Maybe a better question would be: what does not cause distortion in a BJT based amplifier? 
As the device itself is far from ideal (almost everything depends from something other), unfortunately we are forced to adapt...
And the only chance is to embed the devices to a "protected/relaxed"
(electrical+thermal) environment where their imperfections does not have such a role.
Also at first we have to forget about the GNFB as a holy grail as it does not solve all of these problems.
My algorythm would be something like this:
As the device itself is far from ideal (almost everything depends from something other), unfortunately we are forced to adapt...
And the only chance is to embed the devices to a "protected/relaxed"
(electrical+thermal) environment where their imperfections does not have such a role.
Also at first we have to forget about the GNFB as a holy grail as it does not solve all of these problems.
My algorythm would be something like this:
- collect all known BJT errors/imperfections together in a list
- prioritize (but I'm afraid that step in itself could lead to a religion war )
- collect all known solutions for each problem + a collective brainstorming could be interesting
- try to integrate as much as possible based on the prioritized list in one system
- or have multiple paralell versions crosschecked with eachother
- simplify (occam)
- goto 6;
Last edited:
Douglas Self did that kind of analysis more than 20 years ago. Buy his book or read the article on the web.
Short version: power output stages suffer from crossover-related distortion, but are usually low-distortion otherwise. Most of the distortion occur in the voltage gain stage(s).
-JS
Short version: power output stages suffer from crossover-related distortion, but are usually low-distortion otherwise. Most of the distortion occur in the voltage gain stage(s).
-JS
Current SOTA of solid state amplifiers is pushing the magnitude of all harmonics below the noise floor of a very low noise amplifier. Most designers doing this have applied Doug Self's techniques of linearizing each stage of the amplifier, plus a few tricks of their own, while maintaining a very large open-loop gain. This means inaudible. Really.
I contend that the proponents of zero global NFB have no idea what they are talking about. They think if a little feedback is bad, then more is worse. Incorrect. Feedback compensation is a noise/distortion shaping method, that translates nonlinearity and noise to higher frequencies. A small amount of feedback only manages to move the problems to a higher location in the audible band.
A large amount of feedback moves the problems out of band, especially if a high loop gain can be maintained to at least 20kHz.
I contend that the proponents of zero global NFB have no idea what they are talking about. They think if a little feedback is bad, then more is worse. Incorrect. Feedback compensation is a noise/distortion shaping method, that translates nonlinearity and noise to higher frequencies. A small amount of feedback only manages to move the problems to a higher location in the audible band.
A large amount of feedback moves the problems out of band, especially if a high loop gain can be maintained to at least 20kHz.
Here are some quotes from Douglas Self's book.
These quotes show that Douglas Self is not himself an audio subjectivist and considers subjectivism as irrational. They also show, he regards negative feedback as a great way of realising highly performant, and at the same time, cheaply produced amplifiers.
Sincerily, I am perplexed reading posts in these fora very strongly defending subjectivist views with some even going to the extreme of entirely denying objective measurement.
May I ask, is Negative Global Feedback INHERENTLY BAD, or is it only, a delusion? The application of negative feedback in control systems is good as it is a simple way of achieving stability. However, when audio is concerned we have to deal with the perception of hearing, and perceptions are purely subjective as they cannot be shared directly with third parties.
I have seen the amplifier using no global feedback by, if I rememeber well, Valerie. My impression is that it still using feedback from the output, but instead of feeding it into the inverting input, it uses an intermediate opamp to compare the signal differentially with the actual signal at the non-inverting input. It then feeds the opamp output to feed the VAS.
Please, don't tell me it is all opinions, especially, making the claim that if it is my opinion that GNFB is bad, nothing can be done. I do not hold mere opinions highly, I make an effort to be open to facts. This is why I am asking. Douglas Self is a researching amplifier engineer who definitely knows his area of expertise.
D. Self said:
These quotes show that Douglas Self is not himself an audio subjectivist and considers subjectivism as irrational. They also show, he regards negative feedback as a great way of realising highly performant, and at the same time, cheaply produced amplifiers.
Sincerily, I am perplexed reading posts in these fora very strongly defending subjectivist views with some even going to the extreme of entirely denying objective measurement.
May I ask, is Negative Global Feedback INHERENTLY BAD, or is it only, a delusion? The application of negative feedback in control systems is good as it is a simple way of achieving stability. However, when audio is concerned we have to deal with the perception of hearing, and perceptions are purely subjective as they cannot be shared directly with third parties.
I have seen the amplifier using no global feedback by, if I rememeber well, Valerie. My impression is that it still using feedback from the output, but instead of feeding it into the inverting input, it uses an intermediate opamp to compare the signal differentially with the actual signal at the non-inverting input. It then feeds the opamp output to feed the VAS.
