0 dB SPL is just loud enough to be heard. It is roughly equivalent to the sound of a mosquito flying 10 feet (3 m) away."
Interpreting THD Measurements - Think dB not Percent! - Benchmark Media Systems
//
"A sound at 0 dB SPL is just loud enough to be heard. It is roughly equivalent to the sound of a mosquito flying 10 feet (3 m) away."
Interpreting THD Measurements - Think dB not Percent! - Benchmark Media Systems
//
Hi chris,
I actually don't like the culture of suspicion, so I assume all audio companies have at least the tools in the hands of a couple of poor hobbyists.
Unless someone demonstrate the opposite.
But let me go ahead.
I have listed some instruments, do you think they are suitable to make proper measurements?
Andrea
I actually don't like the culture of suspicion, so I assume all audio companies have at least the tools in the hands of a couple of poor hobbyists.
Unless someone demonstrate the opposite.
But let me go ahead.
I have listed some instruments, do you think they are suitable to make proper measurements?
Andrea
Stereophile’s review of the Benchmark AHB2... Benchmark Media Systems AHB2 power amplifier Measurements | Stereophile.com shows rising distortion as output drops, but the distortion level is extremely low, and much lower than most other amps. That’s an amp I would like to own one day.
Look at this amp on the other hand... Mark Levinson No.333 power amplifier Measurements | Stereophile.com Low output distortion is, to put it plainly, terrible.
Here is another one to look at... Sophia Electric 91-01 300B monoblock power amplifier Measurements | Stereophile.com for an example of dropping distortion.
Which one of these amps do you want driving your 100+ DB/W/M speakers? I want the AHB2, but I’ll make do with a DIY 300B amp for now.
Who cares about distortion and other parameters at high wattage? Show me what an amp is doing at the wattages that I use daily. In my case, .0001 to 1 watt.
is that the right approach? With instruments in had you fix problems the instrument shows. However the "when you have a hammer all problems are nails" approach usually gets nowhere. You get an instrument to address and identifiable problem; e.g. jitter is identified so you need a tool to measure the jitter. However not all problems are jitter so you need other tools. Then you need to know how to use the tools to give you information, not just data.
Here is an interesting example- we can measure distortion to a ridiculous degree (-170 dB is my limit). But it doesn't tell us if the frequency response is flat. How flat does it need to be? There are anecdotal references to .1 dB over a range being audible. That can be addressed. I don't see many measurements showing the composite system response if only to the speaker leads. Let alone response variations with level. Lots more about distortion or in some cases unique "unknowable" measurements but rarely frequency response. I would suggest addressing the parameters you know how to address and are recognized as issues with known thresholds of audibility (e.g. frequency response and distortion) before getting lost in special issues of obscure and debatable audibility. Just fix the basics first.
I have an almost absurd collection of instruments able to measure many parameters to near the current state of the art. Most have seen only a few applications and collect dust. I think the value is knowing when to use them and when to move to the next real issue. Real world solutions are the results of lots of compromises.
Here is an interesting example- we can measure distortion to a ridiculous degree (-170 dB is my limit). But it doesn't tell us if the frequency response is flat. How flat does it need to be? There are anecdotal references to .1 dB over a range being audible. That can be addressed. I don't see many measurements showing the composite system response if only to the speaker leads. Let alone response variations with level. Lots more about distortion or in some cases unique "unknowable" measurements but rarely frequency response. I would suggest addressing the parameters you know how to address and are recognized as issues with known thresholds of audibility (e.g. frequency response and distortion) before getting lost in special issues of obscure and debatable audibility. Just fix the basics first.
I have an almost absurd collection of instruments able to measure many parameters to near the current state of the art. Most have seen only a few applications and collect dust. I think the value is knowing when to use them and when to move to the next real issue. Real world solutions are the results of lots of compromises.
It's often not so much distortion at lower levels but the noise contribution (therefore the "+N" in "THD+N), which inevitably leads to a rising figure.
+1
In addition, an open mind is needed; one of the reason why I quite regularly quoted from Frindle's 1997 AES-Presentation/Paper.
It helps a lot if listeners report specific effects, but even unspecific ones "don't know what it is, but ...." can be a good starting point.
The hypothesis "_all_ those people are deluded" is .... well, just a hypothesis.
Those THD+N plots are misleading (Data, not information). They do not differentiate between distortion products and noise. Its not possible to extract the distortion products alone with those plots. Further the relationship between the different harmonics and level. Not to metion that the Levinson and Sophia seem to start around the same midband THD+N at 100 mW but go opposite directions.
I have a friend who has been quite successful designing high end amps. Good solid state stuff. He has a control to add HD2 in controlled amounts. He is convinced of its importance and his customers prefer the "optimized" distortion levels. Do you want accuracy or beauty? Reality is not always pretty.
Perhaps, regards timbre etc, the brain adjusts and finds it pleasing even without the high frequency content due to gradually getting used to the sound and not missing what can no longer be heard. That's why it seems to me that listening to music/audio is not a good test of hearing ability and speech intelligibility is a better one.
No doubt it’s a combination of factors that lead to our listening preferences. The contribution of noise on a THD+N measurement should be taken into account, but more importantly, should be tested for and shown to the degree possible with a given set of measurement equipment. Distortion and noise at plainly audible output levels is relevant, and every attempt should be made to measure and quantify it. Too much emphasis has been put on measurements on the opposite end of the wattage spectrum. It being difficult to measure something doesn’t make it irrelevant or unimportant. When we have difficulty squaring SQ and measurement, it’s not enough to point to the inherent failings of our monkey brains, you also have to look at the ways that existing measurements are failing us.
So true. And not all has experienced it so would not recognise it and probably spontaneously frown upon it. Many (most?) seek "nice/good sound"... (some) distortion is like suger....
//
Thank you, Indra.
Yes, was plotted the signal input vs. signal output, with a synthetic load (R+L, with self resonance at ca. 200KHz). I use two dummy loads: one simulating a woofer or TT with 2.7mH, and a separate one for mids and tweeter (MT nad HT) with 700uH, both 8 Ohms).
I choose +/-80W for 60W just for the sake of having dynamic room. You are right, but I personally prefer to have the peak SPLs well below IP3 (hence keep IMD3 low for transients).
I usually (professionally) choose triangular signal for validations: easier to see that everything is right up to at least H5, compared to square wave or pulses. In particular, phase shift. Also in particular, shows if any stresses at input modules are sufficiently good handled (saturation, linearity, transient dynamics etc).
People evaluate such hi-fidelity systems as flat. Then I thought to solve this by adding a controlled level of even distortion and flair, by tube-based preamp.
What do you think, without the preamp, will be pleasant enough, good image, good transient, good phase?
People here are scaring us: all perfect measuring systems are going to sound "flat", by definition.
Last edited:
- Status
- Not open for further replies.