janneman said:
I know this is off-topic, but indeed, a recent European study where 17.000 cars and their owners took part has showh that Japanese and Korean cars are nowadays the most reliable in the world, despite what all those young car tuners think.
This is a link to a Spanish newspaper mentioning it:
http://www.elmundo.es/elmundomotor/2005/12/19/usuarios/1135019510.html
I have not found an English version, since the study took place in non-english speaking countries (France, Spain, Italy, Portugal and Belgium).
Then again, modifiying your electronic equipment by adding and replacing exotic fashion components is more or less the same as modifying your car by adding and replacing exotic fashion parts.
Greetings from Norfolk
I fully agree with you Jan, I am learning a great deal, and re-learning some things too.
I am very interested in the discussion re. slew rate limiting and FM distortion. I would though consider an amplifier that suffered from slew rate limiting to be fundamentally flawed, as this surely (as I read it) would be a falut in the design of the open loop or forward path. It is a valid distortion to consider, and placing it on the 'menu' of distortions to be evaluated and 'removed' in an amplifier is worth while.
Richard
janneman said:
I fully agree with you Jan, I am learning a great deal, and re-learning some things too.
I am very interested in the discussion re. slew rate limiting and FM distortion. I would though consider an amplifier that suffered from slew rate limiting to be fundamentally flawed, as this surely (as I read it) would be a falut in the design of the open loop or forward path. It is a valid distortion to consider, and placing it on the 'menu' of distortions to be evaluated and 'removed' in an amplifier is worth while.
Richard
Again - greetings from Norfolk
Many aplifiers use a long tailed pair at the input, often with the input signal applied to one transistor and the 'overall' feedback (often both signal and output dc level) applied to the other. This pair of transistors, plus any active loads they may use, are then outside the global feedback loop.
I seem to remember once reading an article which analysed the distortion of a long tailed pair in some detail, but cannot yet find it, but this configuration must be considered as a significant part of any amplifier where it is used.
Richard
Many aplifiers use a long tailed pair at the input, often with the input signal applied to one transistor and the 'overall' feedback (often both signal and output dc level) applied to the other. This pair of transistors, plus any active loads they may use, are then outside the global feedback loop.
I seem to remember once reading an article which analysed the distortion of a long tailed pair in some detail, but cannot yet find it, but this configuration must be considered as a significant part of any amplifier where it is used.
Richard
tHis begs some questions:
What exactly is 'blameless-styled' ? I know what "blameless" is from reading the book, a baseline design used to examine the effect of altering various cicuit configurations. Even the author doesn't actually peddle that specific design.
How much feedback is "high" versus "not-high"?
Who or what is an "audiophile"? Do they carry ID cards? Have secret handshakes?
Gandalph said:
The underlying assumption with fb is that there is a causal relation between input and output. The transfer function may be non-linear, but there IS a reaction from the output when the input changes. When an amp is hard slew limited, that causal relation no longer exists. You can do whatever you like to the input, but it will not effect the output. The fb mechanism fails because the underlying assumption is no longer valid.
Jan Didden
Since I'm not good with math, I try to visualize what is distortion caused by slew rate.
Here I use square wave as signal, because I don't know how to draw a smearing sinusoidal
but the mechanism should be the same.
From what I know, slew rate is like fig A. If slew rate is not enough, when amplifier is fed by perfect square wave (left) it will produce kind of "trapezoidal" output (right), because it needs dT to rise from 0 to 1 and also needed another dT to fall from 1 to 0.
Imagine an amplifier is fed by perfect square like fig B. Because of slew rate limitation, the downscaled of output will be like fig C, trapezoidal with shifted position (notice timing T=1,2,3)
If we go back to the distortion equation, where distortion = downscaled output - input, here if we try to get the distortion, we simply substract fig C (downscaled output) with fig B (input). This results in fig D.
Notice that this substraction result have negative (-) part (where the input is all positive), and notice in fig D, new frequencies (f=1/dT) is formed, the form is even not like square wave at all / triangle (where input is all square wave).
I imagine the same mechanism applied sinusoidal will produce weird artifacts, with higher frequencies too
Here I use square wave as signal, because I don't know how to draw a smearing sinusoidal
but the mechanism should be the same. From what I know, slew rate is like fig A. If slew rate is not enough, when amplifier is fed by perfect square wave (left) it will produce kind of "trapezoidal" output (right), because it needs dT to rise from 0 to 1 and also needed another dT to fall from 1 to 0.
