I was wondering what i was missing with this terminology, and what is a "NP amplifier." Just googling didn't make sense, but looking for references on diyaudio found some in Pass Labs, then it finally hit me: Nelson Pass.
I was going to mention as a child listening to the Beatles through a 50C5 and a too-small output transformer, but thought it too far off-topic.
Its easy. The amplifier people preferred sounded more like the experience of hearing real music. Maybe people liked the amplifier despite its distortion, not because of it. Maybe it was more tolerant of a difficult load, or more tolerant of cable capacitance, maybe it had better dynamics with complex signals, etc.
Because a scientific experiment doesn't say much about your perceived experience. In case of Geddes' data, those data say that, on average, people tend to prefer low harmonic distortion. There will always be individual variability.
There are also many factors that influence your perception when it comes to sighted trials of hifi gear. Olive & Toole uncovered quite a few of them while at Harman-Kardon. The price tag is one of them. Higher price -> higher perceived performance in sighted trials.
I have his book and the ability to read. I did dig it out and read through Ch. 16 before I posted. It's a pretty short chapter. Doug Self wrote about many, if not all, of those distortion mechanisms in his articles in Wireless World back in the 1990s too, so it's not new material to me.
Ah. Ok. So now we're back to "I have experience, you don't".
That doesn't change the fact that Cordell characterize all those effects mentioned in Ch. 16 using THD vs frequency. He makes no mention of any randomness or non-periodic nature of these effects, so that part appears to be a product of your own imagination.
You are free to question THD if you want to, but you are not free to derail any discussion that involves THD. You've done that several times now and I keep wondering what you are trying to gain from that.
Where do you get the idea that THD is only measured at one particular frequency and load? Have you never seen a THD+N vs frequency graph? How about multiple graphs, each showing a different load. That's done all the time.
Tom
Tom,
You're not listening. Someone in this thread made a statement to the effect of, if THD and noise are good enough then...
There was nothing about any other information being available or being necessary.
In my book that is wrong. THD by itself can vary by orders of magnitude, and a higher number may sound better than a lower number to most people simply because 2nd order is less audibly objectionable than 7th order.
I don't see what is so hard to understand about this.
BTW, the derailing was not done by me, it was done by you and others who don't want to understand what I just said above. You keep arguing, and when I say I agree spectra are better than just a THD number but maybe still not enough, you keep on arguing about the value of THD. When I point out that it can't predict reliably without more information, you point to some study where it happened to follow some pattern which was a negative correlation, but the truth is it doesn't have to follow that pattern. And we know very well why it doesn't have to. Once a scientific theory is proven wrong by some counter example, that theory is out the window. It doesn't matter if by some quirk of experimental methodology the theory sometimes appears to be right.
You're not listening. Someone in this thread made a statement to the effect of, if THD and noise are good enough then...
There was nothing about any other information being available or being necessary.
In my book that is wrong. THD by itself can vary by orders of magnitude, and a higher number may sound better than a lower number to most people simply because 2nd order is less audibly objectionable than 7th order.
I don't see what is so hard to understand about this.
BTW, the derailing was not done by me, it was done by you and others who don't want to understand what I just said above. You keep arguing, and when I say I agree spectra are better than just a THD number but maybe still not enough, you keep on arguing about the value of THD. When I point out that it can't predict reliably without more information, you point to some study where it happened to follow some pattern which was a negative correlation, but the truth is it doesn't have to follow that pattern. And we know very well why it doesn't have to. Once a scientific theory is proven wrong by some counter example, that theory is out the window. It doesn't matter if by some quirk of experimental methodology the theory sometimes appears to be right.
Last edited:
Oh, I am listening. But you keep going around in circles moving the goal post along with you.
Yes. THD can vary by orders of magnitude. And high THD is correlated with a poor perceived listening experience according to the data by Geddes which you suggested I go look at. Others have reached the same conclusion including those I referenced earlier.
Tom
Yes. THD can vary by orders of magnitude. And high THD is correlated with a poor perceived listening experience according to the data by Geddes which you suggested I go look at. Others have reached the same conclusion including those I referenced earlier.
You are free to believe that, but that opinion is not supported by science. If you have a link to a scientific study that looks at the perception of, say, low-order vs high-order harmonic distortion I'm happy to read and critique it with you.
Tom
I have said lots of times that high THD is correlated with poor SQ, it just doesn't work the other way around.
IIRC, according to Geddes up to 10% of 2nd order can be inaudible, while other distortion can be audible at .01%.
EDIT: BTW, its not that I am moving the goalposts. I respond to complaints as they are written. As a conversation moves along, sometimes it becomes more clear what each party's complaint is really about, and then the responses can get more focused on the underlying issues. I think I have been consistent in trying to make clear that THD or THD+N by itself is not particularly useful. However, I will agree as I have in other threads that when THD is poor, SQ also tends to be poor. What I also said before is that once THD gets good enough, then its no longer the most objectionable problem with SQ. Its not like THD is all there is, nor like THD+N is all there is.
EDIT: BTW, its not that I am moving the goalposts. I respond to complaints as they are written. As a conversation moves along, sometimes it becomes more clear what each party's complaint is really about, and then the responses can get more focused on the underlying issues. I think I have been consistent in trying to make clear that THD or THD+N by itself is not particularly useful. However, I will agree as I have in other threads that when THD is poor, SQ also tends to be poor. What I also said before is that once THD gets good enough, then its no longer the most objectionable problem with SQ. Its not like THD is all there is, nor like THD+N is all there is.
