Got the feeling that most of you guys measure with just a nice 8 or 4 ohm resistor at the output.
How about a real world dummyload?
How about a real world dummyload?
Interesting, Pavel & thank you for the plots - I really don't have any clue about the topologies but in my naivety I would suspect the purple & blue are feedback designs & the other two are non-feedback. My preference amp would be the one that produced the light blue (Magenta?) plot.
Edit: or are we looking at differences between Class A & Class AB here - I really don't know?
It would be great if plots of other known amplifiers were shown - I presume these were into a 8 ohm dummy load?
Edit: or are we looking at differences between Class A & Class AB here - I really don't know?
It would be great if plots of other known amplifiers were shown - I presume these were into a 8 ohm dummy load?
Yes Pavel.... low power performance is very important to me too
It is unusual to me to reproduce using more than 1 watt RMS average.
This thread is producing interesting results...i am very happy with this thread.
I have to test the 100 Kilohertz clock DAC... i had doubts about.
Carlos
...................................................................................................
How to translate measurements into something we listen... something very subjective someone (maybe Bear) have developed watching square waves... studying effects and results of some waveform...yes, there are traditional analises watching waveforms..but the translation to sound quality is something i could not read, listen or see till now a days.
I was inspecting some "tricky" capacitors into Aksa amplifiers to see the waveform effect...i could not..but i could listen the result...maybe i have used the wrong method to observe into the scope.
Despite i do not think brain can perceive square waves because delay into bio chemical processing, i feel great that someone can perceive things just watching square waves..i think this is possible, something alike virtual tool.... something alike:
" big slew rate numbers results this way "
"square wave showing low frequency losses may result into sonics that way"
"the biggest the bandwidth of an amplifier the best it will reproduce timbre (sound characteristic that let you discover what kind of music instrument is playing).... say...the harshing sound we have in violine..... the scratching sound from Saxophone...maybe into harmonics reproduced you may have the answer"
This can be very productive...it is a pity this is something as result of personal development..something not easy to be transfered to other guys.
Nice that... watching so many amplifiers into the scope with square wave input and listening to music made someone able to translate, at least to himself, what really means those oscilograms
Carlos
It is unusual to me to reproduce using more than 1 watt RMS average.
This thread is producing interesting results...i am very happy with this thread.
I have to test the 100 Kilohertz clock DAC... i had doubts about.
Carlos
...................................................................................................
How to translate measurements into something we listen... something very subjective someone (maybe Bear) have developed watching square waves... studying effects and results of some waveform...yes, there are traditional analises watching waveforms..but the translation to sound quality is something i could not read, listen or see till now a days.
I was inspecting some "tricky" capacitors into Aksa amplifiers to see the waveform effect...i could not..but i could listen the result...maybe i have used the wrong method to observe into the scope.
Despite i do not think brain can perceive square waves because delay into bio chemical processing, i feel great that someone can perceive things just watching square waves..i think this is possible, something alike virtual tool.... something alike:
" big slew rate numbers results this way "
"square wave showing low frequency losses may result into sonics that way"
"the biggest the bandwidth of an amplifier the best it will reproduce timbre (sound characteristic that let you discover what kind of music instrument is playing).... say...the harshing sound we have in violine..... the scratching sound from Saxophone...maybe into harmonics reproduced you may have the answer"
This can be very productive...it is a pity this is something as result of personal development..something not easy to be transfered to other guys.
Nice that... watching so many amplifiers into the scope with square wave input and listening to music made someone able to translate, at least to himself, what really means those oscilograms
Carlos
Yes Carlos,
very interesting thread with some real meat to it rather than continual bickering. Thank you Lumba for starting it.
I made a mistake in my colours in my last email - I meant I prefer the light blue (cyan) plot. We all know men don't know colours - peach is a fruit, jade is a stone, teal is a type of bird - no wonder we're confused
very interesting thread with some real meat to it rather than continual bickering. Thank you Lumba for starting it.
I made a mistake in my colours in my last email - I meant I prefer the light blue (cyan) plot. We all know men don't know colours - peach is a fruit, jade is a stone, teal is a type of bird - no wonder we're confused
Cyan is nice to me too.
What makes an amplifier good?
This is the hell question... i do not know...and i do not know if someone has this knowledge.
So, despite measurements and evoluted technology we do not have answer for the most fundamental question.
well.... i have not... if you have please tell me.
Carlos
................................................................................................
I have some ideas alike the transistor bias point...i felt 575 milivolts into VBE as something very good.
Also i found stable voltage to the input resulting great too.
I have found removing emitter resistances as sound turns better into bass
Something i have found..but about measurements and results i have not the knowledge.
I have noticed good measurements do not means nice sound.... not always...just sometimes.
I have perceived speakers as a big problem too... and crossovers (passive) as a hell thing.
What makes an amplifier good?
This is the hell question... i do not know...and i do not know if someone has this knowledge.
So, despite measurements and evoluted technology we do not have answer for the most fundamental question.
well.... i have not... if you have please tell me.