Imagine an amplifier is fed by perfect square like fig B. Because of slew rate limitation, the downscaled of output will be like fig C, trapezoidal with shifted position (notice timing T=1,2,3)
If we go back to the distortion equation, where distortion = downscaled output - input, here if we try to get the distortion, we simply substract fig C (downscaled output) with fig B (input). This results in fig D.
Notice that this substraction result have negative (-) part (where the input is all positive), and notice in fig D, new frequencies (f=1/dT) is formed, the form is even not like square wave at all / triangle (where input is all square wave).
I imagine the same mechanism applied sinusoidal will produce weird artifacts, with higher frequencies too
Attachments
David,
Your fig D is exactly what you will see at the amp input: it is the missing part between the input and output (but the polarity may be the opposite though).
When I said above that at slew rate limiting the output doesn't change when you change the input, of course the output is moving, it is slewing to the required output level as fast as it can but not fast enough. Now if you (of the fb) double the input level, still the output will continue to slew at the existing rate at not suddenly faster. That is what I mean by saying that there is no longer a causal relation between input and output.
If you recall the famous fb equation with the beta * A factor (A being the amp gain), for varying signals (that is what we are concerned with) the factor 'A' becomes effectively zero, another (mathematical) way to show that the fb relationship breaks down. Vout effectively becomes zero (meaning no longer changing as a result from an input change). Theory, practise and mathematics are come to the same conclusion.
There is nothing so practical as a really good theory...
Jan Didden
Your fig D is exactly what you will see at the amp input: it is the missing part between the input and output (but the polarity may be the opposite though).
When I said above that at slew rate limiting the output doesn't change when you change the input, of course the output is moving, it is slewing to the required output level as fast as it can but not fast enough. Now if you (of the fb) double the input level, still the output will continue to slew at the existing rate at not suddenly faster. That is what I mean by saying that there is no longer a causal relation between input and output.
If you recall the famous fb equation with the beta * A factor (A being the amp gain), for varying signals (that is what we are concerned with) the factor 'A' becomes effectively zero, another (mathematical) way to show that the fb relationship breaks down. Vout effectively becomes zero (meaning no longer changing as a result from an input change). Theory, practise and mathematics are come to the same conclusion.
There is nothing so practical as a really good theory...
Jan Didden
Hi, Janneman,
But, assume we are talking about the whole amp (assume 3 stages), where is this bottleneck?
But what if that mechanism is happening in the differential itself? What would happen? Who will fix the differential?
Ah.. I see now. Thanks
But, assume we are talking about the whole amp (assume 3 stages), where is this bottleneck?You are right, if the assumption is that the differential is very fast, and the "slow" one is the output stage, for example, then fig D will be the error signal seen at the base of the differential.
But what if that mechanism is happening in the differential itself? What would happen? Who will fix the differential?
lumanauw said:
Depends on the amp topology. If you are talking about a run-of-the-mill amp with diff pair, Vas, EF, generally the cause is the fact that the diff amp cannot charge the comp cap fast enough. The diff amp is limited to the tail current so charging the miller cap (which normally is connected to the diff amp output) takes finite time. But there are many topologies that avoid this kind (or any) slew rate. Slew rate is a non-issue for a competent design.
Some suggestions: higher tail current, lower gain (emitter degeneration in the diff pair) so the comp cap can be smaller, different design topology (see Otala for instance) etc. This can be a course in itself.
Jan Didden
PS to my previous post: what is gain? Gain = delta Vout / delta Vin. With a slew rate limited amp, the delta Vout as a result of a delta Vin = 0, so gain = 0. The fb equation assumes a non zero 'A' (you are not allowed to divide by zero in a mathematical equation). Yet another confirmation that Feedback Actually Works. It doesn't promis anything it cannot deliver
MikeB said:
Yes, but the problem is, what else does it deliver than the promised ?
For example: Any change in openloop gain will change the phaseshift...
Mike
Sure. What I tried to convey, in my limited way, that fb isn't magic which can run wild.
