There are those who agree with you, gm doubling or no gm doubling. At minimum, I always prefer to err on the side of over-bias than under-bias.
Cheers,
Bob
I am. It IS a bad thing. Harmonics above the third are innatural.
Anybody is entiled his own opinion.
I think we all agree that distortion products above the third are bad. What we disagree on is what makes those high-order products high or low in amplitude.
Anyone who has done any serious simulations or, better yet, measurements on real amplifiers, while changing amounts of NFB in an apples-apples way, knows that NFB works, and that, when properly applied, REDUCES high-order products. This stuff is readily measureable. It is not about who hears what. Once you make the argument that NFB makes high-order products worse, you are in the realm of hypotheses that are readily tested by measurement or simulation.
Cheers,
Bob
Guys, you've got to stop focusing solely on harmonics. Think IM. The amplifier is not passing a single instrument anyway. Just because a violin makes harmonics does not mean that 5th order nonlinearities are OK. Think IM.
Cheers,
Bob
Lumba,
You are just making intuitive generalizations that are not valid. How much time have you spent looking at a spectrum analyzer? You seem to be just repeating a lot of anecdotal stuff without technical basis. Yes, there are some grains of truth to what you say; yes, closing feedback around many stages requires more care. Just don't generalize and throw the baby out with the bathwater.
Cheers,
Bob
Good morning,
tonality is a delicate issue. The harmonics constitute a oneness, coherence and integrity, any alteration disturbs the tonal balance and will be reflected in the audible range, no matter how high the distortion products are pushed up and no matter how small your frequency range of hearing is.
Unfortunately all amplifiers distort (especially the blameless-types), fortunately, the ear distorts pretty heavily as well. If the patterns coincide, no distortion will be detected, congratulations, you have created a transparent sounding masterpiece (in sharp contrast to the blameless-types).
Sounds made from enormously complex waveforms, consisting of many thousands of harmonics that the ear/brain is able to perceive and interpret with an almost incredible sensitivity and sophistication. Dusty coarse measurement protocols are supposed to match up to that ability. Amplifiers need to handle even more complex music signals under harsh real-life conditions. From that perspective, making steady-state measurements with pure sine waves appears ridiculous. Keep your expectations relatively tiny.
tonality is a delicate issue. The harmonics constitute a oneness, coherence and integrity, any alteration disturbs the tonal balance and will be reflected in the audible range, no matter how high the distortion products are pushed up and no matter how small your frequency range of hearing is.
Unfortunately all amplifiers distort (especially the blameless-types), fortunately, the ear distorts pretty heavily as well. If the patterns coincide, no distortion will be detected, congratulations, you have created a transparent sounding masterpiece (in sharp contrast to the blameless-types).
Sounds made from enormously complex waveforms, consisting of many thousands of harmonics that the ear/brain is able to perceive and interpret with an almost incredible sensitivity and sophistication. Dusty coarse measurement protocols are supposed to match up to that ability. Amplifiers need to handle even more complex music signals under harsh real-life conditions. From that perspective, making steady-state measurements with pure sine waves appears ridiculous. Keep your expectations relatively tiny.
I dont know. I listen to every theory until proved wrong.
My ears are very sensitive to distortion but not much by phase. But i also look at the distortion figures before going to the listening phase. Or after, trying to understand why an amplifier sounds that way.
In the example on a previous post i made, i DID compare the two amplifiers and i DID prefer the one with lowest overall distortion and quite high NFB, in everything but in the HF. Output devices were not BJT in both cases. Apples to oranges maybe, but it made me think, a lot.
And yes, I clearly hear that 0,1% 2nd armonic (For Bob, IMD was extremely low).
I think I disagree on the bolded part above. If the distortions are pushed low enough they do merge into the noisefloor. Why shouldnt they?
Point is that the noise floor must be low for a start, damn low.Hmm this sounds weird
BTW, even if the topic is about BJT, I try to think of any possible output device available.
I hope so
I actually do not know it, so I dont disagree with you.
I do not have your expertise and i candidly admit it
Leaving armonics aside, the time smearing effect of GNFB is still undebated. I find this topic much more interesting and worth digging into it.
Not only that, but I'd like to talk about the slew rate of the output stage and the real (measurable also wih impulses) effects.
I remember somebody said that bipolars are "slow" devices. Does it depends on the transistors themselves or more on the circuit used (i propend on the latter)?
Last edited:
Hello TelstarI hope so
Leaving Harmonics aside, the time smearing effect of GNFB is still undebated. I find this topic much more interesting and worth digging into it.
How can you debate this.