Carlos
................................................................................................
I have some ideas alike the transistor bias point...i felt 575 milivolts into VBE as something very good.
Also i found stable voltage to the input resulting great too.
I have found removing emitter resistances as sound turns better into bass
Something i have found..but about measurements and results i have not the knowledge.
I have noticed good measurements do not means nice sound.... not always...just sometimes.
I have perceived speakers as a big problem too... and crossovers (passive) as a hell thing.
Hi Carlos
My main focus is Tone, Transparency and Dynamics.
Sadly, instruments are not able to measure these.
Re: Back in the 80's
The MOSFET amp you bought surely did not have 0.001% THD at 20 kHz. The spec you are referring to is almost certainly 0.001% at 1 kHz (assuming you recall it correctly). This very different, and much easier to achieve. It sounds like you have possibly fallen into the trap of 1 kHz THD specs.
Cheers,
Bob
mosfets said:
The MOSFET amp you bought surely did not have 0.001% THD at 20 kHz. The spec you are referring to is almost certainly 0.001% at 1 kHz (assuming you recall it correctly). This very different, and much easier to achieve. It sounds like you have possibly fallen into the trap of 1 kHz THD specs.
Cheers,
Bob
Hey guys,
There is something which may well be really important just above your posts from Pavel & then confirmed by Hugh Dean - the "Spectral distortion at low power into reactive loads is crucial."
This may well be one of the keys for unlocking the conundrum of measurement correlation to sound so let's try retain a focus on this until we can make some conclusions.
For instance. Pavel, would you care to give a synopsis of the sound of the 4 different amps in the graph without saying which is which & let people try to match the plot to the synopsis? Then we could later talk about topology that gave rise to each distinctive plot.
Hugh, without revealing any sensitive commercial information, do you have any similar plots that might illustrate the same?
There is something which may well be really important just above your posts from Pavel & then confirmed by Hugh Dean - the "Spectral distortion at low power into reactive loads is crucial."
This may well be one of the keys for unlocking the conundrum of measurement correlation to sound so let's try retain a focus on this until we can make some conclusions.
For instance. Pavel, would you care to give a synopsis of the sound of the 4 different amps in the graph without saying which is which & let people try to match the plot to the synopsis? Then we could later talk about topology that gave rise to each distinctive plot.
Hugh, without revealing any sensitive commercial information, do you have any similar plots that might illustrate the same?
PMA said:
Pavel,
You are quite correct, but that is no reason to throw out THD measurement. It IS reason to have it measured more diligently and in ways that properly include extremely low levels and exclude potentially masking noise.
Quite a number of bipolar class AB power amplifiers leave their class A region at as little as 50 mW into 4 ohms when everything is working right, and at that point are in the region of creating potential crossover distortion. When they are under-biased, either statically or dynamically, the situation is worse.
When reviewers measure THD+N and see it rise at lower power levels, they automatically assume it is noise coming up, and that is unfortunate.
Crossover and other nasty low-level distortion is quite insidious, and may be audible at levels as low as 0.01% THD, even when buried in noise. Noise is less annoying because it is largely uncorrelated with the signal. Our ears are adept at perceiving correlated impairments even when under the noise.
It would indeed be nice if reviewers would feed the distortion residual output of their THD analyzer into a spectrum analyzer under the conditions of these low test levels; they might be surprized at what they find.
Cheers,
Bob
Re: Re: Back in the 80's
The other trick is to so limit the bandwidth of the measurement that the observations are irrelevant -- 20kHz stresses a lot of amplifiers --
I must confess to using the 400Hz High Pass filter which makes it so much easier to separate out "hum" issues from "amplifier" issues. One problem at a time, please.
Bob Cordell said:
The other trick is to so limit the bandwidth of the measurement that the observations are irrelevant -- 20kHz stresses a lot of amplifiers --
I must confess to using the 400Hz High Pass filter which makes it so much easier to separate out "hum" issues from "amplifier" issues. One problem at a time, please.
PMA said:
Hi Pavel,
Nice work with those plots. It would be nice to see what the residual looks like, both in the time domain and in spectra.
Cheers,
Bob
Bob Cordell said:
Hi Bob,
if you do not mind, I'll give a link in my language:
http://hifi.slovanet.sk/bb/viewtopic.php?p=74467#74467
I believe the spectra are self-explanatory. They are for (50mV) 100mV, 200mV, 500mV, 1V and 2Vrms (5V) into 4ohm, 4 different amplifiers. But different measuring equipment and some are unfortunately cut below 1kHz.
Hi Pavel
In their measurements you used a loadspeaker, more the signal level is very low (1W) its speakers have difficulty in reproducing very low level, because is necessary overcome the mechanical strength of the speakers diaphragm. The reproduction of a transient signal would not be perfect as you measured.
Talking to a designer here in my country, he has a theory that to be considered an amplifier "High-End" must have frequency response in 20Khz, phase plane and distortion less than 1% in open loop....logical and low negative feedback too.
