I got this schematic from an expensive power amp.
There are things I don't understand. It uses 2 differential, the K389 and J109 connecting to input and feed back, then it is re-differentiated by MAT02 and MAT03. The MAT's are not VAS, the VAS are BD139 and BD140.
What is the merit of doing differential-ing a differential? (K389+J109 to MAT02+MAT03) ?
If it is to lower the input capacitance of the FETs, I think it should be done by adding the degenerating resistors on the sources, and putting cascode on the drains. But this circuit is not doing that, so there are maybe other purposes of differentiating the differential. What is it?
There are things I don't understand. It uses 2 differential, the K389 and J109 connecting to input and feed back, then it is re-differentiated by MAT02 and MAT03. The MAT's are not VAS, the VAS are BD139 and BD140.
What is the merit of doing differential-ing a differential? (K389+J109 to MAT02+MAT03) ?
If it is to lower the input capacitance of the FETs, I think it should be done by adding the degenerating resistors on the sources, and putting cascode on the drains. But this circuit is not doing that, so there are maybe other purposes of differentiating the differential. What is it?
Attachments
It is just another stage of gain. It isn't really necessary, and will probably cause problems with stability.
It's anybody's guess...here are some possibilities:
1) Current drive to the BD139/140 is insufficient without an extra stage. The extra stage needn't be a diff pair but might as well be as it helps with 2)
2) JFETs are not very linear. The second diff pair produces a more linear differential current for the BD139/140.
3) JFET Vds rating is not high enough for the power rails. The second stage allows you to set a much lower drain voltage.
4) JFET power dissipation. Same point as 3) but regarding power.
5) To reduce common-mode distortion. If the pos and neg JFETs are not perfectly balanced in terms of sensitivity of drain current to Vdg or Vds change you could end up modulating the bias current through the output stage. This is because of pos and neg asymmetry of the devices.
I'm sure there are more.
1) Current drive to the BD139/140 is insufficient without an extra stage. The extra stage needn't be a diff pair but might as well be as it helps with 2)
2) JFETs are not very linear. The second diff pair produces a more linear differential current for the BD139/140.
3) JFET Vds rating is not high enough for the power rails. The second stage allows you to set a much lower drain voltage.
4) JFET power dissipation. Same point as 3) but regarding power.
5) To reduce common-mode distortion. If the pos and neg JFETs are not perfectly balanced in terms of sensitivity of drain current to Vdg or Vds change you could end up modulating the bias current through the output stage. This is because of pos and neg asymmetry of the devices.
I'm sure there are more.
Like John said there no point to adding another gain stage here.
Maybe, this would be ok it you were running a balanced configuration and needed very high command mode rejection. However, that can be accomplished with precision parts in the first stage.
Happy New Years
Maybe, this would be ok it you were running a balanced configuration and needed very high command mode rejection. However, that can be accomplished with precision parts in the first stage.
Happy New Years
some confusion in terminology here, "differential pair", "differential stage" refer to taking the algebraic difference of 2 inputs and providing a proportional output voltage/current
"differentiate" means to provide an output proportional to the rate of change of the input
the example circuit does both! the MAT pairs are "differentiating differential pairs" which have an output proportional to the rate of change of the difference of the two input signals
the caps across the MAT emitters provide a zero that results in a differentiated ouput over a range of frequency, depending on R,C and gm values this region of "differential"(=differentiating or derivative) gain could cancel the roll off of the integrating miller compensation of the following stage over some frequency range - some designers feel that the open loop gain should be constant over the audio frequency range so that the feedback factor is constant for all audio frequencies
a really interesting possibility is that it could be part of the high frequency compensation, implementing a "Bode Step" - not something i recall seeing in audio amplifier circuits
or the zero-pole pair could both be below the audio frequency range and simply part of some DC bias scheme
(I assume the MAT02 bases really connect to the fet drains)
"differentiate" means to provide an output proportional to the rate of change of the input
the example circuit does both! the MAT pairs are "differentiating differential pairs" which have an output proportional to the rate of change of the difference of the two input signals
the caps across the MAT emitters provide a zero that results in a differentiated ouput over a range of frequency, depending on R,C and gm values this region of "differential"(=differentiating or derivative) gain could cancel the roll off of the integrating miller compensation of the following stage over some frequency range - some designers feel that the open loop gain should be constant over the audio frequency range so that the feedback factor is constant for all audio frequencies
a really interesting possibility is that it could be part of the high frequency compensation, implementing a "Bode Step" - not something i recall seeing in audio amplifier circuits
or the zero-pole pair could both be below the audio frequency range and simply part of some DC bias scheme
(I assume the MAT02 bases really connect to the fet drains)
Quite possibly an example of the "Not invented here" syndrome. The design team just couldn't bear to use an approach that someone else had already developed. The rest of the amp looks very conventional and searching this forum you can find numerous versions of mirrored long-tailed pair input stages that do what needs to be done without gatting clever.