Regards
Arthur
Regarded the miller effect, I agree. This I have observed by the Mark Levinson No 23 (see attached schematic, especially the predriver voltage gain stage - schematic created some years ago for repair service and modify).
For reducing these audible distortions (causes through large values for miller caps in the second gain stage), it was create a low-pass filter with too low upper frequency limit. After reduction this capacity at normal values each would say, the idle current through the output stage isn't enough (virtually Class-B by Class-A in the lower output power range)
Hey there are 4 inches of snow on the ground here this morning. Very pretty, very unusual.
A few points to keep us on solid ground:
1) It is very important not to confuse phase shift and time delay. These are completely different things. A phase shift is an inertial effect - the output responds to the input instantaneously, it is just that the relative magnitude and phase of input and output signals, usually when expressed as sinewaves, varies with frequency. A time delay is when the output is shifted in time relative to the input, usually a fixed delay. The two effects are quite different. If you erroneously consider phase shift to be the same as delay then you will have an incorrect conception of the behaviour of NFB. I recommend folks use the terms phase shift/lead/lag rather than phase delay to avoid confusion.
2) The application of NFB to a non-linear circuit, be it around a single component or around 1000 components, will both change the relative amplitudes of harmonics and create new ones. So although stable NFB will usually reduce the total amount of harmonics it will change the amplitudes and spread of the harmonics.
3) Stable NFB reduces total distortion. That is to say it is a ratiometric thing. The distortion is reduced by a ratio which is approximately the 1/loop_gain at a particular frequency. This is not the same thing as cancellation which is a summation thing - subtracting an inverse error signal from the output as in feed-forward. As a result, NFB can never eliminate error. Because NFB is regenerative (output is added to input) it is dogged by stability problems and often saturation effects (clipping, slew limiting). These effects are exaggerated when the circuit contains non-linear elements.
4) In actual circuits on the bench that use NFB it is quite difficult to be sure the circuit is stable. Oh yes. You can measure the phase shift and plot the gain and do all those usual things and a circuit can look perfectly stable. And yet it is not. In real circuits there are many ways in which instabilities can arise that are not at all obvious and are hard to excite with simple signals and loads.
Is NFB a bad thing? Not necessarily. It can be a very good thing. But there are lots of ways in which it can make things worse so it has to be applied with much care. So when someone claims that their circuit sounds worse when NFB is applied don't assume they are lying or deaf...take a moment to consider the possibility that they are right and the application of NFB is doing something unwanted.
A few points to keep us on solid ground:
1) It is very important not to confuse phase shift and time delay. These are completely different things. A phase shift is an inertial effect - the output responds to the input instantaneously, it is just that the relative magnitude and phase of input and output signals, usually when expressed as sinewaves, varies with frequency. A time delay is when the output is shifted in time relative to the input, usually a fixed delay. The two effects are quite different. If you erroneously consider phase shift to be the same as delay then you will have an incorrect conception of the behaviour of NFB. I recommend folks use the terms phase shift/lead/lag rather than phase delay to avoid confusion.
2) The application of NFB to a non-linear circuit, be it around a single component or around 1000 components, will both change the relative amplitudes of harmonics and create new ones. So although stable NFB will usually reduce the total amount of harmonics it will change the amplitudes and spread of the harmonics.
3) Stable NFB reduces total distortion. That is to say it is a ratiometric thing. The distortion is reduced by a ratio which is approximately the 1/loop_gain at a particular frequency. This is not the same thing as cancellation which is a summation thing - subtracting an inverse error signal from the output as in feed-forward. As a result, NFB can never eliminate error. Because NFB is regenerative (output is added to input) it is dogged by stability problems and often saturation effects (clipping, slew limiting). These effects are exaggerated when the circuit contains non-linear elements.
4) In actual circuits on the bench that use NFB it is quite difficult to be sure the circuit is stable. Oh yes. You can measure the phase shift and plot the gain and do all those usual things and a circuit can look perfectly stable. And yet it is not. In real circuits there are many ways in which instabilities can arise that are not at all obvious and are hard to excite with simple signals and loads.
Is NFB a bad thing? Not necessarily. It can be a very good thing. But there are lots of ways in which it can make things worse so it has to be applied with much care. So when someone claims that their circuit sounds worse when NFB is applied don't assume they are lying or deaf...take a moment to consider the possibility that they are right and the application of NFB is doing something unwanted.
Last edited:
Hi Brian, nice to you back again
Yes, this is a point I also have been trying to make. People often think that because of phase shift the voltage on a capacitor occurs some time AFTER the current into the cap starts. That is not true, as you say, that cap voltage occurs INSTANTANEOUSLY when the current starts. However because of the phase shift between current and voltage, the rise of the cap voltage lags the current.
jd