It is all understood and, in principle, calculable. Although if you really want to do the calculation for a closed loop circuit with all the nitty gritty details like harmonics expansion and phaseshift, you quickly get a hurtin' head.
What really gets me is a statement like: "fb around an amp converts the phaseshift to FM modulation, so that must be why a fb amp sounds bad". That is such an exquisite example of bad thinking one wonders whether that person is actually able to figure out how many dollars he has left from 20 after he spends 10.
Jan Didden
Jan Didden
Ouch !
I will try to "calculate" the amount of FM-distortion caused by dynamic phasehift, but i have the feeling that it is negligible.
I did more investigation... PSRR... Completely underestimated ?
It's a well known effect, the sagging of supplylines. Symetric reduction of the supplylines changes the gain of an amplifier, the voltagedrop multiplied by the PSRR will modulate the rest of the signal actual playing. This results in very nice intermodulation, i attached a shot of an fft showing this behaviour. This is a "complex" signal 10khz+12khz. Now estimating a 20% sagging with a 40db PSRR (not really unrealistic) i modulated this signal with a simultaneous playing 1khz tone. Observe the intermodulation products at 0.1%. Perfect smearing...
This would mean an amp with 0.001% THD must have at least 80db PSRR ?
Additionally, the changed gain will cause dynamic phaseshifts through the nfb, giving FM distortion.
The way out ? H-Bridge with pure ClassA ? Did i get paranoid or is my calculation wrong ? Or is this inaudible ?
Mike
I will try to "calculate" the amount of FM-distortion caused by dynamic phasehift, but i have the feeling that it is negligible.
I did more investigation... PSRR... Completely underestimated ?
It's a well known effect, the sagging of supplylines. Symetric reduction of the supplylines changes the gain of an amplifier, the voltagedrop multiplied by the PSRR will modulate the rest of the signal actual playing. This results in very nice intermodulation, i attached a shot of an fft showing this behaviour. This is a "complex" signal 10khz+12khz. Now estimating a 20% sagging with a 40db PSRR (not really unrealistic) i modulated this signal with a simultaneous playing 1khz tone. Observe the intermodulation products at 0.1%. Perfect smearing...
This would mean an amp with 0.001% THD must have at least 80db PSRR ?
Additionally, the changed gain will cause dynamic phaseshifts through the nfb, giving FM distortion.
The way out ? H-Bridge with pure ClassA ? Did i get paranoid or is my calculation wrong ? Or is this inaudible ?
Mike
Attachments
Sam
Blameless-styled means to me "maximum feedback without messing up (very mych) something else" as far as I remember Self's words.
Audiophile- is the term I honestly don't like very much, just the short-cut of thoughts, guessed everyone will know what i am talking about. Many has written they don't like the sound of blameless. If you find my words as a kind of provocation, then yes, it was my intention to say this controversial thing and get feedback.
Jan
Thanks for answer. But don't you believe there are technical reasons for blaming the blameless? Let's forget about psychology for a moment.
Eva
You are statistically right. Placebo and other kinds of self-suggestions do work for most people. But I wouldn't have much pleasure to fool someone. Too many of them do it themselves, so no one needs to bother.
best regards to all
Adam
Blameless-styled means to me "maximum feedback without messing up (very mych) something else" as far as I remember Self's words.
Audiophile- is the term I honestly don't like very much, just the short-cut of thoughts, guessed everyone will know what i am talking about. Many has written they don't like the sound of blameless. If you find my words as a kind of provocation, then yes, it was my intention to say this controversial thing and get feedback.
Jan
Thanks for answer. But don't you believe there are technical reasons for blaming the blameless? Let's forget about psychology for a moment.
Eva
You are statistically right. Placebo and other kinds of self-suggestions do work for most people. But I wouldn't have much pleasure to fool someone. Too many of them do it themselves, so no one needs to bother.
best regards to all
Adam
darkfenriz said:
Adam,
I am not at all convinced that in a controlled blind test the blameless amp (against whatever other amp) will actually be blamed. What I see is a statement "Technically there are a lot of problems so that's why it sounds so bad". Huh? Who said it sounds bad? Compared to what? There's so much jumping to conclusions and flying off the handle, they should make it an olympic sports!
But yes if it really sounds bad it can only be for technical reasons, of course.
Jan Didden
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.