I do not know if this is true ?
In their measurements you used a loadspeaker, more the signal level is very low (1W) its speakers have difficulty in reproducing very low level, because is necessary overcome the mechanical strength of the speakers diaphragm. The reproduction of a transient signal would not be perfect as you measured.
Talking to a designer here in my country, he has a theory that to be considered an amplifier "High-End" must have frequency response in 20Khz, phase plane and distortion less than 1% in open loop....logical and low negative feedback too.
I do not know if this is true ?
Bob,
All the nonlinear distortion measurement standards are very old and have been widely criticized for their many flaws, especially for providing poor correlation with the perceived quality of reproduced sound. Several concrete proposals for improvement have been presented over the years, some of them taking perception and detectability of distortion and signal frequency content into account. These will unlikely be adopted by the audio industry, primarily intending measurement data for marketing purposes, however, there`s no apparent reason for DIY people to fool themselves.
I have posted this before illustrating the problem of THD metric, showing two signals having the same THD level. The distortion characteristics correspond very well with the subjective rating.
I may have done that in more detail elsewhere, not sure.
All the nonlinear distortion measurement standards are very old and have been widely criticized for their many flaws, especially for providing poor correlation with the perceived quality of reproduced sound. Several concrete proposals for improvement have been presented over the years, some of them taking perception and detectability of distortion and signal frequency content into account. These will unlikely be adopted by the audio industry, primarily intending measurement data for marketing purposes, however, there`s no apparent reason for DIY people to fool themselves.
Very valuable, it´s essentially about spectrum, not level, that`s my point.
My understanding says that the value of simulation in that regard is zero, because the conjugate variables, position and momentum of a moving particle cannot be determined simultaneously. In other words, spectral components are not traceable at any given time instant.
I have posted this before illustrating the problem of THD metric, showing two signals having the same THD level. The distortion characteristics correspond very well with the subjective rating.
Attachments
Hi John,
I'd prefer not to post details of my work. I am a practical designer, not much into the theory, and my measurement is quite basic. My work is deductive, based largely on psychoacoustic ideas, and listening tests form a large part of my assessement. To be frank, I like to avoid the fighting that punctuates these discussions and I haven't the time or training to examine the finer points of the math.
I agree emphatically with Lumba on his notions of distortion spectra. I think it's fundamental. Here's an FFT from LTSpice at 20KHz, 45mV peak input sine wave, loaded by 8R//100nF. It shows an interesting and extremely good sounding distortion profile with very low odd order distortion. H3 here, the large dip just to the right of the fundamental, is -73.5dB below 0dB, with all subsequent harmonics very, very low. This is a very low output; gain of the amp is 30.5dB, the output is 1.5V peak, a listening level where fine detail is present.
I'd be happy to discuss things privately.
Hugh
I'd prefer not to post details of my work. I am a practical designer, not much into the theory, and my measurement is quite basic. My work is deductive, based largely on psychoacoustic ideas, and listening tests form a large part of my assessement. To be frank, I like to avoid the fighting that punctuates these discussions and I haven't the time or training to examine the finer points of the math.
I agree emphatically with Lumba on his notions of distortion spectra. I think it's fundamental. Here's an FFT from LTSpice at 20KHz, 45mV peak input sine wave, loaded by 8R//100nF. It shows an interesting and extremely good sounding distortion profile with very low odd order distortion. H3 here, the large dip just to the right of the fundamental, is -73.5dB below 0dB, with all subsequent harmonics very, very low. This is a very low output; gain of the amp is 30.5dB, the output is 1.5V peak, a listening level where fine detail is present.
I'd be happy to discuss things privately.
Hugh
Attachments
Thank you Hugh,
I understand your care in avoiding the quarrelling that drains energy & frankly swamps the often significant nuggets of information that appears here at times often from you and other skilful & thoughtful audio designers that pass by from time to time.
I understand your care in avoiding the quarrelling that drains energy & frankly swamps the often significant nuggets of information that appears here at times often from you and other skilful & thoughtful audio designers that pass by from time to time.
I think this may be useful to those who might want to pursue the psycho-acoustic direction. See Table of Contents and selected text.
http://books.google.com/books?id=tk...a=X&oi=book_result&resnum=5&ct=result#PPP1,M1
http://books.google.com/books?id=tk...a=X&oi=book_result&resnum=5&ct=result#PPP1,M1
Hugh , add this as the main directive in your lt.
.param num_fft_pts 16384
.param freq 20k
.param timestep {1/((num_fft_pts-1)*freq)}
.tran 0 10m 0 {timestep}
.options plotwinsize=0
.four {freq} V(c)
add this to your generator (ac source) "voltage"
{sqrt(.01)}
and "freq"
{freq}
make your output "C"
turn off all "compression" in "control panel/compression"
It seems as through your H3/5/7 are "hiding" beneath the
noise floor of the simulator.
It should look like this..(attached) Your "type" of topology
should be at least -85db H2 or better.. 🙂
Attachments
- Status
- Not open for further replies.