First, Happy New Year to everyone.
Sorry about the drawing, I should draw the connection to MAT02 is from the Drains of K389 below the resistors, not from upper resistors, that is wrong.
The Caps in the emitors of MAT's are only 1nF. All the degeneration resistors (in emitors and sources) are 100ohm.
The VCC1 is 45V, the VCC2 is 35V. The miller cap is 15pF, the current in first differential is 7mA, the second differential is 12mA, the VAS is 35mA (attached in heatsink).
I read that a differential pair do cancel distortion (like left+right K389). Is the same "Distortion Canceling Mechanism" also happened in "up and low" transistors, like left K389 to left J109?
If the "up and low" transistors also cancels distortion, wouldn't it be that the Complementary Differential (like K389 and J109 in this example) have 2 distortion cancelation mechanism, one is from the left and right diffeential transistors, and two, is from upper and lower differential?
If this is correct, would it be adding another complementary differential like the MATs are also adding this "Distortion Canceling Mechanism"?
I have another question. Once I observed an input and output of a power amp. I use a dual trace osciloscope and a signal generator. I sweep from 20hz to 20khz.
From 20hz, the input and output are in phase. But as the frequency rises to 20khz, the output are beginning to shift from the input signal (the trace of the sinusoidal is not the same place between input and output).
What is causing this? Is a good amp don't have this "phase shifting" between output and input? How can we advoid this in design phase?
Sorry about the drawing, I should draw the connection to MAT02 is from the Drains of K389 below the resistors, not from upper resistors, that is wrong.
The Caps in the emitors of MAT's are only 1nF. All the degeneration resistors (in emitors and sources) are 100ohm.
The VCC1 is 45V, the VCC2 is 35V. The miller cap is 15pF, the current in first differential is 7mA, the second differential is 12mA, the VAS is 35mA (attached in heatsink).
I read that a differential pair do cancel distortion (like left+right K389). Is the same "Distortion Canceling Mechanism" also happened in "up and low" transistors, like left K389 to left J109?
If the "up and low" transistors also cancels distortion, wouldn't it be that the Complementary Differential (like K389 and J109 in this example) have 2 distortion cancelation mechanism, one is from the left and right diffeential transistors, and two, is from upper and lower differential?
If this is correct, would it be adding another complementary differential like the MATs are also adding this "Distortion Canceling Mechanism"?
I have another question. Once I observed an input and output of a power amp. I use a dual trace osciloscope and a signal generator. I sweep from 20hz to 20khz.
From 20hz, the input and output are in phase. But as the frequency rises to 20khz, the output are beginning to shift from the input signal (the trace of the sinusoidal is not the same place between input and output).
What is causing this? Is a good amp don't have this "phase shifting" between output and input? How can we advoid this in design phase?
Re Phase shifting: regardless of what may be written somewhere, I think all Lin topology amplifiers (which is probably 75+% of all built) have "phase shift". It is inherent to capacitors and virtually every other component in some dgreee, except resistors (even them too if inductive). The amount increases with frequency but is not normally noticable/measurable below 20k. This topic will lead you in to the mind-numbing discussion of open loop gain, compensation, negative feedback, Nyquist (in)stability and so on.
It's one of those things that just is. There is normally a prejudice against outright phase inversion unless you are deliberately trying to provode a non-inverting/inverting option.
It's one of those things that just is. There is normally a prejudice against outright phase inversion unless you are deliberately trying to provode a non-inverting/inverting option.
Yes, see my comments 2) and 5) above. How well this is achieved depends upon the closeness of matching of the devices.
Caused by electrical inertia. Imagine a yo-yo attached to a length of elastic and your finger. If you lift your finger up and down very slowly the yo-yo will follow it. The faster you move your finger the more the yo-yo falls behind in phase.
ALL amplifiers exhibit this characteristic and it is normal and rare these days that the phase shift, and associated amplitude change, will be big enough to be audibly significant. It is a bit like a tone control effect. This sort of thing is sometimes called "linear distortion" to separate it from distortion that introduces new frequencies into the signal. The latter are the primary villains that you need to eradicate.
So, the phase shift between input and output are "unadvoidable" as long as we have capacitance and inherent inductance inside the whole schematic.
Does anyone have measure/analyze this in tube amps? Is it possible that what makes tube amps is said tobe more "nice" is because it has a certain kind of this phase shift between input and output, that matches our "psycho-acoustic", where SS amps didn't have this? (just guessing).
If we can make the exact phase between input and output of an audio power amp, would it be more "nice" to listen to, or the same as other amp?
Is it posible to pursure the non-phase shift audio amplifier design? How can we do it?