Last edited:
I'd welcome a link if you can find the original reference.
What do you mean that "it just doesn't work the other way around"?
If you're trying to say that high THD is just one of many ways of screwing up sound quality then I'd agree with you. But all that means that THD is a relevant measure along with many other measurements.
That contradicts your message in Post #32, which you have posted in many threads now. In Post 32 you're saying that THD is a useless measurement.
Tom
So what else is there in signal processing, which is what we are doing after all...
I agree with you / your findings 100%.
The test sets need to improve, to analyse data that is most important to us humans - NOT to Audio Precision Anyliser current approach to audio measurement.
Over at audiosciencereview, Amir is buying an AP for each of his blind followers... they can look at graphs all day long, and be content knowing they have the best sound possible 🙂 No need to listen... at all... and I am okay with that... whatever keeps them happy.
I do understand where they are coming from... close-minded engineering egos, unwilling to accept other people's opinions, discarding even the slightest possibility that what humans hear (and what we deem to be sounding good to us), is not presented correctly in the current AP test-sets / approach to audio measurements.
However, I do value their (audiosciencereview) contribution to the electronics / audio industry. I know we have to have a test set and some rules and regulations.... otherwise, total anarchy would prevail and 1% distortions of valve amplifiers would become completely acceptable, and listable... all day long.
Last edited:
I agree that Amir's approach is too simplistic and in some cases invalid, at least from a scientific standpoint. I've tried taking this up with him in the past but haven't gotten far with that. For example, I disagree with his stacked ranking of equipment in cases where the measured difference between the pieces of equipment really only reflects the measurement tolerances. I also don't care much for the dogmatic culture that ASR has (d)evolved into.
That said, I do commend Amir for trying to provide some transparency in the audio market. Without measurements all we have is the marketing team's statement that the equipment is the greatest thing ever.
I would hate to just throw science out with the bathwater as some subjectivists seem to suggest.
Tom
That said, I do commend Amir for trying to provide some transparency in the audio market. Without measurements all we have is the marketing team's statement that the equipment is the greatest thing ever.
I would hate to just throw science out with the bathwater as some subjectivists seem to suggest.
Tom
And I agree with you on all (all three) accounts.
I also tried to point to him the flaw in the USB cable measurements he used... but the armada of bullies over there pretty much made him feel good about ignoring what others had to say.
But, I do go there to see the reviews/measurements he (and a few others) do... there's value in what he is doing...
I also tried to point to him the flaw in the USB cable measurements he used... but the armada of bullies over there pretty much made him feel good about ignoring what others had to say.
But, I do go there to see the reviews/measurements he (and a few others) do... there's value in what he is doing...
Possibly the worst examples of ‘science out with the bath water’ can be seen in What HiFi (I call them the Sun newspaper of audio journalism- UK based members will know what that means and it ain’t a compliment) and HiFi+. Both have no measurements but both want payments for a good review.
This isn't what I am claiming. What I am claiming is that DA can produce time shifted artifacts not necessarily restricted to causing harmonics manifest as distortion. It has equivalency to a speaker in a room with a microphone measuring the SPL of a swept frequency. In other words there is no distortion being measured yet time smeared.
"An amplifier that measures 1 ppm distortion that uses capacitors for signal coupling" can be highly questionable to be degraded more so by DA than by an amplifier capable of 1 ppm distortion. Also, even if the DA of components can be measured this doesn't necessarily translate as remotely predictable in network applications.
Herd is an article by Kenneth Kundert (MIT) on DA iao
https://designers-guide.org/modeling/da.pdf
You can also read about dielectric spectroscopy here
https://en.m.wikipedia.org/wiki/Dielectric_spectroscopy
In short, the mechanisms binding electrical charge up in the dielectric is complex (see the charge mechanisms discussed) and frequency dependent. However, modelling DA as an RC network across the main capacitance is adequate for audio purposes. What appears across the DA parasitic capacitance is the average voltage of the applied signal which will be as near as damn it to 0V and this is why you don’t see a measurable distortion component in AC coupling applications.
However, in systems where charge has to be accurately stored, as for example in a dual or quad slope A-D or a S+H, DA is a critical parameter because there is not the opportunity to average out the charge on the parasitic DA capacitance as is the aforementioned case with an AC audio signal.
🙂
https://designers-guide.org/modeling/da.pdf
You can also read about dielectric spectroscopy here
https://en.m.wikipedia.org/wiki/Dielectric_spectroscopy
In short, the mechanisms binding electrical charge up in the dielectric is complex (see the charge mechanisms discussed) and frequency dependent. However, modelling DA as an RC network across the main capacitance is adequate for audio purposes. What appears across the DA parasitic capacitance is the average voltage of the applied signal which will be as near as damn it to 0V and this is why you don’t see a measurable distortion component in AC coupling applications.
However, in systems where charge has to be accurately stored, as for example in a dual or quad slope A-D or a S+H, DA is a critical parameter because there is not the opportunity to average out the charge on the parasitic DA capacitance as is the aforementioned case with an AC audio signal.
🙂
Just to expand upon the above, with an electrolytic which has very high DA, if you size it right there will only be a few mV across it terminals at LF. Then across that few mV you will then have the DA RC network which will in turn be averaging out the small AC signal. So, at audio frequencies, the DA component will in all likelihood be just a few uVs.
Last edited:
Sorry for the off-topic. But sound quality discussions always remind of this
Bill
Your words do exist in a Dictionary.
Their combination means nothing.
Sorry.
I thought so, thanks for clearing it for us.