Please, don't tell me it is all opinions, especially, making the claim that if it is my opinion that GNFB is bad, nothing can be done. I do not hold mere opinions highly, I make an effort to be open to facts. This is why I am asking. Douglas Self is a researching amplifier engineer who definitely knows his area of expertise.
Last edited:
D. Self only measure output voltages because 0.0005% acoustic distortion at full power is hard to believe. Normal humans can not perceive audio directly from output voltage of amplifiers but hear the acoustic output, so relevant distortion measurement should be obtained from the acoustic output. THD of amplifier output voltage can only show part of the picture but generates a lot of unnecessary confusion from incomplete interpretation on significance of the measured quantity.
This one looks half decent independently measured:
MC2 Audio MC650 Amplifier Test Results :: ABELtronics
It's true. Amplifiers at this point are contributing almost nothing to the total distortion. I have some former co-workers at Purifi Audio in Denmark that have done a great job in reducing loudspeaker distortion. Someone here on diyaudio interviewed Lars Risbo on YouTube. He did a great job of explaining how they have lowered loudspeaker distortion.
They are not actually that complicated, when taken one concept at a time. A diff pair has very low distortion, so long as the input voltage is kept low, and the pair is kept balanced. The concept is easy; the solutions may appear complicated. A VAS stage, at its worst, has a pure exponential relationship between its input and output voltage. But if it's considered as a transresistance amplifier, it can be quite linear. It's a simple idea, but techniques are sometimes elaborate.
Output stages (in B and AB amps) have to hand off to the other half of the output transistors whenever the output current crosses zero. Class A is a simple, but power-wasting solution. Class B distortion can be made ultra-low, but it requires a lot of care, and sometimes some complication.
As I said in my previous post, and you seem to understand some control theory, feedback, of the type applied in power amplifiers, has very strong correction at audio frequencies. An amplifier with a typical gain of 40 with feedback may have an gain of 100,000 without the feedback. Any nonlinearities are divided by 2,500. That ratio my drop as frequency increases, but distortion at 20 kHz (which would be inaudible anyway) is still reduced by at least a factor of 100 by the global feedback. Several people are getting linearity correction far higher.
Real constant current sources do not deliver a constant current, but a voltage dependent current, with a relation like the following:
Code:
output_current = basic_current + voltage * current_incrementThis formula is emperical from my observations.
Cascodes provide more or less a contant voltage, but when this is applied to the collector of a differential pair, the latter would float with the input voltage, with the result, the collectors of the differential pair do not see a constant voltage but a voltage modulated by the input signal. The same can be written about a VAS's transistor with emitter degeneration, which is often the case.
What do you understand by phase margin?
The pertinent question is, how the inherent non-linearity of active devices, is defied to such an extent, of obtaining extreme linearity?
There must be something which is escaping me. Maybe, it can be understood mathematically, but how?
Without this understanding, amplifier design is easily seen as magic, but Douglas Self, does not hold such an opinion. In his book he is categorical in stating amplifier design is accessible to anyone wanting to follow a few rules of thumb.
Last edited:
Hi edbarx,
The new W-function can give exact equations for a few simple circuits in audio amps like the common emitter stage with degeneration and shunt feedback.
The maths is intensive even for simple circuits; it may interest some engineers.
Most diyer's prefer Douglas' rules of thumb.
Which level are you looking for?
BTW Have you check your PM recently?
I differentiate between amps and these are the closest thing to a 'wire with gain' I've come across. I completely forget there existence except that things get louder when I turn the volume knob and that is the best thing I can say about any piece of audio gear.
Amplifier cannot be heard. You can hear how the amplifier works in conjunction with specific speakers or headphones. With a specific source. With specific connecting cables and connectors. With certain conditions of electromagnetic interference and mains interference. etc.
Sometimes you can hear a transformer humming or an SMPS beeping
Sometimes you can hear a transformer humming or an SMPS beeping
Last edited:
Thank you for taking the interest to send me a private message. I am somewhat disappointed with the transistor formula as stated by Ebers-Moll. This only applies to small signals, which is often not the case. Having a formula which can also model large signal analysis is definitely advantageous.
I will have a look at the reading resources you suggested me. If knowledge of vectors in three dimensions, phasors, complex numbers and calculus are enough, I should be able to survive.
I will have a look at the reading resources you suggested me. If knowledge of vectors in three dimensions, phasors, complex numbers and calculus are enough, I should be able to survive.
For large signals, it is required to take characteristics for specific modes on test installations: curve tracer, measurement of harmonic distortion, slope, etc.
There are no good or bad types of transistors. There are more or less suitable for a particular application in terms of the combination of properties.
There are no good or bad types of transistors. There are more or less suitable for a particular application in terms of the combination of properties.
Last edited:
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.