Does anyone have measure/analyze this in tube amps? Is it possible that what makes tube amps is said tobe more "nice" is because it has a certain kind of this phase shift between input and output, that matches our "psycho-acoustic", where SS amps didn't have this? (just guessing).
If we can make the exact phase between input and output of an audio power amp, would it be more "nice" to listen to, or the same as other amp?
Is it posible to pursure the non-phase shift audio amplifier design? How can we do it?
"Does anyone have measure/analyze this in tube amps? Is it possible that what makes tube amps is said tobe more "nice" is because it has a certain kind of this phase shift between input and output, that matches our "psycho-acoustic", where SS amps didn't have this? (just guessing)."
Improbable. It appears pretty well established that for those who think tube amps are nice (and not everyone thinks so!), the most of the "niceness" stems from two factors A: a high ratio of second order harmonics added to the input signal, which some people genuinly enjoy and B: soft clipping behavior when over driven. Undoubtedly there are other factors but these seem to be the main ones.
Improbable. It appears pretty well established that for those who think tube amps are nice (and not everyone thinks so!), the most of the "niceness" stems from two factors A: a high ratio of second order harmonics added to the input signal, which some people genuinly enjoy and B: soft clipping behavior when over driven. Undoubtedly there are other factors but these seem to be the main ones.
I have seen this Kind of input stage and alot of the time the fets are source followers this will buffer the high input bias current of the BJT Diff amps. I think this is why John qualified his asertion with "(I assume the MAT02 bases really connect to the fet drains)" jcx
"So, the phase shift between input and output are "unadvoidable" as long as we have capacitance and inherent inductance inside the whole schematic. "
Correct.
Don't worry about this, your brain is ok with it and there are much bigger fish to fry.
One of which is the reason tubes sound the way they do. It is the non-linear distortion that defines the sound.
Correct.
Don't worry about this, your brain is ok with it and there are much bigger fish to fry.
One of which is the reason tubes sound the way they do. It is the non-linear distortion that defines the sound.
Thanks for all the explenation.
I have soooooo many questions about solid state amps, that I couldn't answer myself.
Sorry, it's a bit out of the main topic, but I have questions
1. About the feedback resistor network (voltage divider). Why this resistor's value range so much. Like one amp uses 100k to output and 10k to ground. Other amp uses 10 ohm to ground.
What is the main consideration to pick the value of this voltage divider? Is it have something to do with how many current have tobe fed to the base of the differential pairs or not? I assume the differential pairs needs some current tobe fed to it (if we use bipolars), so if we use too high value (like 100k) there weren't enough current to fed the base.
2. About differential pairs. It is clearly stated that a differential pair do cancel distortion. But I've seen an input stage that is not built by differential pairs. It is built with 1 n-ch on top and 1 p-ch on bottom. The feedbacks are fed into the emitors (like hiraga amp).
Is this configuration doing the same "distortion canceling mechanism" like the differential pairs? If they do, what is the merit and drawback compared to diff pairs?
3. I look at alephx schematic by Grollins. I try to trace the outputs by feeding a normal input (not balanced) to this schematic. So the +input is fed by signal, but the -input is grounded. It seems that the left amp (the - side) has less gain than the right amp. If it is right, the alephx has a non identical output between its -output and +output (magnitude to ground) if the amp is fed by normal (not balanced) input. If this is wrong, can someone tell me how to simulate the inputs and outputs in alephx to understand how this amp works?
I have soooooo many questions about solid state amps, that I couldn't answer myself.
Sorry, it's a bit out of the main topic, but I have questions
1. About the feedback resistor network (voltage divider). Why this resistor's value range so much. Like one amp uses 100k to output and 10k to ground. Other amp uses 10 ohm to ground.
What is the main consideration to pick the value of this voltage divider? Is it have something to do with how many current have tobe fed to the base of the differential pairs or not? I assume the differential pairs needs some current tobe fed to it (if we use bipolars), so if we use too high value (like 100k) there weren't enough current to fed the base.
2. About differential pairs. It is clearly stated that a differential pair do cancel distortion. But I've seen an input stage that is not built by differential pairs. It is built with 1 n-ch on top and 1 p-ch on bottom. The feedbacks are fed into the emitors (like hiraga amp).
Is this configuration doing the same "distortion canceling mechanism" like the differential pairs? If they do, what is the merit and drawback compared to diff pairs?
3. I look at alephx schematic by Grollins. I try to trace the outputs by feeding a normal input (not balanced) to this schematic. So the +input is fed by signal, but the -input is grounded. It seems that the left amp (the - side) has less gain than the right amp. If it is right, the alephx has a non identical output between its -output and +output (magnitude to ground) if the amp is fed by normal (not balanced) input. If this is wrong, can someone tell me how to simulate the inputs and outputs in alephx to understand how this amp works?
- Status
- Not open for further replies.